#### *Frontiers Copyright Statement*

*© Copyright 2007-2017 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.*

*The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.*

*Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.*

*Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.*

*As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.*

> *All copyright, and all rights therein, are protected by national and international copyright laws.*

*The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use.*

ISSN 2296-7745 ISBN 978-2-88945-126-5 DOI 10.3389/978-2-88945-126-5

#### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

#### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

# Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view.

By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

# **BRIDGING THE GAP BETWEEN POLICY AND SCIENCE IN ASSESSING THE HEALTH STATUS OF MARINE ECOSYSTEMS, 2nd EDITION**

Topic Editors: **Angel Borja,** AZTI-Tecnalia, Spain **Michael Elliott,** University of Hull, UK **María C. Uyarra,** AZTI-Tecnalia, Spain **Jacob Carstensen,** Aarhus University, Denmark **Marianna Mea,** Ecoreach srl, Italy & Jacobs University of Bremen, Germany

DEVOTES researchers activities: collecting environmental data, sampling, analysing samples in the lab and organising training workshops.

Cover photos courtesy of: AZTI-Tecnalia, Spain (photos 1-2, 6-7), CEFAS, United Kingdom (photo 3), CoNISMa, Italy (photo 4), SYKE, Finland (photo 8), and SZN, Italy (photo 5)

**Citation:** Borja, A., Elliott, M., Uyarra, M. C., Carstensen, J., Mea, M., eds. (2017). Bridging the Gap Between Policy and Science in Assessing the Health Status of Marine Ecosystems, 2nd Edition. Lausanne: Frontiers Media. doi: 10.3389/978-2-88945-126-5

# Table of Contents

*08 Editorial: Bridging the Gap between Policy and Science in Assessing the Health Status of Marine Ecosystems*

Angel Borja, Michael Elliott, María C. Uyarra, Jacob Carstensen and Marianna Mea

*11 Grand Challenges in Marine Ecosystems Ecology* Angel Borja

#### **Legal framework of marine activities and management**

*18 A Typology of Stakeholders and Guidelines for Engagement in Transdisciplinary, Participatory Processes*

Alice Newton and Michael Elliott


Samuli Korpinen and Jesper H. Andersen

*75 Uses of Innovative Modeling Tools within the Implementation of the Marine Strategy Framework Directive*

Christopher P. Lynam, Laura Uusitalo, Joana Patrício, Chiara Piroddi, Ana M. Queirós, Heliana Teixeira, Axel G. Rossberg, Yolanda Sagarminaga, Kieran Hyder, Nathalie Niquil, Christian Möllmann, Christian Wilson, Guillem Chust, Ibon Galparsoro, Rodney Forster, Helena Veríssimo, Letizia Tedesco, Marta Revilla and Suzanna Neville

#### **Better monitoring, better assessment**


Joana Patrício, Sally Little, Krysia Mazik, Konstantia-Nadia Papadopoulou, Christopher J. Smith, Heliana Teixeira, Helene Hoffmann, Maria C. Uyarra, Oihana Solaun, Argyro Zenetos, Gokhan Kaboglu, Olga Kryvenko, Tanya Churilova, Snejana Moncheva, Martynas Bucˇ as, Angel Borja, Nicolas Hoepffner and Michael Elliott

#### *116 Implementing and Innovating Marine Monitoring Approaches for Assessing Marine Environmental Status*

Roberto Danovaro, Laura Carugati, Marco Berzano, Abigail E. Cahill, Susana Carvalho, Anne Chenuil, Cinzia Corinaldesi, Sonia Cristina, Romain David, Antonio Dell'Anno, Nina Dzhembekova, Esther Garcés, Joseph M. Gasol, Priscila Goela, Jean-Pierre Féral, Isabel Ferrera, Rodney M. Forster, Andrey A. Kurekin, Eugenio Rastelli, Veselka Marinova, Peter I. Miller, Snejana Moncheva, Alice Newton, John K. Pearman, Sophie G. Pitois, Albert Reñé, Naiara Rodríguez-Ezpeleta, Vincenzo Saggiomo, Stefan G. H. Simis, Kremena Stefanova, Christian Wilson, Marco Lo Martire, Silvestro Greco, Sabine K. J. Cochrane, Olga Mangoni and Angel Borja

*141 Evaluation of Alternative High-Throughput Sequencing Methodologies for the Monitoring of Marine Picoplanktonic Biodiversity Based on rRNA Gene Amplicons*

Isabel Ferrera, Caterina R. Giner, Albert Reñé, Jordi Camp, Ramon Massana, Josep M. Gasol and Esther Garcés


Eva Aylagas, Ángel Borja, Xabier Irigoien and Naiara Rodríguez-Ezpeleta

*177 Historical Data Reveal 30-Year Persistence of Benthic Fauna Associations in Heavily Modified Waterbody*

Ruth Callaway

*190 The Application of Long-Lived Bivalve Sclerochronology in Environmental Baseline Monitoring*

Juliane Steinhardt, Paul G. Butler, Michael L. Carroll and John Hartley

*216 High Frequency Non-invasive (HFNI) Bio-Sensors As a Potential Tool for Marine Monitoring and Assessments*

Hector Andrade, Jean-Charles Massabuau, Sabine Cochrane, Pierre Ciret, Damien Tran, Mohamedou Sow and Lionel Camus

*226 Microplastics in Seawater: Recommendations from the Marine Strategy Framework Directive Implementation Process*

Jesus Gago, Francois Galgani, Thomas Maes and Richard C. Thompson

#### **Indicators to assess the status**

*233 Biodiversity in Marine Ecosystems—European Developments toward Robust Assessments*

Anna-Stiina Heiskanen, Torsten Berg, Laura Uusitalo, Heliana Teixeira, Annette Bruhn, Dorte Krause-Jensen, Christopher P. Lynam, Axel G. Rossberg, Samuli Korpinen, Maria C. Uyarra and Angel Borja

#### *253 A Catalogue of Marine Biodiversity Indicators*

Heliana Teixeira, Torsten Berg, Laura Uusitalo, Karin Fürhaupter, Anna-Stiina Heiskanen, Krysia Mazik, Christopher P. Lynam, Suzanna Neville, J. German Rodriguez, Nadia Papadopoulou, Snejana Moncheva, Tanya Churilova, Olga Kryvenko, Dorte Krause-Jensen, Anastasija Zaiko, Helena Veríssimo, Maria Pantazi, Susana Carvalho, Joana Patrício, Maria C. Uyarra and Àngel Borja

*269 An Objective Framework to Test the Quality of Candidate Indicators of Good Environmental Status*

Ana M. Queirós, James A. Strong, Krysia Mazik, Jacob Carstensen, John Bruun, Paul J. Somerfield, Annette Bruhn, Stefano Ciavatta, Eva Flo, Nihayet Bizsel, Murat Özaydinli, Romualda Chuševe˙, Iñigo Muxika, Henrik Nygård, Nadia Papadopoulou, Maria Pantazi and Dorte Krause-Jensen


Monaca Noble, Gregory M. Ruiz and Kathleen R. Murphy

*325 Long-term Patterns of Eelgrass* **(Zostera marina)** *Occurrence and Associated Herbivorous Waterbirds in a Danish Coastal Inlet*

Thorsten J. S. Balsby, Preben Clausen, Dorte Krause-Jensen, Jacob Carstensen and Jesper Madsen

*339 Predicting the Composition of Polychaete Assemblages in the Aegean Coast of Turkey*

Marika Galanidi, Gokhan Kaboglu and Kemal C. Bizsel


Sergej Olenin, Aleksas Naršcˇius, Stephan Gollasch, Maiju Lehtiniemi, Agnese Marchini, Dan Minchin and Greta Sre˙baliene˙

# **Assessing the status in an integrative way**

*377 What Is Marine Biodiversity? Towards Common Concepts and Their Implications for Assessing Biodiversity Status*

Sabine K. J. Cochrane, Jesper H. Andersen, Torsten Berg, Hugues Blanchet, Angel Borja, Jacob Carstensen, Michael Elliott, Herman Hummel, Nathalie Niquil and Paul E. Renaud

*391 Tales from a Thousand and One Ways to Integrate Marine Ecosystem Components When Assessing the Environmental Status*

Angel Borja, Theo C. Prins, Nomiki Simboura, Jesper H. Andersen, Torsten Berg, Joao-Carlos Marques, Joao M. Neto, Nadia Papadopoulou, Johnny Reker, Heliana Teixeira and Laura Uusitalo

*411 Overview of Integrative Assessment of Marine Systems: The Ecosystem Approach in Practice*

Angel Borja, Michael Elliott, Jesper H. Andersen, Torsten Berg, Jacob Carstensen, Benjamin S. Halpern, Anna-Stiina Heiskanen, Samuli Korpinen, Julia S. Stewart Lowndes, Georg Martin and Naiara Rodriguez-Ezpeleta

#### *431 Integrated Assessment of Marine Biodiversity Status Using a Prototype Indicator-Based Assessment Tool*

Jesper H. Andersen, Karsten Dahl, Cordula Göke, Martin Hartvig, Ciarán Murray, Anna Rindorf, Henrik Skov, Morten Vinther and Samuli Korpinen

*439 Indicator-Based Assessment of Marine Biological Diversity–Lessons from 10 Case Studies across the European Seas*

Laura Uusitalo, Hugues Blanchet, Jesper H. Andersen, Olivier Beauchard, Torsten Berg, Silvia Bianchelli, Annalucia Cantafaro, Jacob Carstensen, Laura Carugati, Sabine Cochrane, Roberto Danovaro, Anna-Stiina Heiskanen, Ville Karvinen, Snejana Moncheva, Ciaran Murray, João M. Neto, Henrik Nygård, Maria Pantazi, Nadia Papadopoulou, Nomiki Simboura, Greta Sre˙baliene˙, Maria C. Uyarra and Angel Borja

#### **Socio-economic aspects and management**

*461 Assessing Costs and Benefits of Measures to Achieve Good Environmental Status in European Regional Seas: Challenges, Opportunities, and Lessons Learnt*

Tobias Börger, Stefanie Broszeit, Heini Ahtiainen, Jonathan P. Atkins, Daryl Burdon, Tiziana Luisetti, Arantza Murillas, Soile Oinonen, Lucille Paltriguera, Louise Roberts, Maria C. Uyarra and Melanie C. Austen


Ibon Galparsoro, Angel Borja and María C. Uyarra

# **Lessons learnt**

*507 From Science to Policy and Society: Enhancing the Effectiveness of Communication*

Marianna Mea, Alice Newton, Maria C. Uyarra, Carolina Alonso and Angel Borja

*524 Bridging the Gap between Policy and Science in Assessing the Health Status of Marine Ecosystems*

Angel Borja, Michael Elliott, Paul V. R. Snelgrove, Melanie C. Austen, Torsten Berg, Sabine Cochrane, Jacob Carstensen, Roberto Danovaro, Simon Greenstreet, Anna-Stiina Heiskanen, Christopher P. Lynam, Marianna Mea, Alice Newton, Joana Patrício, Laura Uusitalo, María C. Uyarra and Christian Wilson

# Editorial: Bridging the Gap between Policy and Science in Assessing the Health Status of Marine Ecosystems

Angel Borja<sup>1</sup> \*, Michael Elliott <sup>2</sup> , María C. Uyarra<sup>1</sup> , Jacob Carstensen<sup>3</sup> and Marianna Mea<sup>4</sup>

*<sup>1</sup> AZTI-Tecnalia, Marine Research Division, Pasaia, Spain, <sup>2</sup> Institute of Estuarine and Coastal Studies, University of Hull, Hull, UK, <sup>3</sup> Department of Bioscience, Aarhus University, Roskilde, Denmark, <sup>4</sup> Ecoreach, Ancona, Italy*

Keywords: environmental status, monitoring, modeling, assessment, marine strategy framework directive

**Editorial on the Research Topic**

#### **Bridging the Gap between Policy and Science in Assessing the Health Status of Marine Ecosystems**

There is a continuing requirement in all environments for science to inform policy and policy to inform science and these interactions have created an expanding and fast-moving field. Furthermore, new research and new policy requirements continually change the demands both on policy makers and scientists and both groups need to be well-informed about their own and other fields. Marine management is no different from that in any other environment, albeit perhaps more complex and interrelated, and as such it requires approaches which bring together the best research from the natural and social sciences. It requires stakeholders to be well-informed by science and to work across administrative and geographical boundaries, a feature especially important in the inter-connected marine environment. It also requires us to be clear regarding the nature and role of stakeholders, especially if all groups are to be engaged to achieve a sustainable marine system which can deliver a healthy ecosystem and the economically-based "Blue Growth" required by society.

Given these demands, marine management must ensure that the natural structure and functioning of ecosystems is maintained to provide ecosystem services. Thus, once provided by ecosystem processes, the ecosystem services can lead to the delivery of societal goods and other benefits as long as society inputs human complementary assets such as its skills, time, money and energy to gather those benefits. The economic benefits obtained from the seas thereby constituting Blue Growth, which is currently demanded by policy-makers and politicians worldwide. However, if sufficient societal goods and other benefits are to be obtained, society requires appropriate administrative, legal and management mechanisms (i.e., the right laws and management agencies) to ensure that exploiting such benefits does not impact on environmental quality, but instead supports the sustainable use of our seas.

Therefore to achieve the goal of "Bridging the Gap Between Policy and Science in Assessing the Health Status of Marine Ecosystems" there is the need to find a common ground in which scientists should advance their science and provide policy makers with the best available and timely knowledge. This cannot be achieved without a sound and detailed knowledge and interpretation of the functioning of marine ecosystems. Hence, policy makers, recognizing the complexity and vulnerability of this system, should, through informed decisions, establish and implement a framework for the environmentally sustainable exploitation of the seas by society. To ensure that this is achieved, adequate, timely and fit-for-purpose monitoring is needed. For that monitoring to be meaningful and effective, it should be carried out against both quantitative and qualitative indicators and using fit-for-purpose methods. This in turn will enable scientists and managers to determine trends in the system and assess whether previously implemented management actions are successful.

Edited and reviewed by: *Stelios Katsanevakis, University of the Aegean, Greece*

> \*Correspondence: *Angel Borja aborja@azti.es*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *05 January 2017* Accepted: *27 January 2017* Published: *08 February 2017*

#### Citation:

*Borja A, Elliott M, Uyarra MC, Carstensen J and Mea M (2017) Editorial: Bridging the Gap between Policy and Science in Assessing the Health Status of Marine Ecosystems. Front. Mar. Sci. 4:32. doi: 10.3389/fmars.2017.00032*

The assessment and management of large marine areas is particularly challenging given the transboundary nature of marine problems and uses but is required especially to deliver Blue Growth and expand the Blue Economy. Despite the importance of this, given current economic restrictions, all of this has to be achieved in a cost-effective and cost-beneficial manner. However, it is particularly notable that many countries do not now have (or are not willing to commit) the sufficient financial resources to fully assess the state of the marine environment.

With all of this in mind, in 2012, EU policy-makers and regulators funded a research project on the "DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status" (DEVOTES: http://www. devotes-project.eu), under the 7th Framework Programme "The Ocean of Tomorrow" Theme. The funders required that the expected impacts from accepted proposals should "contribute to the implementation of the Marine Strategy Framework Directive (MSFD) and associated Commission Decision on Good Environmental Status (GES) and strengthen the knowledge base necessary to address sustainable management of seas and oceans resources." Hence, any selected project was required to contribute to bridge the gap between policy (i.e., the MSFD) and science (in this case, the creation of indicators, models, assessment tools, etc.), by increasing the knowledge necessary to assess the marine environmental status in an effective manner.

Given the interrelated nature of the features of seas and transnational marine management, tackling the marine problems requires a multidisciplinary team covering multinational continental sea areas. In the case of European seas, this needed a focus on strong collaboration among European institutions, regional seas as well as overseas partners, to achieve the much-needed synergies in research. Hence, the DEVOTES project encompassed 295 scientists from 23 institutes and 15 countries, including observers from the United States and an Advisory Board with members from Canada, the European Commission and the European Regional Seas Conventions. Further collaboration with other European and national projects was initiated during the 4 year lifespan of DEVOTES (2012–2016) (see Mea et al. in this eBook). As a measure of its wide reach, this internal and external collaboration has resulted in 32 Ph.D students trained, 4 stakeholder workshops, 9 scientific sessions organized in international conferences, 27 post-graduate training courses, 6 training courses on the tools developed, 4 summer schools, 424 contributions to conferences, and to date over 180 scientific papers, 31 of which are included in this ebook (see details in Mea et al.).

Successful scientific dissemination and the wider use of the science carried out requires a commitment to publishing our research in open access outlets, and in making our results available to scientists, stakeholders, policy-makers and the society at large. As such, all DEVOTES deliverables are publicly available (http://www.devotes-project.eu/deliverablesand-milestones/), as are the software and tools produced under DEVOTES (http://www.devotes-project.eu/software-and-tools), and all our papers are in gold or green open access (http://zenodo. org/collection/user-devotes-project). However, with the aim of bridging the communication gap between science and policy, the scientific knowledge generated in DEVOTES has also been communicated to policy makers through policy briefs, local press releases, fit-for-purpose workshops/webinars and conferences, etc. (see details in Mea et al.).

With the above in mind, we took the view that a Research Topic in Frontiers in Marine Science would be an ideal platform for synthesizing and promoting up-to-date research in marine science and management. Accordingly, this led to this volume giving the results from DEVOTES as well as other projects developing tools to improve marine management, and putting these into a global context. Therefore, this volume allowed the scientific community to contribute their research worldwide to advance the knowledge on assessing health status of marine ecosystems. Hence, this Research Topic is the result of this effort by including investigations from the DEVOTES project published in Frontiers in Marine Science between 2014 and 2016 (Andersen et al., 2014; Borja, 2014; Borja et al., 2014, 2016; Carstensen, 2014; Galparsoro et al., 2014), together with new syntheses and reviews (Cochrane et al.; Smith et al.; Lynam et al.; Danovaro et al.; Heiskanen et al.; Teixeira et al.; Mea et al.; Borja et al., 2016; Borja et al.) and original research (e.g., Newton and Elliott; Korpinen and Andersen; Patricio et al.; Patricio et al.; Ferrera et al.; Aylagas et al.; Aylagas et al.; Queiros et al.; Uusitalo et al.). We have also included studies from external research groups which complement the DEVOTES studies (Chartrand et al.; Callaway; Gago et al.; Dietl et al.; Noble et al.).

Following the production of the First Edition of this eBook, and as a reflection of the fast-moving and innovative nature of the field covered, other contributions have now been added to this greatly expanded Second Edition. As with the First Edition, which has been well-received, with plenty of downloads, the contributions are structured as follows. Firstly, the Introduction explains the background of the Research Topic and introduces the grand challenges in marine ecosystems ecology (Borja, 2014), some of which have been addressed within the DEVOTES project and so are included in this eBook.

Secondly, we give the legal and administrative framework of marine activities and management, including the efforts made in the past 20 years in developing a unified framework for marine management (Patricio et al.); the conceptual models used in managing the marine environment (Smith et al.), and a global review of cumulative pressure and impact assessment (Korpinen and Andersen). This section also includes the first published typology of stakeholders involved in marine environmental management (Newton and Elliott) as well as guidance for stakeholder involvement. Finally, we have included some modeling tools required to implement the MSFD (Lynam et al.).

Thirdly, the need for fit-for-purpose monitoring is especially shown by first understanding and assessing current European Marine Biodiversity Monitoring Networks (Patricio et al.), then developing innovative monitoring methods. The latter includes the use of new molecular methods in monitoring picoplankton (Ferrera et al.) and macroinvertebrates (Aylagas et al.); the use of historical data in studying benthic fauna (Callaway); the application of sclerochronology to monitoring (Steinhardt et al.); the development of a new biosensor as an early warning signal of pollution (Andrade et al.), and the monitoring of microplastics (Gago et al.). All of this is presented with the aim of ensuring that we can monitor the sustainable provision of marine ecosystem services (Carstensen, 2014) and so we have included all innovative methods developed under the DEVOTES project (Danovaro et al.).

Fourthly, there is the need for good monitoring data linked to indicators to assess the environmental status of marine ecosystems. Hence, we present a review of the current use of indicators in Europe (Heiskanen et al.; Teixeira et al.), together with an objective framework to test the quality of candidate indicators of good environmental status (Queiros et al.). However, indicators need adequate thresholds, and so these are described, for example, in a study on thresholds to prevent dredging impacts on seagrasses (Chartrand et al.). With regard to other contemporary marine challenges, the assessment of ballast water exchange compliance is discussed (Noble et al.) as is the food-web assessment in the Baltic Sea (Lehtinen et al.). The link between seagrasses and seabirds is presented (Balsby et al.) together with the prediction of the composition of polychaete assemblages (Galanidi et al.) and mollusc assemblages (Dietl et al.). Furthermore, we have included the development of a new non-indigenous species indicator (Olenin et al.).

Fifthly, a solid framework is required to assess environmental status in an integrative way but by focusing on the central theme of biodiversity protection. The latter requires a good and accepted understanding of biodiversity (Cochrane et al.). The aim to create such an integrated assessment requires us to consider different ways in which multiple ecosystem components can be integrated in holistic evaluations (Borja et al., 2014) and a review of currently available methods to undertake such integrated assessments (Borja et al., 2016). In addition, we present the innovative basis for a new assessment tool (Andersen et al., 2014) and this new tool (Nested Environmental status Assessment Tool: NEAT) was tested in 10 case studies across all European seas (Uusitalo et al.).

Sixthly, we emphasize that the socio-economic perspective of this work deserves attention as well as the ability of marine habitats to provide ecosystem services, which in turn provide societal benefits, as presented by Galparsoro et al. (2014). Furthermore, there is the need to assess the cost and benefits of measures to achieve Good Environmental Status (Börger et al.) and to determine and present the value of marine monitoring (Nygård et al.).

All of the above emphasize that scientific understanding and research is only valuable once disseminated to its users,

#### REFERENCES


especially those beyond the scientific community and because of this we emphasize the need to improve the two-way knowledge transfer between researchers and policy makers. Therefore, we present ways to enhance the effectiveness of research results communication (Mea et al.) and show how DEVOTES has contributed to filling in the gaps between policy and science for assessing the health status of marine systems, including the main challenges for the future (Borja et al.).

This ebook with its extensive collection of papers is aimed at scientists and policy makers and implementers, at educators needing to communicate such up to date aspects to the next generation of scientists and policy makers, and at industry which has to respond to the requirements of marine policy. Although the contributions are the result of a European project with predominantly European workers, we consider that the findings, lessons and messages will be of high relevance to those working in other geographical systems and areas. We hope that all readers of this eBook will find the collection of peer-reviewed papers useful in their daily work, through selecting appropriate indicators, implementing and improving monitoring networks, modeling marine systems, or assessing the status in an integrative way. As such, we hope that this eBook conveys and disseminates the outcome of the DEVOTES collaborative and multidisciplinary work to a broad audience, including scientists, policy-makers, environmental managers, stakeholders and the public in general. Although bridging science and policy will always remain a challenge, our hope is that with this eBook the gap has been reduced. We thank all the contributors and are confident that you will enjoy reading these papers as much as we did writing them!

# AUTHOR CONTRIBUTIONS

AB wrote the first draft and then all co-authors contributed equally to the final draft.

# FUNDING

This editorial is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), http://www.devotes-project.eu.

Carstensen, J. (2014). Need for monitoring and maintaining sustainable marine ecosystem services. Front. Mar. Sci. 1:33. doi: 10.3389/fmars.2014.00033

Galparsoro, I., Borja, A., and Uyarra, M. C. (2014). Mapping ecosystem services provided by benthic habitats in the European North Atlantic Ocean. Front. Mar. Sci. 1:23. doi: 10.3389/fmars.2014.00023

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2017 Borja, Elliott, Uyarra, Carstensen and Mea. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Grand challenges in marine ecosystems ecology

#### *Angel Borja\**

*AZTI-Tecnalia, Marine Research Division, Pasaia, Spain \*Correspondence: aborja@azti.es*

#### *Edited by:*

*Alberto Basset, University of Salento, Italy*

**Keywords: biodiversity, functionality, human pressures, global change, ecosystems health, ecosystem services, conservation and protection, ecosystem-based management**

#### **INTRODUCTION**

The study of marine ecosystems has become a hot research topic in recent times. In fact, the number of manuscripts including the words "marine ecosystems" published since 1970 has immensely increased reaching between 1100 and 1500 articles per year in the past five years (**Figure 1**). Based on the keywords used in these manuscripts, the most frequent topics can be grouped into: (i) marine ecosystems (28.8% of the papers); (ii) biodiversity (26.6%), used as keyword at any level of organization, such as bacteria, phytoplankton, zooplankton, benthos, fishes, mammals, seabirds, etc.; (iii) functionality (10.7%), including aspects such as ecosystem function, biomass, foodwebs, primary and secondary production, etc.; (iv) environmental research (9.7%), including pollution, environmental monitoring, human pressures, impacts, etc.; (v) structural parameters (6.6%) such as abundance, richness, diversity; (vi) climate change (3.4%); (vii) ecology (3.4%); (viii) systems management (3.2%); (ix) genetic and genomic issues (1.6%); (x) protection (1%); (xi) ecosystem modeling (0.9%); and (xii) others (4.5%).

Taking into account the large number of papers published in recent years, several grand challenges can be identified for future research within the field of marine ecosystem ecology and as outlined below.

#### **GRAND CHALLENGE 1: UNDERSTANDING THE ROLE OF BIODIVERSITY IN MAINTAINING ECOSYSTEMS FUNCTIONALITY**

Currently, the global species extinction rate far exceeds that of speciation, this difference being the primary driver for change in global biodiversity (Hooper et al., 2012). The rate of biodiversity loss is one of the 10 planetary boundaries within which humanity can operate safely that has already been exceeded (Rockström et al., 2009). The effects of this global decline in biodiversity provide evidence of its importance in sustaining ecosystem functioning and services and preventing ecosystems

from tipping into undesired states (Folke et al., 2004).

Historically, researchers have investigated ecosystems focusing on individual or few components of biodiversity, i.e., microbes, phytoplankton, zooplankton, macroalgae, macroinvertebrates, fishes, mammals, seabirds, etc., trying to understand individual species' roles. However, it is now recognized that understanding the entire ecosystem requires the study of all biodiversity components, from the genetic structure of populations, to species, habitats and ecosystem integrity, including food-webs and complex bio-physical interrelationships within the system.

Thompson et al. (2012) emphasize that food-web ecology will act as an underlying conceptual and analytical framework for studying biodiversity and ecosystem function, if the following challenges are addressed: (i) relating food-web structure to ecosystem function; (ii) combining food-web and ecosystem modeling; (iii) transitioning from individual traits to ecosystem function; (iv) incorporating space and time in studies; and (v) understanding the effects of biodiversity loss on ecosystem function.

The study of the ecological function of biodiversity is very recent; yet, it has been recognized to have fundamental implications for predicting the consequences of biodiversity loss (Benedetti-Cecchi, 2005). Species in an ecosystem can be functionally equivalent, meaning that they play the same role. As such, these functionally equivalent species can be grouped together as functional types (i.e., guilds, trophic groups, structural groups, ecological groups, traits). Other key attributes of biodiversity organization, such as the density mass–relationship between abundance and body size, have become a major research area. These attributes relate to food webs, determined by the trophic position, predator–prey relationships, and energy balance. Theoretically, a higher number of functional group types will provide higher functional biodiversity organization to the system, and thus, contribute to more stable and resilient ecosystems (Tomimatsu et al., 2013).

Despite the importance of this question, the relationship between diversity and stability is still being resolved. As with many biodiversity-related topics, there are different ways of expressing stability. One way is to define it as the ability of a system to return to its original state after being disturbed (i.e., resilience), so how quickly it can return and how large a disturbance it can return from are key variables (Elliott et al., 2007). Another definition is how resistant to change the system is in the first place. No matter which definition is used, there are definite trends that appear.

Finally, a major issue in maintaining the functionality of ecosystems comes from invasive species, which can dramatically disturb stable systems thereby impacting ecosystem services (Sorte et al., 2010; Vilà et al., 2010). Methods to detect and control this biological pollution are therefore needed (Olenin et al., 2011).

#### **GRAND CHALLENGE 2: UNDERSTANDING RELATIONSHIPS BETWEEN HUMAN PRESSURES AND ECOSYSTEMS**

Global biodiversity is threatened by human activities which are increasingly impacting marine ecosystems (Halpern et al., 2008). These impacts are usually cumulative and can lead to degrading habitats and ecosystem functionality (Ban et al., 2010). In some seas, such as the Mediterranean and Black Sea, less than 1% of the surface is considered unaffected by human disturbance with most of the surface affected by cumulative impacts (Micheli et al., 2013). There is evidence that the likelihood of regime shifts may increase as a result of reduced ecosystem resilience through a decrease in diversity, functional groups of species or trophic levels, thereby impacting ecosystems (with waste, pollutants and climate change) and altering the magnitude, frequency, and duration of disturbance regimes (Folke et al., 2004).

Current socio-ecological theories consider humans as part of the marine ecosystem (Livingston et al., 2011). Hence, understanding the relationships between human activities and their various impacts on marine ecosystems represents another grand challenge to be discussed within the specialty section of Marine Ecosystem Ecology.

#### **GRAND CHALLENGE 3: UNDERSTANDING THE IMPACT OF GLOBAL CHANGE ON MARINE ECOSYSTEMS**

Sea waters are getting warmer, sea-level rise is accelerating and the oceans are becoming increasingly acidic (Stocker et al., 2013). From a database of 1735 marine biological responses to global change, Poloczanska et al. (2013) determined that 81–83% of all observations for distribution, phenology, community composition, abundance, demography and calcification across taxa and ocean basins were consistent with the expected impacts of climate change on marine life (Richardson et al., 2012).

As there is an insufficient understanding of the capacity for marine organisms to adapt to rapid climate change, Munday et al. (2013) emphasize that an evolutionary perspective is crucial to understanding climate change impacts on our seas and to examine the approaches that may be useful for addressing this challenge.

We need also a deeper understanding of the climate change impact on body size and the cascading implications on ecosystem functioning, considering the recent attempt of applying metabolic theory on modeling the biosphere. Hence, organisms often have smaller body sizes under warmer climates, and body size is a major determinant of functionality of the ecosystems, as commented above. Therefore, by altering body sizes in whole communities, current warming can potentially disrupt ecosystem function and services (Edeline et al., 2013).

In addition, our understanding of the linkages between climate change and anthropogenic disturbances needs to be improved. Borja et al. (2013b), investigating the combined effects of human pressures (i.e., exploitation and waste discharges) and environmental variables (i.e., light, waves, temperature) in macroalgae over a long-term series, demonstrated that in impacted areas macroalgae are more vulnerable to environmental changes and that their resilience is reduced. In turn, there is clear evidence that marine reserves enhance resilience of ecosystems to climatic impacts (Micheli et al., 2012).

As determined by Philippart et al. (2011), a better understanding of potential climate change impacts can be obtained by: (i) modeling scenarios at both regional and local levels; (ii) developing improved methods to quantify the uncertainty of climate change projections; (iii) constructing usable climate change indicators; and (iv) improving the interface between science and policy formulation in terms of risk assessment. These factors are essential to formulate and inform better adaptive strategies to address the consequences of climate change.

#### **GRAND CHALLENGE 4: ASSESSING MARINE ECOSYSTEMS HEALTH IN AN INTEGRATIVE WAY**

Assessing the status of the oceans requires tools that allow us to define marine health across different marine habitats. Such tools have been developed in recent years, including ecological indicators to be applied to different ecosystem components (Birk et al., 2012; Halpern et al., 2012). One of the current challenges is to clearly understand what good status or good health is/means in marine systems and how we know when it has been attained (Borja et al., 2013a; Tett et al., 2013). This way, integrating knowledge across different ecosystem components and linking physical, chemical and biological aspects when assessing the status of marine systems is crucial for accurate evaluations (Borja et al., 2009, 2011).

However, one of the most critical issues when assessing the health status of marine ecosystems relates to the setting of adequate reference conditions and/or environmental targets to which monitoring data should be compared (Borja et al., 2012). These targets should be set taking the ecological characteristics of the studied ecosystems into account.

#### **GRAND CHALLENGE 5: DELIVERING ECOSYSTEM SERVICES BY CONSERVING AND PROTECTING OUR SEAS**

Marine ecosystems provide numerous goods and services (Barbier et al., 2012), such as biogeochemical services (e.g., carbon sequestration), nutrient cycling, coastal protection (e.g., provided by coral reefs or phanerogams), food provision (e.g., fisheries), and grounds for tourism, etc. (Costanza et al., 1997). Despite the important role of such goods and services and albeit quickly attracting more attention, their study and their associated monetary value (often demanded to support conservation efforts) is still limited, particularly for the high seas and deep water habitats (Beaumont et al., 2007; Barbier et al., 2011; Braat and de Groot, 2012; Van den Belt and Costanza, 2012; Liquete et al., 2013; Thurber et al., 2013). Furthermore, recent debates have raised the question whether all ecosystem services can or should be quantified in monetary terms, when the public finds such values difficult to relate to.

It has been suggested that ecosystem services of high value critically depend on biodiversity (EASAC, 2009). As biodiversity loss is accelerating, maintaining biodiversity and healthy ecosystem services should be a priority when investigating, conserving and managing marine systems.

In marine management, Marine Protected Areas (MPAs) are an important tool for conserving and protecting biodiversity, by enhancing ecosystem resilience and adaptive capacity (Roberts et al., 2003; García-Charton et al., 2008). They allow for the mitigation of anthropogenic factors, such as overfishing or habitat destruction within their boundaries, by means of management or prohibition (Roberts et al., 2001; Mumby et al., 2006). Not only MPAs, but also the protection of near-natural ecosystems are very good strategies for managing climate change-related stressors and preserving biodiversity (Heller and Zavaleta, 2009).

Additional important issues in marine protection include the reduction of habitat fragmentation (Didham, 2010; Didham et al., 2012), determining the vulnerability of threatened species and habitats (Le Quesne and Jennings, 2012), and the study of connectivity between habitats and species distribution, which is a critical factor in maintaining habitat quality (Berglund et al., 2012).

#### **GRAND CHALLENGE 6: RECOVERING ECOSYSTEM STRUCTURE AND FUNCTIONING THROUGH RESTORATION**

Most estuarine, coastal and offshore waters worldwide have experienced significant degradation throughout the past three centuries (Lotze, 2010) and investments in marine protection have not been totally effective. Hence, ecological restoration is becoming an increasingly important tool to manage, conserve, and repair damaged ecosystems, as stated by Hobbs (2007).

Measuring effectiveness of restoration at habitat, community, or ecosystem level is not easy, and requires a focus on restoration of processes and functionality, rather than studying the recovery of particular species (Verdonschot et al., 2013). Thus, according to Borja et al. (2013c), restoration efforts should rely on what is known from theoretical and empirical ecological research on how communities and ecosystems recover in structure and function through time. Hence, studies on dispersal, colonization dynamics, patch dynamics, successional stages, metapopulations theory, etc., are needed for a deeper knowledge of recovery processes (Borja et al., 2010). This research will provide evidences to enhance restoration success of complex systems (Verdonschot et al., 2013).

#### **GRAND CHALLENGE 7: MANAGING THE SEAS USING THE ECOSYSTEM APPROACH AND SPATIAL PLANNING**

The management of marine systems, including the assessment of their overall health status, is increasingly carried out by applying ecosystem-based approaches (Borja et al., 2008). After all, the protection and conservation of marine ecosystems, together with the sustainable use of the services they provide, are of fundamental importance to the maintenance of global marine health (Tett et al., 2013). The goal of ecosystem-based management is to maintain an ecosystem in healthy, productive, and resilient conditions so that it can provide the services needed for the well-being of society (Yáñez-Arancibia et al., 2013). The guiding principles for ecosystem-based management are founded on the idea that ocean and coastal resources should be managed to reflect the relationships among all ecosystem components, including humans, as well as the resulting socioeconomic impacts (Yáñez-Arancibia et al., 2013).

In addition to the need for better management tools, the increasing anthropogenic impacts on marine waters (e.g., fisheries, aquaculture, shipping, renewable energies, recreation, mining, etc.) has promoted the discussion on how to manage and to conserve marine resources sustainably (Collie et al., 2013). Marine Spatial Planning, as defined by Ehler and Douvere (2009), is a management tool that attempts to balance conservation efforts with increasing demands on marine resources, which, together with the ecosystem-based approach, relies on a multidisciplinary approach integrating sociological, economic and ecological components (Qiu and Jones, 2013; Stelzenmüller et al., 2013).

#### **GRAND CHALLENGE 8: MODELING ECOSYSTEMS FOR BETTER MANAGEMENT**

The specificities of oceans when compared with terrestrial systems (see Norse and Crowder, 2005), and the increasingly complex approaches to investigate ecosystems at an integrative level requires the use of computer models (e.g., hydrodynamic, habitat suitability models, ecosystem models, etc.) for a better understanding of the processes, functioning and interrelationships among ecosystem components (Fulton et al., 2004). As a result, the use of species, ecological niche, habitat and ecosystem models has dramatically increased in recent years (Elith and Graham, 2009; Ready et al., 2010).

To guide conservation actions more effectively, the use of species distribution models has been recommended (Guisan et al., 2013), for example for studies on biological invasions, the identification of critical habitats, etc.

#### **CONCLUSION**

To adequately address the abovementioned grand challenges in Marine Ecosystem Ecology, effective long-term monitoring of populations and communities is required to understand marine ecosystem functioning and its responses to environmental and anthropogenic pressures (Stein and Cadien, 2009). However, monitoring programs often neglect important sources of error (e.g., the inability of investigators to detect all individuals or all species in a surveyed area) and thus can lead to biased estimates, spurious conclusions and false management actions (Katsanevakis et al., 2012). One of the newest ways to get reliable, verifiable, efficient and cost-effective monitoring of biodiversity is metabarcoding (Bourlat et al., 2013; Ji et al., 2013).

In addition to the acquisition of information on a regular basis, complete maps of habitats, ecosystem services, etc., are needed for a better understanding of spatial ecology and marine management (Brown et al., 2011). All this information requires data integration of the different ecosystem components in order to understand large-scale patterns and long-term changes (Stocks et al., 2009; Vandepitte et al., 2010).

Finally, the movement toward open access to scientific data and publications provides greater access to datasets and current research, which has the potential to result in better spatial and temporal analyses, by using existing information in a much more effective way through Information and Communication Technologies (i.e., e-Science). Make data open, accessible online in a standard format available for aggregation, integration, analysis and modeling, is a crucial step to boost the development of marine ecosystem ecology, to address the above highlighted challenges, and to move toward the frontiers of marine science (see Baird et al., 2011). Therefore, Frontiers in Marine Ecosystem Ecology promotes open access to data and information to enhance collaborations, whilst discussing hot marine topics and addressing the grand challenges described here.

#### **ACKNOWLEDGMENTS**

This manuscript writing has been partially supported by DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), www.devotes-project.eu. Naiara Rodríguez-Ezpeleta and María C. Uyarra (AZTI-Tecnalia), and Alberto Basset (Associate Editor of Frontiers of Marine Ecosystem Ecology) have provided interesting comments to the manuscript. This paper is contribution number 659 from AZTI-Tecnalia (Marine Research Division).

#### **REFERENCES**


of biodiversity. *Trends Ecol. Evol.* 27, 689–697. doi: 10.1016/j.tree.2012.08.005


et al. (2010). Data integration for European marine biodiversity research: creating a database on benthos and plankton to study large-scale patterns and long-term changes. *Hydrobiologia* 644, 1–13. doi: 10.1007/s10750-010- 0108-z


agement approach in the Gulf of Mexico. *J. Coast. Res.* 63, 243–261. doi: 10.2112/SI63-018.1

*Received: 19 December 2013; accepted: 04 February 2014; published online: 12 February 2014.*

*Citation: Borja A (2014) Grand challenges in marine ecosystems ecology. Front. Mar. Sci. 1:1. doi: 10.3389/ fmars.2014.00001*

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science.*

*Copyright © 2014 Borja. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# **Legal framework of marine activities and management**

# A Typology of Stakeholders and Guidelines for Engagement in Transdisciplinary, Participatory Processes

#### Alice Newton1, 2 \* and Michael Elliott <sup>3</sup>

*<sup>1</sup> Department of Environmental Impacts and Economics, Norwegian Institute for Air Research, Kjeller, Norway, <sup>2</sup> Centre for Marine and Environmental Research, University of Algarve, Faro, Portugal, <sup>3</sup> School of Environmental Sciences, Institute of Estuarine and Coastal Studies, University of Hull, Hull, UK*

This paper fulfils a gap in environmental management by producing a typology of stakeholders for effective participatory processes and co-design of solutions to complex social–environmental issues and then uses this typology for a stepwise roadmap methodology for balanced and productive stakeholder engagement. Definitions are given of terminology that is frequently used interchangeably such as "stakeholders," "social actors," and "interested parties." Whilst this analysis comes from a marine perspective, it is relevant to all environments and the means of tackling environmental problems. Eleven research questions about participative processes are addressed, based on more than 30 years of experience in water, estuarine, coastal, and marine management. A stepwise roadmap, supported by illustrative tables based on case-studies, shows how a balanced stakeholder selection and real engagement may be achieved. The paper brings these together in the context of several up-to-date concepts such as complex, nested governance, the 10 tenets for integrated, successful, and sustainable marine management, the System Approach Framework and the evolution of DPSIR into DAPSI(W)R(M) framework. Examples given are based on the implementation of the Marine Strategy Framework Directive, the Water Framework Directive, the Environmental Impact Assessment Directive, the Framework Directive for Maritime Spatial Planning, as well as for Regional Sea Conventions. The paper also shows how tools that have been developed in recent projects can be put to use to implement policy and maximize the effectiveness of stakeholder participation.

Keywords: Stakeholders, transdisciplinary research, coastal and marine management, implementation of MSFD

# INTRODUCTION

### Context and Objectives

Successful integrated marine management requires the coordination of many aspects, from an assessment of the source, causes, and consequences of problems, the delivery of ecosystem services and societal benefits, the incorporation of governance from the local to the global, and implementing the ecosystem approach (Elliott, 2014). The success of each of these requires the input from and often agreement with the "stakeholders," defined in Section Definitions below. This paper focuses on participatory processes with examples of marine management in Europe, but the principles can be applied in other, non-European contexts and non-statutory processes.

#### Edited by:

*Marianna Mea, Ecoreach s.r.l., Italy and Jacobs University of Bremen, Germany*

#### Reviewed by:

*Sabine Pahl, Plymouth University, UK Lyndsey Ann Dodds, WWF-UK, UK*

> \*Correspondence: *Alice Newton anewton@ualg.pt*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *20 June 2016* Accepted: *31 October 2016* Published: *16 November 2016*

#### Citation:

*Newton A and Elliott M (2016) A Typology of Stakeholders and Guidelines for Engagement in Transdisciplinary, Participatory Processes. Front. Mar. Sci. 3:230. doi: 10.3389/fmars.2016.00230*

Initiatives deemed to be stakeholder-led, or at least with a high degree of consultation are increasing. For example, the UK Marine Conservation Zone project, which aimed at proposals for Marine Protected Areas, was required by statute to be "stakeholder-led" using local, stakeholder panels (Natural England and the Joint Nature Conservation Committee, 2012). Similarly, planning regulations involving a formal Environmental Impact Assessment, as sanctioned by the EU EIA Directive, are centered on stakeholder consultation. Elliott (2014) therefore briefly proposed a brief, initial typology of stakeholders but this needs to be further explored and refined to ensure it covers all potential bodies. Furthermore, because of the participatory process, it is valuable to assess the types of stakeholders, their role in each part of the marine management process and the influence both on the process and on them personally.

Accordingly, the aim of this paper is to provide guidelines, which can eventually be embedded into a prescriptive method, to support and develop participatory processes by using an appropriate framework for stakeholder definition and engagement. Therefore, the objectives are: (i) to further develop a typology of stakeholders; (ii) to provide guidance for appropriate and equitable stakeholder engagement; and (iii) to illustrate how this can be achieved in the implementation of marine environmental governance, such as the EU (2008) Marine Strategy Framework Directive (MSFD, 2008/56/EC). This Directive, based on a System Approach to management and participatory processes, requires that "all interested parties are given early and effective opportunities to participate in the implementation of this Directive." In turn, this gives rise to set of research questions (**Box 1**) that will be addressed in this paper.

### Definitions

The terms social actors, stakeholders, and interested parties are used throughout environmental management, therefore it is appropriate to define these first. Sociology is a comprehensive science of social action with an analytical focus on individual human actors or **social actors** (Weber, 1991). This may or may not include those with a statutory remit, those who actively influence the course of social action, and/or those passively affected by others' actions rather than actively influencing the outcomes.

#### BOX 1 | Stakeholder-orientated research questions.


There are many definitions of the term stakeholder, several of which are collected in Mehrizi et al. (2009). Using these, we suggest the following inclusive definition that is relevant in a marine management context "a stakeholder is a person, organisation or group with an interest (professional or societal) or an influence on the marine environment or who is influenced directly or indirectly by activities and management decisions."

The MSFD gives a brief indication of what is meant by **interested parties** in that it should be: "involving, where possible, existing management bodies or structures, including Regional Sea Conventions, Scientific Advisory Bodies, and Regional Advisory Councils." However, many other "stakeholders" or "actors" would also be "interested parties" with respect to the implementation of the MSFD, if they have an interest in the outcome or an influence on the outcome. They include the many people whose livelihood and welfare depends on the sea, such as: fishers and shellfish harvesters; aquaculture farmers; offshore extractors of minerals such as oil, gas, sand, and gravel; offshore wind farms, tidal and wave energy; coastal, cruise, and eco-tourism developers; and maritime transport. This also includes the millions of people who choose "sun, beach, and sand" vacations (Semeoshenkova and Newton, 2015). For the purposes of this paper, the term "stakeholder" has thus been adopted as an inclusive term that also incorporates the various interested parties and social actors. This makes it necessary to have some clear definitions and a typology that covers the roles of stakeholders, both of which are provided in this paper.

# WHO ARE THE INTERESTED PARTIES, STAKEHOLDERS AND SOCIAL ACTORS?

#### Interested Parties

The MSFD aims to ensure that EU Member States can achieve Good Environmental Status in their seas by 2020 according to a set of 11 descriptors which encompass and affect all the uses and users of the seas (Borja et al., 2013). Hence, by definition, the detection and achievement of GES has to be a stakeholderled process in order to achieve successful and sustainable marine management. Correctly identifying the stakeholders is fundamental to participatory processes (WMO, 2006). The "interested parties" referred to in the MSFD (**Table 1**) include the four European Regional Sea Conventions and the seven fisheries Regional Advisory Councils. The MSFD focus therefore seems to be limited to fisheries as the main economic sector, and thus, by not specifically indicating other stakeholders, it is not as inclusive as intended. Examples of possible Scientific Advisory Bodies are also listed. Previously (Elliott, 2014), we considered that the pressures affecting the marine environment emanate from three sources—materials (including infrastructure) put into the seas, materials, and space/habitat removed from the seas, and external factors such as climate change. Each of these has its own interested parties, although of course there is extensive overlap between them. In **Table 1**, the interested parties encompass the users, those controlling the users and those

#### TABLE 1 | Interested parties referred to in the Marine Strategy Framework Directive.


affected by or benefitting from the uses. This underlines the need to develop both a typology and a methodology that promotes balanced participation and stimulates meaningful rather than perfunctory engagement. Ideally, this should ensure that the relevant interested parties, stakeholders, and social actors are invited to be involved in a participatory process and that each is aware of the roles of the others, even though this may be difficult to achieve.

## What Types of Stakeholders Are There and What Are Their Roles?

A typology that encompasses all the types of stakeholders is proposed in **Table 2** and has resulted from many years of experience in water, estuarine, coastal and marine management as well as considering and discussing the need for, and role of stakeholders. Successful and sustainable solutions to marine problems range across the so-called 10-tenets that encompass technical, economic, governance, and societal aspects (Elliott, 2013; Barnard and Elliott, 2015), and the proposed typology embraces these. The typology includes six types of stakeholders that have been developed from those in Elliott (2014) and are not dissimilar to those proposed by Lovens et al. (2014). The links to the source, causes and consequences of human activities in the sea use the DAPSI(W)R(M) framework (Patrício et al., 2016; Scharin et al., 2016) in which Drivers (the basic human needs, the individual and societal aspirations) require Activities (by the users, developers, industries, etc.) that in turn create Pressures, which are the mechanisms to cause adverse State changes to the natural (physical, chemical, and biological) environment. If left unchecked, these create an Impact (on human Welfare) that need to be addressed by Responses (involving Measures).

Those creating the pressures in the sea are the "inputters" (of pollution, infrastructure, sediment, etc.) and "extractors" (of fish, water, space) who then are regulated by the "regulators," those statutory bodies with a legislative competency, supported by administrative bodies and given that competency by a very large number of legal instruments (e.g., Boyes and Elliott, 2014, 2015). Those who take or receive advantage of those uses and materials provided by the seas or even who get advantage by reducing their costs due to putting wastes into the seas, are termed "beneficiaries," a group that contains most if not all of society. Next, there is a large group of stakeholders that are affected, possibly adversely, by those using and managing the seas, for want of a better term and in keeping with the labels for the other types these are called "affectees." Finally, there are the "influencers," those politicians, non-governmental organizations, media, academics, and educators who play a part in directing the nature of marine use.

Some stakeholders have a very precise role and this typology. For example, an Environmental Protection Agency or Nature Conservation Agency is a statutory regulator in a defined area of competence, (in this case, water quality and species protection, respectively). However, they may be an "influencer" for topics outside their own jurisdiction and are often a statutory or nonstatutory consultee when other bodies are faced with decisions. In contrast, stakeholders often play a role in more than one of these groups, given that the role of the stakeholder depends on the context and issue. For example, fishers extract a resource and input materials and infrastructure, such as discards, waste,



*The type is given in the context of the Driver-Pressure-State-Impact-Response (DPSIR) framework (Gari et al., 2015) and the DAPSI(W)R(M) framework (see text, Scharin et al., 2016; Patrício et al., 2016).*

garbage, and harbors. They are beneficiaries of the ecosystem provisioning services. They are also affected by other fishers, as well as many other users of the seas, for example the installation of offshore wind farms. They collectively are policy-influencers with strong lobbies and a presence on advisory bodies (such as the National Federation of Fishermen's Organizations in the UK) and they may have a role in the local governance of the activity.

An example of the typology of stakeholders for two Good Environmental Status Descriptors of the MSFD is given in **Table 3** and an indicative list of stakeholders for the UK sector of the North Sea is given in four to show the breadth of bodies involved. General (civil) society is additional to this list.

### Why Are Stakeholders Important in Participatory Processes?

Stakeholder engagement and involvement is the basis of a participatory process and is fundamental to acceptance of management actions and by definition the process is not participatory if stakeholders are not involved. Community-based, adaptive management requires stakeholder engagement and participation from the early planning stage. Stakeholders gain a better understanding of issues and conflicts through participation in the co-design and co-development of the management plan. The process can provide an opportunity for conflict resolution and also increases the ownership, ease of acceptance and uptake of jointly designed solutions. Under the governance principle of subsidiarity, a cornerstone of the European Union, key management decisions should be made as close to the scene of events and the actors involved for the sustainable management of socio-ecological systems and their resources (Ostrom, 2009). Governments therefore strive to engage stakeholders to influence policy and to reach a consensus for sustainable management. There are several examples in which a range of stakeholders is required to be consulted and, in the case of Environmental Impact Assessments, it is often legally required (Glasson et al., 2011).

Finally, stakeholders have an important role in checking whether the outcome of the adaptive management process (e.g., responses of governance, regulations, recommendations, programme of measures, management plans) conform to the 10-tenets proposed by (Elliott, 2013). These require management actions to be Ecologically sustainable, Technologically feasible, Economically viable, Socially desirable/tolerable, Legally permissible, Administratively achievable, Politically expedient, Ethically defensible (morally correct), Culturally inclusive, and Effectively communicable. These facets, if achieved, should cover all parts of the decision-making process and thus, by engaging stakeholders in decision-making, should provide for a sustainable and accepted, consensual solution. It should allow contentious issues to be raised, defined and resolved early on in the process, and thus be used to minimize conflicts. Nevertheless, there is a paradox of stakeholder consensus in reaching a stakeholder-led decision, whereby stakeholder panels may agree collectively on the lowest common denominator, i.e., the least painful solution for each of them.


#### TABLE 3 | Example of typology of MSFD stakeholders for Good Environmental Status Descriptor 3 (commercial species) and Descriptor 5 (eutrophication).

*The listing is general not specific, e.g., RSC is given instead of OSPAR, HELCOM, Mediterranean, and Black Sea Conventions.*

As an example of this, the recent Marine Conservation Zone projects in the UK, in response to government demands for a stakeholder-led process, set up regional stakeholder panels that each aimed to have one representative of each main sector within the region (Jones, 2012). Their role was to use the Ecological Network Guidance provided by the statutory nature conservation bodies in designing and positioning Marine Conservation Zones, (Natural England and the Joint Nature Conservation Committee, 2012). Within such a stakeholder panel for a geographical area, a stakeholder who is a fishing representative is unlikely to agree to site a Marine Protected Area in a fishing area where fishing will be limited. An aggregate (sand and gravel) extractor may agree to site an MPA in areas away from the extractors favored resource. Hence, the stakeholders may all agree to site a Marine Protected Area in a location no one wants for any other purpose, i.e., an area unsuitable for fishing, aggregate extraction, offshore wind, etc. Should this hypothetical example be approved, the regulators could claim that a stakeholder-led solution was reached, but this was only achieved by designating a conservation-poor area.

# SOCIAL EQUITY: ARE ALL STAKEHOLDERS EQUAL?

Some stakeholders are more relevant to particular issues than others. Relevant stakeholders for one issue may be of little or no importance for another issue. For example, while NGOs may encourage developers to create an ecologically sustainable option, the shareholders of the developer are more likely to be responsive to the consequences of statutory regulators threatening legal action with financial penalties or the size of a resource to be exploited. The relevance of a stakeholder in one of the type categories in **Table 2** (and the examples for each type in **Table 4**) thus depends on the issue, for example the MSFD descriptor that is being addressed. For example, in the case of the general implementation of the Marine Strategy Framework Directive, the participation of the competent authorities of the Member States and the Regional Seas Commissions are not just desirable but paramount. If only one particular descriptor is being addressed, then representatives of economic sectors that are significant should definitely be included. A semi-quantitative example of


#### TABLE 4 | List of Marine Stakeholders—Example for the UK Sector of the North Sea (D Burdon, IECS University of Hull, pers. com).

how the weightings of stakeholders may be employed and differ for different issues is shown in **Table 5**. The weightings serve to emphasize important stakeholders to be recruited. This compares an environmental quality issue (eutrophication) with a resource exploitation one (fishing). Both of these have a set of social actors/stakeholders causing the problem as well as those being affected by it socially and economically and those trying to regulate it. The relative weightings will vary, not only according to the societal or business repercussions on the stakeholder in question, but also the prevailing environmental and societal conditions. For example, under economically difficult times, as over the past decade, the financial and economic imperatives may be prioritized (Borja and Elliott, 2013). Nevertheless, each stakeholder may project a different weighting either through forceful argument (the problem of "he who shouts loudest"), or a self-assumed weighting, giving the impression that a stakeholder is considered more important than others. For example, in maritime countries where fishing was historically more important than today, the fishing lobby may assume an exaggerated importance (the boxing analogy of "punching above their weight"!).

An experienced moderator is essential for the success of a participatory process. Ideally, a stakeholder panel should represent all relevant stakeholders in a fair and balanced manner. It is assumed that if all stakeholders are consulted, then a balanced outcome should be guaranteed but of course, this is not always the case. Similarly, while it is hoped that all stakeholders should have an objective and rational view, it is perhaps better to assume that they are all defending their own interests, and hence cancel each other out.

# WHAT ARE THE DIFFICULTIES AND CONFLICTS AND HOW CAN THEY BE RESOLVED?

In addition to the tendency of reaching the lowest common denominator, a participatory process may be hampered by other difficulties, for example: (i) the misinterpretation of scientific guidelines and information by a non-scientific body, and (ii) bypassing the process. The second can arise, for example, if the main fishing lobby decides not to take part in the stakeholder process but, once the panels have finished deliberating, they petition the government minister directly, thus by-passing the process. The minister (regulator) then must decide whether to ignore the views of all the other stakeholders in favor of one stakeholder that is perceived (by itself) to be the most important. This would be regarded by the remaining stakeholders as circumventing the democratic process.

In addition to these difficulties, there are often conflicts between the stakeholders, based on the activities for which they may be responsible or have an involvement. A clear understanding of the nature of these conflicts is important for both the mediator and the stakeholders to better consider the various aspects of an issue and the resulting points of view. The use of a conflict matrix approach can avoid problems, reduce conflicts between stakeholders and encompass all bodies. The

#### TABLE 5 | Example weightings (0 low–3 high) of different stakeholders for two MSFD Good Environmental Status Descriptors (D3 commercial species and D5 eutrophication).


INTERREG project TIDE (see http://www.tide-toolbox.eu) used conflict matrix analysis and stakeholder focus groups followed by multivariate analysis to determine use and user conflicts for a set of North Sea estuaries (the Scheldt (Belgium/Netherlands), Humber (UK), Elbe (Germany), and Weser (Germany). The results in **Table 6** show the links and conflicts between the different activities and their proponents, with a view to determining the priorities for resolving conflicts, either real or imagined. For example, Conservation by protection of an area was viewed by some stakeholders to be negative because access to the area was restricted.

In the marine arena, common problems requiring to be resolved by stakeholder participation include spatial conflicts, such as the access to fishing grounds being limited by offshore development of aquaculture, mineral and oil extraction, and wind farms. These are the topic of the Framework Directive for Maritime Spatial Planning, (2014/89/EU). There are frequent conflicts between types of stakeholders, such as conservationists (influencers) and fishers (extractors), especially for high value species such as blue fin tuna, or to avoid the bycatch of turtles and cetaceans. The most frequent conflicts arise between beneficiaries and the affected (winners and losers). One solution is to use tradeoffs, for instance building a longer road around a protected area rather than through it. These require a thorough cost-benefit analysis that also examines externalities and a developer or member state could cite economic constraints as a reason for not carrying out stakeholder wishes. If successful, however, mutually and stakeholder-led agreement of co-location of activities can help to reduce stakeholder conflicts (Christie et al., 2014).

Conflicts of responsibilities can also occur, for example between the EU and RSCs, (Cinnirella et al., 2014) and between EU and non-EU countries, for example the issues in the Black Sea arising from the Crimea crisis in 2014. The plethora of marine legislation and administrative bodies implementing these has the potential to increase confusion (e.g., Boyes and Elliott, 2014, 2015). For example, the EU Water Framework Directive requires an assessment of Good Ecological Status to 1 nm from the coastal baseline, the Habitats Directive requires Favorable Conservation Status in designated areas, and the Marine Strategy Framework Directive requires Good Environmental Status from the HW mark outwards to the 200 nm line (Boyes et al., 2016). These Directives all rely on stakeholder discussion and agreement thus requiring stakeholders, who are often the same group of individuals, to be familiar with the legislation and their implementation, the differing ecological principles and the science base, and the geographical overlap. This is rarely the case, especially as different organizations meet with different stakeholders and implement different legislation (Boyes and Elliott, 2015). Conflicts of responsibilities arise, even though each legal instrument is designed for a particular role. Thus, there are conflicts between both instruments and bodies that require to be solved by stakeholder involvement. Furthermore, when the same stakeholders are consulted regarding multiple developments from many agencies then "stakeholder-fatigue" can occur. The interest of the stakeholders may wane if they feel that their opinions are not being heard or taken seriously. An experienced moderator will ensure that the opinion of all participants is sought, heard, and treated with due respect.

A stakeholder and governance mapping step is important, based on the issue and the geographical context. Next, stakeholders are approached during the consultation phase called scoping, but it is important not to raise their expectations of being able to direct the result. They also may be confused regarding the scoping and their precise and legally defined role that includes checking, consulting, challenging, and championing. So,


TABLE 6 | Example of results from Conflict Matrix Analysis showing the strong negative and positive associations between uses/users for North Sea estuaries.

*Those in bold are the strongest noted. (from http://www.tide-toolbox.eu).*

it is important to clarify that, by statute or accepted practice, stakeholders are deemed to have a role, or at least be given the opportunity to have a role, in all stages of adaptive management. Usually, the remit is in the planning stage, and then further during each part of the assessment and writing the final report, for example the Environmental Statement (EU EIA Directive 2011/92/EU; 2014/52/EU). Finally, it is important to keep the stakeholders interested and involved in the process but not to over-burden them, again which results in stakeholder fatigue.

Conflicts of responsibilities require stakeholders to have the capacity to take such decisions. These difficulties and conflicts can result in an increasing severity of impediments to achieving sustainable marine management. Increasing severity is shown in **Box 2**: the first column lists "bottlenecks" or minor impediments, which do not require much effort to clear; the second column lists "showstoppers," which require a moderate focus to remove; and the third column includes "train wrecks," which potentially stop everything. Poor scientific understanding (column 1) may be overcome in the stakeholder forum through interaction with the scientists (see Issue Definition in 6a). For example, stakeholders may know that the water quality is poor, but they may have misunderstood the reason why. They may blame point sources, such as the effluent from a very effective waste water treatment plant rather than diffuse sources such as agricultural run-off. However, poor knowledge (column 2) can only be overcome by obtaining more, fit-for-purpose information and using this in an appropriate assessment, but this may not be possible because of lack of funding (Column 3). In the point-source vs. diffuse runoff example, only data from well-designed monitoring will allow the contribution of each source (point or diffuse) to be attributed.

Conflicts are frequent, so there are several available tools that can be used by a skilled mediator to defuse and attenuate them. It is not in the scope of this article to review the many available tools, but as an illustrative example, the SPICOSA project developed the kercoast deliberation-support tool http:// www.spicosa.eu/kercoast/main.htm. This is based on a matrix of stakeholders verus issues and uses a color code of red for "unacceptable," to yellow "maybe," to green "OK." The mediator starts exploring why the green issues are all OK, to build a common understanding between the stakeholders and open the dialogue. Once this has been achieved, the yellow issues are discussed to explore whether these can be made more acceptable and changed to green. Once mutual trust has been built and the stakeholders are more aware of the other points of view, then the more difficult "red" issues can be tackled.

#### WHAT EXAMPLES HAVE WORKED

A good example of stakeholder engagement in developing and implementing Marine Management is the process by which the Norwegian environment agency developed the management plans for the Barents Sea, the Norwegian Sea, the North Sea and Kattegat. The Norwegian environment agency built a very broad stakeholder forum for the design of the management plans. These were then open to public consultation, debated in Parliament and adopted. When several Member States or contracting parties are involved, consensus, and implementation may be difficult if the process is not planned and executed in a timely manner. Another good example involving several countries is the Common Implementation Strategy (CIS) of the


Water Framework Directive (WFD) for coastal waters, which included an inter-calibration process (Goela et al., 2009). This was specifically designed so that neighboring Member States would reach similar results in the assessment of transboundary waters.

# A ROAD-MAP FOR IMPROVING STAKEHOLDER INVOLVEMENT AND PARTICIPATORY PROCESSES

Integrated marine management involving stakeholders can be regarded as consisting of four steps—integration, adaptation, participation, and collaboration (**Figure 1**, adapted from Carvalho and Fidélis, 2013). Stakeholders play a key role in each of these steps and indeed drive those steps through the engagement and participatory process. In turn, there are several existing and useful guidelines for stakeholder engagement in participatory processes, (e.g., Fish et al., 2011; Durham et al., 2014) and so here we set out a road-map for participatory processes and useful stakeholder engagement. The road-map includes simple steps, processes and checklists to navigate through participatory processes. It is derived from earlier examples bringing together tried and tested features from many experiences in different countries, continents and contexts, estuarine, coastal and marine. There are no panaceas, and any method should be locally adapted, but we suggest the following steps, giving examples to clarify as necessary.

# Issue Definition

It is axiomatic that if society perceives a problem, then by definition, there is a problem to be tackled. For example, stakeholders may report "bad water quality," but this can be due to a plethora of issues such as industrial contamination, or just sewage effluent or, in the case of the general public, if they see turbid waters there may be a perception of poor quality, irrespective of the scientific evidence. Scientists working with the authorities and the stakeholders can help to better define the issue and this will help to better identify the system (6b) most the relevant stakeholders (6c).

# System Definition

This step defines "who is in and who is out." For example, is Austria a relevant stakeholder in the MSFD? It may be a minor actor since the Danube flows to the Black Sea. Sometimes, given the extensive connectivity that governs marine and coastal processes, relevant stakeholders may be half a continent away, or even further, contributing to the idea of "unbounded boundaries" in marine systems. An example are the Iowa farmers who input fertilizers into the Mississippi, contributing to the eutrophication and the Dead Zone of the Gulf of Mexico (Rabalais et al., 2002). This is also the case for the atmospheric deposition of long-range pollutants, such as mercury (Pacyna et al., 2006) and Persistent Organic Pollutants (POPs; Bidleman et al., 1990).

#### Stakeholder and Governance Mapping

Once the issue (6a) and the system (6b) have been defined, the mapping of the stakeholders and governance can proceed. For example, in the case of eutrophication (Descriptor 5 of the MSFD), food processing, paper mills, municipalities, farmers, the fertilizer industry, and animal rearers, including aquaculture, are possible stakeholders, some of whom may be irrelevant for another descriptor, such as Descriptor 11, Energy, and Noise. The categories and types of stakeholders are identified to be potentially involved in the decision making process. This is to better understand the stakeholder landscape and is a worthwhile process (see Morris, 2012). The mapping defines the possible participants, consistent with the proposed typology (**Table 2**), that are relevant (**Table 3**) to the issue and system. The mapping should include representatives of relevant economic sectors (extractors and inputters, externalities), and also include relevant influencers (e.g., NGOS, scientists, activitists, media). The governance mapping defines who can act (regulators) on any recommendations, such as decision makers, policy makers, managers at all geographical scales, local, national, regional (sea), European.

The objective is to achieve a representative cross-section of relevant stakeholders. Once more an experienced moderatorfacilitator is useful at this stage of selection, so that the process remains inclusive and balanced. Opinions about who the

relevant stakeholders are may vary, for example, when number of experts on the MSFD were asked "who are stakeholders of the MSFD?" in the context of the DEVOTES research project, the greatest number of answers were influencers, mainly from the sub-category of researchers and academics. Whole sectors, such as maritime transport were omitted. There were also large disparities according to the Member States, some mainly included regulators and influencers. Weightings may be also be attributed to stakeholders, and an illustrative example is given in **Table 5**, but it is not prescriptive.

The three steps above (a, b, c) are part of the System Approach Framework (SAF) methodology, a step-wise approach to transdiciplinary co-design of management plans, (Newton, 2012). In addition, a methodology for successful stakeholder involvement and the design of participatory processes involves assessing what level of engagement is required. Participation is a way of engaging decision makers and approaches vary. It is important to consider distinctions between wishing to inform, learn from or collaborate with stakeholders and to evaluate what is appropriate in particular contexts. Furthermore, it is important to assess resource commitments within any engagement process and technique. For example, will the engagement be only via online questionnaires or will there be frequent face-to-face meetings? What is practically achievable in a given context is hugely dependent on available resources: money, time, and skills.

#### Scoping

At this stage the focus shifts from planning to recruitment of the stakeholders for the implementation of the participatory process. The engagement process should be designed to be appealing, to draw the stakeholders in an engaging manner and to convince them to remain involved throughout the process. The stakeholders should feel that their opinion is being sought and heard by the decision-makers included in the forum, especially when the participatory process is statutory. Particular effort should be made to engage relevant stakeholders of each type, which may require persistence. The invitation to participate should highlight that the contact is being made because they are regarded as a significant stakeholder whose opinion should be considered. Scoping is a defined term within the EIA legislation and requires stakeholders (at least "statutory consultees," such as regulatory bodies, but also "non-statutory-consultees" such as environmental NGOs) to indicate the main areas of concern, which can then be rigorously addressed. There are several considerations regarding barriers to involvement that need to be overcome, as well as issues of good conduct during the consultation.

At this point, it is also valuable to ask the stakeholders to respond to a very brief questionnaire. This should be structured with care, to establish whether the stakeholders are informed about the general issue (e.g., the implementation of the MSFD), and the specific issue (e.g., Descriptor 1, Marine Biodiversity). It should also allow the first version of the conflict matrix to be drafted, and ask for any suggestions for other possible stakeholders. This item allows identifying stakeholders that may have been missed, but also allows the mapping of stakeholder groups and networks. If the stakeholders do not respond to the questionnaire, a second invitation is an opportunity to reiterate that their participation is considered important. It is unlikely that a stakeholder who will not respond to a short questionnaire will participate actively in the stakeholder forum. However, the invitation should be kept open and reiterated, so that stakeholders may join at a later stage. Nevertheless, it should be emphasized to the stakeholders that this is a process that requires sustained participation and jumping in at the end to try to halt the process and acting as a "spoiler" is not an option.

# Establish the Stakeholder Forum

The stakeholder forum (SF) members should be:


An example list of stakeholders for a Member State (UK) and marine region (North Sea) is given in **Table 4**.

# Briefing of the Stakeholder Forum Members and Their Meetings

The level of engagement and commitment partly depends on the invitation to participate and the convivial atmosphere of the meetings. To keep stakeholders interested and engaged, they must be well-briefed at the start and kept well-informed of the steps and developments of the process. The Stakeholder Forum (SF) members should be briefed about the general aim (e.g., contribute to the implementation of the MSFD); the context (e.g., in the Baltic Sea); the specific issue (e.g., Descriptor 5, Eutrophication); their role (e.g., contribute expert opinion); the remit that may include the desired final outcome (e.g., establish a long-standing forum) and output (e.g., a Report of recommendations to HELCOM). An experienced and wellinformed convener will clarify any questions that may arise, particularly about the legitimacy of other members of the forum.

The meetings should be convened sufficiently far in advance to allow busy members of the Stakeholder Forum to attend. Once more, the meeting should be chaired by an experienced and well-informed convener and moderator, with the necessary skills to maintain a balanced debate, so that the views of all participants are heard. Minutes of the meeting, and especially any decisions, should be circulated to ensure that they reflect all points of view and retained for future cross-checking. Tasks and action points should be revised and reviewed at the beginning of each meeting. If the group is large, it may be useful to divide into subgroups, ensuring that there is a balance across the types and that each group has a moderator, and also to avoid groups containing "networks" (Section 6d). There are many existing conflict-resolution methods and tools (Section What Are the Difficulties and Conflicts and How Can They Be Resolved?) that can be used to reach consensus.

# Drafting the Report of Recommendations

First of all it is important to identify appropriate choices from the range that are potentially available to decision makers. It is important to match the choice of the recommended technique to the purpose, and to understand what it will deliver, and what are the limitations. Once a consensus has been reached about the structure of the recommendations and the structure of the report, each stakeholder should be invited to contribute to sections for which their expertise is relevant. If they decline to do so, they may then be invited to review these sections. The lead author or editor of the report will invite and compile contributions and circulate the draft to the Stakeholder Forum for comments. Usually there are several iterations: the outline, the first draft, the second draft, the final draft and the pre-print or pre-submission proofs.

# Evaluate the Process and Its Outcomes

There are different ways of evaluating the success of a participatory process. An important distinction exists between process success and outcome success. The outcome success is short-term, it addresses the issue that was analyzed. A longlasting success is achieved when the process was so wellconducted that a robust stakeholder forum continues to exist after the short-term issue has been resolved. This can result from a successfully run participatory process, where stakeholders become a supportive network of colleagues, allies and in some cases friends. It is emphasized that, for example, in an EIA the process not only has to be carried out but it has to be seen to be carried out.

# CONCLUDING REMARKS

If we are now in the Anthropocene (Crutzen and Stoermer, 2000) and exceeding our planetary boundaries (Rockström et al., 2009; Steffen et al., 2015) there is an urgent need for future earth sustainability (Future Earth, 2014) to deliver on the promise of science for society. This entails "transdisciplinary" research that includes both science-society and science-policy, interfacing throughout the whole research process, from codesign of research to co-production of knowledge. Such a process inevitably relies for success on genuine and successful

stakeholder engagement in truly participatory processes. This paper outlines both a typology and a roadmap that may serve to make co-design and acceptance of solutions a reality.

#### AUTHOR CONTRIBUTIONS

The manuscript is the result of many discussions over several years between the co-authors. The "order" of authors was discussed and it was decided that AN would lead the article. However, many of the ideas and substantial material were contributed by ME, as co-author. AN drafted the outline and most of the roadmap, while ME concentrated on the typology and many of the examples. AN contributed **Box 1**, **Tables 1**, **3**, **5**.

#### REFERENCES


ME contributed **Box 2**, **Tables 4, 6** and **Figure 1**. **Table 2** was the work of both authors. Both authors contributed to the revisions and these were submitted by AN.

#### ACKNOWLEDGMENTS

To SPICOSA, KnowSeas, VECTORS, and DEVOTES colleagues for helping us to form our views. This manuscript is a result of the DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392, http://www. devotes-project.eu).


Document No. 4, Flood Management Policy Series. ISBN: 92-63-11008-5. Geneva, Switzerland. p. 100.

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declared a collaboration with the authors and states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Newton and Elliott. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# DPSIR—Two Decades of Trying to Develop a Unifying Framework for Marine Environmental Management?

Joana Patrício<sup>1</sup> \*, Michael Elliott <sup>2</sup> , Krysia Mazik <sup>2</sup> , Konstantia-Nadia Papadopoulou<sup>3</sup> and Christopher J. Smith<sup>3</sup>

*<sup>1</sup> European Commission, Joint Research Centre, Directorate for Sustainable Resources, D.2 Water and Marine Resources Unit, Ispra, Italy, 2 Institute of Estuarine and Coastal Studies, University of Hull, Hull, UK, <sup>3</sup> Hellenic Centre for Marine Research, Crete, Greece*

Determining and assessing the links between human pressures and state-changes in marine and coastal ecosystems remains a challenge. Although there are several conceptual frameworks for describing these links, the Drivers-Pressures-State change-Impact-Response (DPSIR) framework has been widely adopted. Two possible reasons for this are: either the framework fulfills a major role, resulting from convergent evolution, or the framework is used often merely because it is used often, albeit uncritically. This comprehensive review, with lessons learned after two decades of use, shows that the approach is needed and there has been a convergent evolution in approach for coastal and marine ecosystem management. There are now 25 derivative schemes and a widespread and increasing usage of the DPSIR-type conceptual framework as a means of structuring and analyzing information in management and decision-making across ecosystems. However, there is less use of DPSIR in fully marine ecosystems and even this was mainly restricted to European literature. Around half of the studies are explicitly conceptual, not illustrating a solid case study. Despite its popularity since the early 1990s among the scientific community and the recommendation of several international institutions (e.g., OECD, EU, EPA, EEA) for its application, the framework has notable weaknesses to be addressed. These primarily relate to the long standing variation in interpretation (mainly between natural and social scientists) of the different components (particularly P, S, and I) and to over-simplification of environmental problems such that cause-effect relationships cannot be adequately understood by treating the different DPSIR components as being mutually exclusive. More complex, nested, conceptual models and models with improved clarity are required to assess pressure-state change links in marine and coastal ecosystems. Our analysis shows that, because of its complexity, marine assessment and management constitutes a "wicked problem" and that there is an increasing need for a unifying approach, especially with the implementation of holistic regulations (e.g., European framework Directives). We emphasize the value of merging natural and social sciences and in showing similarities across human and natural environmental health. We show that previous approaches have adequately given conceptual and generic models but specificity and quantification is required.

Keywords: biodiversity, conceptual framework, drivers, pressures, state, impacts, response, environmental assessment

Edited by: *Angel Borja, AZTI, Spain*

#### Reviewed by:

*Angel Pérez-Ruzafa, University of Murcia, Spain Ana Isabel Lillebø, University of Aveiro, Portugal*

\*Correspondence: *Joana Patrício joanamateuspatricio@gmail.com*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *01 June 2016* Accepted: *01 September 2016* Published: *14 September 2016*

#### Citation:

*Patrício J, Elliott M, Mazik K, Papadopoulou K-N and Smith CJ (2016) DPSIR—Two Decades of Trying to Develop a Unifying Framework for Marine Environmental Management? Front. Mar. Sci. 3:177. doi: 10.3389/fmars.2016.00177*

# INTRODUCTION

The highly-complex marine system has a large number of interrelated processes acting between its physical, chemical, and biological components. Many diverse human activities exert pressure on this complex environment and the cumulative environmental effects of these activities on the system varies according to the intensity, number and spatial and temporal scales of the associated pressures. There is an increasing need to demonstrate, quantify predict and communicate the effects of human activities on these interrelated components in space and time (Elliott, 2002). The study and management of marine systems therefore requires information on the links between these human activities and effects on structure, functioning and biodiversity, across different regional seas in a changing world. It also requires the need to merge approaches from natural and social sciences in structuring and solving the problems created by human activities in the seas (Gregory et al., 2013).

Conceptual models are needed to collate, visualize, understand and explain the issues and problems relating to actual or predicted situations and how they might be solved. These models can be regarded as organizational diagrams, which bring together and summarize information in a standard, logical and hierarchical way. Since the early 1990s, Pressure-State-Response (PSR) frameworks have been central to conceptualizing marine ecosystem risk analysis and risk management issues and then translating those for stakeholders, environmental managers and researchers. In this context, the pressures cause the changes to the system, the state changes are the unwanted changes and the responses are what society does to remove, minimize, or accommodate the changes. Hence, it is axiomatic that society has to be concerned about the risks to the natural and human system posed by those pressures (thus needing risk assessment) and then it is required to act to minimize or compensate those risks (as risk management) (Elliott et al., 2014).

It is apparent that one of the key current conceptual frameworks in widespread use, the Drivers-Pressures-State change-Impact-Response (DPSIR) framework (see **Figure 1A** original concept and definitions from EC, 1999), has developed since the 1990s as the basis for most conceptual approaches addressing pressure-state change links. It is policy-oriented and provides a framework for categorizing a problem domain, along the cause-effect chain. The DPSIR framework was developed from the PSR framework initially proposed by Rapport and Friend (1979), and adapted and largely promoted by the OECD (Organization for Economic Cooperation and Development) for its environmental reporting (OECD, 1993). Several international organizations, such as US Environmental Protection Agency (EPA, 1994), UNEP (1994) and the EU have also adopted the framework, the latter noting that this was the most appropriate way to structure environmental information (EC, 1999). Within the EU, Eurostat focuses on Response (the societal mechanisms effecting ecosystem management, in particular, expenditure on environmental protection), Driving forces (environmentally relevant sectoral trends, for example, societal need for and food) and Pressure (e.g., resource exploitation trends). Indicators of State and Impact are the domain of the European Environment Agency (EC, 1999) which is required to communicate the state of the environment for policy-makers. DPSIR has thus been used with increasing frequency for problem solving both by natural and social scientists and they have further refined/defined and applied DPSIR and its derivatives in an on-going process tailored to many different applications.

Gari et al. (2015) recently reviewed 79 published and gray literature sources involving eight DPSIR derivatives for coastal social-ecological systems. More recently, Lewison et al. (2016) reviewed many papers covering 24 relevant DPSIR coastal zone articles. Both publications point out limitations and in particular differences in the terminology or definitions used by different authors. Important differences in definitions particularly concerning States and Impacts, had led to the "modified DPSIR" (mDPSIR) of the ELME EU FP6 project. Within mDPSIR the Impact category was restricted to impacts on human systems thus leading in turn to the definition of the DPSWR framework in the KNOWSEAS FP7 project, where Cooper (2013) replaced Impact with Welfare. However, it has been suggested that it is the "impact on human welfare" rather than "welfare" per se that is important hence leading to the most recent DAPSI(W)R(M) derivative (Wolanski and Elliott, 2015; Scharin et al., 2016) (**Figure 1B**). In another modification, used by social scientists, DPSIR has been related to Goods and Services through EBM-DPSER where Ecosystem Based Management (EBM) is directly related to Driver-Pressure-State-Ecosystem Service-Response (Kelble et al., 2013) or the Ecosystem Services and Societal Benefits (ES&SB) linked-DPSIR approach (Atkins et al., 2011). A further development of DPSIR in the area of human health has been the DPSEEA framework comprising Driving forces-Pressures-State-Exposure-Effect-Action (and sometimes DPSEEAC, where "C" relates to Context), a framework used primarily in risk assessments for contaminants and developed by the World Health Organization (von Schirnding, 2002). Given that such a framework requires indicators to determine whether management actions are effective, successful and sustainable (Elliott, 2011), a further development was in creating indicators such as those of child environmental health using the MEME framework (many-exposures many-effects); this therefore progressed from the linear and pollution-based view of DPSEEA (and other) frameworks (Briggs, 2003).

Given the above history and confusion, as part of the EU funded DEVOTES project (see http://www.devotesproject.eu), we have comprehensively reviewed marine/coastal environmental investigations concerned with the DPSIR framework and its derivatives. We have furthermore assessed its applications, habitats addressed, geographical use, problems and developments, and the general advantages and disadvantages of using the framework to address marine issues. Our aim was to establish the extent to which DPSIR as an overarching framework has been applied to marine and coastal ecosystems and to identify factors which either facilitate or hinder its application. In this way, we focus on the ability and adequacy of the DPSIR framework to analyze and explain the relationships between human uses of the seas and the resulting problems, their management and the communication of these to interested

FIGURE 1 | DPSIR and derivatives development. (A) DPSIR first elaboration, redrawn from the original EU framework (EC, 1999), (B) DAPSI(W)R(M), top of the tree evolution of DPSIR (as defined in Scharin et al., 2016), (C) timeline and development/relationship of DPSIR and derivatives.

stakeholders. To our knowledge, this is the first study that summarizes the use of DPSIR in marine ecosystems.

# MATERIALS AND METHODS

This comprehensive review of the available literature concerned with the DPSIR framework, its "derivatives" and other related frameworks. We used the following search keywords both singly and in combination: DPSIR, PSR, Drivers + Pressures + State + Impacts + Responses, State change, conceptual framework, Marine and Coastal. We conducted primary searches using Web of Science, ScienceDirect, Scopus and Google Scholar and then checked the reference lists of the previous review papers. We only considered publications published in English. We furthermore compiled projects starting with an initial list of European projects where DPSIR was known to be used as a conceptual framework and then we expanded the search using the same keywords used for publications plus the word "project." Our keywordbased screening was narrowed according to the text in the abstracts. We retained documents when the abstract explicitly mentioned the DPSIR framework or any derivative and was linked with coastal or marine ecosystems. Although this review focused primarily on research projects and publications dealing with these ecosystems, the scope broadened to include both projects and publications that present or discuss the framework, regardless of its application to specific case studies and studies that address biodiversity (sensu lato) under the scope of DPSIR.

The 152 studies retained for the review included research papers, review papers, essays, short communications, viewpoint papers, seminar papers, discussion papers, journal editorials, policy briefs, conference long abstracts, monographs, technical reports, manuals, synthesis or final project reports and book chapters (**Figure 2A**). The studies were collated and, after detailed reading, each reference was categorized by "Study site," "Habitat," "Region," "Framework/Model type," "Issue/problem addressed by the study," "Implementation level" and "Type of publication." Appendix 1 in Supplementary Material presents the final list of references and their classification according to the previous categories.

The analysis also considered research projects from 1999 onwards and showed that at least 27-research projects focusing on coastal and marine habitats have used (or are using) the DPSIR framework and/or derivatives as part of their conceptual development phases. Appendix 2 in Supplementary Material shows the final list of projects that were considered, categorized by "Acronym," "Title," "Duration," "Funding institution," "Region," "General objective" of the project, "Framework" used, "Keywords," "Website" and some examples of "Output

references." A further column gives complementary details for the projects where available.

**Box 1** shows the 25 frameworks found in the review and the general components of each conceptual model.

#### RESULTS

#### Published Investigations

Despite the increasing popularity of the DPSIR framework and derivative models among the scientific community since the early 1990s, and the recommendations of OECD (1993), EPA (1994), EEA (1999), and EC (1999) for its application, few studies have focused on the marine habitat (**Figure 2B**). From our comprehensive review, only 26 studies exclusively cover this habitat and from these, only eight illustrate concrete case studies [German Exclusive Economic Zone (Fock et al., 2011); German waters of the North Sea (Gimpel et al., 2013); Baltic Sea, Black Sea, Mediterranean Sea, and North East Atlantic Ocean (Langmead et al., 2007, 2009); Baltic Sea (Andrulewicz, 2005); North and Baltic Sea (Sundblad et al., 2014); Northwestern part of the North Sea (Tett et al., 2013) and Florida Keys and Dry Tortugas (Kelble et al., 2013)]. The remaining 18 studies are either explicitly conceptual or illustrate the framework with generic situations/issues. For example, Elliott (2002) examined offshore wind power and Ojeda-Martínez et al. (2009) studied the management of marine protected areas.

In addition to studies exclusively focusing on marine habitats, 19 others focused simultaneously on marine and coastal habitats (13 of them applied). These cover the Mediterranean region (Casazza et al., 2002), Portuguese marine and coastal waters (Henriques et al., 2008), German North Sea (Lange et al., 2010), West coast of Schleswig-Holstein (Licht-Eggert, 2007), Baltic Sea (Lundberg, 2005; Ness et al., 2010; Lowe et al., 2014), Dutch Wadden Sea region (Vugteveen et al., 2014), UK waters (Rogers and Greenaway, 2005; Atkins et al., 2011), the North East Atlantic (Turner et al., 2010) and the Black Sea (Hills et al., 2013).

Approximately half of the references focus explicitly on coastal habitats (e.g., estuaries, coastal lagoons, entire basins) and half of these are solid case studies where, to a lesser or greater extent, the DPSIR framework or derivatives were applied (for examples, see **Box 2**). The remaining references (N = 29) are not habitat-specific (**Figure 2B**). Approximately 45% of the studies are conceptual (i.e., defining or reviewing the frameworks, using DPSIR and derivatives as reporting outline or as a framework for selecting environmental indicators, assessing biodiversity loss, etc.) (**Figure 2C**).

It is also of note that most publications refer to the use of DPSIR as a framework for specific issues (**Box 2**), for gaining greater understanding, as a research tool, for capturing and


• Tetrahedral DPSIR: Driver - Pressure - State - Impact – Response (adapted)

#### BOX 2 | KEY AND RECENT PUBLICATIONS IN WHICH DPSIR AND DERIVATIVES HAVE BEEN USED. ............................................................................................................................................................................................................................................. Uses of DPSIR framework Indicative references Development and selection of indicators Bowen and Ryley, 2003; EPA, 2008; Espinoza-Tenorio et al., 2010; Bell, 2012; Perry and Masson, 2013; Pettersson, 2015 Assessment of eutrophication Bricker et al., 2003; Cave et al., 2003; Newton et al., 2003; Karageorgis et al., 2005; Lundberg, 2005; Nunneri and Hofmann, 2005; Pirrone et al., 2005; Rovira and Pardo, 2006; Trombino et al., 2007; Zaldívar et al., 2008; Gari, 2010; Garmendia et al., 2012 Assessment of the impact and vulnerabilities of climate change Holman et al., 2005; Hills et al., 2013; Hossain et al., 2015 Fisheries and/or aquaculture management Rudd, 2004; Mangi et al., 2007; Marinov et al., 2007; Viaroli et al., 2007; Henriques et al., 2008; Hoff et al., 2008 in Turner et al., 2010; Knudsen et al., 2010; Ou and Liu, 2010; Nobre et al., 2011; Cranford et al., 2012; Martins et al., 2012 Integrated coastal management Turner et al., 1998b, 2010; EEA, 1999; Licht-Eggert, 2007; Mateus and Campuzano, 2008; Schernewski, 2008; Vacchi et al., 2014; Vugteveen et al., 2014; Dolbeth et al., 2016 Management of marine aggregates Atkins et al., 2011; Cooper, 2013 Assessment of seagrass decline Azevedo et al., 2013 Management of water resources Giupponi, 2002, 2007; Mysiak et al., 2005; Yee et al., 2015 Assessment of wind farm consequences Elliott, 2002; Lange et al., 2010 Ecosystem health assessment Tett et al., 2013; Wang et al., 2013 Framing monitoring activities Pastres and Solidoro, 2012 Synthesis of information related with ecosystem goods and services Butler et al., 2014

communicating complex relationships, as a tool for stakeholder engagement, as the subject of reviews and as the subject for further tool/methodology development linked to policy making and decision support systems. For example, Cormier et al. (2013), using Canadian and European approaches, emphasized DPSIR as a Risk Assessment and Risk Management framework and recommend that ICES (International Council for the Exploration of the Sea) uses this as their underlying rationale for assessing single and multiple pressures.

This review shows clearly that the DPSIR framework and its extensions have mainly been used in a European context (**Figure 2D**). If we consider only those studies that specify a geographical location (N = 100), only 20% of the studies were performed in other regions (e.g., EPA, 1994, 2008; Bricker et al., 2003; Espinoza-Tenorio et al., 2010; Kelble et al., 2013; Perry and Masson, 2013; Cook et al., 2014; Fletcher et al., 2014; Yee et al., 2015 in North America; Bidone and Lacerda, 2004 in South America; Turner et al., 1998a; Lin et al., 2007; Ou and Liu, 2010; Nobre et al., 2011; Wang et al., 2013; Zhang and Zue, 2013; Hossain et al., 2015 in Asia; Walmsley, 2002; Mangi et al., 2007; Scheren et al., 2004 in Africa; Cox et al., 2004; Butler et al., 2014 in Oceania).

# Research Projects

Since 1999, at least 27-research projects focusing on coastal and marine habitats have used (or are using) the DPSIR framework and/or derivatives as part of their conceptual development phases (Appendix 2 in Supplementary Material). Three of these projects had a scope beyond coastal and marine ecosystems, aiming to tackle large-scale environmental risks to biodiversity (e.g., ALARM), to contribute to the progress of Sustainability Science (e.g., THRESHOLDS) and to identify and assess integrated EU climate change policy (e.g., RESPONSES). They have been included in this review as their findings can extend to coastal and marine habitats. One of these projects (ResponSEAble, see Appendix 2 in Supplementary Material for more details) specifically addresses the human-ocean relationship and the need to encourage Europeans to treat oceans with greater respect and understanding (see **Box 3**).

Hence the DPSIR is a framework that several European projects have applied and/or developed but is less commonly the case in non-EU areas. From the many projects that used the framework or derivatives, only one was non-European funded. The USA National Oceanic and Atmospheric Administration Centre for Sponsored Coastal Ocean Research supported the MARES project that developed the EBM-DPSER framework (see Kelble et al., 2013; Nuttle and Fletcher, 2013).

In addition to the scientific context, the role played by the DPSIR framework and/or derivatives also varied markedly from project to project: ELME, KNOWSEAS, ODEMM, DEVOTES, and VECTORS have used the DPSIR framework extensively and some of these projects have developed and further modified the framework (e.g., ELME–mDPSIR and KNOWSEAS–DPSWR). However, this review encountered some difficulties mainly in relation to accessing information (see <sup>∗</sup> in **Box 3**). In other projects, it has been difficult to find specific content even with a careful and thorough examination of websites, lists of deliverables and publications. The lack of easy openaccess acts as a constraint to apply and explore further the knowledge gained by the application of the conceptual frameworks.

# DISCUSSION

The DPSIR framework, as used widely in the literature, aims to act as a tool linking applied science and management of human uses (and abuses) of the seas. Because of this, and as shown here, it is necessary to define the framework and its terms and to show how the framework has been used, to indicate its advantages and benefits, as well as its disadvantages and anomalies. Most importantly there is the need to show whether it fulfills a role and whether it needs modifying and, if so, how it should be modified for future applications in an increasingly complex system of marine uses, users, threats, problems, and management repercussions. In particular, if successful, the DPSIR framework presents a simplified visualization and means of interrogating and managing complex cause-effect relationships between human activities, the environment, and society. It can therefore be used to communicate between disciplines (Tscherning et al., 2012), addressing the different aspects of environmental management (research, monitoring, mitigation, policy, and society) and between scientists, policy makers, and the public (Niemeijer and de Groot, 2008; Tscherning et al., 2012).

# DPSIR—Advantages and Benefits as a Holistic Framework

#### DPSIR—A Wide-Ranging Tool Applicable to All Types of Environmental Problems

Through identifying the progressive chain of events leading to state change, impact, and response, the DPSIR framework and derivatives can potentially be applied to all types of environmental problems. For example, Fock et al. (2011) used PSR to link marine fisheries to environmental objectives concerning seafloor integrity in the German EEZ (Economic Exclusive Zone). Langmead et al. (2007) used mDPSIR to organize information relating habitat change, eutrophication, chemical pollution, and fishing in several European seas. Hills et al. (2013) used DPIVR to assess the impact of, and the vulnerability of marine and coastal ecosystems to, climate change. Lange et al. (2010) used DPSIR to analyse coastal and marine ecosystem changes related with offshore wind farming. Additionally, the framework and its derivatives, have been often used to select and develop indicators for environmental analysis (e.g., Casazza et al., 2002; Andrulewicz, 2005; Rogers and Greenaway, 2005) and inform management decisions (Kelble et al., 2013).

#### DPSIR—A Tool for Risk Assessment and Risk Management

While the DPSIR framework has been used for certain types of problems in the marine environment, the most important aspect is in tackling a set of hazards which, if they adversely affect human assets, economy and safety, become risks to society (Elliott et al., 2014). The hazards may be from natural sources, such as erosion patterns, tsunamis, or isostatic rebound due to geological phenomena. More importantly, from a societal view, they may be anthropogenic such as the over-extraction of material from the sea, the input of chemicals or the building of structures such as windfarms. Human actions may exacerbate the hazards and


lead to greater risks such as the removal of a protective saltmarsh or seagrass bed which otherwise could absorb energy and reduce erosion and the consequences of sea-level rise (Elliott et al., 2016). As such those human-induced hazards and risks emanate from activities and thus lead to the pressures as mechanisms resulting in adverse effects unless mitigated; consequently management responses as measures are required to address, mitigate or reduce those hazards and risks.

Each of those risks requires assessment, both cumulatively and in-combination thus requiring a rigorous framework that can accommodate multiple risks. Cumulative threats and pressures emanate from within one activity whereas in-combination threats and pressures arise from multiple activities occurring concurrently in an area. Therefore, once the risks are identified, by determining the source or cause of the threat and its consequences for the marine system, there needs to be a rigorous risk management framework (Cormier et al., 2013) which has to encompass a suite of measures by covering social, governance, economic, and technological aspects (Barnard and Elliott, 2015). This risk assessment and risk management framework thus especially encompasses the DPSIR approach in which the source and causes of risk are the Drivers and Pressures, the consequences are the State Change and Impacts and the risk is managed through the Responses (see Cormier et al., 2013).

#### DPSIR—A Stakeholder-Inclusive and Communication Tool for Implementing the Ecosystem Approach

DPSIR use has been adopted by and demonstrated to various actors, including research, academia, central and regional policy and decision makers, environmental NGOs, and the wider public. As an example, the EBM-DPSER model for the Florida Keys and Dry Tortugas is the agreed outcome of the joint efforts of over 60 scientists, agency resource managers, and environmental non-governmental organizations (Kelble et al., 2013). Various central administration bodies in Europe have used or are using the framework including, for example, the EEA, UNEP, and the Black Sea Commission (e.g., CLIMBITZ and BS-HOTSPOTS projects). UNEP used the framework as the base for organizing its State of the Environment assessment report (UNEP/MAP., 2012) by including an overview of major drivers in the Mediterranean, an analysis of the pressures, state and known impacts associated with each of the issues addressed by the Ecosystem Approach Ecological Objectives as well as major policy responses. Environmental NGOs have used the framework to present the main issues and to focus their needfor-change message to both the public and policy makers (e.g., WWF, MEDTRENDS project). Despite this, the level of detail depicted in these mostly conceptual applications of the DPSIR framework varies greatly. Most of the publications and projects included in this review do not go beyond the conceptual level although some of the conceptual models do include more details and/or more levels (e.g., Atkins et al., 2011). While O'Higgins et al. (2014) and Scharin et al. (2016) use the framework as a tool to analyse the relationship between human activities and their Impacts or to capture the information needed for marine management, Pettersson (2015) presents a case around eDPSIR and the Port of Gothenburg that includes development of indicators for factors influencing biodiversity and for the assessment of biodiversity itself. Pastres and Solidoro (2012), for the Venice lagoon, emphasize the importance of adopting a DPSIR approach to monitoring strongly supported by modeling tools and mathematical models as these can provide quantitative links between Pressures and State/Impacts. Furthermore, Cook et al. (2014) use detailed conceptual models (EBM-DPSER) together with expert opinion and matrix analyses to explore the direct and indirect relative impact of 12 ecosystem pressures on 11 ecosystem states and 11 ecosystem services.

#### DPSIR—Disadvantages and Anomalies DPSIR—Restricted Coverage and Application

It is emphasized here that there is a widespread and increasing usage of DPSIR-type conceptual framework models in management and issue-resolving. Although many papers are conceptual, there are more case studies over time either used to describe an issue, thereby communicating a problem with an emphasis on the P-S link, where the natural scientists can apply a high degree of detail, or give the framework entirety across the whole cycle, solving problems through management with more involvement of social scientists, but less detail on the P-S links. In a more restricted study, Lewison et al. (2016) noted that only eight of the 24 DPSIR articles that they reviewed actively engaged decision-makers or citizens in their research, thereby completing a full cycle or involving all stakeholders. Bell (2012) emphasized that the challenge for DPSIR is to be both a precise Problem Structuring Method and of wide use to stakeholders.

It is of note that the analysis here clearly shows that the use of DPSIR is primarily European-based, also noted in the Lewison et al. (2016) review, with surprisingly sparse use elsewhere such as in the USA. This should not necessarily be regarded as a less-holistic or integrated approach to environmental issues, although it may be the result of the European framework directives guiding sustainability becoming increasingly complex, inclusive and integrated with respect to ecosystems, humans, and their activities (Boyes and Elliott, 2014). However, driving the European use is not just the institutional organizations of the EU, but also growth through parallel and sequential funding of European projects supporting those EU framework directives, that have used DPSIR as a central pillar in environmental problem-framing. As indicated above, it has been recognized as a valuable problem structuring method, both within scientific circles as well as its adoption by international organizations. It is perhaps less surprising that there is less use in fully marine systems than in coastal systems, where there are greater populations and environmental problems. In our comprehensive review, only 26 studies covered exclusively marine habitats and from these only eight illustrate concrete case studies. It is expected that in future more studies will focus on fully marine ecosystem due to the further implementation of the European Marine Strategy Framework Directive (2008/56/EC) and the European Marine Spatial Planning Directive (2014/89/EU).

#### DPSIR—Non-standard Use of Terms

The wide variety of derivatives is shown in their evolution over time in **Figure 1C**. Most of the frameworks derive directly from DPSIR after 1999, although the DPSEEA-eDPSEEA branch used primarily in health/medicine appeared to diverge earlier. There is some differentiation in use between social sciences and natural sciences, although theoretically DPSIR and close derivatives should cover both types of science. However, more emphasis may be on one or the other depending on the use, where natural scientists may have stronger emphasis on the pressure/state side and the social scientist may have greater emphasis on the impact/response/drivers side. This emphasizes the singular essence of using the DPSIR framework and derivatives in its holistic treatment bridging natural and socio-economic systems and in being a common framework applicable to human and environmental health.

The large number of derivatives indicates that use is wideopen to interpretation and our experience has shown that even specifically within DPSIR there is a high degree of variation in how the major components are interpreted or defined. It thus becomes necessary to define how it is used in every study otherwise there is great confusion in whether a component is ascribed to driver/pressure, pressure/state, or state/impact (Wolanski and Elliott, 2015; Scharin et al., 2016). Under the DPSIR framework (EEA, 1999), there has been longstanding variation in the interpretation and use of various components Drivers-Pressures-State change-Impact-Response, in particular in relation to the P, S, and I components. For example, the term "pressure" is commonly used interchangeably with "activity" or Driving force (Robinson et al., 2008). Similarly, state change and impact are both commonly used in the context of impacts on the environment (Eastwood et al., 2007) whereas impact also commonly refers to the impact on society brought about by a state change to the environment (Atkins et al., 2011). This issue is highlighted by Martins et al. (2012) who also noted variation in the use of indicators between studies (in a fisheries context) as a direct result of this misinterpretation. Whilst there are multiple matrices of the links between sectors, activities and pressures, this has not been carried through to the links between pressures and state changes, state changes and impacts and pressures and impacts, probably due to the large number and complexity of these interactions. Most importantly, the links have not been quantified but remain mostly at the conceptual level.

The recent developments within and between recent EU funded projects (see above), often through their common membership by participants, has helped to standardize definitions and component lists and has given a more rigid structure in starting from concepts and moving to assessments, even though they may have used different definitions.

#### DPSIR—Oversimplifies the Problems

It is emphasized that the concept of DPSIR is well-illustrated to be sound in that it presents a logical, stepwise chain of cause-effect-control events that describe the progression from identification of a problem to its management. However, its application requires a deeper understanding of the relationships between the different DPSIR components (Bell, 2012) before the concept can be effectively applied and its limitations need to be acknowledged. For example, P-S-I components are not mutually exclusive, despite being commonly treated as such. In particular, the P and S components are strongly linked in that Pressure, as the mechanism of change, causes a number of physical state changes that ultimately lead to biological change (hence the variation in the interpretation of that described by the P, S, and I components), or it can cause immediate biological change. The timescale over which this change occurs is variable and, in dose-response terms, can be chronic (subtle over long time periods) or acute (immediate). However, a discrete classification of pressures and state changes does not acknowledge this (Niemeijer and de Groot, 2008; Svarstad et al., 2008) and therefore overlooks an important part of the process leading to state change. Whilst activities are linked to both the D and P components, DPSIR in its current form does not categorically address activities or follow the pathway through pressure, state change, and impact, thus not adequately illustrating clear cause-effect relationships (Carr et al., 2007), which makes it difficult to pinpoint management actions. This problem has been overcome by the DAPSI(W)R(M) model (Wolanski and Elliott, 2015; Scharin et al., 2016), at the top of the "evolutionary tree" in **Figure 1C**, where these relationships are inherently contained with a good balance between natural and social aspects.

The DPSIR approach has to reflect the increasing knowledge of the complexity in the system. It is widely acknowledged that multiple activities occur simultaneously and create incombination effects, that a single activity can give rise to multiple pressures (termed cumulative effects), that a pressure may not necessarily lead to a state change or impact, that a pressure associated with one activity may act differently to the same pressure associated with another activity and that the severity and the potential for state change may differ (Smith et al., 2016). Hence, it will be regarded as being oversimplified if DPSIR focuses on one-to-one relationships, disregarding the complex interactions between multiple pressures, activities, the environment, and society (Niemeijer and de Groot, 2008; Svarstad et al., 2008; Atkins et al., 2011; Tscherning et al., 2012). This can prevent early detection of state changes and impacts and therefore prevent timely, targeted management. Bell (2012) argued that targeted research was necessary to improve understanding of the S and I components of DPSIR (i.e., the state of the environment and its links to social and cultural drivers and impacts on society).

# DPSIR—Solutions and Recommendations for the Way Ahead

The existing models appear to be adequate for depicting the relationships between drivers/pressures and the habitat/biological component that might be affected (or have its state changed) but may be inadequate in addressing state change, what it is or how it arises. The science behind assessments is advancing as new knowledge becomes available, but it still has to deal with ecosystems that are complex, and where pressure-effect relationships on ecosystem components and interrelationships between these components are not fully understood at the quantitative level. This complexity is further highlighted by the 4000+ potential regional seas sector-pressurecomponent "impact chains" identified from the ODEMM project with state change components only identified at the very highest level (Knights et al., 2015). Consequently, whilst DPSIR provides a strong and well-accepted concept, there is room for much more development in refining the concept, methodologies and applications.

#### Clarity of Terms in the DPSIR Framework

It has recently been concluded that the DPSIR approach and its terms have several anomalies and flaws which require it to be revised (Wolanski and Elliott, 2015; Burdon et al., 2015; Scharin et al., 2016). The main discussion is given elsewhere (see for example, Wolanski and Elliott, 2015; Scharin et al., 2016) but in brief, this contends that the terms require more accurate definition. Furthermore, the DPSIR framework does not categorically refer to the human activities which give rise to pressures. The most recent proposal to optimize the DPSIR framework for environmental management (DAPSI(W)R(M)) (pronounced "dapsiworm"), gives a more accurate and complete indication of the DPSIR framework (Wolanski and Elliott, 2015; Scharin et al., 2016, defined in **Figure 1B**). The original components of DPSIR, and their definitions, are retained but clarified by the inclusion of activities within the framework (**Figure 3**). The term **Driver** thus needs to refer to the basic human needs such as food, shelter, security, and goods. In order to obtain these, society carries out **Activities** (fishing, aggregate extraction, infrastructure building) which in turn create **Pressures** which are defined as the mechanisms whereby an Activity has an effect, either positive or negative. These effects, when on the natural system (the physico-chemical and ecological system) then need to be referred to as **State Changes** to separate them from State, a description of the characteristics at one time. These State changes thus encompass alterations to the substratum, the water column and their constituent biota.

Once these effects occur on the natural system then society is concerned that there will be a resulting change on human welfare and on the ecosystem services which ultimately produce societal benefits (Turner and Schaafsma, 2015). Hence, this **Impact** is on the human Welfare. Those Impacts on human Welfare and State Changes on the natural system then need to be addressed using **Responses**. As the EU Directives refer to these responses as **Measures** then we can use the final term as Responses (using Measures). Those measures then include economic and legal instruments, technological devices, remediation agents, and societal desires (Barnard and Elliott, 2015).

#### Expansion of DPSIR—Coping with Complexity

Although as indicated above, DPSIR cannot remain merely a very good concept dealing with a single driver/activity/pressure,

given that ecosystems are rarely affected in this single mode and, from the point of view of effects on State Change, a single activity may cause more than one pressure (or mechanism of pressure) or multiple activities might cause a similar single pressure. Further difficulties may exist with different levels of the same pressure from different activities or differing activity spatial or temporal (timescale) footprints in a defined area. At another level, an impact from several pressures or activities might require a single or integrated response or measure. However, this has been recognized and there have been a number of developments to try and deal with more real-world and complex systems. Atkins et al. (2011) used the first nested-DPSIR approach where their marine case study area had many activities that required multiple DPSIRs nested to provide a more holistic view of complex ecosystems. The individual activity DPSIRs could be grouped with their Response components linked within one common Response area, which would comprise an integrated management plan of the case study area. Scharin et al. (2016) have also used this approach in a Baltic Sea case study with the more-evolved DAPSI(W)R(M) framework, where different sectoral activity chains each produce a state change where their sum total is the current state of the ecosystem. They also re-grouped the activities chains around Response to propose an integrated management plan. DAPSI(W)R(M) also can be nested spatially and sequentially, for example across ecosystem boundaries from a river catchment area through an estuary into the sea. Dolbeth et al. (2016) also used the Atkins et al. (2011) approach with nested DPSIR cycles grouped around a central management response, but with possible interactions between the different independent activity cycles and also beyond single area ecosystems at a pan-European level for lagoonal ecosystems. Smith et al. (2016) have also taken the Atkins et al. (2011) concept forward by rotating common grouped DPSIR cycles around a common pressure (for example seabed abrasion caused by individual DPSIR cycled marine activities) and then building up a three dimensional picture of an area affected by many different pressure cycles. All these developments have shown the adaptability of a simple DPSIR concept to a more complete ecosystem approach.

The essence of any framework which is to be successfully and widely applied is that it should be adaptable and, as emphasized here, have an ability to deal with generic and sitespecific problems. It must encompass the inherent complexity and connectivity in all environments but especially marine, estuarine, and coastal systems. That adaptability resulting from complexity has been described by Gregory et al. (2013), using terms more common in social rather than natural sciences, as the need for the use of Problem Structuring Methods (PSMs) which enable us to learn from Complex Adaptive Systems (CAS) theory. In particular, both in general terms and specifically for marine environmental management this then encompasses and tackles what are regarded by social scientists as "wicked problems," a particular challenge in marine systems (Jentoft and Chuenpagdee, 2009; Gregory et al., 2013). While such "wicked problems" have been long-acknowledged in social sciences (Rittel and Webber, 1973), and regarded as problems that are "difficult or impossible to solve because of incomplete, contradictory, and changing requirements that are often difficult to recognize" (https://en.wikipedia.org/wiki/ Wicked\_problem), they are only now being acknowledged in the natural sciences. Here, we emphasize that we do now have the approaches, framework and background to tackle those problems.

#### Overall Approach

The analysis here has emphasized that, based on a long and extensive use, the DPSIR framework, its large number of derivatives and its recent expanded derivative DAPSI(W)R(M) has the potential as a holistic and valuable tool for analysing cause-effect-response links, determining management measures and communicating these aspects as long as it is used in its entirety. It is required to cover the complexity of coastal and marine systems, the competing and conflicting uses and users and their effects and management but in particular all steps from identifying the source of the problems, their causes and consequences and the means by which they are addressed. It has the potential as a visualization tool for complex interactions and so is valuable for the many stakeholders involved in managing the marine system.

The framework also has the flexibility to be applied across many systems and geographical, it can link marine systems and it can show the connectivity between adjacent systems. In particular, it shows the way in which environmental management is not only embracing complex systems analysis but is very well suited to it because of the many competing aspects. Similarly, to be effectively used it requires effectively merging natural and social science and cooperation between natural and social scientists and thus requires multi- and cross-disciplinary approaches. Hence, it has the ability to solve what may be the seemingly "wicked problem" of integrated marine assessment and management, but with the proviso that we need to keep moving from conceptual and generic models to those which are specific and quantified.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

JP carried out the literature review, analysis, and prepared the figures. CS and KP set up the study and contributed data. All authors contributed to the text. All of the individuals entitled to authorship have been listed, contributed substantively to the research, read, and approved the submission of this manuscript.

#### ACKNOWLEDGMENTS

This manuscript is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), http://www.devotes-project.eu. A preliminary version of this work is given as Smith et al. (2014) found on-line at http://www.devotes-project.eu/wp-content/uploads/2014/06/ DEVOTES-D1-1-ConceptualModels.pdf. The authors would also like to thank two reviewers for helping to improve the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00177

Implementation. Mar. Pollut. Bull. 86, 39–47. doi: 10.1016/j.marpolbul. 2014.06.055


from the Lagoon of Venice. Estuar. Coast. Shelf Sci. 96, 22–30. doi: 10.1016/j.ecss.2011.06.019


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declared a collaboration with the authors and states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Patrício, Elliott, Mazik, Papadopoulou and Smith. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Managing the Marine Environment, Conceptual Models and Assessment Considerations for the European Marine Strategy Framework Directive

Christopher J. Smith<sup>1</sup> \*, Konstantia-Nadia Papadopoulou<sup>1</sup> , Steve Barnard<sup>2</sup> , Krysia Mazik <sup>2</sup> , Michael Elliott <sup>2</sup> , Joana Patrício<sup>3</sup> , Oihana Solaun<sup>4</sup> , Sally Little2, 5, Natasha Bhatia<sup>2</sup> and Angel Borja<sup>4</sup>

*<sup>1</sup> Hellenic Centre for Marine Research, Heraklion, Greece, <sup>2</sup> Institute of Estuarine and Coastal Studies, University of Hull, Hull, UK, <sup>3</sup> European Commission, Joint Research Centre (JRC), Directorate for Sustainable Resources, Ispra, Italy, <sup>4</sup> Marine Research Division, AZTI-Tecnalia, Pasaia, Spain, <sup>5</sup> School of Animal, Environmental and Rural Sciences, Nottingham Trent University, Southwell, UK*

#### Edited by:

*Jacob Carstensen, Aarhus University, Denmark*

#### Reviewed by:

*Katherine Dafforn, University of New South Wales, Australia Michelle McCrackin, Stockholm University, Sweden*

> \*Correspondence: *Christopher J. Smith csmith@hcmr.gr*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *03 June 2016* Accepted: *28 July 2016* Published: *25 August 2016*

#### Citation:

*Smith CJ, Papadopoulou K-N, Barnard S, Mazik K, Elliott M, Patrício J, Solaun O, Little S, Bhatia N and Borja A (2016) Managing the Marine Environment, Conceptual Models and Assessment Considerations for the European Marine Strategy Framework Directive. Front. Mar. Sci. 3:144. doi: 10.3389/fmars.2016.00144* Conceptual models summarize, visualize and explain actual or predicted situations and how they might be tackled. In recent years, Pressure-State-Response (P-S-R) frameworks have been central to conceptualizing marine ecosystem issues and then translating those to stakeholders, environmental managers and researchers. Society is concerned about the risks to the natural and human system posed by those Pressures (thus needing risk assessment) and then needs to act to minimize or compensate those risks (as risk management). This research relates this to the DPSIR (Drivers-Pressure-State(change)-Impact-Response) hierarchical framework using standardized terminology/definitions and lists of impacting Activities and Pressures affecting ecosystem components, incorporating the European Marine Strategy Framework Directive (MSFD) legal decision components. This uses the example of fishing activity and the pressure of abrasion from trawling on the seabed and its effects on particular ecosystem components. The mechanisms of Pressure acting on State changes are highlighted here as an additional refinement to DPSIR. The approach moves from conceptual models to actual assessments including: assessment methodologies (interactive matrices, ecosystem modeling, Bayesian Belief Networks, Bow-tie approach, some assessment tools) data availability, confidence, scaling, cumulative effects and multiple simultaneous Pressures, which more often occur in multi-use and multi-user areas. In defining and describing the DPSIR Conceptual Framework we consider its use in real-world ecosystems affected by multiple pressures or multiple mechanisms of single pressures, and show how it facilitates management and assessment issues with particular relevance to the MSFD.

Keywords: DPSIR, risk, pressure mechanisms, exogenic pressures, endogenic pressures, assessment, benthic trawling

# INTRODUCTION

Determining the cause and consequence of marine environmental problems entails risk assessment, and the responses entail risk management (Cormier et al., 2013). Conceptual models help to summarize, explain and address the identified risk by deconstructing each aspect being assessed, prioritized and addressed (Elliott, 2002). In risk management, these models communicate relevant knowledge to managers and developers as well as having an educational value (Mylopoulos, 1992), to increase awareness of the environmental risks through ocean literacy (Uyarra and Borja, 2016). This enables the development of quantitative and numerical models, hypothesis generation or for indicating the limitation of such models and the available scientific knowledge (Elliott, 2002).

Conceptual models are simple to complex diagrams which collate and summarize relevant information and so by their nature they may become increasingly complex, hence the term "horrendograms" (Elliott, 2002), but they are the pre-requisite for all numerical models.

A key current conceptual framework in widespread use, the Driver-Pressure-State-Impact-Response (DPSIR) framework (OECD, 1993), has developed over the last few decades and is used as the basis for many conceptual approaches addressing Pressure-State change links (Elliott, 2014; Gari et al., 2015). It structures and standardizes conceptualizing complex issues although at present it provides an overly simplistic representation of the relationship between Pressures and State changes, merely indicating that Pressure leads to State change (which may not necessarily be the case). It takes no account of the interaction between different Activities and their associated Pressures occurring simultaneously (Gari et al., 2015). Furthermore, it does not highlight the difference in the nature, severity, timescale or longevity of State changes in relation to pressure intensity, frequency or duration.

Today the DPSIR framework has produced many derivatives and refinements (e.g., Gari et al., 2015; Lewison et al., 2016) with the most extensive review undertaken by Patrício et al. (2016), covering some 152 studies and 27 major projects based around DPSIR, noting more than 23 derivative acronyms, with one further derivative recently being published (DAPSI(W)R(M)— Wolanski and Elliott (2015) and Scharin et al. (2016)). In this manuscript we use the terminology of the "DPSIR framework" rather than any one specific derivative, with emphasis on defining and clarifying components.

An improved understanding of the interactions between Drivers, Pressures and States (or, more particularly, the Pressure-State change (P-S) linkage) is important to help consider possible risk management responses. Pressures are the mechanisms that lead to State changes (and Impacts on human welfare). Hence a Pressure may be analogous to hazard as the cause of risk to an element. In turn, the risk is the probability of effect (likely consequences) causing a disaster or assets affected by the hazard (as human consequences) (Elliott et al., 2014). Smith and Petley (2009) consider that hazard, as a cause, and risk, as a likely consequence, relate especially to humans and their welfare. In the discussion here, the consequence may be regarded as relating to the Impact (on human Welfare) part of the DPSIR cycle (Cooper, 2013). Therefore, we can emphasize the links between the DPSIR approach and risk assessment and risk management.

European Union (EU) Member States must ensure no significant risks to, or impacts on marine biodiversity, marine ecosystems, human health or legitimate uses of the sea. This is enshrined in the Marine Strategy Framework Directive (MSFD; 2008/56/EC), an ambitious legislative instrument for the EU and indeed global marine environmental management which extends control of EU seas out to 200 nm (EC, 2008). Boyes and Elliott (2014) show its importance linking with other holistic and EU framework directives such as the Water Framework Directive (2000/60/EC), Habitats Directive (92/43/EEC), and the Maritime Spatial Planning Directive (2014/89/EU). The MSFD links the causes of marine environmental changes, human Activities and Pressures to their consequences leading to controlling and managing those causes and consequences. If successful, it will protect the natural system while also allowing the seas to produce ecosystem services and deliver societal benefits (Borja et al., 2013). The MSFD focuses on the assessment and monitoring of the functioning of marine ecosystems rather than just its structure. It aims to achieve Good Environmental Status (GES) by 2020 to ensure marine-related economic and social activities and via a roadmap for each Member State to develop an iterative strategy for its marine waters including assessments, determination of GES, establishment targets, indicators and monitoring with a programme of measures to achieve or maintain GES (EC, 2008, 2010; CSWP, 2011; CSWD, 2014). This structured approach allows each EU Member State to ensure there are no significant risks to marine ecosystems, human health or legitimate uses of the sea. Three Member States (Estonia, Denmark and Greece) used DPSIR in their MSFD initial assessments (CSWD, 2014), primarily in their socio-economic analyses.

This review focuses on the relevance of the DPSIR framework to the MSFD to organize and focus assessments in real marine situations including the linkages between multiple Activities exerting multiple Pressures and leading to State changes through multiple mechanisms (i.e., beyond simplistic single DPSIR chains). This ensures the DPSIR approach becomes more usable and a first choice starting approach to addressing marine issues. We standardize the approach incorporating ecosystem characteristics/components to allow ease of use in marine assessments, the movement from concepts to assessments and different assessment methodologies.

# THE DPSIR FRAMEWORK

Rapport and Friend (1979) proposed the first Pressure-State-Response (PSR) framework which was then promoted by the Organization for Economic Cooperation and Development (OECD, 1993) for its environmental performance monitoring. This framework assumes causality that human Activities exert Pressures on the environment (marine and terrestrial), which can induce changes in the State/quality of natural resources. Society addresses these changes through environmental, governance, economic and sectoral responses (policies and programmes). Highlighting the cause-effect relationships can help decision makers and the public see how those issues are interconnected. The OECD (1993) re-evaluated the PSR model, whilst initiating work with environmental indicators. Its use has been extended widely and with many iterations (Patrício et al., 2016). The US Environmental Protection Agency (EPA, 1994) extended it to include the effects of changes in State on the environment (Pressure-State-Response/effects), UNEP (1994) further developed the Pressure-State-Impact-Response (PSIR) framework and the UN Commission on Sustainable Development proposed the Driving Force-State-Response framework (DSR). Here, Driving force replaced the term Pressure in order to accommodate more accurately the addition of social, economic and institutional indicators. Through agencies such as the European Environmental Agency and EUROSTAT, the EU adopted the Driving Force-Pressure-State-Impact-Response framework (DPSIR), as an overall mechanism for analyzing environmental problems (EC, 1999). The EU scheme (**Figure 1A**) shows that Driving forces (e.g., basic economic sectors) exert Pressures (e.g., carbon dioxide emissions), leading to changes in the State of the environment (e.g., changes in the physico-chemical and biological systems, nutrients, organic matter, etc.), which then lead to Impacts on humans and ecosystems (e.g., decreased fish production) that may in turn require a societal Response (e.g., research, building water treatment plants, energy taxes). The Response can feed back to Driving forces, Pressures, State or Impacts directly through adaptation or remedial action (e.g., policies, legislation, restrictions, etc.).

Interpretation of DPSIR has been variable and there has been the need to clarify terms which are often defined/used differently by natural and social scientists. For example, where either:


This lack of clarity has mostly led to further re-definition of one element of the model for example DPSWR where Impact has been replaced/clarified with Welfare (Cooper, 2013) or taking this further to DAPSI(W)R(M) [Driver-Activity-Pressure-State change-Impacts (on Welfare)-Responses (through Measures), (Wolanski and Elliott, 2015; Scharin et al., 2016)]. A clearer terminology (**Figure 1B**), is based on Borja et al. (2006), Robinson et al. (2008), and Atkins et al. (2011), for the DPSIR framework in natural ecosystems:

• **Drivers:** at the highest level, "Driving Forces" are the overarching economic and social policies of governments, and economic and social goals of those involved in industry. At a mid-level they may be considered to be Sectors in industry (e.g., fishing) and at a lower level, Activities in the Sector (e.g., demersal trawling).


# DPSIR CYCLES

Whilst a single DPSIR model or cycle (**Figure 1B**) greatly oversimplifies the "real world," it can conceptualize the relationships between environmental change, anthropogenic pressures and management options. However, to be of value, the model does need to be bounded (e.g., Svarstad et al., 2008), for example, by defining its spatial limits (usually the management unit such as a particular area of sea or length of coast). Furthermore, while a simple DPSIR cycle relates to the Activity or Sector to which it applies, the marine environment is a complex adaptive system (Gibbs and Cole, 2008) with areas subject to several Drivers. Accordingly, this requires to be visualized as several interlinked DPSIR cycles (each representing different interacting Activities or Sectors which compete for the available resources). Atkins et al. (2011) linked separate systems by the Response element, arguing that the effective management of anthropogenic impacts requires integrated actions (involving many types of response) affecting all relevant Activities; in contrast, Scharin et al. (2016) linked DAPSI(W)R(M) in similar cycles around State changes. Separate DPSIR cycles, each relating to a different Activity, can also be linked by Pressures and reflect the concept that several different Activities can create the same environmental pressure (**Figure 2A**). Following Atkins et al. (2011), **Figure 2A** illustrates how a single Pressure (the central blue circle) provides a common link between five separate DPSIR cycles, which represent five separate Activities. For clarity, the links within each individual DPSIR cycle have been simplified (e.g., by omitting the direct R-P link within each cycle and the links between other D, S, I, and R elements for different cycles a la Atkins et al., 2011). Linking separate DPSIR cycles in this way, and placing Pressure at the heart of the model, focuses attention on the Pressure as the system element that needs to be managed, thus supporting the assessment of Pressure-State change linkages. Hence, any such single Pressure may bring about a State change across a number of different ecological components. In essence, we assess State

FIGURE 1 | Driving forces-Pressure-State-Impact-Response evolution. (A) DPSIR redrawn from the original EU framework (EC, 1999). (B) DPSIR as used by the authors. Reproduction is authorised provided the source is acknowledged.

FIGURE 2 | Multispace DPSIR cycles. (A) Separate DPSIR cycles linked through a common Pressure element (e.g., abrasion pressure from the activities of benthic trawling, anchoring, dredging, etc.). (B) Example of linked DPSIR cycles in a particular ecosystem with individual separate Pressures (P1-P3), each associated with discrete Activity types (A1-A4). It is important to acknowledge that, just as the same Pressure may be generated by more than one Activity, so the same Activity may give rise to a number of different Pressures. For example, (P1) might represent abrasion, and may link DPSIR cycles relating to three Activities—benthic trawling (A1), anchoring (A2), and dredging (A3); (P2) might represent marine litter, linking benthic trawling (A1), the development of both non-renewable energy facilities (A4) and renewable energy facilities (A5); whilst (P3) might represent substratum loss, linking DPSIR cycles relating to the development of both non-renewable energy facilities (A4), renewable energy facilities (A5), and dredging (A3). For simplicity and clarity, Responses are shown here as having limited, within-cycle effects. In practice however, Responses within one DPSIR cycle may affect one or more of those other DPSIR cycles that are linked by a common Pressure or, indeed, DPSIR cycles that are not directly linked.

changes and Impacts but we manage the Drivers, the Activities and the Pressures, and in some cases State changes. Having a series of nested and linked DPSIR cycles, and linking these across ecosystems, accommodates many Pressures within one area (Atkins et al., 2011). Thus, a nested DPSIR cycle in a nearshore area, for example, has to link with those in the catchments, estuaries and at sea. This overcomes some of the difficulties in applying the framework to dynamic systems, cause-consequence relationships, multiple Drivers and only linear unidirectional causal chains.

It is necessary for the framework to accommodate multiple pressures and state changes which can lead to cumulative, synergistic or antagonistic impacts (Nõges et al., 2016; Teichert et al., 2016; **Figure 2A**). For example, the different cycles in **Figure 2B** representing different Pressures or classes of Pressure, P1, P2, and P3 acting on an ecosystem [for example, (P1) might represent abrasion, and may link DPSIR cycles for three Activities—benthic trawling (A1), anchoring (A2), and dredging (A3); (P2) might represent marine litter, linking benthic trawling (A1), the development of non-renewable energy facilities (A4) and renewable energy facilities (A5); whilst (P3) might represent substratum loss, linking DPSIR cycles relating to the development of non-renewable energy facilities (A4), renewable energy facilities (A5), and dredging (A3)]. Hence there are many links between DPSIR chains across the different levels; for example, where the Responses and Drivers for one Activity interact with or affect the Responses and Drivers for a different Activity.

#### DPS CHAINS IN THE MSFD

The MSFD lists indicative characteristics, pressures and impacts to be taken into account during assessments (EC, 2008). There is some ambiguity in terms where the Directive presents "pressures" and "impacts" together, when pressures (P, Pressures in the DPSIR framework) should be distinguished from Activities, and Pressures should be distinguished from adverse effects on the natural system (i.e., S, State changes in the DPSIR framework). These lists have evolved since first publication, for example, in DIKE (2011) and CSWP (2011), but some ambiguities still remain.

## Activities

In addition to clarifying the terminology, we also advocate alternative tables that list Activities and Pressures based on the work of a number of MSFD-related EU funded projects, particularly ODEMM (https://www.liverpool.ac.uk/odemm), VECTORS (www.marine-vectors.eu) and DEVOTES (www.devotes-project.eu). A list of possible and/or existing Activities is needed from which a subset can be extracted that may contribute to a greater number and/or more detrimental pressures for risk assessment and risk management and used to fulfill programmes for monitoring and response measures. **Table 1** shows a complete Activities list contributing to Pressures, refined from the ODEMM project (White et al., 2013) where Activities had been separated into Sector and sub-sectors. To avoid duplication with either Driver or Activity, we consider that the term "Sector" is unnecessary, meaning that only an Activity is required to produce Pressures. Overall 13 major Activities characterize the wide range of sea uses.

#### Pressures

The MSFD Pressures list (EC, 2008) identifies eight Pressure themes with 18 individual Pressures or mechanisms. Robinson et al. (2008) listed further Pressures, which were later updated by White et al. (2013). Except for Pressures from climate change, Pressures predominantly relate to anthropogenic Activity, also referred to as endogenic managed Pressures (Atkins et al., 2011; Elliott, 2011; Elliott et al., 2014), i.e., emanating from within the system to be managed. Exogenic unmanaged Pressures, in contrast, are from outside of the system and mostly relate to climate change, isostatic/eustatic change, or seismic activity. Elliott (2011) emphasizes that whereas the causes and consequences of endogenic managed Pressures are addressed within a management scheme for a marine area, only the consequences (as opposed to the causes) of exogenic unmanaged Pressures can be addressed at management scales; for example, the consequences of climate change can be addressed locally whereas the causes require global action.

As the MSFD only refers to an incomplete list of endogenic Pressures, we have revised both the MSFD and the White et al. (2013) lists to give 26 managed Pressures of which 18 were listed in the Directive (**Table 2**) and 7 are unmanaged Pressures (**Table 3**). The unmanaged Pressures allow climate change to be considered as it has been omitted in MSFD implementation and barely mentioned in the Directive (Elliott et al., 2015). The latter concluded that shifting baselines, resulting from climate change, need to be accommodated and revised during monitoring, environmental status assessment and in management actions (i.e., programmes of measures). The spatial and temporal variation in the response of the various biological components to climate change needs to be understood, as well as their ability (or lack of it) to adapt and reach equilibrium. Climate change may also exacerbate other Pressures and changes in the Descriptors (11 broad qualitative environmental descriptors for which GES must be assessed) for example the movement of nonindigenous species by increased shipping, but these effects may be indistinguishable from those arising from other anthropogenic Activities. Long-term, spatially extensive data sets will be needed to identify changes in ecological indicators. Although such data sets are not widely available for all Pressures, some efforts have been made to solve this gap. For example, for non-indigenous species, several databases hosting and sharing such information have been gathered in the European Alien Species Information Network (EASIN, http://easin.jrc.ec.europa.eu/) (Katsanevakis et al., 2015).

# Using the DPS

As a major example of the complexity of interactions we consider just one Activity, extraction of living resources from benthic trawling and its multiple individual Pressures affecting the seafloor environment (see Blaber et al., 2000, and conceptual models in Gray and Elliott, 2009). In terms of Pressures, benthic trawling targets and results in the selective extraction of species but also brings about the non-selective extraction of other living resources and causes abrasion, scouring and turning over the sediment as well as causing compaction and other changes in the seabed. Fishing vessels can also input various objects/elements into the marine environment (e.g., noise, synthetic compounds, non-synthetic compounds, other substances, litter), and cause death by collision. Benthic trawling includes some 12 individual primary and lesser Pressures (**Table 2**) each with differing effects.

In turn, the trawling Pressures may be site-specific, acting on specific habitats and ecosystems; **Table 4** shows the European Commission MSFD-provided ecosystem components, with the first part highlighting habitats potentially impacted by benthic trawling—predominantly shallow to shelf sublittoral sedimentary habitats. The habitats in turn define and link with the potential

#### TABLE 1 | Activities contributing to Pressures (modified extensively from White et al., 2013).


biological components present (e.g., shallow sublittoral muddy sand supporting seagrass).

Within any one habitat, the different Pressures may affect several environmental characteristics (**Table 4**, highlighted) which also define/affect the niches of species groups (**Table 4**, highlighted) such that following a Pressure, the environmental characteristics may no longer be suitable for that species group. Each of those species groups has structural and functional characteristics (**Table 4**, highlighted) that may be affected to various extents. Although most of the effects that have been highlighted are direct, there are indirect effects for example through damage or habitat modification or changes to predatorprey relationships.

The situation is further complicated as different Pressure levels create different State change trajectories; for example, a Pressure causing large scale direct mortality will immediately reduce species, abundance, biomass, diversity, community structure, etc., and the duration of this depends on the nature of the habitat and its recovery potential (Duarte et al., 2015). The degree of Pressure then determines the severity and timescale of wider effects (e.g., at higher trophic levels) or on individuals (e.g., crushing, loss or damaged limbs or shells through collision with fishing gear) so that energy is allocated to individual recovery rather than growth/reproduction etc. In the long term, biomass, some components of population and community may be compromised with wider effects at the ecosystem level.

# REFINING DPSIR PRESSURE-STATE CHANGE RELATIONSHIPS

Whilst it is well understood that Pressures on environmental systems can result in varying degrees of State change causing, for example, a loss of biodiversity and ecosystem services, the process by which those Impacts occur is complex. For a single, specific Pressure, the relationship between Pressure and Impact varies according to the degree of Pressure (e.g., spatial extent, duration and/or frequency, intensity), the habitat type upon which the Pressure is acting, the component species and those species in the wider ecosystem which they support. This produces many potential Pressure-State change trajectories that increase in complexity with concurrent potentially synergistic or antagonistic combinations of Activities and Pressures (Griffen et al., 2016). Hence the need to move from a conceptual framework to "nested horrendograms" to encompass the interlinked complexity (e.g., Elliott et al., 2015). Thus, generic processes leading to Impacts for a selection of Activities, Pressures, habitat types and biological components, then require specific, detailed trajectories that are site/system specific and specific to the nature of the Activities and their associated Pressures.

Current attempts to link Pressure with State change assume Pressure to act as a single mechanism leading to State change

#### TABLE 2 | Endogenic managed Pressures in the marine environment.


\**Notes original pressure listed in the MSFD. #Whilst extraction is clearly an Activity, the specific extraction of non-living resources or species is considered here as a Pressure, as extraction is the mechanism of State change.*



(Knights et al., 2011; Robinson and Knights, 2011; White et al., 2013). Hence, Pressure is the cause of physico-chemical and biological State changes which, through lethal or sublethal processes, compromise the performance or survival of one or more level of biological organization (cell, individual, population, community) (see **Figure 3**, overall organization). For example, the physical environment may be unsuitable to support the existing biological community, thus changing species

#### TABLE 4 | Marine Strategy Framework Directive (MSFD) components, highlighted for those impacted by benthic trawling.

#### Habitats (predominant habitats related to monitoring)



• Chemical levels/contaminants

Environmental characteristics

*Bold highlighted: strongly impacted by benthic trawling; light highlights indicates lesser influence by benthic trawling. Components adapted from (EC, 2008, 2010) and CSWP (2011, 2012). Benthic habitats: littoral (approximately 0–1 m – intertidal zone), shallow sub-littoral (approximately 1–60 m), shelf sub-littoral (approximately 60–200 m), upper bathyal (approximately 200–1100 m), lower bathyal (approximately 1100–2700 m), abyssal (approximately* >*2700 m).*

composition and relative abundance (O'Neill and Ivanovic, ´ 2016).

Achieving State change can be a progressive process and whilst changes to the physico-chemical and biological structure may be classed as State changes, paradoxically they may also be viewed as the mechanisms through which a Pressure acts to cause a biological State change (i.e., not mutually exclusive as the DPSIR model suggests). For example, a substratum change during an Activity is a physico-chemical State change and at the same time is a mechanism (and hence a Pressure) resulting in a biological State change in the benthos (see examples on trawling impacts in Clark et al., 2016). Hence, whilst most Pressures are associated with physical State changes (e.g., hydrodynamic changes, substratum changes), the direct removal of species, the introduction of non-indigenous species and the input of microbial contaminants represent biological mechanisms of change.

These physico-chemical and biological modifications to the environment lead to a series of biological State changes, which can occur at any level of biological organization (Solan and Whiteley, 2016). Responses may be lethal (referring to loss) as a result of direct mortality associated with the Pressure, direct removal (e.g., by fishing gear) or emigration, or sublethal. Lethal responses can have immediate, direct effects on an individual, population and community (and ultimately ecosystem) in terms of the species composition, their relative abundance and biomass, total population and community biomass, trophic interactions and other functional attributes such as primary and secondary production and biogeochemical cycling. Sublethal responses relate to physical, chemical or biological damage caused by the Pressure at an individual level, whereby the organism survives but its performance and, therefore, contribution to ecosystem processes is compromised. Hence, biological State changes to the lower levels of organization (individual, population) will, if unchecked, lead to higher level (community, ecosystem) changes (Borja et al., 2015). The ultimate degree of State change at a community or ecosystem level associated with lethal and sublethal mechanisms of State change may be broadly similar but their severity, extent and duration will differ (Amiard-Triquet et al., 2015).

Despite this, the inherent variability and complexity throughout the levels of biological organization may mean that an effect at a lower level does not necessarily manifest itself at higher levels, i.e., stressors at lower levels (e.g., cellular, individual) may get absorbed so that the higher levels (e.g., population, community, ecosystem) do not show any deleterious ecological effects. The ability to absorb that stress has been termed environmental homeostasis (Elliott and Quintino, 2007).

The severity and sequence of biological State changes will vary according to:


• sensitivity of the balance of interactions within and between habitats and biological components.

Using abrasion from benthic trawling as a specific worked example (**Figure 3**, fine detail), and assuming a sublittoral sedimentary (mud/sand) habitat, there are several physical State changes that may arise and which may, in turn, lead to a series of biological State changes (O'Neill and Ivanovic, 2016 ´ ).

The physical State changes associated with abrasion can be divided into those that cause immediate biological State change at higher biological levels (population/community/ecosystem), for example, by direct mortality, and those that cause a progressive State change over an extended time period (Eigaard et al., 2016; O'Neill and Ivanovic, 2016 ´ ). This leads to two different trajectories of State change (lethal and sub-lethal), which act over different timescales and may ultimately differ in severity and longevity (Gilkinson et al., 2005) or require a different intensity of stressor.

With respect to sub-lethal effects, "abrasion" can lead to various sedimentary changes (**Figure 3**, Physico-chemical State Change box). Since the benthic inhabitants are intimately linked to the substratum (Snelgrove and Butman, 1994), such changes, if of sufficient severity or duration, will physically impair biological community structure and its long term survival, larval settlement and recruitment (Alexander et al., 1993). Similarly, the removal of species will affect a feedback loop whereby the organisms modify the sedimentary conditions through bioturbation, bioengineering, biodeposition, etc. (e.g., Gray and Elliott, 2009). Additionally, those organisms that are more mobile may simply relocate to other areas. Whilst sedimentary changes can lead to species loss, it also presents opportunities for colonization by new species leading to an overall change in community structure. Coupled with this may be a change in community function, if species are replaced by functionally different species (Koutsidi et al., 2016). Abundance, biomass and secondary production would be influenced (and perhaps species richness and diversity), which may impact on wider ecosystem processes (Hiddink et al., 2006; Queiros et al., 2006). Whilst this impact would be more gradual than in the second (lethal effects) scenario, and may be partly counteracted by colonization by new species, overall community structure and function may nevertheless be altered.

Additionally, sub-lethal effects may arise through (for example) morphological damage (caused by interaction with fishing gear) and the associated physiological stress, changes in the physico-chemical parameters of the water column (e.g., dissolved oxygen, suspended solids), clogging of respiratory structures, inability to feed or burrow and behavioral modifications (Tillin et al., 2006). Subsequently, somatic growth and reproductive capacity may be compromised as a result of, for example, increased respiration rate, increased ammonia production in response to stress, re-allocation of resources to survival and recovery (e.g., Widdows et al., 1981) or evolutionary adaptations that enable accelerated maturation and early reproduction at the expense of ultimate body size (Mollet et al., 2007; Elliott et al., 2012). These effects may

habitat and links to the MSFD descriptors (e.g., through physico-chemical, structural or functional indicators at different levels from individual to ecosystem for

descriptors D1 biological diversity, D3 commercial fish species, D4 food webs and D6 seafloor integrity).

initially be apparent at the individual or population level but, if sustained, will ultimately change abundance, biomass and function at community and ecosystem levels (Thrush et al., 2016).

Lethal effects will create immediate State changes at the population and community level, including biomass and abundance declines in both target and non-target species (Hiddink et al., 2006; Koutsidi et al., 2016). In the longer term, and particularly with frequent benthic trawling, a sustained reduction in species richness and diversity may occur, coupled with changes to community structure and function (Bremner et al., 2003). Population structure in disturbed habitats may also be altered, particularly in longer-lived species, as certain size classes are selectively removed or where species of a more opportunistic nature allocate resources to reproductive output rather than somatic production resulting in a population dominated by small and or/young individuals. Ultimately, these State changes will reduce secondary production which, coupled with altered predator-prey interactions, will alter higher ecosystem processes (Thrush et al., 2016).

In terms of timescale, and regarding the ability of MSFD indicators to detect State change, such sub-lethal population and community level changes are likely to be relatively acute (and rapidly detectable) processes. The duration would depend on the sensitivity of the species and habitats, their resilience (or their potential to recover to an alternative state which supports wider ecosystem processes) and the intensity of the Pressure (or causative Activity). It also depends on the processes in the first (sub-lethal) scenario, since the two do not occur in isolation, whereby physical and biological changes to the environment will influence recovery rates and trajectories (Foden et al., 2010; Lambert et al., 2014).

The above changes in these scenarios (lethal and sub-lethal) have the potential to ultimately produce overall negative effects at higher trophic levels and wider ecosystem processes. The difference between the scenarios lies in the complexity/detail trajectory between the application of a Pressure and the resultant State change. Finally, the effects of trawling can result in human welfare being affected through the reduction in the provision of ecosystem services (Muntadas et al., 2015) and societal benefits (Atkins et al., 2011). The resulting changes compromise the performance or survival of an ecological component and so may bring about State change detected by MSFD descriptors [e.g., at the population, community or ecosystem level for descriptors D1 (biological diversity), D3 (commercial fish species), D4 (food webs) and D6 (seafloor integrity)].

Whilst the scenario above relates only to a single Pressure, abrasion, this Pressure may potentially arise as the result of a number of different Activities (**Table 5**).

### ISSUES IN MOVING FROM CONCEPTS TO ASSESSMENTS

Environmental management issues involve many challenges in moving from a conceptual framework to a data-based or expert judgment-based assessment. This involves identifying all the components and their linkages (e.g., D-P-S chains), and data/indicators and their quality or thresholds, etc.

#### Regional Seas

The European regional seas cover approximately 11,220,000 km<sup>2</sup> (EEA, 2014) with a wide range of environmental conditions and different ecosystems, which vary in diversity and sensitivity. This affects the repercussions of human Activities and their resultant Pressures. Pressures in one regional area may not have the same footprint (type, extent, or duration) in another area because of differing conditions (see examples in the Baltic Andersen et al., 2015, Mediterranean Claudet and Fraschetti, 2010, and Black Seas Micheli et al., 2013). For example, the Mediterranean Sea is characterized by high salinity, high temperature, predominantly wind-driven or water mass difference-driven currents, deep water, oligotrophic conditions with a fauna exhibiting low abundance and biomass. In contrast northern waters have opposing characteristics where, for example, tidallydriven mixing may create a different footprint of a Pressure (Andersen et al., 2013). The regional seas also have contrasting developmental and socio-economic issues producing complex and fragmented governance systems (Raakjaer et al., 2014). Although each of the regional seas have their own conventions (North-East Atlantic, Oslo/Paris Convention; Baltic Sea, Helsinki Convention; Mediterranean Sea, Barcelona Convention; Black Sea, Bucharest Convention) with similar objectives and targets, there are differences in the cohesiveness of each regional seas EU Member States and state of developed/stability of the related bordered countries. Geographically differing stages of development influence the status, quality and quantity of monitoring programmes producing data for assessments of Drivers, Pressures and State change (Patrício et al., 2014).

## Data Availability

Within a causal link framework and to provide the route for and efficacy of management, indicators and their component indices/metrics are needed to determine the level of Pressure, and changes in State and Impact (e.g., Aubry and Elliott, 2006). The trajectory of State change can be used to determine targets or reference conditions for the assessment of the indicators (see Borja et al., 2012) which requires developing assessment methods or indices such as those within the Water Framework Directive (Birk et al., 2012). However, they need to be validated and calibrated against independent abiotic datasets (Birk et al., 2013). As some of the MSFD descriptors are related to Pressures, whilst others are related to State change, data analysis is needed to assess the effects that Activities have on marine physical, chemical and biological quality. Consequently all the relevant Activities, Pressures, States and their indicators need to be identified together with the linkages (cause-effect interactions) between them. The ODEMM Project linkage framework (Knights et al., 2013; White et al., 2013), for example, provides a means to fully evaluate all components that can affect the achievement of GES in a fully integrated ecosystem assessment. Applying a framework relies on having not only indices of change but also baselines, thresholds and targets against which to judge that change. In addition, there is the need to define the inherent

#### TABLE 5 | Activities (related to Table 1) that may give rise to abrasion Pressures on the seabed.


*(Continued)*

#### TABLE 5 | Continued


variability ("noise") against which the "signal" of change is measured (Kennish and Elliott, 2011). Each of these requires a fit-for-purpose data background for each biological and physicochemical component relevant to a particular stressor. Given that for many Activities, the amount of Pressure required to produce a given State change and thus Impact on human welfare is unknown, then the amount of data required to determine and assess the State change is also unknown. Furthermore, although this could be determined through power analysis, it cannot be used unless the inherent variability in the components is known. Hence, it is likely that the approaches advocated here will continue to be semi-qualitative at best and reliant on expert judgment (see below).

#### Cumulative/In-Combination Effects

As single Activities exert multiple Pressures and the marine ecosystem usually supports multiple Activities, we need to consider cumulative/co-occurring (within an Activity) and incombination (between Activities) effects. The multiple Pressures will rarely be equal and will lead to cumulative and incombination effects which may be synergistic or antagonistic (Griffen et al., 2016). To indicate some of difficulties in assessing cumulative impacts, Crain et al. (2008) analyzed 171 multiple stressors studies in marine and coastal environments and found effects to be 26% additive, 36% synergistic, and 38% antagonistic, while interaction type varied by response level, trophic level, and specific stressors. In another meta-analysis of 112 experimental studies Darling and Côté (2008) found similar combined effects of two stressors with 23% additive, 35% synergistic and 42% antagonistic. Despite the lack of knowledge at the community and ecosystem level elucidating or predicting effects of combinations of individual Pressure impacts, we can measure the status of an ecosystem that is impacted by multiple Pressures (Griffen et al., 2016; Nõges et al., 2016; Teichert et al., 2016). Hence, while we can identify some elements, we are unclear regarding the precise changes at a sub-species, species, population or community level.

Co-occurring multiple Activity/Pressure impacts, as cumulative and in-combination threats or impacts, have been investigated according to the footprints of a particular Driver/Activity and their overlap with habitats using spatial mapping/modeling (Nõges et al., 2016). Cumulative impacts (including both overlap and weighted cumulative methods) have been investigated at a global level by Halpern et al. (2008) producing global impacts maps but also at the European level, for example in the Baltic (Korpinen et al., 2013; Andersen et al., 2015), eastern North Sea (Andersen et al., 2013) and the Mediterranean-Black Seas (Claudet and Fraschetti, 2010; Coll et al., 2011; Micheli et al., 2013). These techniques may not be of direct use in assessing State changes, but may nevertheless be of value in spatial planning applications, for example, in identifying areas where high levels of protection may be necessary. It should be noted that an Activity does not always have to lead to an impact especially if mitigation measures are employed.

# Assessment Scales and Scaling Up to Regional Seas

The connections between ecosystem features and human Activities (and their related Pressures) should determine the appropriate scale at which the ecosystem approach should be implemented (Borja et al., 2016). Defining these scales and their boundaries is imperative for any ecosystem-based management (EEA, 2014). For a well monitored small bay, a comprehensive assessment can be normally made, because the Drivers, Pressures, and State changes could be well understood, mapped and assessed. However, at a larger scale, not all issues may be well known; some areas have quantitative data, some have no data, and a more widespread range of very differing habitats may be included. Borja et al. (2013) suggest that the fundamental challenge of obtaining a regional quality status is by either having a broad approach and omitting or down-weighting point-source problems or summing the point-source problems (which may cover only a very small area) to indicate the quality status of the whole area. State change becomes much more complicated and diverse. An important issue is the mismatch between the quantitative information from the pressures and different descriptors and biodiversity components at a large scale (i.e., regional or sub-regional sea), making difficult the large-scale assessment of the response of indicators of change. During the first phase of MSFD implementation, the baseline assessment of the EU marine area in 2012 gave a very broad understanding of Pressures and impacts from human Activities. Although most Member States have reported on most descriptors, providing an overview of their marine environment, the quality of reporting varies widely between countries, and even within individual Member States, from one descriptor to another (EC, 2014; Palialexis et al., 2014). In addition, when different countries are involved in the assessment, the relevant information may come from many different sources, which each have their own assessment timescales, aims, indicators, criteria, targets and baseline values thus limiting not only direct comparison, but also coherence in implementation (Cavallo et al., 2016).

### Levels of Confidence

A conceptual framework such as DPSIR aims to encompass all key components and interactions of an ecosystem problem. However, when moving to the next step of assessment, incorporating many types of data, confidence in the outcome becomes an issue for both the assessors and the users of the assessment. The level of confidence in an assessment depends on the degree of uncertainty associated with the method of assessment, data availability and adequacy, and knowledge and understanding. This requires distinguishing between the lack of knowledge and natural variability (Hoffman and Hammonds, 1994), and uncertainty in the future forecasted state (due to lack of long-term data sets and historical data and/or spatiotemporal variability of a biological indicator) as well as in the resulting ecosystem state post-management action, present challenges in target setting (Knights et al., 2014). Uncertainty is mostly addressed through monitoring programmes that have adequate spatio-temporal coverage (Borja et al., 2010), although the absence of reference conditions or clear targets makes it difficult to establish an accurate assessment (Borja et al., 2012). However, confidence can also be given through a range of methods from cumulative qualitative assessment of each metric and, for example, a traffic-light overall confidence assessment to a separate quantitative confidence metric (e.g., Andersen et al., 2010; Carstensen and Lindegarth, 2016). Despite this, most uncertainty is due to poor definition in the determination of deviation from that expected, in a physico-chemical or biological component. If the agreed targets against which indices and metrics are judged are not sufficiently well defined, then it is not possible to judge the efficacy of management measures.

### MOVING TO THE NEXT STEP: ASSESSMENT

The intricacy and complexity of Driver-Pressure interactions, and the relationship of Pressures to State changes makes it difficult to undertake high level or quantitative assessments for management purposes. It requires knowledge of all the potential causal chains and State changes. The possible methodologies are broadly either a matrices approach or as a form of ecosystem modeling but the assessment is only as good as the knowledge and detail applied (Borja et al., 2016).

#### Simple Matrices Approach

Matrices are simple tables where Drivers (or, more specifically, the Activities resulting from them) can be related to Pressures, and where Pressures can be related to ecosystem components. These allow the identification of chains formed by particular causal links and permit linear analysis of the impact chain (Knights et al., 2013). The matrices record relationships between Activity and Pressures, and between Pressures and ecosystem components. The relationships that are represented are complex with, for example, any single Activity potentially causing many Pressures, and any single Pressure being caused by more than one Activity (i.e., a many-to-many relationship). The matrices can be linked simply by an overlap (Pressure X affects component Y) or through more detailed information on potential levels of interaction, for example showing high/low or increasing/decreasing changes to a component. The degree of State change caused by a Pressure on a habitat can be assessed in terms of: Activity area or footprint, frequency, persistence, and characteristics of the habitat/ecosystem component impacted, including sensitivity and resilience (ability for recovery) (Knights et al., 2015). Matrices and Pressure assessment approaches were used extensively in the ODEMM project (Knights et al., 2011; Robinson and Knights, 2011; White et al., 2013) and in the DEVOTES project (Barnard et al., 2015). They have also been used as standard tools for Pressure assessments by, for example, HELCOM and OSPAR (Johnson, 2008). Complex matrices and linkages can be compiled through databases where programming can be used to analyze and filter data, for example, to highlight Activities that need to be managed or sensitive ecological components that might be at risk of State change [e.g., the PRISM and PISA Access database tools developed through the U.K. Net Gain project (Net Gain, 2011)]. The accuracy and value of the matrix approach depends on identifying and parameterizing components and linkages for a particular area. They are valuable for assessments, and depending on how comprehensive they are, will show impacted components which may then allow prediction of the State changes under given circumstances.

# Ecosystem Models

With the move toward ecosystem-based management, much attention has been devoted to ecosystem modeling. These models may be conceptual, deterministic (in which there is underlying theory or embedded mathematical relationships) or empirical in which the links are described statistically even when there is no apparent underlying theory. Some relate to the management of particular aspects of the ecosystem (e.g., Robinson and Frid, 2003; Plagányi, 2007) whilst other, more recent, models concern the whole natural ecosystem or socio-ecological system (i.e., "end-toend" models) model development/application (e.g., Rose et al., 2010; Heath, 2012). Piroddi et al. (2015), in reviewing whole ecosystem models with respect to MSFD assessments, note that they are more relevant as they may better represent interactions with biodiversity components, for example, ECOPATH with ECOSIM, ATLANTIS or coupled lower trophic and high trophic models (Rose et al., 2010). The ability to apply models to Drivers and Pressure effects relies on knowledge of Activities/Pressures and being able to parameterize the elements accordingly. For example, if trawling causes a 30% reduction in suspension feeders in a modeled area, this figure can be applied to that biological component (according to temporal or spatial scales) (Petihakis et al., 2007). A specific model may not have the resolution to apply a precise mechanism or be applied at individual habitat scale. Whilst pelagic habitats may be defined by salinity, temperature, depth, nutrients, oxygen, etc., benthic habitats as different spatial entities (an important setting for all species groups) are generally not parameterized in models. Nevertheless, such models may well be able to accommodate indirect effects such as changes in predator-prey relations or be used in a predictive manner where climate change could be de-coupled from anthropogenic impacts.

# Bayesian Belief Networks

Bayesian Belief Networks (BBNs; also referred to as belief networks, causal nets, causal probabilistic networks, probabilistic cause effect models, and graphical probability networks) offer a pragmatic and scientifically credible approach to modeling complex ecological systems and problems, where substantial uncertainties exist. A BBN is a graphical and probabilistic representation of causal and statistical relationships across a set of variables (McCann et al., 2006). It consists of graphically represented causal relationships (for example, the DPSIR D-P-S chain links) comprised of nodes that represent component variables and causal dependencies or links based on an understanding of underlying processes/relationships/association. Each node is associated with a function that gives the probability of the variable dependent on the upstream/parent nodes. As each variable is set with best data available and can include expert opinion, simulation results or observed data, this is flexible and also allows the information to be easily updated with improved data (from Hamilton et al., 2005; Pollino et al., 2007). Notwithstanding their potential, BBNs represent a relatively new modeling approach. They have only been applied to marine assessments in a limited way (e.g., Langmead et al., 2007; Stelzenmuller et al., 2015; Uusitalo et al., 2015). However, BBNs are becoming an increasingly popular modeling tool, particularly in ecology and environmental management. This is largely because they can be used in a predictive capacity and also, because they use probabilities to quantify relationships between model variables, while explicitly allowing uncertainty and variability to be accommodated in model predictions (Barnard and Boyes, 2013). They show high promise in adaptive management being iterative and especially in being able to combine both empirical data and expert knowledge, a necessary feature given the often poor data for those empirical relationships.

# The Bow-Tie Approach

The Bow-tie method was initially presented as a conceptual model; whilst its original application was mostly in relation to industrial risk assessment and management (de Ruijter and Guldenmund, 2016), it is now increasingly being used in a qualitatively manner to explore the natural and anthropogenic causes of change, and the associated consequences and responses (e.g., Cormier et al., 2013; Smyth and Elliott, 2014; Burdon et al., in press). It facilitates analysis or assessment of a defined problem by focusing attention onto the areas of a system where the consequences of a potentially damaging event can be proactively managed. The Bow-tie method provides for a graphical representation of the expansion of the initial DPSIR environmental cause-and-effect pathway (Cormier et al., 2013). More specifically, it can be used to focus on the pathway between Pressure and State change, and provides a means of identifying where controls can be put in place either to control the occurrence of a particular event, or to mitigate for the effects of the event should it occur (see **Figure 4**). It comprises several components:


There may be several top events in any one area as the result of the Drivers and hazards such that nested Bow-ties are required in any assessment of cumulative impacts (Cormier, 2015). Similarly, the consequence of the loss of control in one Bow-tie sequence may become the top event in another (Smyth and Elliott, 2014). For example, the threat of the introduction of non-indigenous species may be a top event, the consequence of which may be that an area fails GES under the MSFD. In turn, the failure to meet GES will then become the Top Event which has legal and financial consequences, each requiring mitigation (Smyth and Elliott, 2014). It is of note that ICES (2014) has recommended that the Bow-tie framework be used to address cumulative and in-combination Pressures and their consequences.

center). It also details the related Prevention Measures (left of center), Mitigation Measures (right of center), and Escalation Factors.

The DPSIR framework can be superimposed on the Bow-tie structure given that the threats to the top-event will be Drivers and/or Pressures and the top-event and consequences are likely to be the State changes and/or Impacts (**Figure 4**). The barriers both as prevention measures and as mitigation or compensation measures, constitute the Response within DPSIR. As such, this links to a risk assessment and then risk management (RARM) framework as the need for responses to human Pressures. Burdon et al. (in press) have directly linked the DAPSI(W)R(M) concept with Bow-tie in integrating natural and social sciences in a case study for the management of the Dogger Bank in the North Sea.

# Nested Environmental Status Assessment Tool

The Nested Environmental status Assessment Tool (NEAT) is a recent tool for biodiversity assessments based on State indicators (Borja et al., 2016; and see therein for older, similar assessment tools). NEAT is a specialized user-friendly desktop application developed recently within the EU DEVOTES project (Berg et al., 2016) specifically targeted toward MSFD biodiversity assessments for defined spatial areas. It does not relate to Activities or specific Pressures, rather levels of State in relation to targets/thresholds. Assessments are indicator based with a large library of available indicators, habitats and ecosystem components. It allows different rules to be used for aggregating indicators, is fully customizable and will determine uncertainty values based on data inputs. The environmental status of a spatial assessment unit is obtained by choosing the marine region, entering the assessment values for the indicators chosen (along with an uncertainty measure and the classification scale) allowing the software to calculate and show the resulting status assessment. The algorithms and intermediate calculations are based on weighted average normalized indicators within specific groups. The NEAT weighting procedure avoids the dominance of certain indicators or habitats or spatial units. Thus, no bias is introduced into the assessment by the choice of the indicators. The tool is being trialed with many different user groups and national authorities. It is freely available for a number of different platforms at http://www.devotes-project.eu/neat/.

# CONCLUDING REMARKS

In defining and describing the DPSIR Conceptual Framework, we show how it facilitates management and assessment issues and, through the detailed worked examples, show its particular use with respect to the MSFD. By showing the predominant use of the DPSIR framework and its derivatives as a generic approach to risk assessment and risk management, we emphasize the practical limits of conceptual models and diagrams. Whilst they are of value in an abstract or generic application, the underlying complexity of marine systems means that specific applications cannot be easily shown diagrammatically. Hence, following simple Pressure-Impact linkages, the most straightforward option for assessing specific examples of this conceptual model is to record relationships between successive stages by means of matrices. Subsequently, matrices and linkages can be compiled within a database and interrogated and analyzed by means of interactive data filters. Such an approach facilitates the extraction of information for specific stages of the overall process, which can then be used as the input to other techniques, such as Bow-tie analysis.

In emphasizing the complexity of the marine system, here we show that although creating a system which covers all eventualities (all Activities, Pressures, State changes and Impacts on human welfare and the links between these) is a laudable aim, it is more profitable to focus on a problem-based approach. Hence for any specific area (e.g., a Regional Sea, eco-region, or sub-ecoregion) to determine the ranked priority Pressures based on the number of Activities. Each of these can then be addressed through the proposed DPSIR-Bow-tie linked approach in which we can address the main risks and hazards creating Pressures, and thus the Main Event of concern (Smyth and Elliott, 2014). The challenge for marine management, as shown here, is to apply a linked DPSIR approach for the area being managed. By focusing on the risk assessment approach, i.e., the Pressures as mechanisms causing the State changes and Impacts on Human Welfare (and so ultimately impacting on Ecosystem Services and Societal Benefits, sensu Atkins et al. (2011)), then by definition management measures for prevention and mitigation/compensation can be implemented; hence the latter being the Responses under DPSIR and the means by which

#### REFERENCES


the Responses address the Drivers and Pressures (and State changes) becomes the risk management framework (see Elliott, 2014).

A further challenge, again given the complexity of the marine system, its uses and users, is its ability to respond to exogenic unmanaged Pressures as well as the endogenic managed Pressures where current assessments rarely consider climate changes, although its effects may be implicit in the measurement of indicators. Hence management not only has to provide the Responses to the causes and consequences of change due to system internal Pressures but also the Responses to the consequences of external Pressures. Because of this, the application of the proposed scheme to cumulative and incombination Pressures, as discussed here, is also an imminent challenge.

### AUTHOR CONTRIBUTIONS

CS led the work on this manuscript. CS, KP, SB, KM, ME comprised the core writing team with inputs from JP, OS, SL, NB, AB. All authors were involved in reviewing and editing the manuscript.

### ACKNOWLEDGMENTS

This study was undertaken as part of the DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), www.devotes-project.eu. A preliminary version of this work was available in Smith et al. (2014) found online at www.devotes-project.eu/wp-content/uploads/2014/06/ DEVOTES-D1-1-ConceptualModels.pdf. The authors would also like to thank two referees for helping to improve the manuscript.


ability to attain Good Environmental Status for marine biodiversity? Mar. Pollut. Bull. 95, 7–27. doi: 10.1016/j.marpolbul.2015.03.015


supporting European policies and scientific research. Manag. Biol. Invasions 6, 147–157. doi: 10.3391/mbi.2015.6.2.05


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declared a collaboration with the authors and states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Smith, Papadopoulou, Barnard, Mazik, Elliott, Patrício, Solaun, Little, Bhatia and Borja. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A Global Review of Cumulative Pressure and Impact Assessments in Marine Environments

Samuli Korpinen<sup>1</sup> \* and Jesper H. Andersen<sup>2</sup>

*<sup>1</sup> Marine Research Center, Finnish Environment Institute, Helsinki, Finland, <sup>2</sup> NIVA Denmark Water Research, Copenhagen, Denmark*

Ever more extensive use of marine space by human activities and greater demands for marine natural resources has led to increases in both duration and spatial extent of pressures on the marine environment. In parallel, the global crisis of decreasing biodiversity and loss of habitats has revitalized scientific research on human impacts and lead to methodological development of cumulative pressure and impact assessments (CPIA). In Europe alone, almost 20 CPIAs have been published in the past 10 years and some more in other sea regions of the world. In this review, we have analyzed 40 recent marine CPIAs and focused on their methodological approaches. We were especially interested in uncovering methodological similarities, identifying best practices and analysing whether the CPIAs have addressed the recent criticism. The review results showed surprisingly similar methodological approaches in half of the studies, raising hopes for finding coherence in international assessment efforts. Although the CPIA methods showed relatively few innovative approaches for addressing the major caveats of previous CPIAs, the most recent studies indicate that improved approaches may be soon found.

Keywords: human activities, pressures, multiple stressors, cumulative effects, impacts, ecosystem-based management

# INTRODUCTION

Globally, the marine environment is at risk from multiple human activities such as overfishing, chemical contamination by hazardous substances, inputs of nutrients, physical modification, etc., in addition to climate change, leading to impaired environmental conditions (Lotze et al., 2006). Increasing human pressures leads to decreasing biodiversity and loss of habitats. A greater awareness of these problems has revitalized the scientific research on human impacts and led to an increasing number of laws, strategies and commitments to reduce human impacts on the ecosystem. The challenge for the scientific community lies in showing evidence of the causalities between human activities, the pressure they cause and the associated impacts on species and habitats, including humans and the human society. In the marine environment, the global assessment of human impacts by Halpern et al. (2008) fostered a wave of impact assessments in the world's seas (e.g., Selkoe et al., 2009; Ban et al., 2010; Korpinen et al., 2012). Although many of these assessments followed the same methodology as in the global assessment, new approaches were also found (e.g., Andersen and Stock, 2013; Knights et al., 2013), old approaches were re-assessed (e.g., van der Wal and Tamis, 2014) and spatial accuracy of the assessments increased

#### Edited by:

*Angel Borja, Tecnalia, Spain*

#### Reviewed by:

*Stelios Katsanevakis, University of the Aegean, Greece Vanessa Stelzenmüller, Thuenen Institute of Sea Fisheries, Germany*

#### \*Correspondence:

*Samuli Korpinen samuli.korpinen@ymparisto.fi*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *15 June 2016* Accepted: *12 August 2016* Published: *30 August 2016*

#### Citation:

*Korpinen S and Andersen JH (2016) A Global Review of Cumulative Pressure and Impact Assessments in Marine Environments. Front. Mar. Sci. 3:153. doi: 10.3389/fmars.2016.00153* (e.g., Ban et al., 2010). In this review, we have assessed 40 recent marine assessments of cumulative pressures and impacts and focused on the methodological approaches. We were especially interested in discovering methodological similarities, identifying good practices and proposing areas in need of more robust scientific input.

So-called cumulative pressure and impact assessments (CPIA) aim to cover additive, synergistic and antagonistic effects of multiple pressures on selected features of the ecosystem. In their fullest form, they attempt to cover all existing anthropogenic pressures and estimate their impacts on a wide spectrum of ecosystem components (e.g., Korpinen et al., 2012). More focused CPIAs assess specific species (Certain et al., 2015; Marcotte et al., 2015), communities (Giakoumi et al., 2015) or are limited to specific human activities (Benn et al., 2010) or pressures (Coll et al., 2016). The selection of ecosystem components in CPIAs is an important step, at least in case of selecting characterizing species to represent ecosystems, food webs or habitats, and hence, this review will also analyse the assessment methods in this respect.

The complexity of CPIAs has led to simplistic assumptions in the methods. Halpern and Fujita (2013) listed many of those and discussed the consequences of the assumptions for the overall assessment conclusions. For instance, many methods assume additivity of impacts, while meta-analytical studies indicate strong roles by synergistic and antagonistic effects (Crain et al., 2008). Similarly, the CPIAs analyzed typically assume that the impacts increase linearly with increasing pressures, while nonlinear responses seem to be more common in nature (Hunsicker et al., 2016). Despite these assumptions the CPIAs have provided robust outcomes which seem to correlate with the state-ofthe-environment assessments (Andersen et al., 2015) and have potential to inform management decisions.

CPIAs are primarily meant to inform decision-makers and guide management decisions. Therefore, the impacts should be traceable all the way to the human activities at sea, on the coast or in in the upstream catchments. Established links between human activities, pressures and ecosystem components are essential for effective and reliable CPIAs. These links are formed on the basis of causality (i.e., which human activities cause which pressures and which ecosystem components do they affect?), spatial overlap, or exposure (i.e., where are the activities, pressures, and ecosystem components located? Is uncertainty considered? How do the pressures decay from their source?) and sensitivity (i.e., how sensitive is a given ecosystem component to a specific pressure?). So far, only a few attempts to link these in a generalized and systematic way have been published (Knights et al., 2013, 2015) but some linkage frameworks have been in use by regional sea conventions for years (e.g., the North-East Atlantic, the Baltic Sea). Solid basis and transparent communication of these links is crucial for taking the message from the scientific community to the decision-making level. The progress in spatial data tools and online map services will certainly help in that task. Nonetheless, this review critically evaluated the activity-pressure-impact links of the CPIAs.

The cornerstone of CPIAs is the estimation of the potential impact of a specific pressure on a specific ecosystem component. This has been estimated numerically on the basis of spatial damage or loss of individuals (e.g., Giakoumi et al., 2015; Coll et al., 2016) or categorically on the basis of literature reviews and expert panels (e.g., Halpern et al., 2007; Eno et al., 2013). The potential concerns with such a variety of approaches are, firstly, if the different estimate variables are comparable, and secondly, whether the validation (referred to by some authors as "groundtruthing") of the CPIAs to realistic "effect scales" is reliable. To our knowledge this is the first scientific review of the CPIAs and as such its general aim is to lay down an overview of the existing methods and practices.

# MATERIALS AND METHODS

### Scope of the Review

This review has the general aim to provide an overview of the methods and practices that are used to produce CPIAs in marine environments. It will not evaluate input data or assessment practices outside the methods, even though these may, nonetheless, have important functions in communication, transparency, and confidence of the assessments.

This review has five specific objectives: (1) To compare and find similarities in the structures of the CPIA methods; (2) to evaluate the selection of ecosystem components included in CPIAs; (3) to evaluate the links between human activities, pressures and associated impacts in the CPIAs; (4) to compare the methods in estimating potential impacts; and (5) to find good practices in validating the CPIAs. Each of these objectives is met by defining a number of research questions to be answered for each of the reviewed studies. The research questions are given in **Table 1**.

#### Selection of Studies

We reviewed CPIA studies which have been published after 2000 and included integration of at least two different pressures. We accepted studies which assess cumulative pressures or cumulative impacts but did not include concept papers unless they piloted a case study or gave an operational method formulation. We performed this search globally by the Google Scholar engine with key words "cumulative effect [/impact] on marine environment [/ecosystem]," "marine cumulative impact assessment," and "Halpern impact assessment of marine pressures." The search was limited to the period 2000–2016 and the results were asked in the order of relevance. The search gave thousands of matches, but we analyzed only 750 first hits and applied the above-mentioned exclusion/inclusion criteria. We also included studies which were cited in the found CPIAs and matched with our search criteria. In total, 35 peer-reviewed CPIA studies were found. However, we also noticed that many CPIA studies have been published as project reports or in institutions' report series due to the nature of this assessment field and those assessments included interesting methodological development. Therefore, we included five additional studies. Hence, our review included altogether 40 studies. Global distribution of the studies is given in **Figure 1** and full references to the studies are given in the Supplementary Material (Appendix A).

#### TABLE 1 | Specific research questions in the review.


*The questions are categorized under the five objectives of the review.*

#### Evaluation Criteria

Each of the studies were analyzed to find answers to the five specific objectives and research questions (**Table 1**). The five objectives were evaluated generally following the descriptions of the reviewed study methods but also a more specific analysis of the methods was made in order to see tabular summary information of the recent CPIAs and compare them against major assumptions of the CPIAs as listed by Halpern and Fujita (2013). In case of the cumulative pressure studies, we evaluated only the general structure (objective 1) and links between activities and pressures (objective 3), as the other objectives require an impact assessment. Full results of the analyses are annexed as Supplementary Material (Appendix A).

#### Defining the Terms

The scientific literature provides a wide range of terms for CPIAs. An extensive discussion on this is given by Judd et al. (2015), who also provide definitions for the whole pathway from sources (e.g., human activities) to pressures, effects, receptors (e.g., ecosystem components), and impacts. In this study, we use the term "human activity" instead of "source" and define "pressure" (following Judd et al., 2015) as "an event or agent (biological, chemical, or physical) exerted by the source to elicit an effect." Although an effect and an impact can be defined as different steps on the pathway, we have chosen to use the term "impact" in this review. This is a pragmatic solution as our reviewed literature uses both these terms in justifiable way (sensu Judd et al., 2015).

### RESULTS

# Similarities in the Structures of the CPIA Methods

Of the 40 studies reviewed, 33 had assessments of cumulative impacts and seven assessed cumulative pressures. Most of the assessments (n = 35, 88%) assumed cumulative pressures or impacts as additive and five assessments included synergistic or antagonistic effects (**Figure 2**). The synergistic and antagonistic effects were mainly assessed in those CPIAs which used ecosystem models, but in one study synergistic effects were inserted into an additive model by defining pressures enhancing the effects of other pressures (Certain et al., 2015). Most of the methods (93%) also assumed linear relationships between activities, pressures and impacts (**Figure 2**). In one assessment the relationship was not clear and in two assessments the relationship was categorical. With the exception of four studies (Aubry and Elliott, 2006; Foden et al., 2011; Giakoumi et al., 2015; Knights et al., 2015), all the others made the assessments with varying spatial resolution (often by 0.2–2.5 km grid cells).

The CPIAs showed relatively similar structures. More specifically, 50% of the studies claimed that they follow the same method as in Halpern et al. (2008) or had a similar method (without directly referring to the Halpern study) (see Appendix A in Supplementary Material). These assessments consisted mainly of three components: (1) intensity of pressures (>1 layers), (2) occurrence of ecosystem components (>1 layers, only if impacts were assessed), and (3) some types of weighting factors to express impacts or to weight pressures. In those studies, where impacts were assessed, a weighting factor was produced for each specific pressure–ecosystem component combination, whereas in the pressure assessments the weighting factors were produced to balance threats between the pressures. The impact weighting factors were sometimes called "vulnerabilities" or "sensitivities" of the ecosystem components to pressures.

In addition, there were a few other methods which relied on similar additive-type models and will likely produce comparable assessment results (e.g., Zacharias and Gregr, 2005; Stelzenmuller et al., 2010; van der Wal and Tamis, 2014). Thus, there seems to be a mainstream approach in the CPIAs which is used worldwide (**Figure 1**), but where small adaptations have been applied in treating of input data and ecosystem sensitivity and in integrating these into the score of cumulative pressures or impacts.

# Selection of Species and Habitat Data into the CPIAs

Cumulative impacts were assessed for benthic habitats in 76% of the impact assessments, but also species (41%) and pelagic habitats (38%) were included in the studies (**Figure 3**). Species, benthic habitats and pelagic habitats together were included in only 12% of the studies. Only two studies assessed an entire community, including all the major components to the model (sea grass ecosystem: Giakoumi et al., 2015; 3 exploited fish

FIGURE 1 | Map with studies included in this review. Global studies and sea areas with several studies are shown in separate text boxes. Key: 1, Eastern North Sea (Andersen and Stock, 2013); 2, U.K. (Aubry and Elliott, 2006); 3, Canada's Pacific (Ban et al., 2010); 4, Portugal (Batista et al., 2014); 5, NE Atlantic (Benn et al., 2010); 6, North Sea (Certain et al., 2015); 7, New Zealand (Clark et al., 2016); 8, Canada's Pacific (Clarke Murray et al., 2015); 9, Mediterranean Sea (Claudet and Fraschetti, 2010); 10, Mediterranean Sea (Coll et al., 2012); 11, Mediterranean Sea (Coll et al., 2016); 12, Netherlands (de Vries et al., 2011); 13, UK (Eastwood et al., 2007); 14, UK (Foden et al., 2011); 15, Mediterranean (Giakoumi et al., 2015); 16, North Sea (Goodsir et al., 2015); 17, SE Australia (Griffith et al., 2012); 18, Global (Halpern et al., 2008); 19, California Current (Halpern et al., 2009); 20, Global (Halpern et al., 2015); 21, Washington US (Hayes and Landis, 2004); 22, French Mediterranean (Holon et al., 2015); 23, Massachusetts (Kappel et al., 2012); 24, Scotland (Kelly et al., 2014); 25, European seas (Knights et al., 2015); 26, Baltic Sea (Korpinen et al., 2012); 27, Baltic Sea (Korpinen et al., 2013); 28, Netherland (Lindeboom, 2005); 29, Hong Kong (Marcotte et al., 2015); 30, California Current (Maxwell et al., 2013); 31, Puget Sound, Canada's Pacific (McManus et al., 2014); 32, Mediterranean Sea and Black Sea (Micheli et al., 2013); 33, Spain (Moreno et al., 2012); 34, Liguarian Sea (Parravicini et al., 2011); 35, Mediterranean Sea (Rodríguez-Rodríguez et al., 2015); 36, Hawaii (Selkoe et al., 2009); 37, North Sea (Stelzenmuller et al., 2010); 38, Noth Sea (van der Wal and Tamis, 2014); 39, Jiaozhou Bay, North Yellow Sea (Wu et al., 2016); 40, (Zacharias and Gregr, 2005). Map from Natural Earth (free vector and raster map data @ naturalearthdata.com).

FIGURE 2 | Summary information of the 40 cumulative pressure and impact assessments (CPIA) included in the review. The CPIA type is divided into pressure and impact assessments. The integration was additive, synergistic or synergistic, and antagonistic. The scale of the pressure-impact relationship is divided into categorical, linear and linear and non-linear (with some uncertainty of this indicated by ?-mark).

species: Coll et al., 2016). Obviously all of the CPIAs had a limited number of ecosystem components in the assessments, but 21% of them had focused only on a species group (e.g., Zacharias and Gregr, 2005; Coll et al., 2012) or a single species only (Marcotte et al., 2015). However, many of the studies claimed to be demonstration studies and, hence, the selection of ecosystem components was made on practical grounds. Only in one study, a specific justification was given on the grounds of cultural, biological and legal arguments (Hayes and Landis, 2004). Nevertheless, there seemed to be a common lack of precise justification in the reviewed CPIAs, why some species or habitats were selected and others not.

#### Have the CPIAs Defined Linkages between Activities, Pressures and Impacts?

Ten studies (25%) had defined all the linkages between human activities, pressures and impacts and made a framework to support the CPIA. All of the 10 CPIAs were assessments and not demonstration studies (see Appendix A in Supplementary Material). Additionally, nine more studies had covered all the human activities or all the pressures in the area but not linked them in a systematic way. However, in many cases, it was not possible to estimate whether the systematic framework was made outside the study and used in a more limited way. The review showed that the actual CPIA have taken seriously the linkages between activities, pressures and ecosystem components, often consulting local experts or making extensive literature surveys (e.g., Selkoe et al., 2009; McManus et al., 2014).

In summary of the review results, human activities were included in 31 studies (78%), 26 studies (65%) linked pressures to the human activities and 30 studies (75%) had defined the human pressures into general pressure categories, for instance according to the EU Marine Strategy Framework Directive (MSFD).

Only one study had considered the maximum potential value of pressures (Clark et al., 2016). This is a necessary step in the CPIA procedure if pressures are quantified. Hence, almost all of the reviewed studies assumed that the maximum pressure value in the assessment area is the maximal intensity of that pressure. Moreover, while the majority of studies had normalized the pressure intensities (e.g., 0–1), none of the studies had benchmarked the pressures in order to estimate the impacts in a comparable way (i.e., defined the level of pressure where the impacts occur; see Halpern and Fujita, 2013). One of the studies asked experts to estimate impacts on a "typical level of pressures" (Andersen and Stock, 2013). The lack of definite benchmarks is especially problematic in case of non-linear relation of pressures and impacts. If the relation is non-linear, for instance logarithmic, a relatively low level of pressure can cause high impacts and the magnitude of impact does not increase much at higher pressure levels. However, most of the reviewed CPIAs assumed a linear increase of impacts as a pressure increases. This simplifies the impact formula, where each pressure can be given a single sensitivity score (for each ecosystem component combination).

#### Estimation of Impacts

We analyzed whether the CPIA studies estimated impacts from anthropogenic pressures by expert judgment or based on scientific literature. Of the 35 studies giving some kind of a weight factor (for impacts or pressures), 23 CPIAs (66%) relied on expert judgment, and 14 (40%) on literature (**Figure 4**). In two studies, the experts were informed by a review of scientific literature (See Appendix A in Supplementary Material).

Impact estimates were most often (69%) categorical expressions of the sensitivity of the ecosystem components to the pressures or severity of the pressures on ecosystem components (**Figure 4**). Continuous impact scales were used in 31% of the studies and in these CPIAs the impacts were often estimated either from a few known parameters, such as mortality (e.g., de Vries et al., 2011), biomass change (Coll et al., 2016), or loss of habitat area (e.g., van der Wal and Tamis, 2014). In these studies, the scope of the CPIA was more limited, focusing on a few ecosystem components (a single species or a species group), of which the impact parameter (e.g., mortality) could be estimated. The more diverse ecosystem components there were in the CPIA studies, e.g., both species and habitats, the more the studies relied on categorical or semi-quantitative impact/sensitivity categories.

Five of the 33 studies (15%), which assessed cumulative impacts, used meta-analyses or an ecosystem model to estimate impacts. The ecosystem models included, for instance, fishing effects on commercially exploited fish species (Coll et al., 2016) and main threats to the seagrass food web (Giakoumi et al., 2015). In one study, pressures were linked to biological quality indicators and the relationship was modeled (Parravicini et al., 2011). This model was used to predict impacts when the pressures were changed.

FIGURE 4 | Differences in estimating and expressing impacts of anthropogenic pressures. The impact estimates are based on expert judgment or literature (including models where the interactions are literature-based). The impacts are expressed on categorical scales and on continuous scales. Note that the numbers also include those studies where "impacts" are not specific to ecosystem components but used to weight pressures. Two of the studies used both literature and expert judgment as the basis.

#### Validation of the Impacts

Only 8 of the 40 studies (20%) had validated the results, i.e., compared the cumulative impact (or pressure) scores with observed environmental status and then re-categorized the impact gradient into a realistic scale (Appendix A in Supplementary Material). However, three of the eight validated CPIAs used a scale obtained from another study and made no reanalysis in their own study. Thus, in reality, only five studies had really validated their impact scores with environmental status assessments. In addition, two more studies indicated how the validation should be made but did not apply it (Zacharias and Gregr, 2005; Claudet and Fraschetti, 2010).

The best description of validation was given by Clark et al. (2016) who compared the cumulative impact scores (on benthic habitats) with benthic fauna data. They found significant relationships between the benthic community composition based on Bray-Curtis similarities and the cumulative impacts by using non-parametric regression (DISTLM). This was also used to test the relation of individual standardized pressures to macro fauna data, without including the habitat sensitivity information to the pressure data. Clark et al. (2016) argue that validation may result in relatively weak relationships if the range of stressor levels is small, which is often the case in local studies. A large-scale validation was applied by Andersen et al. (2015) on a Baltic Sea-wide scale, where cumulative impact scores for sub-basins were compared with integrated state of marine biodiversity. In that scale, the relationship was significant, but due to the small number of sub-basins (N = 9), it was not possible to make conclusions about thresholds or tipping points.

# DISCUSSION

Identification of marine areas that are sensitive and vulnerable to human activities is not a novelty; environmental sensitivity indices were launched already in the 1970s (Gundlach and Hayes, 1978). Cumulative assessments of multiple pressures and their impacts were carried out already in 1990s (e.g., Wiegers et al., 1998). Methodological development did not, however, receive wide attention until the 2000s when series of CPIAs were produced after the global impact assessment (Halpern et al., 2008). As shown in this review of 36 CPIAs in 2000s, more than half of them were based on the method by Halpern et al. (2008). However, similar research threads had already been started elsewhere (e.g., Lindeboom, 2005; de Vries et al., 2011; van der Wal and Tamis, 2014; Certain et al., 2015) and in comparison to these earlier methods, it is interesting to note that the method presented in Halpern et al. (2008) has allowed wider assessments in terms of human activities, pressures and ecosystem components than the other methods which tend to produce more focused (and sometimes more detailed) assessments in terms of activities, pressures and ecosystem components. Also various ecosystem models have this same limitation.

The review showed that the CPIAs have, in general, three essential components: spatial data on intensity of pressures, spatial data on occurrence of ecosystem components, and factors estimating impacts. In all of the three components, many of the reviewed CPIAs used simplified assumptions (see Halpern and Fujita, 2013) and had small differences in the approaches. Nonetheless, the majority of the studies, at least the ones based on additive integration and estimates of habitat sensitivity, can be expected to produce relatively comparable results and one can see potential improvements to the general method in the most recent studies. Although the 40 reviewed CPIAs were published between 2004 and 2016, 30% of them were from the 2 most recent years and these contained novel approaches more often than the earlier CPIAs. Such approaches were, for instance, the use of fuzzy logic for impact occurrence (Marcotte et al., 2015), building on a fixed linkage framework (Goodsir et al., 2015), separating habitat recovery to a specific assessment (Knights et al., 2015), using food web models (Coll et al., 2016) or other statistical methods (Wu et al., 2016) and describing good practices in validation and pressure quantification (Clark et al., 2016).

#### Treatment of Spatial Input Data

In the pressure data sets, the main assumptions relate to the spatial extent of pressures from their sources, quantification of the pressures (often on the basis of underlying human activities) and the normalization of the pressures. Spatial extent of pressures has often been treated as a linear decaying model from the source, whereas e.g., Andersen and Stock (2013) produced five alternative models which were used for different types of pressures. The quantification of pressures on the basis of human activities is an assumption which is difficult to replace by real pressure data. No monitoring programme can be expected to measure, e.g., resuspension from bottom-trawling and, hence, fishing activity data is used to estimate the pressure. The pressures are then normalized to a dimensionless scale in order to make them comparable with other pressures, measured in other units. The most frequently used approach was to scale the pressure values linearly such that the highest value is equal to 1.0. Obviously, the main problem with this method is the assumption that the data set contains the maximum value of that pressure. In reality, the pressures in the assessment period may be much lower than the long-term maximum if management measures have been implemented. Among the studies in this review, Clark et al. (2016) was the only CPIA setting a theoretical maximum value for each of the pressure data sets. In addition, Halpern et al. (2015) normalized the pressures according to the highest value of two data sets to allow temporal comparison of two assessment periods.

Occurrence of ecosystem components—species and habitats—in the assessment units determines whether an impact can take place in that area. The occurrence of the habitats was in all the cases reported as presence/absence, whereas for species occurrence probabilities were also applied (Andersen and Stock, 2013). Even though no CPIA used a probability scale for habitat presence, this could be applied if the habitat presence is uncertain due to the low confidence in the input data. Only a few of the reviewed CPIAs (9%) targeted the entire marine ecosystem, i.e., species, benthic, and pelagic habitats. The majority of the studies (55%) focused solely on benthic and pelagic habitats and 21% included species only. Because of the additive approach in most of the CPIAs, a major difference is also the choice to use only benthic habitat layers over the entire assessment area with only one habitat type in a grid cell (e.g., Korpinen et al., 2013) or, alternatively, to use several overlapping layers of ecosystem components and several ecosystem components per grid cell (e.g., Halpern et al., 2008). In the former, the resulting cumulative impacts are relatively simple to interpret, because all the impact scores indicate the amount of pressures, whereas in the latter case one needs to consider also the diversity of ecosystem components in an area when interpreting the cumulative impacts. Both of the approaches are conceptually correct, but they tell slightly different stories from the anthropogenic pressures.

# How Vulnerability Is Assessed?

There are basically two types of differences in integrating impacts from multiple pressures: using similar endpoints (same variables) from all the pressures or integrating categorized impacts of different types of variables. In this review, these two basic categories were found and further divided to more detailed sub-types: (1a) categorical expressions of potential impacts on ecosystem components, where the impacts have been usually defined by 3–5 criteria (e.g., functional impact, resistance, recoverability and frequency; e.g., Halpern et al., 2007); (1b) categorical expressions of habitat sensitivity, which has been defined by resistance and resilience (e.g., Stelzenmuller et al., 2010, see also Eno et al., 2013); (2a) numeric estimate of impact by a measurable variable (e.g., proportion of disturbed sea floor; van der Wal and Tamis, 2014, or change in biomass in Coll et al., 2016); and (2b) effect sizes of impacts in a metaanalysis (e.g., Claudet and Fraschetti, 2010). The two former methods are comparable, both considering categorical estimates of sensitivity of the ecosystem component, while the two latter ones use data-based approaches. These latter approaches share the limitation that common parameters are difficult to find for multiple pressures. So far, the quantitative, data-based CPIAs have not been applied to more than a few pressures or ecosystem components, which has limited their usefulness for getting a wider view of human impacts on marine environment.

There has been considerable progress in recent years in developing sensitivity estimates for species and benthic habitats. Zacharias and Gregr (2005) defined the terms sensitivity and vulnerability in an explicit and quantifiable manner with the aim to produce a tool that can predict and quantify vulnerable marine areas (VMA). Using the same or similar definitions, Tyler-Walters and Jackson (1999), Tillin et al. (2010), Eno et al. (2013), and La Rivière et al. (2016) have defined parameters for sensitivity estimates and procedures how these can be assigned to broader habitat types, which are usually the only available mapped marine habitats. Also the meta-analytical approach has been used by Claudet and Fraschetti (2010) to produce datadriven impact estimates for the Mediterranean Sea. Despite the progress, these were used very little, if at all, in the reviewed CPIAs.

# Needs for Further Progress in CPIA Methodology

The review showed that none of the CPIAs had benchmarked the pressures (i.e., a quantitative definition of a certain level of pressure, for which the impact or sensitivity is estimated). This is especially problematic for CPIAs which assessed very different types of activities causing same types of pressures. For example, siltation of seabed is caused by laying cables on sea floor, bottom-trawling, dredging and disposal of dredged material (to name a few activities), but the amount of sedimentation varies between the activities, i.e., a low pressure for each activity, if measured by different parameters, may mean different amounts of sediment and, hence, different impacts. This difference in activities was normally addressed in the reviewed CPIAs by giving different sensitivity scores for the pressures from different activities. This is an adequate "fix" if the impacts from pressures increase linearly. However, in non-linear cases, this assumption is no longer valid. This challenge was addressed by Tillin et al. (2010) who proposed to divide pressures to 2–3 sub-pressures based on their magnitude and define benchmarks for these pressures in order to give sharper and more comparable estimates of habitat sensitivity. For example, sea-floor abrasion was sub-divided to "penetration of the seabed surface," "shallow abrasion/penetration of the seabed surface" and "surface abrasion," and benchmarks to these were defined as ">25 mm penetration," "≤25 mm penetration," and "surface damage." The approach by Tillin et al. (2010) was taken up by La Rivière et al. (2016) and gives an easily approachable method for CPIAs where habitat sensitivity is defined by expert judgment.

The element of time was not very visible in the reviewed CPIAs. As data sources of human activities and pressures are often imprecise with regard to time of occurrence and duration, the CPIAs assume that pressures are long-lasting and overlap in time. This may well be the case with long-lasting impacts, i.e., with long recovery times, but many of the pressures and impacts are relatively short-lived (e.g., noise, siltation in exposed shores). Such an assumption can be considered as a conservative approach, but some realism could be introduced by specifying impacts seasonally (de Vries et al., 2011) or assessing the potential recovery separately (Knights et al., 2015). A more difficult aspect is the potential accumulation of effects in time (Eastwood et al., 2007). Although difficult to quantify, this was addressed by at least Korpinen et al. (2012) by summing certain pressures over the assessment period when preparing the input data.

An issue in regard to assessing vulnerability which has not been addressed by any of the reviewed studies is the question of historical impacts which have already modified the marine environment. This is especially problematic for the spatial ecosystem data, which only reflects the current situation. In addition, the question of how to assess extinct species or significantly reduced habitat coverage was not addressed by any of the reviewed studies. This specific weakness is something that needs to be solved.

# Criticism against the Major Assumptions in CPIAs

Five years after the global map of human impacts (Halpern et al., 2008), a paper was published criticizing the major assumptions in CPIAs (Halpern and Fujita, 2013). The authors listed nine major assumptions in the CPIAs, which are: (1) Stressor layers are of roughly equal importance, (2) Uniform distribution of stressors within a pixel, (3) Habitats either exist or are absent in a pixel, (4) Transforming and normalizing stressors, (5) Linear response of ecosystems to stressors, (6) Consistent ecosystem response, (7) Vulnerability weights sufficiently accurate, (8) Additive model, and (9) Linear response of ecosystems to cumulative impacts. For more detailed description and examples of these assumptions, readers are invited to read the full paper, but here we can briefly analyse how well the studies of this review, especially those published after 2013, have addressed these assumptions.

In this review, we saw that fairly few studies had included the full array of pressures in the assessment. Those that did this had commonly built a linkage framework between activities and pressures and aimed to aggregate pressures from several activities (addressing assumption #1). This is a tedious task if done properly, as described by Tillin et al. (2010). Assumptions #2 and #3 deal with the spatial resolution of input data and these aspects were not included in this review. However, assumptions #4 and #5 relate directly to the core of this review and may cause under- or overestimation of cumulative impacts, as they are related to the estimation of impacts at different pressure magnitudes. According to our review results, none of the studies addressed non-linear responses between pressures and impacts (as far as we were able to interpret the methods). Assumption #6 is about consistent impacts in different areas and within the definitions of the ecosystem components. Although being a critical assumption, none of the reviewed studies really addressed this in their methodology. However, some of the CPIAs were geographically limited and local experts were involved in making the impact estimates (e.g., Selkoe et al., 2009; McManus et al., 2014), which may mitigate the potential error. This does not, however, answer the other side of the assumption that impacts should be consistent within broad habitat definitions (which is definitely a bold assumption). In case of the broad-scale benthic or pelagic habitats, Tillin et al. (2010) and La Rivière et al. (2016) suggest the use of "characterizing species" as targets of the sensitivity estimation, but this has not, to our knowledge, been applied in any published CPIA. Assumption #7 raises the concern that expert-based impact estimates are not coherent or accurate. According to our review, 40% of the studies based these estimates on literature while 66% used expert elicitation. None of the studies claimed any comparison between the two approaches but two studies used both the approaches. Assumptions #8 and #9 have already been discussed in this study, but briefly, 88% of the studies assumed additivity and after 2013 only 3 of the 15 studies included synergistic and/or antagonistic effects. Nevertheless, this can be seen as an improvement in CPIA development, as before 2013 only one of the reviewed studies addressed these effects. The inclusion of non-linear responses to the pressure—impact relationship had not, according to our results, progressed at all.

The current CPIA practices are obviously limited by the scientific knowledge we have today, but there are theoretically unlimited possibilities of impacts on diverse marine environment. To tackle the challenge the methods should focus on keystone species and habitats and build on uncertainty assessment principles and a structured approach to filter and prioritize pressures, impacts and ecosystem components (see Wiegers et al., 1998; Judd et al., 2015). In this review we saw still diverse approaches and non-structured methods but also some positive signs.

#### CONCLUSIONS AND OUTLOOK

Our review showed that despite rapid method development and several recent publications of CPIA around the world, the assessments still rely on major assumptions which may potentially bias the results (Halpern and Fujita, 2013). Only the most recent studies had started developing methods to address the caveats.

We also showed that the assessment published by Halpern et al. (2008) is gradually developing into a global standard, especially taking some of the recent assessments into consideration. Recalling the concerns raised by Halpern and Fujita (2013), this standard would, however, need new openings such as the inclusion of non-linearity to the models or the use of other types of broad modeling frameworks, e.g., Bayesian Belief Networks, in CPIAs (Uthicke et al., 2016). The direction in the most recent studies indicates that this may indeed be the case in the near future.

In the light of this review, there are currently, in our understanding, no other methods capable to assess the whole range of human impacts than the ones similar to Halpern et al. (2008). Hence, we call not only for a further development of the methodology but also a sharing of tools or codes, such as the open access EcoImpactMapper (Stock, 2016), as this will encourage and support both a short term process focusing on the tools and a long-term process supporting CPIA-based marine ecosystem health assessment as well as evidence-based management.

### AUTHOR CONTRIBUTIONS

SK: the main author responsible for the analysis and the results. JA: building the study database, supporting the analysis and text, responsible for visual presentation.

#### ACKNOWLEDGMENTS

We are grateful to Johnny Reker, Andy Stock, and Ciaran Murray in supporting us in the work leading to this review. Also the ongoing work within the European Topic Center for Inland, Costal and Marine waters provided a fruitful context for the authors. Financial support from the HELCOM TAPAS project (07.0201/2015/717804/SUB/ENVC.2) and the FP7 DEVOTES project (n◦ 308392) helped the authors in carrying out this review.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00153

#### REFERENCES


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Korpinen and Andersen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Uses of Innovative Modeling Tools within the Implementation of the Marine Strategy Framework Directive

Christopher P. Lynam<sup>1</sup> \*, Laura Uusitalo<sup>2</sup> , Joana Patrício<sup>3</sup> , Chiara Piroddi <sup>4</sup> , Ana M. Queirós <sup>5</sup> , Heliana Teixeira<sup>3</sup> , Axel G. Rossberg<sup>6</sup> , Yolanda Sagarminaga<sup>7</sup> , Kieran Hyder <sup>1</sup> , Nathalie Niquil <sup>8</sup> , Christian Möllmann<sup>9</sup> , Christian Wilson<sup>10</sup>, Guillem Chust <sup>7</sup> , Ibon Galparsoro<sup>7</sup> , Rodney Forster <sup>11</sup> , Helena Veríssimo<sup>12</sup>, Letizia Tedesco<sup>2</sup> , Marta Revilla<sup>7</sup> and Suzanna Neville<sup>1</sup>

#### Edited by:

*Maria C. Uyarra, AZTI Tecnalia, Spain*

#### Reviewed by:

*Rodrigo Riera, Atlantic Environmental Marine Center (CIMA SL), Spain Marcos Llope, Instituto Español de Oceanografía, Spain; University of Oslo, Norway*

> \*Correspondence: *Christopher P. Lynam chris.lynam@cefas.co.uk*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *17 June 2016* Accepted: *06 September 2016* Published: *28 September 2016*

#### Citation:

*Lynam CP, Uusitalo L, Patrício J, Piroddi C, Queirós AM, Teixeira H, Rossberg AG, Sagarminaga Y, Hyder K, Niquil N, Möllmann C, Wilson C, Chust G, Galparsoro I, Forster R, Veríssimo H, Tedesco L, Revilla M and Neville S (2016) Uses of Innovative Modeling Tools within the Implementation of the Marine Strategy Framework Directive. Front. Mar. Sci. 3:182. doi: 10.3389/fmars.2016.00182* *<sup>1</sup> Centre for Environment, Fisheries and Aquaculture Science, Lowestoft, UK, <sup>2</sup> Finnish Environment Institute (SYKE), Helsinki, Finland, <sup>3</sup> European Commission, Joint Research Centre (JRC), Directorate for Sustainable Resources, D.2 Water and Marine Resources Unit, Ispra, Italy, <sup>4</sup> Institute of Marine Science (Consejo Superior de Investigaciones Científicas), Barcelona, Spain, <sup>5</sup> Plymouth Marine Laboratory, Plymouth, UK, <sup>6</sup> School of Biological and Chemical Sciences, Queen Mary University of London, London, UK, <sup>7</sup> Marine Research Division, AZTI, Pasaia, Spain, <sup>8</sup> Centre National de la Recherche Scientifique, UMR Biologie des ORganismes et Ecosystèmes Aquatiques, Caen, France, <sup>9</sup> Institute for Hydrobiology and Fisheries Science, University of Hamburg, Hamburg, Germany, <sup>10</sup> OceanDTM Limited, Lowestoft, UK, <sup>11</sup> Institute of Estuarine and Coastal Studies, University of Hull, Hull, UK, <sup>12</sup> Faculty of Sciences and Technology, Marine and Environmental Sciences Centre, University of Coimbra, Coimbra, Portugal*

In Europe and around the world, the approach to management of the marine environment has developed from the management of single issues (e.g., species and/or pressures) toward holistic Ecosystem Based Management (EBM) that includes aims to maintain biological diversity and protect ecosystem functioning. Within the European Union, this approach is implemented through the Marine Strategy Framework Directive (MSFD, 2008/56/EC). Integrated Ecosystem Assessment is required by the Directive in order to assess Good Environmental Status (GES). Ecological modeling has a key role to play within the implementation of the MSFD, as demonstrated here by case studies covering a range of spatial scales and a selection of anthropogenic threats. Modeling studies have a strong role to play in embedding data collected at limited points within a larger spatial and temporal scale, thus enabling assessments of pelagic and seabed habitat. Furthermore, integrative studies using food web and ecosystem models are able to investigate changes in food web functioning and biological diversity in response to changes in the environment and human pressures. Modeling should be used to: support the development and selection of specific indicators; set reference points to assess state and the achievement of GES; inform adaptive monitoring programs and trial management scenarios. The *modus operandi* proposed shows how ecological modeling could support the decision making process leading to appropriate management measures and inform new policy.

Keywords: ecosystem modeling, good environmental status, marine strategy framework directive, indicators, assessment cycle, marine management

# INTRODUCTION

The Ecosystem Based Management (EBM) approach to marine conservation and sustainable use of the natural environment has been promoted by international conventions (e.g., Convention on Biological Diversity UNEP, 1998; CBD, 2014), national legislation across Europe and beyond (Kidd et al., 2011) and global scientific organizations such as the International Council for the Exploration of the Sea (ICES) that provide evidence and advice. EBM recognizes the need to take a holistic approach to understanding ecosystem level change, including explicitly accounting for the governance structures involved in interpreting, enacting and enforcing legislation (Borgström et al., 2015). Tightly linked to these aims, the Marine Strategy Framework Directive (MSFD, 2008/56/EC; European Commission, 2008) aims to achieve Good Environmental Status (GES) for the marine waters within the EU by 2020. During the first cycle of the MSFD (2012–2018), EU Member States prepared initial assessments of their marine waters (Article 8), determined characteristics for GES (Article 9), established environmental targets and associated indicators (Article 10), and established monitoring programs for the ongoing assessment of the environmental status of their marine waters (Article 11). Programs of measures have been identified that will provide the mechanism for changing the system to achieve the individual targets, and the overall aim of GES. EU Member States are required to review each element of the marine strategy every 6 years after their initial establishment (Article 17).

Although the fundamental statistical mechanics of ecosystems are an area of ongoing research (Rodríguez et al., 2012; Rossberg, 2013) many developments have been made by the scientific community in terms of modeling and indicator development (Shin et al., 2012; Piroddi et al., 2015a). Ecological models (hereafter "models"), including a range of conceptual, mathematical, and statistical representations of ecosystem components and processes (e.g., Peck et al., 2016), have an important role to play in the assessment and management cycle. Models provide the means to test how different trophic levels and the biogeochemistry of marine systems respond under specific scenarios of management (e.g., fisheries, Allen and Clarke, 2007) and environmental change (e.g., climate, Artioli et al., 2014). Further, ecosystem models can be used to drive distribution models for species in higher trophic levels, allowing the exploration of management and change scenarios (e.g., fisheries, harmful algal blooms, Gilbert et al., 2010; Sumaila et al., 2015).

A number of studies have shown how ecosystem modeling could support the assessment of different ecosystem components and pressures in several marine regions and also where models require further development (Hyder et al., 2015; Piroddi et al., 2015a; Tedesco et al., 2016; Rossberg et al., 2017). Here we advance on this body of work and demonstrating that modeling is not only useful for the assessment of components, but throughout the entire assessment cycle of the MSFD (**Figure 1**). Within the MSFD there are 11 themes or features that describe GES, termed descriptors within the directive (hereafter "D," Annex I, MSFD). Four of these are strongly linked to biological diversity: biological diversity (D1), non-indigenous species (D2), food webs (D4), and seafloor integrity (D6), all of which have the potential to be addressed using models (Piroddi et al., 2015a). The impacts of human activities on the ecosystem can also be addressed using models, particularly those linked to the pressure descriptors commercial fish and shellfish (D3), eutrophication (D5), and hydrological changes (D7). Although developments vary considerably between MSFD descriptors and assessment regions, some descriptors are well-addressed by ecosystem models in all regions (e.g., D4 food webs), and some remain poorly addressed (e.g., D2-non-indigenous species and D6-seafloor integrity; Piroddi et al., 2015a). Models may not always address biological diversity in a traditional sense (species richness and evenness, Tedesco et al., 2016), but they can be used to address simplified representations of natural biological diversity in relation to seafloor integrity and ecosystem functioning (Queirós et al., 2015).

Through a selection of case studies, this paper demonstrates how modeling can be used throughout the MSFD assessment cycle. Specifically, in the development and selection of indicators, identification of reference points, informing monitoring programs, assessing ecosystem state, and changes in functioning and trialing management scenarios. We propose a modus operandi through which ecosystem modeling can support the decision-making process leading to appropriate management measures and inform new policy.

# INNOVATIVE MODELING TO SUPPORT THE MSFD ASSESSMENT CYCLE—CASE STUDIES

The MSFD follows an adaptive management approach with Marine Strategies that must be reviewed every 6 years (**Figure 1**). Assessing and maintaining GES requires an understanding of the link between pressures on the marine environment and the state of the environment. Marine systems, however, are subject to multiple pressures and the resulting functioning of the system is also influenced by long term climate change such that the expected outcome of management actions is difficult to project. Integrative modeling tools allow researchers to investigate the processes operating in the system and the likely responses of ecosystem components to potential management measures given the prevailing climate.

# Development of Novel Indicators for Routine Assessment

Indicators are metrics used to determine the state of the ecosystem and to detect changes that occur due to anthropogenic or environmental impacts on the ecosystem. In the specific case of the MSFD, indicators need to be applicable to the descriptors (e.g., biological diversity) and pressures (e.g., fishing, pollution, etc.) that are explicitly listed in the directive (European Commission, 2008). Monitoring programs must be designed to provide data for indicator assessments (ICES, 2016a; Patrício et al., 2016). This is a fundamental step in measuring progress toward targets and evaluating the

effectiveness of measures employed to achieve or maintain GES. In addition to measuring the current characteristics of the ecosystem, monitoring programs should consider the wider context within which indicators are measured (such as climate change and the risk of invasions of nonindigenous species). However, ecosystems are complex and we cannot measure "everything everywhere." Rather, evaluation of trade-offs in monitoring different ecosystem state indicators (Kupschus et al., 2016) and potentially additional pressures (not listed in the MSFD) that modulate or confound these changes is needed (Queirós et al., 2016a). By explicitly linking ecosystem model development to

MSFD monitoring program development, modeling can assist by:


(Möllmann et al., 2008), improving our ability to give suitable scientific advice in future.

An important step toward EBM is the selection of suitable indicators from a range of proposed options according to objective criteria regarding their scientific standard and applicability to the MSFD (Queirós et al., 2016a). Some of these criteria, such as the quantification of pressure-response relationships, can be evaluated using models. For example, indicators addressing disturbance of marine fish community structure through exploitation (i.e., fishing, a physical pressure) have been investigated (Houle et al., 2012). The indicators evaluated were derived from simulated catch or survey data, for example, for the body mass or current length of individual fish, their expected length at first maturation, or trophic level, which were aggregated according to a number of formulae proposed in the literature (see Houle et al., 2012 and references therein). The model used described interactions between size-structured fish populations with various maturation body sizes. Specificity to fishing was evaluated in comparison to indicators responses to small random model parameter variation, representing, e.g., environmental change. While not identifying a unique "winner," this analysis revealed clear differences in both sensitivity and specificity among proposed indicators. These results later informed indicator selection (ICES, 2012).

Assessments based on indicators are required by the MSFD at multiple levels: ecosystem, habitats (including their associated communities) and species, and the spatial scale of these assessments must be ecologically meaningful and relevant to the pressures on the ecosystem. Ideally, indicator assessments would support simple advice on status in relation to reference points (thresholds). However, when such assessment reference points are not available, identification of desirable directions of change can be useful (i.e., "reference directions" Jennings and Dulvy, 2005. For indicators where the pressure-state relationship is unknown but the property is considered important to monitor indicators can be used for "surveillance purposes" (Shephard et al., 2015a). Examples are given in the following section for indicators of biological diversity, food webs, sea-floor integrity, and non-indigenous species. There is a risk that the pressure to implement and fulfill legislative requirements could affect the entire process of assessment. Acknowledgment of uncertainty (in both data and models; Carstensen and Lindegarth, 2016; Payne et al., 2016; Peck et al., 2016), recognition of coupled socialecological systems and that decisions reflect societal choice, and the acknowledgment of these trade-offs are therefore needed in GES indicator development (Long et al., 2015). Models can support these aims, as exemplified below.

# Modeling Habitats and Ecosystem Components

#### Pelagic Habitats and Lower Trophic Levels

The value of coupled biogeochemical-physical ocean models with realistic simulations of phytoplankton responses can be seen in many examples of applied ecology. Aldridge et al. (2012) used a variant of the European Regional Seas Ecosystem Model (ERSEM, Butenschön et al., 2015) to examine the effects of a large-scale seaweed farm in the northern North Sea. Compared to control runs, the model with a seaweed farm displayed altered phytoplankton composition at distances of up to 100 km. The impacts of other large-scale marine engineering projects such as the construction of wind farms at multiple sites in the North Sea can also be probed using biogeochemical models (van der Molen et al., 2014).

#### Remote Sensing and Bio-Optical Models for Assessing Pelagic Habitat

Within the DEVOTES project, a case study has been performed in the Bay of Biscay to investigate the potential to estimate chlorophyll-a by each of two bio-optical models applied to MODIS-AQUA imagery for the assessment of status and trends, and to support the definition of reference values and targets for chlorophyll concentration in the water column, as an indicator that responds to eutrophication and suitable for the MSFD D5 (human-induced eutrophication). These data may also be potentially used to support other indicators under other descriptors, such as those relating to pelagic habitat structure within D1 or Harmful Algal Blooms (HABs) within D5. This study revisits and updates the work performed by Novoa et al. (2012) by extending the dataset to 2014. However, satellite data also have limitations and uncertainties (Hooker and McClain, 2000). Firstly, only surface layers are sensed so subsurface peaks of chlorophyll may be missed (Jacox et al., 2013). Biooptical algorithms perform well in waters where the main optical constituent is phytoplankton, but the accuracy decreases in waters with more optical constituents such as dissolved or suspended matter.

The 90th percentile of chlorophyll-a values were evaluated over the defined growing season in a 6 year sliding window. This indicator is already used by the Water Framework Directive (WFD, 2000/60/EC) and is a candidate for the MSFD (Ferreira et al., 2010). The modeled values were compared, using MODIS-AQUA data with the OC5 (Gohin et al., 2002) and OCI (Hu et al., 2012) bio-optical algorithms, to traditional in situ datasets from the Basque Littoral Monitoring Network (Revilla et al., 2009; **Figure 2**). The reference values applied are those set by the North Eastern Atlantic geographical inter-calibration group for the coastal water type "Spain North East Cantabrian," as these target chlorophyll concentrations/ranges are determined locally for different water types and water categories, based on the results of the inter-calibration exercises (European Commission, 2013). Despite the improvements of the algorithms to better estimate chlorophyll-a values, the overestimation of satellite estimations (particularly in values higher than 1 mg per m<sup>3</sup> ) result in significant differences in assessments based on the 90th percentile indicator: where the in situ dataset classifies water bodies as "high quality" class, the MODIS-AQUA OC5 satellite dataset classifies them in "good quality" and the MODIS-Aqua OCI dataset algorithm) in "poor quality" class. However, statistics other than the 90th percentile such as the median or averages are less sensitive to the inaccuracies and classifications from these statistics are agree with the in situ data.

FIGURE 2 | Location of the area of study. (A) Western Europe overlaid with the P90-chla calculated with the MODIS-Aqua OC5 dataset between 2003 and 2013. (B) Detailed information about the location of the sampled stations from the Basque Littoral Monitoring Network, overlaid with main rivers pouring into the coastal zone, WFD water bodies and the P90-chla calculated with the MODIS-Aqua OC5 dataset between 2003 and 2013.

#### Lynam et al. Innovative Modeling Tools for MSFD

#### Remote Sensing and Bio-Optical Models for Estimating Production at the Base of the Food Web

The rate of production of new cells or carbon by phytoplankton is an important indicator for food webs, and as such, has been proposed for use under the MSFD for food webs (D4). Whilst the instantaneous rate of carbon fixation can be measured directly, the daily, seasonal or annual integral of production can be modeled as a function of phytoplankton biomass, underwater light, and photosynthetic activity (Smyth et al., 2005; Carr et al., 2006). Many models of primary production are available, from relatively simple empirical functions through to complex equations describing the underwater light field and plankton response in detail. Marine biogeochemical models such as the ERSEM and the Biogeochemical Flux Model (BFM, Vichi et al., 2015) contain complex productivity calculations in their core code, which, when driven by high quality atmospheric forcing data and accurate physical ocean responses, provide the ability to dynamically generate realistic inputs of new carbon in space and time.

Changes in marine primary production over annual to decadal periods may be driven by changes in underwater light availability, caused for example by increased sediment loading (Dupont and Aksnes, 2013; Capuzzo et al., 2015), or by changing nutrient concentrations or stoichiometry. A long-term study of modeled annual primary production in the eastern Scheldt estuary showed a decreasing trend between 1991 and 2011 (Smaal et al., 2013), which could not be related to changes in the dissolved nutrients or the concentration of suspended matter, but rather indicated an overgrazing of the larger, more active phytoplankton due to expanding aquaculture activities. Biogeochemical models such as ERSEM and food webs tools such as Ecopath with Ecosim (EwE, Christensen and Walters, 2004) could be deployed to investigate the wider ecosystem effects of such a prolonged decrease in overall production, and shift in prey size at the base of the food chain.

The links between primary production and fisheries production are now becoming well-established. Following pioneering work by Ryther (1969) and others, the recent availability of satellite-based primary production estimates for the global ocean has allowed size-based fisheries production models to be constructed for many large marine ecosystems (Jennings et al., 2008; Jennings and Collingridge, 2015; Fogarty et al., 2016). Further, work is required to regionalize satellite production algorithms for European seas, and to establish methods for the automated analysis of phytoplankton sizestructure for use in size-based models suitable for marine policy purposes.

#### Mapping Benthic Habitats and Species Distributions

Modeling of physical habitats, their associated species and connectivity between them can contribute toward the identification of ecologically important areas in need of protection (Baker and Harris, 2012), and form the basis for designing cost-effective monitoring programs (De Jonge et al., 2006). A range of modeling tools (Piroddi et al., 2015a; Peck et al., 2016) have been developed to map habitats and species assemblages and include: distribution modeling techniques to predict the spatial patterns in species distribution, abundance and habitats using observations of environmental variables (e.g., bathymetric and seabed types distribution; Stephens and Diesing, 2015) and new techniques to model connectivity between communities (Chust et al., 2016). Statistical models, in particular, can generate outputs that are easy to communicate (Reiss et al., 2014) and provide information on the uncertainty in the estimates (**Figure 3**). Such uncertainty information can indicate where monitoring is required in order to reduce the variance in the distribution model, or if multiple indicators are supported by one monitoring program this can be optimized by minimizing a weighted average of the indicators' variances (Carstensen and Lindegarth, 2016). Representing model uncertainty spatially (**Figure 3**) is especially useful as the MSFD relies on spatial assessments and species distribution indicators will be directly affected by the quality of the data used to model distributions. The location and frequency of multiple-objective monitoring programs can be modeled and the power needed to detect change in given indicators can be assessed leading to operational decisions on how many data types can be collected whilst maintaining sufficient overall precision and accuracy (Shephard et al., 2015b). As an example the Cefas integrated ecosystem survey program in the western Channel collected multibeam data from which seabed conditions were inferred for the entire area. The modeling process revealed areas of high heterogeneity and low predictability which can be prioritized in future surveys to reduce uncertainty and improve the reliability of species distribution modeling and make actual changes in distribution and extent (related MSFD D1 indicators) more reliable.

# Linking the Prevailing Climate and Pressures to Food Web Responses

Anthropogenic and environmental sources are major threats to marine ecosystems throughout the world (Naylor et al., 2000; Pauly et al., 2005; Diaz and Rosenberg, 2008). Effective marine resource management must take into account a variety of both current and future pressures on marine ecosystems, including fishing, eutrophication, climate change, and ocean acidification. This is explicit to the MSFD which considers that GES must be achieved with consideration for prevailing climatic conditions. Up to now, a large body of work has focused on the impact of single pressures on specific components of the marine environment, while the assessment of cumulative and synergetic effects of these threats remains poorly studied and such studies are now emerging (Link et al., 2010; Hobday and Pecl, 2014; Queirós et al., 2016b).

#### Food Web Responses to Ocean Acidification

van Leeuwen et al. (2016) applied a modeling approach to examine the potential higher level effects of the impacts of climate change and ocean acidification on marine ecosystems. Ocean acidification research has been focused largely on individual species and changes in their local environment, and less frequently considered wider ecosystem and societal impacts (Doney et al., 2009; Griffith et al., 2012; Le Quesne and Pinnegar, 2012; Queirós et al., 2016b). Understanding the combined effects

of direct (species level) and indirect (abiotic environment level) changes due to ocean acidification across the food webs are thus also critical to support the evidence base for management decisions. van Leeuwen et al. (2016) applied a coupled ecosystem model (consisting of a hydro-biogeochemical model and a higher trophic level size-based model) in the North Sea in three hydro-dynamically different sites: seasonally stratified, transition waters, and permanently mixed. Three different impacts affecting fishing yields were studied separately and in combination: climatic impacts (medium emission scenario), a proxy for abiotic impacts of ocean acidification (reduced pelagic nitrification), and a description of potential biological impacts of ocean acidification (reduced detritivore growth rate). Results showed a high regional variability and an overall shift toward more pelagic-oriented systems. Fisheries yields appeared to increase due the climate effects in large areas of the North Sea, but results indicated that ocean acidification could severely mediate this impact for permanently mixed areas. Although there is already evidence for a physiological response to ocean acidification, this does not necessarily lead to an ecosystem level response (Le Quesne and Pinnegar, 2012). Modeling tools used in this case study have enabled an indication of individual and combined effects of direct and indirect impacts of climate change and ocean acidification in a marine food web, and highlights that interactions between pressures can lead to less than or more than the additive response of the system.

#### Food Web Responses to Cumulative Impacts

Piroddi et al. (2015b) used an ecosystem modeling approach for a small area of the Mediterranean Sea (Amvrakikos Gulf, Greece) to assess temporal structural and functional changes of its ecosystem under the combined effect of anthropogenic pressures such as river runoff, fish farms, and fisheries. The model derived indicators highlighted a general degradation of the demersal compartments of the food web but a relative stability of the pelagic compartments. Since the model has showed a marginal role of local fishery in the Gulf's food web and on its dynamics, as also observed by other studies (Koutsikopoulos et al., 2008), eutrophication was considered the only major pressure affecting the system. Specifically, the model suggested that fish farms represented a secondary contribution to nutrients and organic matter to the Gulf, whereas the two major rivers were the main drivers of the Gulf eutrophication. Contrasting results were observed by Piroddi et al. (2010) for another area of the Mediterranean Sea (the Inner Ionian Sea Archipelago, Greece), which is extremely oligotrophic, not influenced by river run off and with a marginal low impact of fish farms. Here, model derived indicators showed a consistent decline with time while the demersal/pelagic biomass ratio and the mean trophic level of the catches have increased linearly. The model pointed to decline of small pelagic fish biomass, particularly sardines, the main target of the local fisheries, and an increase in biomass of demersal species as the likely cause of the change in the ecosystem. Despite the fact that changes in ocean productivity were observed in the area, the model suggested that the degradations of the system were mainly caused by intensive overexploitation of marine resources as suggested also by other studies (Tsikliras et al., 2013; Gonzalvo et al., 2014). Here, the modeling tool pointed strongly to the underlying causes for ecosystem level change and would be useful to managers attempting to improve the environmental state.

Single and combined effects in the North Sea food web were also studied by Lynam and Mackinson (2015) in this case focusing on the response of indicators to direct impacts of fishing and climate change. In the observation based model projections, community composition indicators (Large Species Index, mean maximum length) were found to respond to fishing. In contrast, the trophic level of fish and elasmobranchs was responsive to climate with a marginal effect of fishing only. Importantly, the modeled temperature effect suggested that the biomasses of certain trophic guilds (piscivores and bentho-piscivores) may be suppressed by warming and, if not taken into account during the setting of assessment thresholds, these indicators could conceivably not reach their desired levels due to climate effects. Modeling tools here facilitate scientific advice on the combined effects of fishing and climate impacts on the food web and can be used to demonstrate the likelihood of an indicator reaching its assessment threshold in the future given the prevailing climate and pressure.

#### Uncertainty in Climate Change Projections

Models can be used to investigate the environmental status of a system when prevailing conditions are far removed from those at present, as is expected to occur in the future ocean under global stressors such as warming and ocean acidification. By forecasting future ocean conditions, some models can thus help overcome traditional hurdles in forecasting ecosystem state based on observational data alone, which are bound to historical conditions (Barnsley, 2007; Szuwalski and Hollowed, 2016). Complexity in language choice in reporting modeling results and of the uncertainty associated with such projections has, at times, limited the uptake of the wealth of information generated by models by policy around the world (Hyder et al., 2015). The scientific community is now addressing this issue, for instance, through the use of lay language, more accessible to policy makers, in the expression of confidence attributed to modeling results (Pörtner et al., 2014). Further to this, the partitioning of sources of uncertainty in climate change impact projections, and the explicit assessment of their contributions, are paramount to improve the perception of confidence in modeling results in research-policy communication (Payne et al., 2016). For instance, though explicit recognition and quantification of how physical and biogeochemical model structure, initialization, internal variability, parametric, and scenario uncertainties are carried forward into fish distribution models (Gårdmark et al., 2013; Cheung et al., 2016; Payne et al., 2016) used to derive GES indicators such as those described here. These are important steps toward breaking down of uncertainty propagation and the attribution of confidence to modeling forecasts used to support policy. This effort is key to the uptake of modeling studies within the MSFD process too.

#### Modeling the Risk of Change in Ecosystem Function Due to Species Invasions

Ecological impacts of non-indigenous species (NIS) range from single-species interactions and reduction in individual fitness of native species to population declines, local extinctions, changes in community composition, and effects on entire ecosystem processes and wider ecosystem function (Blackburn et al., 2014; Katsanevakis et al., 2014). One of the MSFD requirements is to assess the consequences of pressures arising from NIS, through measurements of their impacts on the natural systems. A riskbased approach has been generally adopted by Regional Sea Conventions (RSC) and also by EU Member States in their MSFD initial assessments, but usually without clarification of the type of adverse effects in biological diversity or the magnitude of impacts observed (Micheli et al., 2013; Palialexis et al., 2014; Berg et al., 2015). Understanding, quantifying and mapping the impacts of invasive non-indigenous species across the seascape is a prerequisite for the efficient prioritization of actions to prevent new invasions or for developing mitigation measures (Katsanevakis et al., 2016). A new index CIMPAL (Katsanevakis et al., 2016) for measuring the cumulative impact of invasive alien species in the ecosystems provides a spatially explicit quantification of cumulative impacts. To illustrate the potential of ecological-niche modeling (ENM) in the cumulative impact index, DEVOTES used species distribution models (Kaschner et al., 2013) to create a vulnerability map for the whole the Mediterranean Sea under the current conditions, taking a trial group of 17 species (DEVOTES Deliverable 4.2 and Teixeira et al. unpublished). The CIMPAL index calculated using the future projections of these species distributions predicted an increase up to two and a half times the area likely to suffer the effects of cumulative impacts from multiple invasive NIS, with respect to the currently impacted area. Such trends can be easily linked to specific habitats, species or pathways of introduction, facilitating identification of ecosystem components, processes, and services more at risk. Early-warning indicators can be of utmost importance to identify vulnerable spots or preferential pathways of introduction (Thuiller et al., 2005; Hulme et al., 2008; Essl et al., 2015) and to anticipate a joint set of actions in target areas or sectorial activities. These ENM approaches are also effective tools to forecast changes in distribution of invasive NIS under large scale scenarios of climate change or addressing cross sectorial policies to better manage invasions pressures in the marine environment. Still, there are challenges to the use of these modeling approaches for effectively predicting distribution patterns of NIS in conservation and policy related contexts. For example, to obtain meaningful risk maps of the cumulative impact of invasive NIS it is required to consider the complete set of species targeted but, only recently, developments on multispecies distribution models are overcoming limitations of modeling for a large number of species (e.g., Fitzpatrick et al., 2011). Other relevant modeling developments aim at incorporating species co-occurrence data into a species distribution model (e.g., Pollock et al., 2014) or by integrating traits, namely dispersal strategies, into the modeling (e.g., Miller and Holloway, 2015).

#### Detecting Change in Function through Ecosystem Network Analyses

Food web functioning can be investigated through the use of models that capture the complexity and diversity of trophic flows in an ecosystem. The ecological properties of a network of trophic flows can be characterized through Ecological Network Analysis (ENA, e.g., Ulanowicz, 1997). ENA aims to characterize the structure and the functioning of a food web through a set of indices that describe the connections between compartments through an analysis of the inputs and outputs of a compartment, the trophic structure (based on a linearization of the network), the rates of recycling, and the topology of the flows (how redundant/specialized the flows are). Numerical methods recently developed to allow the evaluation of ENA indices and their uncertainty, so that statistical tests can be made to compare changes in observed states or between simulated scenarios (Lassalle et al., 2014; Chaalali et al., 2015, 2016; Guesnet et al., 2015; Tecchio et al., 2016).

ENA indices have been proved useful to evaluate the impacts of human pressures on ecosystem functioning and to simulate likely impacts given climatic change scenarios. A change in ecosystem functioning was modeled by ENA when comparing the ecological network before and after the extension of the Le Havre harbor in the Seine Estuary (Tecchio et al., 2016). Adjacent to the harbor, the food web demonstrated increased detritivory and recycling and the likely cause was a combination of pressures, as human direct effects were associated with hydrological changes induced by climatic conditions.

Scenarios are particularly useful to study cumulative effects and disentangle effects from various pressures. For example, the Bay of Biscay ecosystem was studied to investigate the effect of climate change on the distribution of small pelagic fish and its consequences on food web functioning (Chaalali et al., 2016). Here, ENA analysis suggested that the ecosystem would adapt to the simulated increased production of small pelagic fish in the Bay of Biscay within 100 years and suggested that this fish group would transport carbon toward higher trophic levels. Model derived ENA indices can offer a unique view on change in the ecosystem as a whole and demonstrate promise as food web indicators.

#### Detecting Abrupt Changes and Regime Shifts

Multiple stressors such as climate, fishing, eutrophication, and invasive species, have caused major reorganization of the aquatic ecosystem, and these have been interpreted as regime shifts in many areas including the Baltic, Black, and North Seas (Alheit et al., 2005; Möllmann et al., 2009; Diekmann and Möllmann, 2010; Lindegren et al., 2010, 2012; Llope et al., 2011). This reorganization is likely to be reflected in multiple MSFD descriptors, such as those of biological diversity, food webs, commercial seafood production, and seafloor integrity.

Anticipating regime shifts is difficult since these abrupt changes usually come as surprises (Doak et al., 2008). Recently, Big Data analytics has been employed to evaluate whether regime shifts could be predicted based on unexpected patterns in the data, i.e., anomaly detection. Models such as non-stationary dynamic Bayesian networks (Tucker and Liu, 2004; Robinson and Hartemink, 2009; Ceccon et al., 2011) can be employed to learn from past data but adapt to the fact that the relationships between the ecosystem components may change. This approach has its challenges in the ecological domain, where data is often relatively scarce, but some examples already exist (Trifonova et al., 2015). These models could help identify upcoming regime shifts based on data patterns, and could be used to inform models describing the underlying ecosystem processes.

It is important to note that systems that have experienced regime shifts often show hysteresis effects, i.e., reduction of external drivers need to have substantial stronger driver forcing to recover to the original state (Beisner et al., 2003; Scheffer et al., 2009). Although the existence of alternative ecosystem states is contentious (Cardinale and Svedäng, 2011; Möllmann et al., 2011), it is assumed that ecosystems that have experienced regime shifts have reorganized into novel states (e.g., in terms of species composition, population size, and species interaction strength), and the altered environmental and anthropogenic conditions may limit their recovery potential (Lotze et al., 2011). A recent example from the Baltic Sea is the apparent recovery of Eastern Baltic cod (Gadus morhua) predicted by linear, steadystate models (Eero et al., 2012), but challenged by food web models incorporating threshold dynamics (Blenckner et al., 2015a). This emphasized the need of constant evaluation and development of models, but also highlights that these processes can be incorporated into modeling frameworks.

# Modeling to Evaluate Management Scenarios

When choosing management measures to attain GES, decision makers need to have a strong evidence base to understand the consequences of management options and make informed decisions given a cost-benefit analyses of the options. However, food web interactions are fundamental to any ecosystem such that food web models could be required to fully evaluate changes due to management. For example, Piroddi et al. (2011) used a higher trophic level model of the Inner Ionian Sea Archipelago (Greece) to assess reduction in fishing effort or total closure (e.g., no-take zone) for the main fleets operating in the area as a measure to recover a resident population of common dolphins. Results from forecast scenarios highlighted that closing the area only to the industrial sector would lead to an increase in forage fish and thus a gradual recovery of common dolphins, but by closing the entire area to fisheries (industrial and artisanal) a recovery of common dolphin would be more pronounced. Lynam and Mackinson (2015) modeled the response of the North Sea food web, and a suite of ecological indicators, given a climate change scenario and a strategy in which fisheries management measures may be implemented, in order to achieve maximum sustainable yield targets for fishing mortality associated with the Common Fisheries Policy (CFP; European Commission, 2013). The authors demonstrated that a reduction in fishing effort consistent with CFP targets, would contribute to the attainment of GES as measured by improvements in indicators of biological diversity and food webs, thus linking the MSFD pressure descriptor D3-commercial fish and shellfish to the state descriptors D1-biological diversity and D4-food webs. Given the need for managers to consider environmental targets for indicators alongside traditional fisheries mortality targets for stocks, scientific advice is required on the combined effects of fishing and climate impacts on the food web (Brown et al., 2010). Modeling is one of the only tools able to provide this evidence base to facilitate management decisions on which measures to take.

Lack of consideration of uncertainty and the use of single model parameterizations can be seen as a common limitation to some of the above studies (Jones and Cheung, 2015). Thorpe et al. (2016) used an ensemble approach with 188 plausible parameterizations of a size-based multispecies model (Thorpe et al., 2015) with four fishing fleets to assess the effects of 10,000 alternate fishing scenarios of the ecosystem. They demonstrated that the risk of stock depletions could be related to the value of indicators of biological diversity and food webs (i.e., the Large Fish Indicator and Size Spectrum Slope, respectively) and this approach can be particularly useful for identifying assessment thresholds for indicators. Thorpe et al. (2016) also demonstrated a way to present risks (i.e., of stock depletion and thus loss of biological diversity) and potential rewards (value of the catch) associated with the scenarios tested. Similarly, a management strategy evaluation tool has been developed for the EwE software, capable of exploring the complete parameter space, and multiple fisheries management strategies, having been tested using 1000 model configurations (STECF, 2015). Ecosystems are difficult to model and project so that model uncertainty is also important to capture in addition to parameter uncertainty. The STECF workshop (STECF, 2015) approached this by using four differing models (EwE, Fcube, Simfish, and Fishrent) and contrasting the outcome of the fishing strategies. In the assessment of impacts of climate change on marine ecosystems, large scale intercomparisons of models and configurations are now standard practice particularly to inform global studies such as assessment reports from the Intergovernmental Panel on Climate Change (Coupled Model Intercomparison Project of the World Climate Research Programme).

Predicting the outcomes of the management actions with precision becomes progressively more challenging as the number of major forcing factors and pressures increase since they can occur in previously unseen combinations (Dickey-Collas et al., 2014). Uusitalo et al. (2016) approached this problem in the Baltic Sea case by using three distinct modeling approaches to evaluate how different combinations of fisheries management and nutrient abatement can be expected to affect the ecosystem status of the Baltic Sea, thus linking the MSFD pressure descriptors D5-eutrophication and D3-commerical fish and shellfish. The modeling approaches they chose, (1) a spatial model for cumulative impacts (additive approach), (2) a food web simulation model, and (3) a Bayesian model harnessing expert knowledge, have all been used for management strategy design or evaluation, and all have their strengths and weaknesses in predicting the effects of the management scenarios (Uusitalo et al., 2016). While all of these models were at least implicitly based on the abundant research on the effects of nutrient loading and fishing pressure on the Baltic Sea ecosystem (see e.g., Gårdmark et al., 2013; Tomczak et al., 2013; Korpinen and Bonsdorff, 2014; Blenckner et al., 2015b; Fleming-Lehtinen et al., 2015), these three models were all based on a different logical construct, had different mathematical formulations, and very different specifications in terms of how explicitly they accounted for spatial and temporal aspects and the different ecosystem types of the Baltic Sea. Therefore, the authors concluded that any agreement between the models could be interpreted as representing relatively well-known, or robust, management response, while disagreement between the models imply that the management response may be more uncertain. This highlights the usefulness of multiple, mutually different modeling frameworks in discerning the uncertainties in future predictions.

Alternatively, uncertainty in ecosystem response can be reduced by focusing indicator studies on high-level ecosystem properties known to be more predictable than future projections. Either models are built directly for the high-level properties, or models describing systems at a lower level are used but analyses focus on emergent properties, i.e., high-level responses. Both approaches have been applied to predict recovery of fish community size structure. Examples of the former, direct approach are size spectrum models. The Species Size Spectrum Model (Rossberg, 2012), for example, is sufficiently simple to be solved analytically and this pinpointed key mechanisms slowing recovery: competition for food among fish species of very different size and predator-prey reversal. This was confirmed using a much more detailed, species-resolved food webs model (Fung et al., 2013), which was then used to predict recovery processes a range of different indicators of fish-community size structure. Applying this method to the Celtic Sea, Shephard et al. (2013) predict that recovery of the Large Fish Indicator to proposed target levels would require drastic reductions in fishing pressures and may yet last 30–50 years.

# Models Embedded within Assessments by Regional Sea Conventions

GES is defined in the MSFD for ". . . seas which are clean, healthy and productive within their intrinsic conditions...." Intrinsic conditions are not clearly defined and it is not all clear, what sort of ecosystem would occur in the absence of human interference in a given physical environment. If we considered that the major physical regimes in the sea (i.e., short-term/seasonal/permanent stratified conditions/mixed conditions) promote a particular life form of primary producer over others and thus structure the food web then these regimes are the relevant level in which to assess change in the plankton. Within the DEVOTES project, we modeled physical processes with a coupled hydro-biogeochemical model (GETM-ERSEM-BFM) and determined Ecohydrodynamic zones that capture differing intrinsic conditions (van Leeuwen et al., 2015). The spatial stability of the Ecohydrodynamic zones suggests that carefully selected monitoring locations can be used to represent much larger areas. As such these zones are being used as the spatial basis for plankton (D1 and D4) and oxygen (D5) indicators in the OSPAR region (Greater North Sea and Celtic Seas).

HELCOM engages in modeling to define its nutrient reduction schemes (i.e., with reference to eutrophication D5) through use of the coupled physical-biogeochemical model BALSEM (BAltic sea Long-Term large Scale Eutrophication Model, Gustafsson, 2003; Savchuk et al., 2012) to calculate maximum allowable inputs (MAI, **Table 1**). MAI are the maximal level of annual inputs of water- and airborne nitrogen and phosphorus to Baltic Sea sub-basins that can be allowed while still achieving GES in terms of eutrophication, that is given GES boundaries for eutrophication indicators like nitrogen, phosphorous and chlorophyll-a concentrations, water transparency and oxygen debt. BALTSEM is a time-dependent ecosystem model available through the Nest Decision Support System (www.balticnest.org/nest). It has been used also as the major scientific tool for the development of the HELCOM Baltic Sea Action Plan (HELCOM, 2007). The HELCOM Contracting Parties annually report atmospheric emissions and waterborne inputs of nitrogen and phosphorous from rivers and direct point sources to the Baltic Sea sub-basins. Nutrient input data are compiled in accordance with specific HELCOM guidelines for nine Baltic Sea sub-basins, whose boundaries coincide with the main terrestrial river basin catchments. The BALTSEM model has instead divided the whole Baltic Sea into seven sub-basins in accordance with natural marine boundaries and the MAIs are calculated accordingly through an optimization technique: finding the highest possible inputs that will satisfy given eutrophication targets. A revised HELCOM nutrient reduction scheme was adopted in the 2013 HELCOM Ministerial Declaration (HELCOM, 2013) in which reduction requirements for nitrogen inputs to the Baltic Proper, Gulf of Finland, and Kattegat and for phosphorus inputs to the Baltic Proper, Gulf of Finland, and Gulf of Riga were set (**Table 1A**). The progress of countries in reaching their share of the country-wise allocation of nutrient reduction targets (CART, **Table 1B**) is then assessed separately.

# DISCUSSION

Assessments of GES within the MSFD are made as part of an adaptive management approach with 6 year cycle (**Figure 1**). In this review, we demonstrate the important role that modeling has throughout the MSFD assessment cycle from generating understanding, underpinning assessments, and investigating the impact of changes in prevailing climatic conditions, invasions of non-indigenous species and multiple human pressures, and in the exploration of potential impacts through projections that consider management scenarios. Only through modeling can such scenarios be tested in order to help select appropriate management measures to maintain or recover ecosystems. We have reported on case studies, many of them resulting from DEVOTES project work, to illustrate how models should be used in the MSFD implementation cycle, and suggest that in many areas this is already happening but it is not always recognized nor is it considered a matter of routine.

# Modeling and the MSFD Implementation Cycle—A Modus Operandis

Prior to each new assessment cycle, indicators should be reevaluated given new data and refined where necessary. During TABLE 1 | (A) Maximum Allowable Inputs (MAIs) to the Baltic Sea and (B) country allocated reduction targets (CARTs) as revised by HELCOM using the BALSEM model (HELCOM, 2013).


(B)


\**Reduction requirements stemming from:*

− *German contribution to the river Odra inputs, based on ongoing modeling approaches with MONERIS.*

− *Finnish contribution to inputs from river Neva catchment (via Vuoksi river).*

− *these figures include Russian contribution to inputs through Daugava, Nemunas and Pregolya rivers.*

*The figures for transboundary inputs originating in the Contracting Parties and discharged to the Baltic Sea through other Contracting Parties are preliminary and require further discussion within relevant transboundary water management bodies.*

the 6 year period since the previous selection of indicators, new indicators may have been generated and proposed to support indicator assessment, particularly if new pressures on the system have emerged. Thus, at the beginning of each cycle, a review of the indicator set and a model based study to select the most appropriate indicators in relation to pressures (Houle et al., 2012) would be prudent along with the evaluation of their scientific standard and applicability (Queirós et al., 2016a). Once an assessment of state for GES has been made, the next question likely to be faced by mangers is: are we moving in the right direction to maintain or recover GES? Here again, modeling studies are key since trends in state and indicators can be projected incorporating the prevailing climatic conditions and a range of anthropogenic pressures (Lynam and Mackinson, 2015; Piroddi et al., 2015b; Uusitalo et al., 2016). Such studies can inform the policy makers on the likelihood or reaching previously agreed reference points given the prevailing climate. In some instances, prevailing conditions (such as trends in temperature or changes in storminess) may alter the trajectory of change in the system such that managers may wish to alter targets to account for this (e.g., Lynam and Mackinson, 2015; van Leeuwen et al., 2016). Once these decisions have been made, monitoring programs must be steered to ensure that data are collected to support those areas of the assessment that are most uncertain, and/or showing the strongest degradation (Shephard et al., 2015b). Adaptive monitoring in this way should be most cost-effective and lead to information being generated where it is most needed.

If the ecosystem in question has not met GES, or if the trends assessment suggests the system is likely to be degrading given the range of current pressures, policy makers will wish to implement management measures to maintain biological diversity and food web functioning. However, given the range of pressures in any system and the differing options of measures that managers may consider implementing, further evidence is required to enable an informed choice to be made. Here, once again, modeling can assist through environmental impact assessments and management strategy evaluation. Ecosystem and higher trophic level modeling, through which projections are made for indicators given a range of scenarios that include a suite of management strategies coupled to climate change trajectories, can be used to estimate risk and reward of each potential option. From this set, managers can then choose the most socially acceptable solution.

# Perspectives: Further Development of Novel Indicators

DEVOTES assessed the capabilities of state-of-art models to provide information about current and candidate indicators outlined in the MSFD, particularly on biological diversity, food webs, non-indigenous species and seafloor integrity (Piroddi et al., 2015a; Tedesco et al., 2016). We demonstrated that models are largely able to inform on food webs, but that non-indigenous species habitats and seafloor integrity are often poorly addressed. Notably, however, mechanistic models such as ERSEM have been explicitly designed to represent benthic processes associated with seafloor functioning (Butenschön et al., 2015; Queirós et al., 2015; Lessin et al., 2016). In this project, we used modeling tools to refine some of the existing indicators and to develop novel indicators. To address the gaps related with indicators on non-indigenous species, we developed the CIMPAL index (Katsanevakis et al., 2016). To help refine existing seafloor integrity indicators, we employed benthic trait analysis (van der Linden et al., 2016) to specifically understand benthic community function in relation to habitats of the Bay of Biscay and in the North Sea, and identified typological groups of benthic macroinvertebrates, based on response and effect traits as potential indicators for MSFD D6 and D1 (Lynam et al., 2015).

Understanding the maintenance of the relationship between biological diversity and environmental disturbance is simultaneously challenging and key to supporting an ecosystembased management approach. Traits-based approaches emphasize the functional characteristics of species to study this relationship, and availability of such information for marine species has rapidly increased in recent years, particularly in Europe (Costello et al., 2015). Mechanistic modeling approaches often utilize functional classifications to represent marine organisms (Queirós et al., 2015) providing a route to investigate how ecosystem processes may change under future environmental conditions, despite the complexity inherent to the process (Bremner, 2008; Queirós et al., 2015; van der Linden et al., 2016). Several studies have suggested model developments are needed to support a more comprehensive use of traits-based approaches (e.g., Savage et al., 2007; Webb et al., 2010; Verberk et al., 2013).

Understanding the connectivity attributes of each species is central to establishing effective management and conservation strategies such as the creation of networks of Marine Protected Areas (MPAs). For instance, Webb et al. (2010) suggests that a quantitative framework combining Bayesian multilevel models, dynamical systems models and hybrid approaches has the potential to meaningfully advance traits-based ecology. Reiss et al. (2014) stresses the importance of considering multiple biological traits and benthic ecosystems functions in Distribution Modeling techniques and their high potential to assist in a marine management context (e.g., MPA designations). The fragmentation of habitats is a threat to the maintenance of biological diversity, thus dispersal traits of species and the connectivity within and between population and communities are important attributes of species distributional patterns (Chust et al., 2016). As multiple species traits are likely to influence trophic interactions and functioning, approaches which seek to integrate trait-based methods with the food webs framework are also emerging with many recent advances stemming from modeling work (e.g., Thompson et al., 2012; Eklöf et al., 2013; Poisot et al., 2013; Nordström et al., 2015). These have shown to be successful at predicting network structure (Eklöf et al., 2013), in determining the strength of individual trophic links (Klecka and Boukal, 2013), highlighting that multiple traits are needed for more complete descriptions of interactions (Eklöf et al., 2013) and that functional and trophic attributes should be assessed in an integrated manner to provide accurate assessments in a changing environment (Boukal, 2014).

# Acceptance of Model Information by Decision Makers

There are significant challenges still surrounding the uptake and use of complex models by decision makers. Many of these relate to understanding of models, outputs in the right currency, treatment of uncertainty, rigorous quality standards, and availability of user-friendly model products (Hyder et al., 2015). There are few direct examples of how outputs have led to decisions either related to policy or management, but this is not surprising since decision making is normative and incorporates societal values alongside the evidence base (Fletcher, 2007). Models can contribute to the evidence base that underpins decision making, but this is at an early stage with many other factors accounted for after compilation of the scientific knowledge (see e.g., van den Hove, 2007). However, it is clear that models have a vital role to play in decision making, as many policy or management options cannot be tested experimentally or in real ecosystems. The key to improving the uptake of models by decision makers is to build understanding both of the methods and issues through multidisciplinary communities that co-develop models (Hyder et al., 2015). Decision making timescales are often at odds with model development, so it is important to be able to adapt existing models to address these needs at short notice, and to provide outputs in the right currencies (monetary, stocks, natural capital, ecosystem services) understandable by policy makers (Hyder et al., 2015; Queirós et al., 2016b). For instance, many of the examples above focus on the biological aspects of decision making, whereas many decisions are based on economic or social outcomes (Watson, 2005; Papathanasopoulou et al., 2013). However, including the human dimension is very important as it may represent the largest source of uncertainty (Fulton et al., 2011). Integration of human dimensions like governance into models is vital to increase understanding of the likely outcome of management decisions, and has driven the development of social-ecological models (e.g., Griffith et al., 2012). There is a constant drive to resolve technical challenges around complexity, uncertainty and model skill (Allen and Somerfield, 2009; Payne et al., 2016) but models can provide useful insight despite being wrong (Box, 1979). Hence, modelers need to understand the needs of decision makers and work closely with them to build trust in their models, but ensure that uncertainty and key assumptions are highlighted through quality statements (ICES, 2016b; Queirós et al., 2016b). Without this trust, models will not be included in the decision making process or the implementation cycle.

# CONCLUSIONS

Our overview shows that modeling can support the review of objectives, targets, and indicators for the MSFD. Modeling is the only option to evaluate different management strategies and thus help select appropriate management measures. We recommend that indicator assessments are supported by modeling studies, so that linkages between descriptors and global pressures on the marine environment (such as climate change and ocean acidification) and cumulative impacts are more fully grasped. Models can be used to highlight recovery trajectories of indicators and a range of management strategies should be explored through scenarios to provide support to decision makers. Specifically, the likely synergistic and antagonistic effects of management measures and concurrent changes in prevailing climatic conditions should be investigated at each assessment cycle of the MSFD. Sources of uncertainty (measurement error, uncertainty in pressure-state relationships, model uncertainty) must be considered and communicated during indicator assessments and model studies (Cartwright et al., 2016). Utilizing modeling support as a routine in the assessment cycle would ultimately improve long term planning for the marine environment.

# REFERENCES


There is still a wide gap between modelers and decision makers, and the full utility of models has not yet been realized. To enable models to better support marine environmental management and the MSFD, it is important to ensure that communities of policy makers and scientists are set up to co-develop ecosystem models. At a national level, interdisciplinary groups are required to support assessments and policy making, including internationally, groups of modelers compare approaches and harmonize methods across regional seas to support MSFD assessments. Internationally, independent technical reviews of national groups' progress should be made to ensure high quality advice and promote harmonization between regional seas.

# AUTHOR CONTRIBUTIONS

CL conceived of the paper and all authors contributed to the paper writing.

# ACKNOWLEDGMENTS

This manuscript is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biological diversity and assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), www.devotes-project.eu.


reflectance difference. J. Geophys. Res. 117, C01011. doi: 10.1029/2011JC 007395


and L. L. White (Cambridge; New York, NY: Cambridge University Press), 411–484.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Lynam, Uusitalo, Patrício, Piroddi, Queirós, Teixeira, Rossberg, Sagarminaga, Hyder, Niquil, Möllmann, Wilson, Chust, Galparsoro, Forster, Veríssimo, Tedesco, Revilla and Neville. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# **Better monitoring, better assessment**

# Need for monitoring and maintaining sustainable marine ecosystem services

#### *Jacob Carstensen\**

*Department of Bioscience, Aarhus University, Roskilde, Denmark*

#### *Edited by:*

*Susana Agusti, The University of Western Australia, Australia*

#### *Reviewed by:*

*Jesus M. Arrieta, Instituto Mediterraneo de Estudios Avanzados, Spain Juan-Carlos Molinero, GEOMAR Helmholtz Centre for Ocean Research Kiel, Germany Dolors Vaque, Marine Sciences Institut (Consejo Superior de Investigaciones Científicas), Spain*

#### *\*Correspondence:*

*Jacob Carstensen, Department of Bioscience, Aarhus University, Frederiksborgvej 399, DK-4000 Roskilde, Denmark e-mail: jac@dmu.dk*

Increases in human population and their resource use have drastically intensified pressures on marine ecosystem services. The oceans have partly managed to buffer these multiple pressures, but every single area of the oceans is now affected to some degree by human activities. Chemical properties, biogeochemical cycles and food-webs have been altered with consequences for all marine living organisms. Knowledge on these pressures and associated responses mainly originate from analyses of a few long-term monitoring time series as well as spatially scattered data from various sources. Although the interpretation of these data can be improved by models, there is still a fundamental lack of information and knowledge if scientists are to predict more accurately the effects of human activities. Scientists provide expert advices to society about marine system governance, but such advices should rest on a solid base of observations. Nevertheless, many monitoring programs around the world are currently facing financial reduction. Marine ecosystem services are already overexploited in some areas and sustainable use of these services can only be devised on a solid scientific basis, which requires more observations than presently available.

**Keywords: biodiversity, ecosystem trends, eutrophication, food-webs, global change, ocean acidification, ocean governance, overfishing**

#### **INTRODUCTION**

The last 10,000 years, known as the Holocene, have been a relatively stable period in earth's climate history (Petit et al., 1999), but recently human activities have become the main driver of environmental change at the local as well as global scale (Rockström et al., 2009). Humans have significantly altered the biogeochemical cycles on earth (Vitousek et al., 1997); something thought impossible just a few decades ago. Burning of fossil fuels, deforestation, mining, and other activities have increased the concentration of CO2 in the atmosphere and ocean, elevating the greenhouse effect with rising temperatures as consequence. So far, the oceans have managed to store three times as much heat as the atmosphere (Levitus et al., 2001) and absorb about one third of the human-induced CO2 emitted into the atmosphere (Steffen et al., 2007). However, recent studies suggest that the ocean's buffer capacity might decrease with further warming (Gruber et al., 2004).

Industrial nitrogen fixation and phosphate mining as well as fossil fuel burning have mobilized nitrogen and phosphorus (Vitousek et al., 1997). Humans have almost doubled the supply of nitrogen from the atmosphere to land, leading to an increased release of the greenhouse gas N2O (Gruber and Galloway, 2008). Phosphate demands for agriculture have increased phosphorus inputs to the biosphere by factor of almost four (Falkowski et al., 2000). Nutrients applied to land as fertilizers are partly lost to the aquatic environment, eventually the ocean, where they stimulate production of organic matter, a process known as eutrophication (Nixon, 1995). One of the most deleterious effects of eutrophication is the development of hypoxia (Carstensen et al., 2014), having strong ramifications on nutrient biogeochemical processes (Diaz and Rosenberg, 2008; Conley et al., 2009).

Human demand on fish has significantly reduced populations of marine top predators (Pauly et al., 1998), altering the flow of energy through food-webs and eventually leading to ecosystem collapses (Jackson et al., 2001). Fisheries landings have increased by more than 50% from 1970 to 2005 (Duarte et al., 2009) and the number of unsustainable fisheries is growing (Vitousek et al., 1997). In addition to reducing the overall population of marine top predators, overfishing has also selected toward smaller populations by removing the largest individuals (Jackson et al., 2001). It is possible that overfishing may exacerbate effects of eutrophication through trophic cascades, disrupting the normal flow of energy through marine food-webs (Scheffer et al., 2005). Another facet of altered energy flows is the global loss of biodiversity caused by overfishing, pollution, and habitat destruction reducing ocean ecosystem services (Worm et al., 2006).

Human pressures on marine ecosystems have increased recently to an extent where every area of the oceans is affected to some degree, although the human footprint is largest in the coastal zones with a high population density (Halpern et al., 2008). The multiple pressures of human activities have eroded the capacity of marine ecosystems to provide services benefitting humans. The oceans no longer constitute an infinite reservoir of natural resources that humans can exploit unconcerned. Therefore, science has an important role in identifying problems as well as their solutions, and conveying this knowledge broadly to the public and particularly, decision makers (Levin et al., 2009).

#### **ASSESSING HUMAN IMPACTS ON MARINE ECOSYSTEMS**

Our knowledge on human impacts on marine ecosystems has mainly been driven by observations supported by models for extrapolation. However, there is a significant lack of data on human pressures and marine effects, particularly in the open ocean. Data are often scattered in time and space, because they mostly arise from various research cruises and ships-ofopportunity; uncoordinated activities not aimed at assessing changes over time. Therefore, models are needed to integrate these data (e.g., Boyce et al., 2010; Halpern et al., 2012), but for many components of ocean health such models do not exist or they are so coarse that the reliability of the output may be disputable (Mackas, 2010; McQuatters-Gollop et al., 2010; Rykaczewski and Dunne, 2011).

Remote sensing data from satellites overcome the problem of spatial and temporal sampling heterogeneity and can be used for assessing changes in sea surface temperature and ocean color from which proxies for phytoplankton biomass and productivity can be derived (Behrenfeld et al., 2006), but they also have their limitations. Remote sensing applies to the upper surface layer only, and satellites cannot assess processes taking place at deeper depths. Algorithms for processing remote sensing data have mainly been developed for the open ocean, and the algorithms produce biases in shallower coastal waters. The proxy information obtained from satellite imagery provides only a small fraction of information needed to assess human impact on marine ecosystems.

Autonomous sensors typically placed on fixed buoys or floatable undulating devices such as Argo floats complement remote sensing by providing subsurface information on salinity, temperature, oxygen, and bio-optical properties (Roemmich et al., 2009). For instance, Argo float data with the support of global climate models revealed that the deep ocean (*>*300 m) was taking up more heat during the recent surface-temperature hiatus period (Meehl et al., 2011). At present, only the most basic physicalchemical variables are measured using these autonomous devices, since other measurements of interest (e.g., nutrient concentrations) typically require more regular maintenance, increasing the operating costs substantially.

Monitoring programs providing more consistent time series across a wide range of different physical, chemical and biological variables are found in certain coastal areas, e.g., the Chesapeake Bay and the Baltic Sea. These were typically initiated in the 1970s and 1980s, when pollution effects became clearly visible, to assess the efficiency of management actions to alleviate human pressure on overstressed marine ecosystems (Carstensen et al., 2006). In addition to assessing physical-chemical status, different organism groups from phytoplankton to top predators in the marine ecosystems were monitored. These monitoring programs have contributed substantially to our present understanding of trophic interactions in coastal areas and the disturbance of these imposed by human activities.

Understanding of long-term variations in ocean waters has so far been based on a few observatories, some of these organized within the Long Term Ecological Research (LTER) Network (www*.*ilternet*.*edu). Long-term decreases in pH and aragonite saturation from the Hawaiian Ocean Time-series (HOT) and Bermuda Atlantic Time Series (BATS) have highlighted another problem associated with increased emission of CO2, namely ocean acidification (Doney et al., 2009), which may alter ocean biogeochemistry (Beman et al., 2011). Long-term time series in coastal waters have revealed that pH is governed by changes in inputs from land rather than CO2 in the atmosphere (Duarte et al., 2013). The Continuous Plankton Recorder (CPR) survey has been in operation since 1931 and has provided valuable insights into how climate oscillations affect plankton communities (Edwards et al., 2009). Since 1949 the California Cooperative Oceanic Fisheries Investigations (CalFOCI) program has investigated distributions of phytoplankton, zooplankton and fish distributions off Southern California and showed how changes in the Pacific Decadal Oscillation (PDO) can precipitate sudden shifts in these distributions (McGowan et al., 2003). Nevertheless, despite the value of these unique time series there is a need to establish and maintain ocean time series of high research quality, particularly in subtropical and tropical waters that are severely understudied at present.

#### **DIRECTIONS FOR THE FUTURE**

"We know more about the surface of the Moon and about Mars than we do about the deep sea floor, despite the fact that we have yet to extract a gram of food, a breath of oxygen or a drop of water from those bodies." This statement by Dr. Paul Snelgrove clearly articulates the need for improving our understanding of how marine ecosystems function, particularly as they provide essential ecosystem services to humans and because expanding human activities are putting these services under threat.

Our current understanding of marine ecosystem responses to human activities is limited by the availability of data, particularly long-term time series of physical and chemical conditions as well as biological properties. Moreover, efforts should be made to improve the accessibility and comparability of existing time series. Further development of models integrating monitoring data is needed to better assess changes over time and predict future trends, but models cannot stand alone without data. The lack of data is partly technical, as current measurement techniques may not necessarily provide the needed information, and partly financial, as costs of ocean sampling are indeed excessively expensive. Technological developments are expected to contribute more accurate, precise and cost-effective measurements over time. However, many marine monitoring programs are facing budget reductions, which have led to discontinuation of monitoring stations and abandoning sampling of biological components as well as decreasing monitoring frequencies. A possible consequence is loss of invested capital for establishing such long-term time series, simply because their value has to be written down. There is a growing discrepancy between the need for better understanding of human impact on marine ecosystems and the basis for addressing these scientific questions.

Ducklow et al. (2009) have identified seven key elements that will help science address critical issues on marine ecosystem services in times when human pressures on these are intensifying: (1) maintain existing monitoring programs and expand these with additional biological components, (2) establish new monitoring programs in under-sampled regions, (3) increase the use of remote sensing and autonomous monitoring devices, (4) establish targeted research program (process studies) in connection to long-term monitoring sites, (5) improve the integration of monitoring activities with ships-of-opportunity, (6) modify current funding for ecological research to balance consistent long-term research and short-term targeted studies, and (7) improve data access and synthesis using models. If these are recommendations are pursued we may eventually know more about our oceans than the surface of the Moon and Mars. The growing human imprint on marine ecosystems may, if left unmonitored and unattended, result in significant losses of ecosystem services that are crucial to support a globally growing population.

#### **ACKNOWLEDGMENTS**

This manuscript is a contribution from the DEVOTES project (DEVelopment Of Innovative Tools for understanding marine biodiversity and assessing good Environmental Status; www. devotes-project.eu), funded by the European Union under the 7th Framework Programme (grant agreement no.308392) and the WATERS project (Waterbody Assessment Tools for Ecological Reference conditions and status in Sweden).

#### **REFERENCES**


Worm, B., Barbier, E. B., Beaumont, N., Duffy, J. E., Folke, C., Halpern, B. S., et al. (2006). Impacts of biodiversity loss on ocean ecosystem services. *Science* 314, 787–790. doi: 10.1126/science.1132294

**Conflict of Interest Statement:** The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 23 May 2014; paper pending published: 29 June 2014; accepted: 23 July 2014; published online: 11 August 2014.*

*Citation: Carstensen J (2014) Need for monitoring and maintaining sustainable marine ecosystem services. Front. Mar. Sci. 1:33. doi: 10.3389/fmars.2014.00033 This article was submitted to Global Change and the Future Ocean, a section of the journal Frontiers in Marine Science.*

*Copyright © 2014 Carstensen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# European Marine Biodiversity Monitoring Networks: Strengths, Weaknesses, Opportunities and Threats

Joana Patrício<sup>1</sup> \*, Sally Little2, 3, Krysia Mazik <sup>3</sup> , Konstantia-Nadia Papadopoulou<sup>4</sup> , Christopher J. Smith<sup>4</sup> , Heliana Teixeira<sup>1</sup> , Helene Hoffmann<sup>1</sup> , Maria C. Uyarra<sup>5</sup> , Oihana Solaun<sup>5</sup> , Argyro Zenetos <sup>6</sup> , Gokhan Kaboglu<sup>7</sup> , Olga Kryvenko8, 9 , Tanya Churilova8, 9, Snejana Moncheva<sup>10</sup>, Martynas Bucas ˇ <sup>11</sup>, Angel Borja<sup>5</sup> , Nicolas Hoepffner <sup>1</sup> and Michael Elliott <sup>3</sup>

*<sup>1</sup> European Commission, Joint Research Centre (JRC), Directorate for Sustainable Resources, Ispra, Italy, <sup>2</sup> School of Animal, Rural and Environmental Sciences, Nottingham Trent University, Southwell, UK, <sup>3</sup> Institute of Estuarine and Coastal Studies, University of Hull, Hull, UK, <sup>4</sup> Hellenic Center for Marine Research, Institute of Marine Biological Resources and Inland Waters, Heraklion, Greece, <sup>5</sup> Marine Research Division, AZTI, Pasaia, Spain, <sup>6</sup> Hellenic Center for Marine Research, Institute of Marine Biological Resources and Inland Waters, Athens, Greece, <sup>7</sup> Institute of Marine Sciences and Technology, Dokuz Eylül University, Izmir, Turkey, <sup>8</sup> A. O. Kovalevsky Institute of Marine Biological Research of RAS, Sevastopol, <sup>9</sup> Marine Hydrophysical Institute of RAS, Sevastopol, <sup>10</sup> Institute of Oceanology-Bulgarian Academy of Science, Varna, Bulgaria, <sup>11</sup> Marine Science and Technology Center, Klaipeda University, Klaip ˙ eda, Lithuania ˙*

#### Edited by:

*Marianna Mea, Jacobs University of Bremen, Austria*

#### Reviewed by:

*Patrick Georges Gillet, Université Catholique de l'Ouest, France José Lino Vieira De Oliveira Costa, University of Lisbon, Portugal*

#### \*Correspondence:

*Joana Patrício joanamateuspatricio@gmail.com*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *16 June 2016* Accepted: *24 August 2016* Published: *08 September 2016*

#### Citation:

*Patrício J, Little S, Mazik K, Papadopoulou K-N, Smith CJ, Teixeira H, Hoffmann H, Uyarra MC, Solaun O, Zenetos A, Kaboglu G, Kryvenko O, Churilova T, Moncheva S, Bucas M, Borja A, Hoepffner N and ˇ Elliott M (2016) European Marine Biodiversity Monitoring Networks: Strengths, Weaknesses, Opportunities and Threats. Front. Mar. Sci. 3:161. doi: 10.3389/fmars.2016.00161* By 2020, European Union Member States should achieve Good Environmental Status (GES) for 11 environmental quality descriptors for their marine waters to fulfill the Marine Strategy Framework Directive (MSFD). By the end of 2015, in coordination with the Regional Seas Conventions, each EU Member State was required to develop a marine strategy for their waters, together with other countries within the same marine region or sub-region. Coherent monitoring programs, submitted in 2014, form a key component of this strategy, which then aimed to lead to a Program of Measures (submitted in 2015). The European DEVOTES FP7 project has produced and interrogated a catalog of EU marine monitoring related to MSFD descriptors 1 (biological diversity), 2 [non-indigenous species (NIS)], 4 (food webs), and 6 (seafloor integrity). Here we detail the monitoring activity at the regional and sub-regional level for these descriptors, as well as for 11 biodiversity components, 22 habitats and the 37 anthropogenic pressures addressed. The metadata collated for existing European monitoring networks were subject to a SWOT (strengths, weaknesses, opportunities, and threats) analysis. This interrogation has indicated case studies to address the following questions: (a) what are the types of monitoring currently in place? (b) who does what and how? (c) is the monitoring fit-for-purpose for addressing the MSFD requirements? and (d) what are the impediments to better monitoring (e.g., costs, shared responsibilities between countries, overlaps, co-ordination, etc.)? We recommend the future means to overcome the identified impediments and develop more robust monitoring strategies. As such the results are especially relevant to implementing comprehensive and coordinated monitoring networks throughout Europe, for marine policy makers, government agencies and regulatory bodies. It is emphasized that while many of the recommendations given here require better, more extensive and perhaps more costly monitoring, this is required to avoid any legal challenges to the assessments or to bodies and industries accused of causing a deterioration in marine quality. More importantly the monitoring is required to demonstrate the efficacy of management measures employed. Furthermore, given the similarity in marine management approaches in other developed systems, we consider that the recommendations are also of relevance to other regimes worldwide.

Keywords: Marine Strategy Framework Directive (MSFD), biodiversity, Good Environmental Status (GES), regional sea, pressures, SWOT analysis

#### INTRODUCTION

By 2020, European Union Member States should achieve GES (Good Environmental Status) for their marine waters to comply with the Marine Strategy Framework Directive (MSFD; 2008/56/EC). By the end of 2015, in coordination with the Regional Seas Conventions (RSC), each EU Member State was required to develop a marine strategy for their waters, together with other countries within the same marine region or subregion. Under the MSFD, reporting on GES should be carried out at a Regional Sea level (although marine sub-regions and subdivisions may be used to take into account the specificities of a particular area), which thus requires broad-scale monitoring with the potential to account for ecosystem level changes in response to both anthropogenic and natural pressures. In order to achieve this, assessment of GES under the MSFD is divided into 11 qualitative descriptors that collectively aim to cover the threats, pressures, and status of the whole marine ecosystem to give a complete picture of environmental status (Borja et al., 2013). Some of those descriptors relate to background conditions, some to pressures and some to impacts on the natural or social systems. Specific requirements of the MSFD include: (i) coordination of monitoring between EU Member States, (ii) that monitoring must be compatible with the EU Water Framework Directive (WFD), and Birds and Habitats Directives, and (iii) that monitoring must incorporate physical, chemical and biological components. It is necessary to consider the fundamental niches (i.e., sea bed, water column, and ice) to which each of these 11 descriptors relate, as well as the biological components (e.g., microbes, fish, see below). The assessment of each aspect of the marine environment requires an indicator (or usually a suite of indicators) to inform on state, and these indicators require data collected through monitoring (Shephard et al., 2015) although existing indicators may potentially leave gaps in current monitoring as new needs arise through the MSFD (Teixeira et al., 2014; Berg et al., 2015). Borja and Elliott (2013) describe monitoring sensu stricto as "the rigorous sampling of a biological, physical and/or chemical component for a well defined purpose, against a well defined end-point" and state that this may be in relation to the detection of trends away from an accepted starting point, non-compliance with a legal threshold, and/or comparison to standards, baseline or trigger points. However, current environmental management refers to different types of monitoring, all of which serve different purposes, with differing methods and analysis of the results. For example, Elliott (2011) identified 10 types of monitoring, two of which are of specific relevance to the MSFD: (1) Surveillance monitoring which enables the detection of spatial and temporal trends and, where necessary, leads to management action (for example, the detection of climate change trends), and (2) Condition monitoring to determine the present status of an area, and to detect change in condition over time (for example the health of the environment). However, once any deleterious change has been detected then investigative or diagnostic monitoring will be required to determine the cause-effect relationship, again linking to management actions.

The results of these types of monitoring, which each cover a spatial extent and/or a temporal duration and frequency, then requires feedback into management and policy decisions (Gray and Elliott, 2009). It is axiomatic that a system cannot be managed unless it is monitored thus giving data to show the status of the system and the results of the management measures implemented, hence taking all these elements together then requires and produces a monitoring program. Zampoukas et al. (2014) defined a Monitoring Program as "all substantive arrangements for carrying out monitoring, including general guidance with cross-cutting concepts, monitoring strategies, monitoring guidelines, data reporting and data handling arrangements. Monitoring programs include a number of scheduled and coordinated activities to provide the data needed for the ongoing assessment of environmental status and related environmental targets." A monitoring program can include one or several monitoring activities, defined as "the repeated sampling and analysis in time or space of one or more ecosystem components and carried out by an individual agency or institution. Data and marine information are obtained on a routine or specific basis, using sea surveys, remote sensing (i.e., teledetection), ferry boxes, data mining, or any other way." By expanding the comments of Zampoukas et al. (2014), monitoring programs should have an adequate coverage, in terms of accounting for current pressures and impacts on both natural and social systems but should also be adaptable to address environmental variability associated with emerging issues (see also Scharin et al., 2016). For the purposes of the MSFD, monitoring also needs to be coherent and coordinated, whereby EU Member States within the same region or sub-region follow agreed methods and focus on agreed biotic and abiotic components. This ensures that reporting is comparable across sea areas and can be incorporated into assessing GES at a Regional Sea level (Cavallo et al., 2016).

The nature and scale of marine environmental monitoring within Europe, was assessed within the DEVOTES FP7 project (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing GES, www.devotes-project.eu). This assessment involved compiling a catalog of marine biodiversity monitoring programs at the regional sea level (focusing on MSFD Descriptors where biodiversity is relevant: D1, biological diversity, D2, non-indigenous species (NIS), D4, food webs, and D6, sea-floor integrity). The catalog highlights:


To meet the requirements of the MSFD in terms of demonstrating GES, a detailed understanding of the above requires answering the following questions: (a) what are the types of monitoring currently in place? (b) who does what and how? (c) is the monitoring fit-for-purpose for addressing the MSFD requirements? and (d) what are the impediments to better monitoring (e.g., costs, shared responsibilities between countries, overlaps, co-ordination, etc.)?

By identifying current monitoring, this exercise aimed to highlight omissions in descriptors, biological components and habitats in particular marine regions or sub-regions and provide a broad overview of the spatial distribution and temporal intensity of monitoring activities. In particular, it aimed to identify programs or combinations of programs that will address the requirements of the MSFD, thus enabling decisions to be made about the cost-effectiveness of future monitoring. This high level assessment of the adequacy of current monitoring, in terms of spatial and temporal scale, in turn will allow the identification of components requiring inclusion in existing monitoring programs or the requirement for the development of entirely new monitoring programs. All of these aspects together constitute what is regarded here as a fit-for-purpose monitoring program.

# MATERIALS AND METHODS

#### Devotes Catalogue of Marine Biodiversity Monitoring Networks

Information was compiled regarding the current status of marine biodiversity monitoring, and in particular of the MSFD descriptors 1, 2, 4, and 6. In order to have an adequate spatial coverage of monitoring networks throughout the European Regional Seas, we first identified monitoring activities within the EU Member States or Regional Seas covered by the DEVOTES partners and then circulated the catalog outside that partnership for completion. Several other countries (EU Member States and third countries) voluntarily and enthusiastically provided information to this catalog. However, those areas with which DEVOTES has a stronger link have a more comprehensive coverage in the catalog (**Figure 1**). The catalog and Patrício et al. (2014) form the basis and common authorship of this manuscript. It is however recognized that monitoring programs in EU Member States are subject to regular amendment/change and as such the catalog requires regular updating to reflect the current status of monitoring activities throughout Europe. The catalog is publicly available at http://www.devotes-project.eu/devotes-release-new-versioncatalogue-monitoring-networks/. Despite the slightly incomplete nature, we consider that the catalog provides sufficient coverage to give the main lessons to be learned from this first, broad overview of European monitoring activities. It enables detailed analysis to support the harmonization of monitoring throughout Europe.

The focus of the catalog was on monitoring solely related to biodiversity (i.e., relating to MSFD D1, D2, D4, and D6) and not on determinants for human food provision or quality or physicochemical aspects (unless the latter are collected as supporting data for biotope characterization and biological parameters).

The catalog is presented in the above site as an EXCEL file containing two main tables:—"MONITORnetworks catalogue" and the parameters table "Param & physico-chemical data." The database is structured into three levels:

(1) Monitoring program level: this describes the general features of each monitoring activity, including the program name, the website and the time-series of the monitoring to enable users to find the full details (where available) of monitoring activities, methods, indicators, and parameters associated with a specific program. The geographical scope of each program is indicated through participation at national, EU, Regional Sea or local scale (e.g., for research or a single organization operating in a small area) together with information on the Regional or sub-regional seas to which the program applies.

The MSFD descriptor, the biodiversity component and the specific habitat type targeted by each program were identified to allow an assessment of the extent to which current monitoring practices address the ecological components. The biodiversity components include Microbes, Phytoplankton, Zooplankton, Angiosperms, Macroalgae, Benthic Invertebrates, Fish, Cephalopods, Marine Mammals, Reptiles, and Birds. The choice of biodiversity components was based on official MSFD documents and a related Commission Staff Working Paper (EC, 2012). The habitats (fundamental niches) include Seabed, Water column and Ice habitat. The categories adopted for habitat types followed the EU Commission Decision (EC, 2010) and EU Commission Staff Working Papers (EC, 2011, 2012) where it was agreed that the "use of these types provides a direct link between habitats assessed under Descriptor 1 and the substratum types to be assessed for Descriptor 6) and the European EUNIS habitat classification scheme" (EC, 2011, p. 18). In each case, the associated physico-chemical data collected (in the Param &

FIGURE 1 | Countries that have information reported in the DEVOTES Catalogue of Monitoring Networks (green) by June 2014 (country borders from Natural Earth database, http://www.naturalearthdata.com).

physico-chemical data table) and details of analytical quality control and quality assurance (AQC/QA, e.g., Gray and Elliott, 2009) were highlighted. Including this information broadly indicates the level of detail, confidence in and quality of a monitoring program, giving information on the nature of the explanatory variables, which may be linked to changes in environmental status. In addition, the information contained in these fields provides the opportunity to link the monitoring activities reported in this catalog to the "Data requirements" fields of the DEVOTES Catalogue of Indicators (Teixeira et al., 2014; available at http://www.devotes-project.eu/devotool/).

The extent to which each program accounts for specific pressures (either directly or indirectly where the biological and physico-chemical parameters indicate environmental change associated with those specific pressures) was identified. Here a pressure was defined as "the mechanism through which an activity has an actual or potential effect on any part of an ecosystem," (Robinson et al., 2008; Scharin et al., 2016). There was a list of 37 pressures, several of which were categorized as local and/or manageable if they were considered to occur as a result of human activities taking place on a localized scale and within the management unit (i.e., a discharge, a specific dredge disposal or aggregate extraction site). The causes and consequences of these pressures can be managed through permits/consents and monitoring. They are referred to as "Endogenic Managed Pressures" where the causes are managed as well as the consequences (Elliott, 2011). In contrast, other pressures were categorized as widespread and/or unmanaged, i.e., those that are beyond the control of direct management that are occurring at regional scales and often outside the management unit. For example, temperature and hydrological changes associated with climate change, or pH change due to volcanic activity (which may be local, but is not manageable). These are referred to as "Exogenic Unmanaged Pressures" where the consequences are managed rather than the causes (Elliott, 2011; Scharin et al., 2016). The MSFD only refers to an incomplete list of endogenic pressures and so the DEVOTES pressures list was produced as a revision from the MSFD and Koss et al. (2011). This adds in the managed and unmanaged pressures, thus allowing climate change to be considered as it has been omitted in MSFD implementation despite the wording of the Directive (Elliott et al., 2015).


The rationale behind gathering information at the network and web-platform level was to be able to infer whether and if so how EU Member States are optimizing their monitoring plans and efforts.

#### Data and Information Analysis

The metadata collated in the catalog were subject to a gap analysis to determine missing aspects and whether the current monitoring is fit-for-purpose both in terms of addressing the MSFD requirements but also wider issues within the marine environment such as providing information for maritime spatial planning, blue growth and industrial marine uses. The monitoring programs undertaken within each Regional Sea (and marine sub-region) were collated and assessed against the descriptors, biodiversity components, habitat types, and pressures to identify any gaps in provision. This led to a SWOT (Strengths, Weaknesses, Opportunities, Threats) analysis to better understand the monitoring networks in Europe, thus allowing us: (1) to explore possibilities for new efforts or solutions to problems specific to the MSFD; (2) to identify opportunities for success in the context of threats to success, clarifying directions and choices, and (3) to make recommendations to overcome the identified impediments and develop more robust monitoring strategies for the future.

Both the gap and the SWOT analyses were performed per marine sub-region (where applicable), marine region and at the Pan-European scale (i.e., considering all the activities reported in the catalog). This comprehensive compilation and interrogation allows us to present the main findings that are illustrated by appropriate case studies. More details regarding Regional Sea specific results are given in Patrício et al. (2014, 2015).

### RESULTS AND DISCUSSION

# What Are the Types of Monitoring Currently in Place?

A total of 57 Institutes (including a significant number from outside the DEVOTES project) provided information on monitoring activities. The catalog considers the depth and extent of marine monitoring in 16 EU Member States (Bulgaria, Croatia, Cyprus, Denmark, Finland, France, Germany, Greece, Italy, Lithuania, Malta, Portugal, Romania, Slovenia, Spain, and United Kingdom) and 15 non-EU countries (Albania, Algeria, Egypt, Georgia, Israel, Lebanon, Libya, Montenegro, Morocco, Norway, Russia, Syria, Tunisia, Turkey, and Ukraine) that share European Regional Seas boundaries. The catalog contains 865 entries (i.e., monitoring activities) and >298 monitoring programs (some of them with several activities). These activities covered four marine regions (Baltic Sea, Black Sea, Mediterranean Sea, North Eastern Atlantic), 23 sub-regions (as they appear in the MSFD Guidance documents e.g., Bay of Biscay and the Iberian Coast, Greater North Sea, Ionian Sea and the Central Mediterranean Sea, Levantine Sea, etc.), 83 ecological assessment areas (as they appear in national and regional documents e.g., Celtic Sea North, Kattegat and Skagerrak, Northern Adriatic) and also included 37 entries for non-EU waters.

Despite biological monitoring in the Baltic Sea starting in 1979 and being carried out annually in all nine surrounding countries, it was not possible to have an adequate coverage of these monitoring activities in the DEVOTES catalogue. Hence, data reported for the Baltic Sea were deemed insufficient to allow a robust analysis of regional biodiversity monitoring networks. This was mainly due to the low number of partners from the Baltic region in the DEVOTES project, whereas at the same time representatives from the Baltic countries were also involved in another regional pilot project (BALSAM, http://www.helcom. fi/helcom-at-work/projects/balsam) for enhancing the capacity of the Baltic Sea Member States to develop their monitoring programs. The BALSAM project was led by HELCOM, the Regional Sea Convention responsible for coordinating monitoring and assessment of the marine environment in the Baltic Sea. The HELCOM Monitoring and Assessment Strategy (MAS) was endorsed by HELCOM HOD 41/2013 and was adopted by the HELCOM Ministerial Meeting in 2013. A review of monitoring programs resulted in the report and publications (HELCOM, 2013, 2015) and so to complement the scarce regional information obtained from the DEVOTES catalogue, we also used data compiled by HELCOM (2013, 2015). We acknowledge the methodological inconsistency in respect to other European marine regions but we considered that it was more acceptable to use these comprehensive reports on the monitoring programs in the Baltic Sea, rather than excluding it. Given the large degree of coordination by the HELCOM countries, in assessing the monitoring activities we assumed that there would be a maximum number of national monitoring programs performed by all Baltic countries (i.e., nine programs) for any element monitored by all states.

Regarding monitoring types, most monitoring reported in the catalog comes under the term surveillance monitoring, ranging from 88 to 94% in the North Eastern Atlantic (NEA), Mediterranean and Black Sea (**Figure 2**). There is less condition monitoring which ranged from 6 to 10% in these three regional seas.

The date at which monitoring started varies widely throughout the catalog (**Figure 3**) but in general the number of monitoring activities has increased over the last 100 years, with most over the last three decades. Important triggers for monitoring were the Regional Sea Conventions and associated Action Plans. However, there are large differences between Regional Seas, for example, compared to the Baltic Sea and North East Atlantic which had monitoring from the 1970s, there are few monitoring activities in the Mediterranean Sea prior to the 1990s and most Black Sea monitoring programs were initiated in the 2000s.

Throughout the catalog, very different monitoring frequencies are reported, varying from minute to sub-hour, hourly, daily, weekly, twice a month, monthly, bi-monthly, 3–6 times a year, seasonally, 2/3 times a year, twice a year, annual, bi-annual, every 6 years, and up to every 10 years to sporadic, depending on which biodiversity component is the target, the national and international environmental regulations and the budgetary constraints.

#### Who Does What and How?

The catalog identified 298 monitoring programs that are suitable to address GES of the MSFD descriptors (i.e., directly or indirectly target the biodiversity-related descriptors). In the NEA, 60% of monitoring programs are undertaken to fulfill the objectives of European Directives, the OSPAR Convention and other International Conventions (**Table 1**). Thirty-one percent of these programs address two or more of these legislative drivers and 18% additionally address national monitoring obligations (**Table 1**). Most (83%) of these monitoring programs are undertaken by government agencies and institutions, but 17% are also undertaken by charities, Non-Governmental Organizations (NGOs) and research institutes (e.g., SAHFOS in Plymouth coordinates the Continuous Plankton Recorder scheme, which has been monitoring plankton since the 1920s and produces most of the data required for plankton in the UK; **Table 1**). Most of the programs are surveillance monitoring programs (80%) and generally employ common monitoring protocols, particularly where these programs are undertaken within collaborative monitoring networks [e.g., in the UK the Clean Seas Environment Monitoring Program (CSEMP) previously the National Monitoring Plan (NMP) and the National Marine Monitoring Plan (NMMP)]. In the NEA, 38% of monitoring programs are undertaken as part of research programs (e.g., MESH–Mapping European Sea beds Habitats, MISTRALS and French POPEX research programs) and/or to address national monitoring obligations (**Table 1**). These are undertaken by both government agencies (53%) and NGOs and research institutes (46%) and are all surveillance monitoring programs (**Table 1**).

In the Mediterranean Sea, 55% of the monitoring programs are undertaken because of European legislation [e.g., DCR (Data Collection Framework for the EU Common Fisheries Policy) and WFD; **Table 1**]. Of these, 13% addressed two or more legislative drivers and/or research projects. Most programs (66%) are undertaken by government agencies and institutes (**Table 1**). The remaining programs are undertaken by NGOs and research institutes and address basin wide issues or more local research projects (e.g., JellyWatch—CIESM Monitoring jellyfish blooms along Mediterranean coasts and in the open sea or NETCET— Network for the conservation of Cetaceans and Sea Turtles in the Adriatic) and national monitoring (**Table 1**).

In the Baltic Sea, all of the monitoring programs are undertaken to fulfill the objectives of European Directives, the HELCOM Convention and other International Conventions (**Table 1**). Most programs (93%) address two or more of these legislative drivers in addition to national monitoring obligations

and, in two cases, research programs. As such, most programs are part of monitoring networks and employ standard monitoring and QA protocols (i.e., HELCOM COMBINE, available at http://www.helcom.fi/action-areas/monitoring-and-assessment/ manuals-and-guidelines/combine-manual). These programs are mainly undertaken by government agencies.

In the Black Sea, most monitoring programs (78%) address the objectives of European Directives, the Bucharest Convention and other International Conventions in addition to national monitoring and research programs (e.g., World Ocean–in Russia; **Table 1**). Seventy percent of the monitoring programs are undertaken by governmental agencies and institutes, however 30% of monitoring is carried out by NGOs and research institutes (**Table 1**).

### Is the Monitoring Fit-For-Purpose for Addressing the MSFD Requirements?

In the context of the MSFD implementation, as a first step in the preparation of programs of measures, EU Member States across a marine region or sub-region should analyze the characteristics, pressures and impacts in their marine waters (see MSFD Annex III and Commission Decision 2010/477/EU). The second step toward achieving GES should be to establish environmental targets and monitoring programs for ongoing assessment, enabling the state of the marine waters to be evaluated on a regular basis. Hence, it is necessary to question how the monitoring fitness-for-purpose should be assessed. Monitoring has to provide the data to classify a marine area as reaching or failing to reach GES. To do so, the monitoring programs have to accommodate the descriptors, indicative characteristics, pressures, impacts and ideally should be able to provide data for the calculation of the indicators on which GES should be defined. Overall, our analysis showed several areas where current monitoring might not be fit for purpose to address the MSFD requirements.

#### GES Descriptors

Monitoring programs which address the descriptors D1– biological diversity and D4–food webs are the most numerous in all Regional Seas when taken as a whole, whereas monitoring associated with D2–NIS and D6–seafloor integrity are the least numerous (**Figure 4**). The distribution of monitoring programs that address these descriptors, however, varies both within and between Regional Seas. In the NEA for example, all descriptors are covered by a large number of monitoring programs in the Greater North Sea and Celtic Seas, however monitoring programs in the Bay of Biscay and the Iberian Coast are less numerous and the limited number of monitoring programs in the Macaronesian biogeographic region is of concern. In the Mediterranean, most of the 35 cataloged activities addressing descriptor D4 have been carried out in the Western Mediterranean, whilst only a limited number of monitoring activities currently addresses this descriptor in the Central and Eastern Mediterranean. In the Black Sea, descriptor D4 is the least monitored descriptor and only three monitoring activities cover it. Regarding monitoring of descriptor D2, few monitoring activities have been reported in all Regional Seas apart from the Greater North Sea and Celtic Seas of the NEA.

Some of the above highlighted gaps were expected. For example, monitoring for non-invasive species was not explicitly required by EU law before the MSFD entered into force although some EU Member States have been collecting data on non-invasive species and using them for coastal water quality assessment. The lack of D2 monitoring agrees with Vandekerkhove and Cardoso (2010) that most monitoring programs fail to detect some indicative NIS. Zampoukas et al. (2014) recommended that existing monitoring programs (e.g., for the WFD) should be complemented to explicitly record NIS and to include high priority sampling sites. Descriptor D6 is covered in all Regional Seas and sub-regions, apart from the Maraconesia biogeographic sub-region where D6 monitoring is lacking, which represents a major gap. Until recently, technical difficulties associated with deep sea sampling (Diaz et al., 2004) and a lack of tradition arising from the absence of effective international measures for assessing and protecting those habitats (Davies et al., 2007) explain why these habitats lag behind in established and complete monitoring programs. This explains why regions dominated by open sea and deep-sea ecosystems may have a poor data availability and hence face a greater difficulty in addressing MSFD D6 requirements.

#### Biodiversity Components

In general, monitoring programs which address high trophic level biodiversity components (such as reptiles, mammals, and birds) are lacking or limited in some Regional Seas (e.g., Black Sea and Mediterranean Sea) compared to the NEA (**Figure 5**). Cephalopod monitoring is limited in all Regional Seas. Monitoring programs addressing fish were not identified as lacking or limited in any Regional Seas although that monitoring is not evenly distributed throughout the sub-categories, with



TABLE 1 | Continued

monitoring for deep sea fish, deep sea elasmobranchs, and iceassociated fish lacking or limited to a small number of programs. This pattern is mirrored in the corresponding habitats which lack or have limited monitoring (i.e., deep sea and ice-associated habitats). In addition, most of the fish monitoring focuses on commercial species and less on non-commercial species or is focused on the fish in transitional waters (e.g., estuaries, fjords) as required by the WFD. The limited monitoring for reptiles, mammals, and birds in most Regional Seas was not expected since such monitoring is required in the Habitats and Birds Directives. The same applies to the identified gaps in cephalopod monitoring, expected to be already operational for the Common Fisheries Policy (CFP). Whilst these gaps could be due to incomplete reporting, they may indicate that the implementation of the EU environmental and fisheries related acquis has been limited. However, since some of these components (e.g., mammals) are indeed monitored under the Habitats Directive and regular status updates (every 6 or more years) are freely available through the Article 17 portal for that Directive, it is the lack of access to the monitoring information that represents a problem.

Monitoring programs that address microbes are limited in the NEA and Mediterranean Sea or lacking in the Black Sea (**Figure 5**). With the exception of microbes, biodiversity components that belong to low trophic levels are generally well addressed by monitoring programs in all Regional Seas, however with a smaller number of offshore stations in all relevant components compared to coastal stations, particularly in the Baltic Sea. Zooplankton monitoring also appears limited in the Mediterranean Adriatic and Central Mediterranean Sea. The lack of microbial diversity monitoring is expected as, with the exception of pathogens in the Bathing Water Directive, it was not previously addressed at the European level. Nevertheless, the overall rather good coverage of low trophic level monitoring could be related to the long European tradition of eutrophication monitoring and to the similar requirements of monitoring eutrophication under the WFD (Ferreira et al., 2011). Similarly, and against a declining trend in monitoring effort, de Jonge et al. (2006) emphasized both the lack on monitoring on these lower trophic components and the lack of monitoring on functioning rather than on structure in marine systems.

#### Quality Assurance (QA) and Supporting Physicochemical Data

For a number of biodiversity components QA is lacking. The BEQUALM (Biological Effects Quality Assurance in Monitoring Programmes) and UK NMBAQC (National Marine Biological Analytical Quality Control) schemes respectively for contaminants and benthic invertebrates do provide Analytical Quality Control and QA (Gray and Elliott, 2009) in some Regional Seas (e.g., NEA and Black Sea). However, approximately half of the monitoring activities do not collect supporting physicochemical data which thus provides a major drawback in having sufficient information to explain the ecological findings.

#### Habitats

With respect to the seabed and water column, most monitoring activities have been reported to cover "others" instead of a specific habitat from the list. This indicates that these activities cover several habitats and in many instances notes were added including coverage in multiple habitats. The monitoring activities that cover a specific seabed habitat are most numerous for "littoral sediment" in the NEA, and Mediterranean and Black Seas. In total, 10 seabed habitats have not been reported to be covered by monitoring activities. Nevertheless, these habitats might be covered by the monitoring activities which have been reported to cover "others" (i.e., 256 activities in the NEA, 22 in the Mediterranean and four in the Black Sea). In the water column, the NEA monitoring activities cover all five habitats and the Mediterranean activities cover four habitats (i.e., "variable salinity (estuarine) water" is not covered). In the Black Sea only "marine water: coastal" and "marine water: shelf " are indicated to be covered by monitoring activities but these may be regarded as "catch-all" terms. As with seabed habitats, the water column habitats which do not seem to be covered could be monitored

through activities that include "others" (i.e., 383 in the NEA, 32 in the Mediterranean and three in the Black Sea), however, as stated above this could not be verified. Monitoring programs addressing ice-associated habitats are recorded as completely lacking on those Regional Seas where these habitats occur (NEA and Baltic), which could be partially attributed to the monitoring activities targeting this habitat indirectly through monitoring focusing in the ice-associated species or communities (e.g., seals; Teixeira et al., 2014), but also to a lack of input from more Northern countries.

#### Pressures

In the Greater North Sea and Celtic Sea (NEA), all 37 pressures are covered by monitoring activities (**Figure 6A**). In the Baltic Sea, 26 pressures are covered. Although there are between 11 and 25 pressures covered in the Mediterranean, in the Black Sea, and in the NEA sub-regions Bay of Biscay and Iberian Coast and Macaronesia, the actual number of monitoring activities covering these pressures is limited when compared to the Greater North Sea and Celtic Sea (North Eastern Atlantic; **Figure 6A**).

Despite it being an individual MSFD descriptor (D11– introduction of energy), monitoring programs addressing the pressure "underwater noise" are limited in the Baltic Sea and Black Sea (i.e., only one monitoring activity reported) and lacking in the Bay of Biscay and Iberian Coast and Macaronesia (both NEA) and the Mediterranean (**Figure 6B**). At present, the impact of noise on many biodiversity components is not well understood (e.g., Roberts et al., 2015) and the outputs of such monitoring cannot be used effectively. Also the pressures "marine litter," "noise," and "introduction of non-indigenous species" are mainly monitored in the NEA and coverage is limited in other regional seas. The limitation in monitoring activities for the first

two of these pressures in the catalog represents a partial gap as they are directly linked to MSFD descriptors not targeted by this catalog (i.e., D10–marine litter and D11–introduction

of energy). In the Baltic Sea, until systematic non-indigenous species (NIS) monitoring programs (Lehtiniemi et al., 2015) and port biological sampling (HELCOM, 2013) are routinely

living resources.

established with wider Baltic coverage, the primary sources for NIS occurrence, their distribution and population size estimates remain non-systematic and include "inherent uncertainty" as this information depends on data collection for other purposes than NIS surveillance. Therefore, one of the major issues still to be solved is the establishment of an internationally coordinated monitoring system for NIS/Cryptogenic Species in the Baltic Sea and in other areas (Olenin et al., 2011). However, because of the high degree of concern regarding NIS emanating from the Suez Canal into the Mediterranean Sea, then this has resulted in more information available for parts of the Mediterranean Sea (Galil et al., 2014).

Monitoring programs addressing the pressures "water flow rate changes (widespread-unmanageable)," "change in wave exposure (widespread-unmanageable)," and "electromagnetic changes" are also lacking in the Black Sea and the Mediterranean Sea. Similarly, the pressure "introduction of radionuclides" is generally limited or lacking in all regional seas although this is incorporated into compliance monitoring (as conditions under their license to operate) carried out by nuclear power and reprocessing authorities and industries.

Monitoring for the "selective extraction of living resources," the pressures "catch," "bycatch," and "discards" is covered in the NEA, Baltic Sea and Mediterranean Sea, but lacking or limited in the Black Sea. The coverage of these pressures could be due to the fact that they are also being monitored through the EU Common Fisheries Policy and Data Collection Framework. Activities monitoring the pressures "maerl extraction" and "seaweed extraction" are limited in the NEA and lacking in the Mediterranean and Black Sea (there is limited commercial extraction and production in those areas).

# What Are the Strengths and Weaknesses of the Existing Marine Biodiversity Monitoring in Europe?

#### Strengths

As indicated above, there is a long history of monitoring in the European Regional Seas which has enabled the standardization of techniques and the development of best practice. For example, in the NEA and the Baltic Sea, monitoring starts from the early 1900s and in all Regional Seas at least some monitoring has taken place since the 1950s, with the number of programs increasing to the present day. Monitoring started to become more coordinated in the 1970s with the formation of HELCOM for the Baltic Sea and the Oslo and Paris Conventions (now OSPAR) for the NEA. Within each Regional Sea, it is generally common practice to collect supporting physico-chemical data simultaneously with biological data in order to explain biological change and several programs have associate formal QA guidelines to ensure validity of the data. Furthermore, for the four MSFD descriptors considered, all biodiversity components, habitats and pressures are addressed to some extent in all Regional Seas, with some programs addressing multiple descriptors. This provides a strong basis for the implementation of the MSFD and the assessment of GES. In most Regional Seas, the 11 biodiversity components are being covered and several are monitored simultaneously. Similarly, most monitoring programs concurrently address more than one seabed and water column habitat, thus optimizing the sampling efforts and providing an holistic approach to environmental monitoring. In general, most monitoring programs address more than one pressure. Although these are exceptions, some monitoring activities assess 18–20 pressures simultaneously (e.g., Celtic Sea sub-region), suggesting the potential for monitoring programs to become more efficient.

#### Weaknesses

Whilst the information in the catalog has enabled a broad spatial and temporal assessment of monitoring throughout Europe, it cannot be used to assess completely the adequacy of monitoring although it does identify areas which require further development. For example, whilst it is apparent that all descriptors, habitats and biodiversity components are being addressed, this is only the case for certain areas of some Regional Seas (e.g., in the territorial waters of a single nation). Detailed analysis at the individual Regional Sea level highlights this uneven distribution of monitoring activities at a spatial (sampling sites and stations) and temporal (sampling interval and frequency) level. Additionally, in a number of sub-regions, marine biodiversity monitoring programs address a specific target only (e.g., a particular habitat, species, pressure, etc.) resulting in an uneven distribution of monitored components (i.e., not all components are monitored in all sub-regions). For example, the NEA sub-regions Greater North Sea and Celtic Sea have the most reported monitoring activities of all Regional Seas; in contrast the NEA sub-region Macaronesia has a limited number of monitoring activities and contains several major gaps (e.g., no monitoring activities of D2 and D6). This may be partially an artifact of an incomplete coverage of the catalog, but it still reflects significant imbalances. It may also reflect the fact that monitoring historically has been driven by the presence or societal perception of problems, i.e., if society considers there to be an environmental problem then the authorities are more likely to respond and similarly pristine areas are not deemed to require extensive (if at all) monitoring (de Jonge et al., 2006).

In this broad-scale assessment, the number of monitoring programs that simultaneously address biodiversity components, descriptors, habitats and pressures (managed and unmanaged) is used in our study as a measure of the robustness of ongoing monitoring to potentially meet the requirements of the MSFD (to achieve GES) in all Regional Seas. However, whilst there is much information indicating the presence/absence of supporting physico-chemical data and QA to support this, detailed information on sampling design, sampling frequency, methodology, and the status of the QA programs (e.g., is it a national/international programs which includes assessment of the performance of participants?) is required to assess whether or not the monitoring is fit-for-purpose. Indeed, whilst monitoring in some areas is well developed, the associated indicators for the MSFD assessment of some descriptors are still under development indicating a weakness that needs to be addressed before the requirements of the MSFD can be fully met (Teixeira et al., 2014). Integrated monitoring is more likely to capture intricate ecological relationships, while at the same time the identification of anthropogenic cascade effects and cumulative or in-combination effects may be better identified if monitoring is coordinated in time and space. This includes bottom-up processes and top-down responses, and thus an analysis of ecosystem functioning as well as ecosystem structure, which underpins the Ecosystem-based approach, a central pillar of the MSFD and marine management (Elliott, 2014). Several monitoring programs both within and between regional seas address single or a limited number of components, habitats and pressures and although not explicitly investigated within the catalog, may be limited in terms of spatial (e.g., geographic area, sampling locations) and temporal (time-series, sampling frequency) scale. There is a need for more efficient and robust monitoring programs, integrating several biodiversity components, habitats and pressures through simultaneous monitoring, especially where pressures emanate through the whole ecosystem. Additionally, despite the extensive system of monitoring programs in most Regional Seas, a number of biodiversity components (e.g., microbes), descriptors (e.g., NIS), habitats (e.g., ice or deep sea habitats) and pressures (e.g., noise, introduction of radionuclides, selective extraction of living resources such as seaweed and maerl) are poorly or not addressed. Furthermore, most monitoring is focused on ecosystem structural aspects (the number of species, size of population, cover by a species) rather than on functional aspects (rate processes) even though the MSFD may change this emphasis (Borja et al., 2010; Hering et al., 2010).

The weaknesses identified are not trivial, as they concern some of the most relevant and elemental attributes of sound biodiversity monitoring schemes, recently identified by Pocock et al. (2015), for example, articulate objectives, standardized methodology, suitable field sampling methods, taxonomic literature, national, or regional coordination, data entry systems, QA of data, or/and scientific sampling design. Similarly, monitoring has to provide the 18 attributes for creating sound indicators and monitored elements given by Elliott (2011). Nevertheless, these findings can be used to reassess priorities when planning development or adjustment of the biodiversity monitoring programs in the future.

# What Are the Threats and Opportunities of the Existing Marine Biodiversity Monitoring in Europe?

#### Threats

Budgetary constraints are the most significant and obvious threat to monitoring within EU Member States (e.g., Borja and Elliott, 2013) thus giving rise to what has been termed the "monitoring requirement paradox," that there is an increasing amount of governance requiring monitoring while at the same time monitoring budgets have been cut (Borja et al., 2016; Strong and Elliott, accepted). For example, even where monitoring is undertaken within networks with standardized protocols (e.g., MEDITS and MEDPOL), budgetary constraints can result in countries suffering from data gaps over several years (see also de Jonge et al., 2006).

As identified above, achieving GES through the implementation of the MSFD is only attainable if the current and future monitoring of marine biodiversity is improved in all European Regional Seas. The number of ecosystem components monitored needs to be increased and specific monitoring programs developed to analyze pressures and pressure-impact relationships (Scharin et al., 2016). It may also be necessary to standardize sampling methods, increase sampling frequency and intensify sampling design in some regional seas. In order to ensure successful integrative monitoring schemes within and between Regional Seas, it may be necessary to establish a sustainable funding scheme and/or research budget and a rapid response/intervention framework. In the current economic climate it is difficult to envisage that EU Member States would be able to provide an appropriate budget for this but at present there is no pan-European or EU mechanism for funding monitoring across Member States. It is likely to remain the case that funding within a given area is the responsibility of that Member State.

The integration and holistic assessment of monitoring data at the Regional Sea level may be difficult, time consuming and economically restrictive due to methodological differences between EU Member States. This is also partly due to some EU Member States having a long history of monitoring and where many programs have been expanded, modified and developed over time. Hence, rather than establish new monitoring programs which specifically address MSFD objectives, EU Member States may rely on existing programs, which may be inadequate or not suitable, particularly where these have been designed for other purposes. Hence, when required to submit their MSFD monitoring proposals in 2014, EU Member States appear to report what they were doing rather than what they were required to do additionally for the directive (Boyes and Elliott, 2014), an approach which is expected to lead to anomalies and gaps (Boyes et al., 2016). In addition, differences in methods between countries which then need to produce a uniform assessment, will then need inter-calibration and inter-comparison exercises as has been carried out during the implementation of the EU WFD (e.g., Hering et al., 2010; Lepage et al., 2016).

Regional cooperation is required between EU Member and Non-Member States to implement the MSFD (Cavallo et al., 2016), although Non-Member States are under no legislative requirement to achieve GES in their respective regional seas. However, sea areas controlled by a combination of Member States and Non-Member States will still require coordination to tackle transboundary problems; this will certainly be the case for any current Member State which leaves the EU (Boyes and Elliott, 2016). If agreements with Non-Member States are not in place, achieving this cooperation may put undue additional pressure on EU Member States and may mean that infractions (proceedings in the European Court that a Member State has failed to meet a Directive) cannot be prosecuted and GES in the Regional Sea may not be achieved. For example, Norway is a non-Member State of the EU therefore not implementing the MSFD but it is still performing many of the aspects required by the Directive as well as being a leading member of OSPAR and following its monitoring and assessment protocols. Accordingly Regional Sea Conventions

have an important role in this coordination, for example through the OSPAR, HELCOM, and UNEP/MAP monitoring and assessment programs.

#### Opportunities

Several inadequacies have been identified in the monitoring currently undertaken in the Regional Seas. This presents a number of opportunities to modify and/or expand existing monitoring programs, develop new programs and to collaborate between EU Member States to develop standardized and robust programs and networks. These can occur both within and between Regional Seas that maximize the use of the best available data.

This would mean, for example, standardized verification of analyses and species identification, inter-calibration exercises for hazardous substance concentrations in biota, introduction and/or integration of validated external QA protocols, and a focus on upgrading the spatial and temporal resolution of monitoring and inter-calibration procedures. Introducing the simultaneous monitoring of descriptors, biodiversity components, habitats and pressures within single, large monitoring programs and ensuring that monitoring is designed to address specific pressures would increase the robustness of monitoring. This may also give an opportunity to create an online bank of all monitoring program data, accessible to all EU Member States, which should include information collected under different Directives and research programs (e.g., CFP, WFD, MSFD, EU funded projects, etc.). Creating such a uniform data storage system is being accomplished both at an EU scale, e.g., through the European Environment Agency and EMODnet (http://www.emodnet.eu/), and at an EU Member State level such as MEDIN in the UK (see http://www.oceannet.org/).

Those EU Member States that are members of Regional Sea Conventions (RSC) with a long history of marine monitoring and assessment, such as OSPAR and HELCOM, which have had joint monitoring programs since the 1970s, can provide valuable experience to states with a lesser history. The opportunity for collaborative work afforded by the implementation of the MSFD enables EU Member and Non-Member States to improve and/or develop monitoring programs to achieve GES in some regional seas (i.e., the Black Sea, Mediterranean Sea). This regional cooperation may prove essential for achieving GES in Regional Seas that border non-EU nations. For example, non-EU Member States of the Black Sea and Mediterranean Sea, respectively, should be encouraged through the Black Sea Commission and UNEP/MAP to develop more integrated monitoring programs (especially for the descriptors related to biodiversity monitoring). However, costs associated with activities required by RSC are borne by the country and so each country is required to fund its own commitments. Despite this, funding is becoming increasingly available from the EU, for example, to develop Integrated Regional Monitoring Implementation Strategies in the Mediterranean and Black Seas and basin-wide promotion of MSFD principles (PERSEUS and IRIS-SES projects). However, funding the development of strategies and principles may not be the same as funding the monitoring. Accordingly, our findings regarding the inadequacies in the monitoring currently undertaken in the European regional seas form the basis of further research proposals and requirements.

# Conclusions and Recommendations to Overcome the Identified Impediments and Develop More Robust Monitoring Strategies for the Future

The MSFD explicitly spells out that the assessment strategy is to be implemented at the regional or sub-regional level with both the individual EU Member States and, whenever possible, third countries (sharing the regions/sub-regions) acting together coherently and in a coordinated fashion through regional institutional cooperation structures. The success of the MSFD depends on a high level of cooperation between EU Member States, third countries and regional bodies mandated with environmental protection responsibilities (Long, 2011; Cavallo et al., 2016). Monitoring programs are to be compatible within marine regions or sub-regions and monitoring methods are to be consistent so as to facilitate comparability of monitoring results (Karydis and Kitsiou, 2013). The MSFD further specifies that standardized methods for monitoring and assessment be adopted (Zampoukas et al., 2013) thus putting the onus on the activities of the EU Member States, through coordination by the Regional Sea Conventions and even between RSC. Although there is some detail as to the descriptors or types of biological and other components that should be monitored, given that the MSFD is a Framework Directive, then the method of monitoring is left to the EU Member State level. This can create a large variation and incompatibility between, for example, two EU Member States that share marine borders within the same Regional Sea. There has always been a North and West compared to South and East difference within Europe with the former areas having more developed regional governance and organization, more detailed and long-standing administrative/legislative frameworks, a longer history and culture of environmental management and greater resources. A complicating feature is in the make-up of the Regional Seas, where in northern Regional Sea areas EU Member States comprise more than 80% of the participants compared to the Mediterranean and Black Sea regional areas where EU Member States make up less than 40% of the participant states. In the latter cases therefore, reaching GES for the whole region would require substantive support from the non-EU Member States, the relevant RSC and the EU Member States. The northern RSC (HELCOM and OSPAR) have a much longer experience of coordinated monitoring than the southern ones (UNEP/MAP, Black Sea Commission) and the western Member States have a longer history of compliance with EU environmental Directives than the eastern states. Hence, as all Member States have to implement and comply with the Directives then the intent of the MSFD needs to be reinforced to provide a much stronger level of clear coordination and standardization in the southerly Regional Seas (Zampoukas et al., 2013).

We acknowledge that the database on which the analysis here is based has some omissions and that the regional and national monitoring effort is changing annually. Despite that, we consider that the major and general lessons learned from its interrogation are robust and will hold even for a more complete database. As such, we strongly recommend the following:

	- a) Choice of indicators for each of the descriptors, including coverage of under-monitored components (e.g., microbes) and to monitor functional as well as structural aspects;
	- b) Developing methods which can cover large sea areas more efficiently (e.g., landers, gliders and seabed scanning) and provide the surveillance monitoring against which future investigative and diagnostic monitoring is carried out when marine environmental adverse effects are detected;
	- c) Design of spatial and temporal coverage for indicator measurement (including replication) which includes reconciling the compromise between monitoring effort and the capability of detecting impacts;
	- d) More effective methodologies for sampling, sample processing and analysis to produce data for the selected indicators (or proxies for those indicators), and
	- e) Quality assurance/control of the sampling and analytical process and using inter-comparison and inter-calibration exercises where necessary and where possible.

States. However, Member States can cooperate and one can even carry out monitoring on behalf of another.

Monitoring programs under the MSFD must be compatible with assessment obligations arising from other regional and EU or international instruments for reasons of continuity and efficiency (Long, 2011). While it is easy for an EU Member State to adapt or extend an existing monitoring program, they need to be fitfor-purpose with at least the minimum requirement to ensure adequate, defendable and meaningful assessments. Given that there is likely to be an increasing litigious framework, where assessments may be challenged legally in infraction proceedings (e.g., see Elliott et al., 2015), then the monitoring and resultant data have to be robust to those challenges.

As some of the descriptors or components may be new to established traditional monitoring [e.g., D2–NIS (e.g., see Olenin et al., 2011) and D11—introduction of energy—underwater noise (e.g., Roberts et al., 2015)] or as trends move from structural to functional ecosystem aspects (e.g., Strong et al., 2015), there is the need to develop/adopt cost-effective and innovative methods for monitoring including both state-of-the-art methods/tools and citizen science. Complementing existing monitoring programs for example for the EU WFD to explicitly deal with gap-filling on invasive species is recommended (Zampoukas et al., 2014) as well as integrating assessments made under the Habitats and Birds Directives for mammals, reptiles and birds.

Although, as shown here, there is a good basis on which to build, several EU Member States will need to broaden the scope and expand monitoring coverage and intensity to comprehensively assess the environmental status of their waters. Integrated monitoring programs taking into account a common vision on operational objectives and on indicators and targets for GES, are needed to achieve and maintain a particular or minimum desired level of environmental quality (Cinnirella et al., 2012) at the regional level. In addition, as the protection of the environment and the conservation of marine ecosystems functioning are now rooted in the EU regulatory code as binding legal obligations (Long, 2011; Boyes and Elliott, 2014), standards and protocols also need to be enacted to make the assessments strong, robust and legally defensible if challenged. It is emphasized and acknowledged that while many of the recommendations given here require better, more extensive and perhaps more costly monitoring, this is required to avoid any legal challenges to the assessments or to bodies and industries accused of causing a deterioration in marine quality.

Finally, it is emphasized that the detection of GES rest wholly on the adequacy of monitoring and the ability to detect a signal of change against a background of inherent variability, and conversely that inadequate monitoring will not be able to determine either such a change or determine whether management measures have had the desired effect.

# AUTHOR CONTRIBUTIONS

JP led the work on the manuscript. JP, SL, KM, KP, CS, HH compiled data, analyzed data (status, gaps, and SWOT), conceived, structured and wrote the manuscript. JP and HH constructed the figures. HT and KM built the catalog, compiled data, wrote the manuscript. MU, OS, AZ, GK, OK, TC, SM, MB collected data to fill in the catalog, did the regional data analysis and status/gaps/SWOT and contributed to the manuscript. AB, NH contributed to and revised the manuscript. ME envisioned the need for the monitoring catalog, wrote and revised the manuscript.

#### ACKNOWLEDGMENTS

This manuscript is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and

#### REFERENCES


assessing Good Environmental Status) project, funded by the European Union under the 7th Framework Program, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), www.devotes-project.eu. María C. Uyarra is partially funded through the Spanish program for Talent and Employability in R+D+I "Torres Quevedo." The authors gratefully acknowledge the help and metadata information received from the Regulatory Authorities within each EU Member State, the DEVOTES partnership and numerous non-DEVOTES experts. A list of non-DEVOTES experts that have contributed for the DEVOTES Catalogue of Monitoring Networks (June 2014 version) is available at http://www.devotes-project.eu/wp-content/uploads/ 2014/10/list-of-experts\_june2014.pdf.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declared a collaboration with the authors and states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Patrício, Little, Mazik, Papadopoulou, Smith, Teixeira, Hoffmann, Uyarra, Solaun, Zenetos, Kaboglu, Kryvenko, Churilova, Moncheva, Buˇcas, Borja, Hoepffner and Elliott. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

Edited by: *Jacob Carstensen, Aarhus University, Denmark*

#### Reviewed by:

*Matthias Obst, University of Gothenburg, Sweden Jo Høkedal, Østfold University College, Norway*

#### \*Correspondence:

*Roberto Danovaro r.danovaro@univpm.it Laura Carugati l.carugati@univpm.it*

*† These authors have contributed equally to this work.*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *14 June 2016* Accepted: *14 October 2016* Published: *23 November 2016*

#### Citation:

*Danovaro R, Carugati L, Berzano M, Cahill AE, Carvalho S, Chenuil A, Corinaldesi C, Cristina S, David R, Dell'Anno A, Dzhembekova N, Garcés E, Gasol JM, Goela P, Féral J-P, Ferrera I, Forster RM, Kurekin AA, Rastelli E, Marinova V, Miller PI, Moncheva S, Newton A, Pearman JK, Pitois SG, Reñé A, Rodríguez-Ezpeleta N, Saggiomo V, Simis SGH, Stefanova K, Wilson C, Lo Martire M, Greco S, Cochrane SKJ, Mangoni O and Borja A (2016) Implementing and Innovating Marine Monitoring Approaches for Assessing Marine Environmental Status. Front. Mar. Sci. 3:213. doi: 10.3389/fmars.2016.00213*

# Implementing and Innovating Marine Monitoring Approaches for Assessing Marine Environmental Status

Roberto Danovaro1, 2 \* † , Laura Carugati <sup>1</sup> \* † , Marco Berzano<sup>1</sup> , Abigail E. Cahill 3, 4 , Susana Carvalho<sup>5</sup> , Anne Chenuil <sup>3</sup> , Cinzia Corinaldesi <sup>1</sup> , Sonia Cristina6, 7, Romain David<sup>3</sup> , Antonio Dell'Anno<sup>1</sup> , Nina Dzhembekova<sup>8</sup> , Esther Garcés <sup>9</sup> , Joseph M. Gasol <sup>9</sup> , Priscila Goela6, 7, Jean-Pierre Féral <sup>3</sup> , Isabel Ferrera<sup>9</sup> , Rodney M. Forster <sup>10</sup> , Andrey A. Kurekin<sup>11</sup>, Eugenio Rastelli 1, 2, Veselka Marinova<sup>8</sup> , Peter I. Miller <sup>11</sup> , Snejana Moncheva<sup>8</sup> , Alice Newton<sup>6</sup> , John K. Pearman<sup>5</sup> , Sophie G. Pitois <sup>12</sup>, Albert Reñé<sup>9</sup> , Naiara Rodríguez-Ezpeleta<sup>13</sup>, Vincenzo Saggiomo<sup>2</sup> , Stefan G. H. Simis 11, 14 , Kremena Stefanova<sup>8</sup> , Christian Wilson<sup>15</sup>, Marco Lo Martire16, 17, Silvestro Greco<sup>18</sup> , Sabine K. J. Cochrane<sup>19</sup>, Olga Mangoni <sup>20</sup> and Angel Borja<sup>13</sup>

*<sup>1</sup> Dipartimento di Scienze della Vita e dell'Ambiente, Università Politecnica delle Marche, Ancona, Italy, <sup>2</sup> Stazione Zoologica "A. Dohrn", Napoli, Italy, <sup>3</sup> Centre National de la Recherche Scientifique, Institut Méditerranéen de Biodiversité et d'Ecologie Marine et Continentale, Aix Marseille Université, IRD, Avignon Université, Marseille, France, <sup>4</sup> Biology Department, Albion College, Albion, MI, USA, <sup>5</sup> King Abdullah University of Science and Technology, Red Sea Research Center, Thuwal, Saudi Arabia, <sup>6</sup> Centre for Marine and Environmental Research (CIMA), FCT, University of Algarve, Faro, Portugal, <sup>7</sup> Sagremarisco Lda, Vila do Bispo, Portugal, <sup>8</sup> Institute of Oceanology, Bulgarian Academy of Sciences, Varna, Bulgaria, <sup>9</sup> Institut de Ciències del Mar-CSIC, ICM-CSIC, Pg Maritim de la Barceloneta, Barcelona, Spain, <sup>10</sup> Institute of Estuarine and Coastal Studies, University of Hull, Hull, UK, <sup>11</sup> Plymouth Marine Laboratory, Prospect Place, The Hoe, Plymouth, UK, <sup>12</sup> Fish Behaviour Team, CEFAS Laboratory, Suffolk, UK, <sup>13</sup> AZTI, Marine Research Division, Pasaia, Spain, <sup>14</sup> Finnish Environment Institute (SYKE), Marine Research Centre, Helsinki, Finland, <sup>15</sup> OceanDTM, Riverside Business Centre, Suffolk, UK, <sup>16</sup> Consorzio Nazionale Interuniversitario per le Scienze del Mare (CoNISMa), Ancona, Italy, <sup>17</sup> EcoReach SRL, Ancona, Italy, <sup>18</sup> Istituto Superiore per la Protezione e la Ricerca Ambientale, Roma, Italy, <sup>19</sup> Akvaplan-niva AS, Fram Centre, Tromsø, Norway, <sup>20</sup> Dipartimento di Biologia, Università degli Studi di Napoli Federico II, Napoli, Italy*

Marine environmental monitoring has tended to focus on site-specific methods of investigation. These traditional methods have low spatial and temporal resolution and are relatively labor intensive per unit area/time that they cover. To implement the Marine Strategy Framework Directive (MSFD), European Member States are required to improve marine monitoring and design monitoring networks. This can be achieved by developing and testing innovative and cost-effective monitoring systems, as well as indicators of environmental status. Here, we present several recently developed methodologies and technologies to improve marine biodiversity indicators and monitoring methods. The innovative tools are discussed concerning the technologies presently utilized as well as the advantages and disadvantages of their use in routine monitoring. In particular, the present analysis focuses on: (i) molecular approaches, including microarray, Real Time quantitative PCR (qPCR), and metagenetic (metabarcoding) tools; (ii) optical (remote) sensing and acoustic methods; and (iii) *in situ* monitoring instruments. We also discuss their applications in marine monitoring within the MSFD through the analysis of case studies in order to evaluate their potential utilization in future routine marine monitoring. We show that these recently-developed technologies can present clear advantages in accuracy, efficiency and cost.

Keywords: marine monitoring, marine strategy framework directive, marine biodiversity, molecular approaches, in situ monitoring

### INTRODUCTION

Marine ecosystems are subject to a multitude of direct human pressures, such as overexploitation, eutrophication, pollution and species introductions (Halpern et al., 2008; Hoegh-Guldberg and Bruno, 2010; Burrows et al., 2011), including the effects of global impacts, namely ocean acidification and climate change (Doney et al., 2012). These stressors can have synergistic effects on marine ecosystems (Mora et al., 2013; Griffen et al., 2016), altering their functioning and ability to provide goods and services (Worm et al., 2006; Crain et al., 2008). Their impact is expected to be even stronger in enclosed and semi-enclosed basins with high population density, tourism flow and maritime activities (Danovaro, 2003). Improved knowledge on the consequences of the effects of multiple stressors on marine biodiversity and ecosystem functioning is urgently required (Danovaro and Pusceddu, 2007; Zeidberg and Robison, 2007; Danovaro et al., 2008; Nõges et al., 2016; Zeppilli et al., 2016). In 2008, the European Commission enacted the Marine Strategy Framework Directive (MSFD; 2008/56/EC), which aims to manage the European seas by using an ecosystem-based approach in order to gain a healthy and productive state (so called good environmental status; GES; see **Box 1** for the list of acronyms) (Borja et al., 2013).

The MSFD particularly aims at investigating the functioning of ecosystems (Cardoso et al., 2010; Borja et al., 2011), making a shift from structural, site-specific approaches to a functional, whole-sea system of monitoring (Borja and Elliott, 2013). An overarching aim is to promote regional harmonization of monitoring methods, used to assess marine environmental health and to obtain complete and long-term datasets from multiple ecosystem components, ranging from microbes to large marine mammals (Caruso et al., 2015).

Traditional methods applied to analyse marine biodiversity (e.g., morphological species identification, laboratory culture, toxicological analyses) are based on morphological identification and observational surveys, which are costly, time consuming and characterized by low upscaling potential to resolve change. One of the most evident limitations of traditional approaches is the identification and quantification of rare species and the ability to distinguish morphologically close or identical species (i.e., cryptic species), or poorly characterized juvenile stages of known species. Recently developed technologies present a wide variety of advantages including a higher taxonomic resolution and the capability to rapidly provide, often in near real time, information regarding wide geographic areas (remote sensing) or large temporal scales (e.g., autonomous observation platforms—buoys, moorings, ships-of-opportunity). As a result, the technological advancement is evolving in two main directions: (i) innovative molecular approaches for rapid biodiversity assessment (Bourlat et al., 2013); and (ii) autonomous and sensitive (optical) sensor systems, which allow us to operate and collect data in situ over wide spatial and temporal scales (She et al., 2016). Methods able to combine both requirements are thus highly desirable.

Innovative molecular technologies have fundamentally changed our understanding of biodiversity, particularly for microbes, rare species, "soft-species" or extremely small specimens, which are difficult to identify and cryptic species (to be studied combining molecular and morphological information; e.g., Derycke et al., 2005; Sogin et al., 2006) and new sensors and in situ technologies have already been applied to identify new forms of life in remote deep-sea habitats (Danovaro et al., 2014). However, most of the approaches/tools still need to be tested prior to their application in routine marine monitoring (e.g., EU project DEVOTES DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status). In this overview, we investigate the potential applications of various innovative tools and approaches in order to evaluate their applicability to routine marine monitoring, with a special focus on three main categories, which seem to be the most promising: (i) molecular approaches; (ii) innovative systems for in situ analysis; and (iii) remote sensing.

### MOLECULAR APPROACHES TO ASSESS MARINE BIODIVERSITY: FROM MICROBES TO MACROFAUNA

Morphological identification of species is heavily dependent on taxonomic experts, who are generally specialized on some specific groups of organisms (McManus and Katz, 2009; Bacher, 2012), and in some cases, the identification is impossible (e.g., cryptic and microbial species). Moreover, traditional taxonomy is generally time-consuming (Bourlat et al., 2013; Carugati et al., 2015), making large-scale and intense monitoring programs difficult to be undertaken. Molecular techniques are more universal (e.g., can target a broader range of taxa in a single analysis) and are less influenced by taxonomic expertise. Hence, molecular approaches have the potential to contribute to a large number of MSFD Descriptors (**Table 1**) and are promising tools to analyse the biodiversity of different biotic components (e.g., from prokaryotes, micro-eukaryotes to metazoans; **Table 2**), to identify species with different phenotypes or through the


different stages of the life cycles (still unknown for the majority of marine species).

# Use of Metabarcoding to Study Marine Biodiversity

The term "metabarcoding" refers to large-scale analyses of biodiversity through the amplification and sequencing of marker genes (e.g., 18S and 16S rDNA, Creer et al., 2010) and may also apply to capture-enrichment approach (Taberlet et al., 2012). Originally, most of the studies based on metabarcoding focused on prokaryotes (e.g., Sogin et al., 2006; Gilbert et al., 2009; Brazelton et al., 2010; Salazar et al., 2016), but, more recently, eukaryotes have also been investigated, including marine protists (e.g., Amaral-Zettler et al., 2009; Stoeck et al., 2010; Logares et al., 2014a; de Vargas et al., 2015; Massana et al., 2015) and metazoans (Thomsen et al., 2012; Lindeque et al., 2013; Hirai et al., 2015; Pearman and Irigoien, 2015). The development of highthroughput sequencing (HTS) technologies and of standardized procedures could allow metabarcoding analyses to be included in routine monitoring programmes (Visco et al., 2015; Zaiko et al., 2015a,b).

Morphology-based studies target a limited range of taxa (e.g., meiofauna or macrofauna). These biotic components host a potentially large number of cryptic and rare species (Ainsworth et al., 2010), which could be contextually detected using universal primers, targeting a broad range of taxa at the same time. This could lead to the incorporation of novel candidates for indicator species. For example, Chariton et al. (2010) suggested that phyla such as Kinorhyncha could be sensitive to contamination and used as an indicator. Metabarcoding could also be applied to assess changes in community structure along a disturbance gradient (Hewitt et al., 2005), or to detect non-native transient species (Jerde et al., 2011; Dejean et al., 2012; Cowart et al., 2015; Viard et al., 2016), allowing for better planning and implementation of conservation approaches. An interesting potential development of molecular techniques is the detection of sequences of eukaryotes from ancient DNA, or from the extracellular DNA pools, which enable the comparison between living species and species that were present in the same area in the (even recent) past (Corinaldesi et al., 2008, 2011, 2014; Pearman et al., 2016b). In addition, the progressive reduction of the costs of sequencing over time makes large-scale metabarcoding more feasible (e.g., de Vargas et al., 2015; Salazar et al., 2016).

Although metabarcoding can represent a useful tool for the census of marine biodiversity, there are still different shortcomings and pitfalls that prevent its extensive use in marine monitoring programmes. Metabarcoding can indeed provide an inaccurate or wrong estimation (under/over estimation) of the actual biodiversity of the sample due to variability in primers, PCR conditions, sequencing technology and bioinformatics pathways used.

The use of different marker genes could give different results in terms of taxonomic composition. Different gene regions vary in both taxonomic coverage and species-resolving power, leading to the introduction of errors in the identification and estimates

TABLE 1 | List of monitoring tools and MSFD Descriptors covered in this review.


*D1, Descriptor 1. Biodiversity is maintained; D2, Descriptor 2. Non-indigenous species do not adversely alter the ecosystem; D3, Descriptor 3. The population of commercial fish species is healthy; D4, Descriptor 4. Elements of food webs ensure long-term abundance and reproduction; D5, Descriptor 5. Eutrophication is minimized; D6, Descriptor 6. The sea floor integrity ensures functioning of the ecosystem; D7, Descriptor 7. Permanent alteration of hydrographical conditions does not adversely affect the ecosystem; D8, Descriptor 8. Concentrations of contaminants give no effects; D9, Descriptor 9. Contaminants in seafood are below safe levels; D10, Descriptor 10. Marine litter does not cause harm; D11, Descriptor 11. Introduction of energy (including underwater noise) does not adversely affect the ecosystem.*

of taxon relative abundance (Bik et al., 2013). The mitochondrial gene encoding for the cytochrome oxidase c subunit 1 (COI), is one of the preferred candidate loci for standard DNA barcoding projects (e.g., the International Barcode of Life, http://ibol.org). However, alternative genomic regions (e.g., nuclear 16S/18S rRNA genes, 12S mtDNA) characterized by more conserved priming sites have been identified as more appropriate for "metabarcoding" studies allowing to broader scale amplification of biodiversity across the eukaryotic taxa (Deagle et al., 2014). Nevertheless, for some taxa, these markers provide little resolving power at the species level. A possible alternative is represented by D2–D3 "diversity loop" region of 28S rRNA. A possible way forward to address this issue is represented by the multi-barcode approach (i.e., using a cocktail of gene markers for the same sample), which could help to improve taxonomic coverage and resolution.

Setting the best PCR conditions to recover the organisms present in an environmental sample is crucial for a successful application of metabarcoding to routine marine monitoring. A recent study demonstrated that different PCR conditions could affect the final taxonomic assignment in metabarcoding studies. A constant low annealing temperature (46 or 50◦C) provides more accurate taxonomic inferences compared to the touch down profile (Aylagas et al., 2016). Conversely, increasing the number of PCR cycles leads to the increase in the number of spurious sequences and chimeras formed (Haas et al., 2011). Chimeras can inflate the overall biodiversity estimates and be eliminated by comparing the length of matched bases from the top hit in a MEGABLAST search to the length of the query sequence. As long as the database sequence is longer than the query sequence and a portion of the 3′ end does not match, it is likely that the query is a recombinant. Chimeras can be removed also by using other algorithms, including Perseus (Quince et al., 2011), UCHIME (Edgar et al., 2011) and USEARCH (Edgar, 2010).

The choice of the sequencing platform is strictly linked to the aim of the research (Carugati et al., 2015). Recently Illumina platforms have become more appealing than the Roche 454 to assess metazoan biodiversity, because of their increasing read lengths, lower per base cost, production of tens to thousands times more sequences, and lower error rates (0.1% vs. 1%, Glenn, 2011).

Metabarcoding is not exempt from errors: i) during the processing of the samples (e.g., DNA amplification steps producing "chimeras," see above; Cline et al., 1996; Smyth et al., 2010), (ii) during sequencing (Glenn, 2011), and/or (iii) presence of multi-copy genes within a single species (e.g., Telford and Holland, 1997; Alverson and Kolnick, 2005; Bik et al., 2012). Metabarcoding based on PCR cannot yet provide reliable biodiversity indices since, especially for eukaryotes, it does not supply information on the abundance of every single species detected (Lindeque et al., 2013; Hirai et al., 2015). Most of the studies aimed at evaluating the relationships between species abundance and metabarcoding data obtained looser associations (Carew et al., 2013; Zhou et al., 2013; Hirai et al., 2015). Conversely, stronger relationships have been reported between biomass and read proportions (Elbrecht and Leese, 2015). Measure of relative abundance within metabarcoding samples need to be carefully considered. Nevertheless, in the absence of primer bias, a species characterized by larger biomass should be reflected by a greater proportion of sequence reads. Conversely, if the species is smaller or rarer, then fewer reads are likely to be obtained (Creer et al., 2016).

We are at the very beginning of applying this approach to analyse marine eukaryotic biodiversity. Further studies associated with the recent progress made in DNA sequencing technologies will allow elimination of DNA amplification steps and could open new perspectives to use metabarcoding in marine monitoring programmes. A recently developed approach, which could avoid PCR biases is based on the Illumina-sequencing of environmental metagenomes (mitags) (Logares et al., 2014b). We suggest that this method could represent, in the future, a powerful alternative to 18S rDNA amplicon sequencing and a useful tool to obtain simultaneously information on taxonomic and functional diversity.

An additional limitation of metabarcoding is that it does not differentiate between life stages, and thus juvenile stages and adults are pooled together. Further, species lists produced through metabarcoding currently are presence-absence based, and lack relative abundance data. Thus, traditional community analyses used for impact detection cannot be applied in the traditional manner, and instead the focus will be on overall species richness and presence of indicator species.

Another issue is represented by the still limited availability of sequences in public databases (Carugati et al., 2015). In some cases, operational taxonomic units (OTUs) can not be taxonomically assigned to a species, or even to a genus, due to the paucity of data in reference databases and the lack of taxonomic resolution at the species level of the marker gene (Dell'Anno


et al., 2015; Leray and Knowlton, 2016). Thus, exploiting the data will require the continued refinement of database resources and bioinformatic pipelines (Minster and Connolly, 2006; Hajibabaei et al., 2011; Bik et al., 2012; Radom et al., 2012).

Consequently, the collaboration between molecular ecologists and taxonomists is required for the accurate characterization of species and for the deposition of quality assured barcode sequences in public databases (Jenner, 2004). The improvement of reference databases and thus the ability to assign OTUs to known species will enable metabarcoding techniques to be more reliably used in monitoring surveys, with high potential for the detection of non-indigenous species. It is also important to underline that relating sequences to taxonomically described species is not a necessity for many applications since in monitoring the focus is in pattern changes, not on taxonomic composition per se. We suggest that, in order to apply metabarcoding for the purposes of the MSFD (e.g., Descriptor 1), an attempt could be made using the overall species richness. For instance, significant changes in the species richness of the community can be a useful warning indicator and assessing such changes does not require that each molecular OTU is assigned to a precise taxon. The Biodiversity Descriptor of the MSFD does not explicitly require that species are all taxonomically identified. Furthermore, molecular barcodes of a species, even when the species is not in the reference database, generally allow identification at the genus or family level if other species of the same genus or family are present in the reference database.

#### Case study 1. Microbes

HTS approaches have been recently applied to study the biodiversity of marine viruses (Tangherlini et al., 2012), bacterioplankton (Bacteria and Archaea) (e.g., Sogin et al., 2006; Gilbert et al., 2009; Brazelton et al., 2010), eukaryotic pico- (0.2–3 µm) (e.g., Shi et al., 2009; Massana et al., 2015), nano- (3–20 µm) (e.g., de Vargas et al., 2015; Massana et al., 2015), and microplankton (20–200 µm) (e.g., de Vargas et al., 2015). Data on their abundance and diversity may provide useful information on the impact of human pressures. Protists have been recurrently proposed as bioindicators (Payne, 2013). Nevertheless, the bacterioplankton component is still neglected by the MSFD (Caruso et al., 2015). The use of HTS allows the analysis of microbial biodiversity at an unprecedented scale, greatly expanding our knowledge on the microbiomes of marine ecosystems (Caporaso et al., 2011). These approaches provide relatively fast and cost efficient observations of the microbial component, and thus, may be suitable tools in biodiversity monitoring programs (Bourlat et al., 2013). Application of recently developed sequencing methodologies (e.g., Illumina technologies) to the analysis of the 16S rRNA gene for bacteria and of the 18S rRNA gene for eukaryotes in samples taken along the Barcelona coast (NW Mediterranean Sea) suggests that certain taxa (i.e., members of the Gammaproteobacteria) as well as the ratio between some phylogenetic groups may be good indicators of ecosystem health status. However, the robustness of these indicators needs to be explored by gathering data on plankton diversity in coastal areas subjected to different degrees of anthropogenic pressure over various temporal and spatial scales. Seasonality seems to play a major role in shaping bacterioplankton biodiversity and community structure (Gilbert et al., 2012; Cram et al., 2015) which could overwhelm the effects of human-induced pressures. Thus, despite being extremely promising, the suitability of incorporating prokaryotic/eukaryotic biodiversity into MSFD descriptors needs to be further explored in order to discriminate between changes resulting from human activities and the natural variability of the marine environment (Ferrera et al., 2016).

#### Case Study 2. Meiofauna

Small metazoans belonging to the meiofauna are sensitive to environmental changes and are increasingly used in monitoring studies for the assessment of environmental quality (Moreno et al., 2011; Pusceddu et al., 2011). However, meiofaunal diversity is so large that the analysis of a single phylum, such as Nematoda, requires huge investments of time from highly specialized taxonomists. Metabarcoding could facilitate the census of biodiversity, especially for meiofauna, for which morphological identification is difficult. The typical metabarcoding workflow used to study meiofaunal biodiversity in marine benthic ecosystems is reported in **Figure 1**. Recent investigations of shallow and deep-sea nematodes based on 454 sequencing and classical morphological identification revealed that, at the order-family level, metabarcoding assignments matched

FIGURE 1 | Standardized workflow to study meiofaunal biodiversity in marine benthic ecosystems using high-throughput sequencing. Sediment samples (from shallow to deep-sea environments) are collected and subsequently frozen (−20◦C or −80◦C). In the laboratory, meiofaunal organisms are recovered from the sediments and their DNA extracted and purified. Following the PCR amplification of marker genes (e.g., 18S rRNA), high-throughput sequencing can be conducted on Roche 454 or Illumina platforms. Raw reads are processed and then clustered into operational taxonomic units (OTUs) under a range of pairwise identity cutoffs. After the BLAST-match of the obtained OTUs against public nucleotide databases, analysis of α- and β-diversity and phylogeography are performed. Image of Illumina MiSeq platform: Source: Wikipedia, Author: Konrad Förstner (Carugati et al., 2015).

the results obtained by morphological techniques, but some OTU's remained unassigned (Dell'Anno et al., 2015). Although metabarcoding is a useful tool to explore the diversity of marine meiofaunal organisms, it still presents some gaps. Indeed, not all species in a sample are detected and a certain percentage remains unidentified due to the limited coverage of public sequence repositories for meiofaunal taxa (Carugati et al., 2015). This applies particularly to the deep sea, where most of the taxa are still unknown (Appeltans et al., 2012). Thus, we suggest to continue combining morphological identification performed though light microscopy with molecular analyses, in order to feed or even create local database, at least for marine protected area or high priority areas. To more accurately delineate species in metabarcoding datasets major efforts should be devoted to understanding the actual variability of the 18S rRNA gene amongst individuals of the same species and amongst different species taking into account the contribution of potential biases due to PCR and sequencing steps in such variability. There is also the urgent need to identify alternative single copy markers, nuclear or mitochondrial, less subjected to such intra-specific variability. Finally, alternative solutions can be the use of non-PCR-based metabarcoding approaches, using capture probes, which are much less sensitive to mismatches between probe/primer and target and may replace PCR-metabarcoding. Future investigations are needed to address these issues in order to facilitate the inclusion of meiofaunal diversity in marine monitoring programs.

#### Case Study 3. Macrofauna

Marine benthic macroinvertebrates are commonly used as indicators of ecosystem health; yet, calculation of biotic indices based on macro-invertebrate taxonomic composition (e.g., AMBI) requires each sample to be sorted and each specimen to be taxonomically identified by an expert taxonomist. This is a tedious, expensive and time-consuming process, which has limitations, particularly when cryptic species, damaged specimens or immature life stages are present (Ranasinghe et al., 2012). Metabarcoding is a promising alternative to overcome the limitations of traditional taxonomy and can help in ensuring the accomplishment of temporarily and spatially comprehensive monitoring. However, before routine implementation of this approach, the development of standardized practices at each step of the procedure (Aylagas and Rodríguez-Ezpeleta, 2016) and the increase of the reference libraries for taxonomic assignment are required (Aylagas et al., 2014). Additionally, in order to ensure accurate biotic indices derived from metabarcoding, the ability to detect the majority of organisms representing the full gradient of tolerance to pollution is necessary. With the aim of benchmarking metabarcoding against traditional taxonomy in the context of biotic index calculation, Aylagas et al. (2016) performed a thorough experiment comparing taxonomic inferences and biotic indices derived from samples of known species composition analyzed using alternative metabarcoding protocols. The work resulted in a series of guidelines for the application of metabarcoding for macroinvertebrate monitoring.

# The Application of Microarrays for the Detection of Harmful Algal Blooms

Microarrays have been applied for in situ detection of harmful algal bloom (HAB) species (Descriptors D1, D2, D5 in the MSFD; see **Table 1** for more details). This method is especially useful for the rapid identification of toxic algae (**Table 2**) that can have serious consequences on human health (Bricker et al., 2007). The European project MIDTAL (Microarrays for the detection of toxic algae) has developed a microarray to target major HAB species including toxic dinoflagellates, raphidophytes, prymnesiophytes, Dichtyocophyceae and the diatom Pseudonitzschia (Lewis et al., 2012). Microarrays are made of coated solid surfaces onto which a large number of selected DNA probes (specific for a taxon) can be spotted. Each probe is fluorescently labeled and when the probe hybridizes with a sample, the sample/probe complex fluoresces in UV light. An advantage of this approach is that no PCR step is required when total RNA is selected and this reduces the bias of any unknown inhibitors in the sample. Because microarrays rely on DNA probes for detection of HAB species, the potential for new indicators could be nearly unlimited. This chip has been tested on selected seawater samples previously morphologically identified. Microarrays have shown high sensitivity and several species not identified under light microscope have been recognized by the probes on board the microarray. Thus, microarray could be a potentially useful tool to provide quick evaluation on the presence of toxic algae. However, the use of microarray presents a series of limits. Some of the algal species morphologically identified in a sample could not be detected by the molecular probes. Moreover, the sensitivity of selected probes was confirmed at genus level, but at species level the results were less satisfactory. The costs of the MIDTAL microarray chip plus reagents and consumables is still high. Thus, further attempts are needed to make convenient and accurate the results provided by the use of the microarray approach and we recommend the use of the microarray in monitoring programs only if combined with microscopy analyses. The combined approach between current monitoring practices and microarrays could be applied in the MSFD (e.g., Descriptor 5) in order to provide quick and reliable information on the presence of algae potentially toxic for human health.

# Quantification of Pathogens by Means of Real Time Quantitative PCR (qPCR)

Real-time polymerase chain reaction (qPCR) consists of the amplification and quantification of a gene sequence specific to the organism(s) of interest. The correlation of the amount of DNA obtained with the number of individuals allows the quantification of the investigated organisms in a given sample. This procedure could be applied only to unicellular organisms that contain a known number of copies of the gene under study. Exponential amplification of the target sequence is followed in real-time by means of a fluorescent dye or a fluorescently labeled DNA probe. Quantification is performed by comparison to a standard curve, which is run concurrently with samples using reference material consisting of pre-enumerated cells or DNA. qPCR has been recently tested to evaluate the quality of the freshwater and marine environment (Descriptors D1, D2, D5 in the MSFD, **Table 1**; Newton et al., 2011; Harwood et al., 2014; Lu et al., 2015). Traditionally, the classical microbiological analyses include the investigation, by using cultivation techniques, of the abundance of fecal indicator bacteria such as Escherichia coli and Enterococci in water samples, and E. coli, Enterococci and Salmonella in sediment samples (**Table 2**). The determination of total prokaryotic abundances could be also performed through epifluorescence microscopy. Such a technique allows the determination of the whole quantitative relevance of marine microbes contrary to the cultural techniques, which can only detect less than 1% of the actual abundance of prokaryotes (Staley and Konopka, 1985). Epifluorescence microscopy could be utilized in combination with qPCR of the prokaryotic 16S rRNA genes. The combined use of qPCR and metabarcoding could open new perspectives to investigate the biodiversity of the microbial community in seawater and sediment samples and in particular the relevance of human pathogens, going beyond the limits of the traditional approaches.

## IN SITU INSTRUMENTS TO MONITOR MARINE ABIOTIC AND BIOTIC VARIABLES

Some of the best approaches to meet current demands in marine monitoring are represented by novel in situ technologies, which provide high-frequency (continuous or semi-continuous) observations. So far, most of in situ instruments have been developed to monitor marine hydrological and physico- chemical variables, whereas the monitoring of the biotic variables is still mostly dependent on non-remote or automatic devices. An example is the system of SmartBuoys, which house a range of instruments for measuring salinity, temperature, turbidity, chlorophyll fluorescence, oxygen saturation and nitrate concentration. Such instruments enable the creation of wide-scale international networks of environmental data acquisition and sharing, as implemented in the framework of the ongoing S&T Med European project (http://stmedproject. eu/). Nonetheless, technological limitations are at the base of the presently scarce modeling capacity regarding population/stock and biodiversity assessments as well as ecosystem functioning.

#### Chemical Sensors

There are few sensors currently in use for monitoring concentrations of heavy metals, organic pollutants and algal toxins. An in situ analyzer has been developed to measure nitrate plus nitrite and total sulfide in deep-sea areas close to hydrothermal vents (Le Bris et al., 2000). More recently, Vuillemin et al. (2009) developed an in situ analyzer (the CHEMINI system) which measures analytes at even greater depths. However, as for any instrument deployed at sea, especially in nutrient rich environments, it is subjected by a rapid biological colonization (biofouling), which can limit overall deployment times (Mills and Fones, 2012).

#### Seabed Observatories

Marine observatories allow the collection of long-term time series of environmental parameters, but have yet not been commonly used. It is widely recognized that underwater technology could open new and interesting opportunities to ensure continuous, long-term, execution of monitoring. In particular, during the last decades, underwater video technologies have gained considerable importance in all fields of marine science. They represent a powerful, non-destructive and useful tool to study the dynamics and the interactions between benthic organisms, especially on hard-bottom sediments where traditional grab methods are ineffective. The use of underwater visual surveillance is becoming increasingly accessible for monitoring activities since it is versatile, serving as an "underwater eye" for researchers. Video cameras can be mounted on various vehicles ranging from simple towed platforms, Remotely Operated Towed Vehicles (ROTVs), to more advanced systems such as Remotely Operated Vehicles (ROVs) or Autonomous Underwater Vehicles (AUVs). Stills photos can be acquired using drop cameras, mounted on ROVs or by diver at shallow depths, and long-term data series can be used to study the links between biodiversity and climatic variations, for example correlating changes in biodiversity related to the North Atlantic Oscillation (NAO) index (Beuchel et al., 2006). In coastal benthic and pelagic systems at shallow depth, SmartBuoys equipped with underwater cameras can enable such time-series studies, contextually monitoring multiple environmental parameters to complement visual information. In general, video surveys produce indicators of overall sediment conditions and frequency of occurrence of the most visible taxa. Indicators from stills images focus on small-scale observations and automated image recognition techniques can be employed to quantify both presence and abundance of organisms but also extent of coverage or various proxies for biomass (Beuchel et al., 2006).

The increasing use of ROVs, AUVs and non-permanent camera stations have provided new insights on the biodiversity and ecosystem functioning of continental margin and deep-sea ecosystems (Solan et al., 2003; Stoner et al., 2008). However, challenges emerge in that inherently qualitative information needs to be converted into quantitative data from which indicators can be developed. ROV technology is available at all offshore petroleum installations, and biological visual seabed surveys frequently are carried out in potentially sensitive habitats both before and after the drilling event. Using a set of customized visual indicators, the extent of seabed smothering can be quantified and appropriate mitigation measurements planned based on the information collected during these surveys. Autonomous and cabled observatories are receiving increasing attention in marine science and have been demonstrated as capable platforms for collecting data remotely, and increasing insight into the functioning of remote marine ecosystems (Taylor, 2009; Best et al., 2013). Such cabled systems are expected to become an important tool in marine monitoring and management (Aguzzi et al., 2012a).

A possible limit of the use of video-imaging systems is that the lights necessary to acquire the images may influence the behavior of the organisms being observed. Operational lifetimes of remotely deployed instruments are often limited by the available power supplies. Cabled observatories can provide the power to operate for long-term periods. However, the establishment of the infrastructure is still expensive and therefore limited in scope. Many in situ instruments still rely on commercially available batteries, which could limit they autonomy. Small wireless autonomous devices, such as remote marine sensors can be less energy consuming thus allowing longer deployments (Mills and Fones, 2012). Another challenge is represented by the large amount of data generated, which need to be stored and processed. Cabled multiparametric seafloor observatories are usually connected to the shore to transmit data in real-time. Data could be delivered via cable, automatically streamed to an internet socket, uploaded onto the website and automatically processed (Aguzzi et al., 2012b).

# Underwater Autonomous and Integrated Monitoring

An interesting, recently developed technology is the CLEAN SEA (Continuous Long-term Environmental and Asset iNtegrity monitoring at SEA; **Figure 2**), which uses a commercially available AUV, upgraded with technologies enabling off shore monitoring of seafloor integrity and pollution (**Table 1**). This vehicle is characterized by a set of sensors able to measure both physical and chemical parameters and carry out in situ analysis of trace pollutants (**Table 2**). The CLEAN SEA system can also collect discrete water samples in situ. It is developed to perform acoustic surveys of the seabed and pipelines/flowlines as well as to detect hydrocarbon leakage. The CLEAN SEA system can also perform benthic community survey with detailed photographic/video coverage of the investigated area in order to determine the abundance and biodiversity of benthic assemblages and their temporal variations (**Table 2**). CLEAN SEA is characterized by wireless underwater communication for mission data downloading and wireless power recharge for increased autonomy. This may enable a "permanent" operation subsea independently of support from surface. CLEAN SEA seems to be a powerful technology for future environmental

FIGURE 2 | The CLEAN SEA (Continuous Long-term Environmental and Asset iNtegrity monitoring at SEA). The Clean Sea system, launched by Eni E&P and its subsidiary Eni Norge, in cooperation with Tecnomare, is a commercially available AUV, properly upgraded with key enabling technologies, for the execution of environmental monitoring and asset integrity in offshore fields.

monitoring around oil and gas infrastructures and to gain longterm data on abiotic and biotic variables.

#### Biosensors

High frequency non-invasive (HFNI) valvometers have been utilized as a potential tool for long-term marine monitoring and assessments (Andrade et al., 2016). The principle of the method is based on the regular gaping behavior (closing and opening of the valves) of bivalve molluscs and the fact that physical or chemical stressors disrupt that gaping reference pattern. Bivalve gaping behavior is monitored in the natural environment, remotely, continuously over a long-time period (e.g., years), requirements that must be fulfilled if bivalve behavior is to be a useful biomonitoring tool. We here suggest the potential application of the HFNI valvometry as a biosensor to monitor and provide early-warning alerts of changes in water quality, such as temperature increase, releases of contaminants and toxic algal blooms. Finally, HFNI valvometry could be used in the MSFD for routine monitoring of areas impacted by anthropogenic activities such as bathing beaches and harbors, oil platforms and aquaculture installations.

# Acoustic Monitoring

An alternative method for studying marine organisms is a noninvasive acoustic approach. Active and passive hydroacoustics have explored a wide range of ecological subjects, such as pelagic communities, behavior, predator–prey interactions, and fish standing stock. The use of passive acoustic technologies (e.g., hydrophones) may solve problems of photic disturbance or limitation and provide useful results for the Descriptor 11 of the MSFD (**Table 1**). Most marine organisms produce sounds (marine mammals, fishes, invertebrates) to accomplish important ecological processes (e.g., communication, reproduction, foraging, predation, detection of predators and habitat selection; Van Opzeeland and Slabbekoorn, 2012). Understanding normal levels of variations in acoustic complexity is crucial for conservation efforts, enabling managers to decide whether changes in acoustic dynamics need further investigation. However, quantifying and characterizing the acoustic production of animals in marine soundscapes can sometimes be a challenging task to address. Active acoustic scattering techniques have potential to study the zooplankton and fish distributions, as they provide remote and non-intrusive samples at high resolution over large ranges (**Figure 3**), which is difficult to achieve using traditional net or other underwater systems alone. Multiple frequency scientific echosounders with split-beams and resulting echo-trace analysis (using frequency responses) can provide information on the sizes of animals, thus allowing some distinctions to be made. Despite the fact that the underwater acoustic instruments do not allow species classification (Knudsen and Larsson, 2009), they could be useful to gain information on pelagic and semi-demersal species as well as on zooplankton assemblages (Trenkel et al., 2011; **Table 2**). The Acoustic Complexity Index (ACI) (Pieretti et al., 2011) coupled with a software dedicated to soundscape analysis (Farina et al., 2011) can be used to elaborate collected acoustic files, in order to track the various biological signals, their daily and nightly dynamics

and distinguish them from noise pollution. Anthropogenic noise usually has specific frequency ranges (typically <1 kHz) which overlaps with the frequencies used by fishes for communication and other processes. We suggest that the ACI seems to be a promising tool to analyse marine soundscape filtering out noises and biological sounds.

# NEW METHODOLOGIES FOR MARINE MONITORING

# Comparison of Methods for Identifying Phytoplankton Diversity

Considering the objectives of the MSFD, it becomes important to evaluate emerging methods to enhance the efficacy and costefficiency of monitoring approaches, in particular non-intrusive, relatively low-cost methods based on optics. The optical metrics of phytoplankton include the size, shape, dimensions and complexity of the phytoplankton cell, as well as its light absorption, scattering and fluorescence characteristics, which are influenced by cell size, material and pigmentation. Each optical method shows some degree of selectivity or bias, either for a cell size range, pigment concentration range, or the ability to discern individual cell characteristics vs. a population of cells in a volume as a whole. Furthermore, it is recognized that the optical attributes of phytoplankton taxa are subject to natural variability regarding pigmentation, cell size, and colony formation within species.

Light microscopy is precise with regard to taxonomic determination, but less sensitive to rare species and practically limited to cells larger than 1–2 µm. Both fresh and stored samples can be analyzed, even if for some protists, fixatives deform the cells, making difficult their identification. The main limitation of this method is the time spent by an expert analysing a single sample, which is in the order of 1/day.

Flow cytometry analysis can be considered a combination of particle based and pigment analysis methods. The taxonomic distinction of each investigated particle is dependent on the number of lasers (usually 1 or 2 in benchtop instruments), detectors (4–8 in modern configurations) and is limited to those pigments that exhibit autofluorescence (chlorophylls and phycobilipigments). Besides fluorescence, flow-cytometers record forward- and side-scattering parameters, allowing basic size and shape characterization. Direct comparison of phytoplankton biodiversity obtained by using light microscopy, HPLC pigment and flow cytometry resulting from a multi-year sampling campaign in the productive season in the Baltic Sea revealed no meaningful correlation between the three methods (**Figure 4**). In this case, the lack of correspondence between the three methods can be explained by different lag times in the response of pigmentation, particle size distribution, or species composition to environmental changes. In other two studies a relatively good correspondence has been observed between the various methods (Casamayor et al., 2007; Christaki et al., 2011).

Pigment high-performance liquid chromatography (HPLC), has been for a long time a useful tool for obtaining information on taxonomic composition of phytoplankton, based on presence/absence of diagnostic pigments (Smith et al., 2010; Roy et al., 2011). Computational approaches, such as the statistical fitting tool CHEMTAX, have been used to determine phytoplankton biodiversity by estimating the relative contribution of different taxa to the total chlorophyll a (TChla) concentration in a sample (Mackey et al., 1996; Gibb et al., 2001; Goela et al., 2015). Although the software is fully developed, an a priori knowledge of the classes existent in the samples is required, as well as an appropriate choice of the ratios of pigment:Chla, considering the characteristics of the investigated geographical region (i.e., light availability; Higgins et al., 2011). As the inferences of this technique are based on the chemical composition of a sample and not on the direct observation of the phytoplankton cells, it has an improved capability to differentiate among organisms in smaller size classes, which in traditional methods such as microscopy fall into the category of unidentified flagellates (Goela et al., 2014). A recent application of this approach to oceanic regions, where populations of small organisms can be dominant, has proven to be particularly useful to distinguish the contribution of cryptophytes, prymnesiophytes, and prasinophytes to TChla concentration (Goela et al., 2014). Thus, the use of chemotaxonomic methods in combination with the classical methods (e.g., microscope enumeration, phytoplankton size-structure) would be useful to evaluate and characterize Descriptor 5 of the MSFD (Mangoni et al., 2013; Cristina et al., 2015; Goela et al., 2015; **Table 1**). Once the HPLC methodology is implemented and running, CHEMTAX offers a rapid and cost-effective way to assess the taxonomic composition of a sample, used as a first assessment of the phytoplankton assemblage. It might provide valuable insights on the potential presence of specific groups (e.g., harmful species), especially when there is previous knowledge of the classes that are likely to contain HAB species (Mangoni et al., 2011; Liu et al., 2014).

Marine Laboratory, Finnish Environment Institute).

The major caveats applied to the use of the method are often observed in phytoplankton classes which contains no diagnostic pigments or in which the diagnostic pigment is not present in all the species of the class. That is the case, for example, of dinoflagellates. Often, the marker pigment used in CHEMTAX for dinoflagellates class is peridinin, which is only present in some of the auto- or mixotrophic species of dinoflagellates (Throndsen, 1997). This might lead to the underestimation in areas where most of the dinoflagellates are heterotrophic (e.g., Goela et al., 2014). In those cases, a more reliable CHEMTAX analysis would involve a careful examination of the typical pigment profiling of the local dinoflagellates community, namely the combinations between different diagnostic pigments, or the search for species specific diagnostic pigments (e.g., Örnólfsdóttir et al., 2003; Smith et al., 2010; Roy et al., 2011). The versatility of the method, that is, the possibility to run the software with different combinations and values of pigment:Chla ratios is, in fact, one of the major advantages of the method, allowing easily to locally adapted pigment profile schemes. Recently, several studies have focused on the effective and successful use of CHEMTAX to detect HABs (e.g., Örnólfsdóttir et al., 2003), although pigment profiling studies, such as Liu et al. (2014), in other areas of the globe would be beneficial to the fulfillment of this objective.

# Analysis of Planktonic Microbial Diversity by Flow Cytometry

In plankton microbial flow cytometry, small sample volumes are circulated in front of a laser with a fluidics system that forces each cell to pass in front of the laser, which is typically blue, red or UV. The instruments can observe thousands of cells per second, so a few minutes of operation enables inspection of several hundred thousand cells. Both the cells and the abiotic particles disperse the laser light and generate fluorescence after the excitation. Since scattered light is proportional to cell size (and cell internal rugosity) and fluorescence is proportional to pigment content, it is possible to differentiate various groups of phototrophic oxic (Marie et al., 2005) and anoxic (Casamayor et al., 2007) microorganisms according to their average cell size, types of pigments and pigment ratios. In addition, it is possible to stain the nucleic acids of heterotrophic prokaryotes (Gasol and del Giorgio, 2000), heterotrophic eukaryotes (Christaki et al., 2011) and viruses (Brussaard et al., 2000) and simple activity probes can be used to obtain indication of the relative physiological state of prokaryotes and phytoplankton (del Giorgio and Gasol, 2008). This method allows easy fingerprinting of the microbial assemblages and a fast indication of how they respond to disturbances.

Besides the cost of instrumentation, which is progressively decreasing in recent years, the total cost is on the order of a few euros per analysis and can be done and processed in less than an hour. Moreover, sample collecting, processing, flow cytometry and data analysis can be automated (Besmer et al., 2014) and even commercial (Dubelaar et al., 1999) and non-commercial (Olson and Sosik, 2007; Swalwell et al., 2011) instruments can be submerged and send the data via cabling or radio. This allows their inclusion in environmental monitoring systems such as SmartBuoys, whose multiple sensors provide complementary information of the environmental settings in which cytometry data are acquired.

There are at least four different ways in which flow cytometric data can be used to infer ecosystem properties or environmental status (Gasol and Morán, 2015): (i) Presence/absence of specific microbial assemblages (e.g., presence of red-fluorescing cyanobacteria is generally associated with turbid low-light environments, whereas high abundances of Prochlorococcus or dominance of pico-eukaryotes with nutrient-rich environments; Stomp et al., 2007); (ii) Estimates of cytometric diversity (Li, 1997) of either pico-phytoplankton and heterotrophic prokaryotes; (iii) Population size and pigment content (e.g., temperatures lead to total phytoplankton and bacterioplankton decreases in cell size; Morán et al., 2010, 2015); and (iv) Ratios between populations abundance (e.g., the ratio between picocyanobacteria and eukaryotic picophytoplankters has been used to indicate nutrient levels as cyanobacteria are more likely to be abundant in low nutrient oligotrophic environments while eukaryotes tend to dominate in high nutrient conditions; Calvo-Díaz et al., 2008).

While the potential for these methods to work exists and a cost-savings potential is clearly demonstrated, additional testing is needed to determine how robust the methods are to detect physiological changes, such as those caused by nutrient and light availability. Sensitivity of these methods to cell physiological constrains may for example introduce undesirable seasonal or geographical bias which traditional (e.g., microscopy) methods would not show. Further studies are therefore needed to derive robust indicators of environmental status, preferably based on a multitude of complementary methods. Gathering data over various temporal and spatial scales in order to distinguish natural variability from that resulting from anthropogenic pressures will help validate these indicators, in order to subsequently develop highly automated tools for rapid assessment of marine environmental status.

# REMOTE SENSING

Remote sensing of optical, thermal and radar images from airborne and satellite sensors offers many new opportunities for the direct monitoring of biodiversity, for observing patterns in the land and sea which relate directly to biodiversity, or for the provision of environmental data layers which are needed in order to build predictive models of species and habitat distributions (Turner et al., 2003; Pettorelli et al., 2014). A new impetus has been given to the field of satellite remote sensing by the European Union's Copernicus programme in which the first of a series of Earth-observing sensors on the Sentinel satellites have been successfully launched. Sentinel 1 is a radar satellite with cloud-penetrating ability, in orbit since April 2014, and now delivering images that relate to marine and maritime needs, such as sea-ice extent, oil-spill monitoring and ship detection for maritime security. Radar images are very useful for determining the extent and composition of intertidal and salt-marsh habitats (Van Der Wal and Herman, 2007). Sentinel-2 for high resolution optical images of the coastal zone, as with Sentinel-1, will greatly enhance our ability to detect changes in intertidal and shallow subtidal habitats (Van der Wal et al., 2008). The final recent launch was that of Sentinel-3 for widefield ocean color viewing, altimetry and sea surface temperature on 16th February 2016. Sentinel-3 will continue the progress made by other ocean-viewing satellites such as SEAWIFS, MERIS and MODIS and ensure continuity of ocean color measurements (Le Traon et al., 2015). The use of remote sensing represents a cost-effective tool supplementing conventional in situ sampling. The in situ measurements are typically based on oceanographic cruises that provide discrete data sets with often spatial and temporal coverage, which could limit the analysis of the dynamics of the phytoplankton in relation to human activities (Rivas et al., 2006). Remote sensing can provide highly valuable data bridging the spatial and temporal gaps in observations complementing the in situ measurements. These are the major advantages of remote sensing as compared to in situ observation systems (Blondeau-Patissier et al., 2004). However, ocean color remote sensing also present some limitations as: (i) satellite-derived Chla concentrations estimates of phytoplankton biomass content are based on conversion factors (Rivas et al., 2006); (ii) information about the surface parameters can be obtained only during cloud free conditions, limiting spatial and temporal coverage, especially

in high latitudes and the tropics (Blondeau-Patissier et al., 2004; Peters et al., 2005); (iii) the confidence of the estimated values based on global algorithms has to be validated with in situ observations, which are essential to ensure the optimal quality of the data retrieved by satellite remote sensing, in particular in coastal and estuarine systems due to the optical complexity of such waters (Aurin and Dierssen, 2012).

Selected uses of satellite Earth observation in the field of marine biodiversity are presented in the sections below.

### Satellite Data for the Implementation of MSFD with Respect to Eutrophication (D5)

The use of remote sensing allows a cost-effective and synoptic monitoring of extensive oceanic and coastal areas (IOCCG, 2009). The products acquired by ocean color remote sensing can be quantified by bio-optical algorithms that retrieve the concentration of Chlorophyll a (Chla), Suspended Particulate Matter (SPM) and the absorption of the Colored Dissolved Organic Matter (CDOM). These indicators of the status of the marine ecosystems give information about the phytoplankton biomass (Chla), the water transparency or turbidity (SPM) and about the terrestrial inputs of freshwater (CDOM) (Vantrepotte and Mélin, 2010; **Table 2**).

Several studies have been carried out in European waters for the validation of remote sensing satellite products in a wide range of geographical areas (Sørensen et al., 2007; Antoine et al., 2008; Kratzer et al., 2008; Petersen et al., 2008; Cristina et al., 2009, 2014; Zibordi et al., 2013). These studies demonstrate the accuracy and the precision of the technique to provide good quality data and to identify what are the main sources that influence the complexity of these waters.

The advantages of this tool are evident for countries that have limited resources to monitor one of the largest marine zones of regional seas (Cristina et al., 2015). An ocean color remote sensing product (Chla) can be used to detect and track the development of algal blooms in coastal and marine waters. Thus, this tool can support the implementation of the MSFD with respect to Descriptor 5: eutrophication, as demonstrated in Sagres, southwest Iberia (Cristina et al., 2015, **Table 1**). Furthermore, it allows distinguishing whether the eutrophication is natural, driven by upwelling, or due to land-based inputs. The implementation of a regional algorithm increases the accuracy of the remote sensing data produced to retrieve the Chla, particularly during upwelling events when the highest concentrations of Chla occur (Cristina et al., 2016). This is supported by studies in the Baltic Sea (Harvey et al., 2015), also showing the advantages of using satellite remote sensing for monitoring and eutrophication assessment and for the status classifications of water basins. These studies show that this tool can be applied for both national, European and Regional Seas monitoring plans as well as the implementation of the MSFD and the Water Framework Directive (Gohin et al., 2008; Novoa et al., 2012). In summary, the use of remote sensing can be an efficient tool providing a synoptic view of the products (e.g., phytoplankton biomass), showing their distribution over an extended period, identifying seasonal patterns and showing the effect of changes in marine ecosystems promoted by human pressures and by environmental changes.

However, the eutrophication of the benthic compartment and its effects on the biota, which have been investigated repeatedly in the last decade (Danovaro et al., 2000, 2004; Danovaro and Gambi, 2002; Dell'Anno et al., 2002; Pusceddu et al., 2007, 2009) cannot be assessed through remote sensing.

# Satellite Imaging of Harmful Algal Blooms

Harmful algal blooms (HABs) adversely affect the marine environments by releasing toxins, decreasing food availability for higher trophic levels, and reducing oxygen levels in water, potentially causing mass mortality of marine organisms (Silke et al., 2005). HAB species may dominate the phytoplankton community, with very high chlorophyll concentration that can be detected from satellite sensors (Miller et al., 2006). Hence satellite monitoring of HABs is a novel method to detect undesirable (reduced biodiversity) water quality events, which may sometimes be related to eutrophication as described above. The remote sensing of chlorophyll concentration product has been successfully used to identify algal bloom events in the marine and coastal waters (Babin et al., 2008). However, the algal bloom of potentially harmful species could not be identified from analysis of chlorophyll concentration (Babin et al., 2008).

The method developed at Plymouth Marine Laboratory (PML), UK, uses measurements of water reflectance and inherent properties (IOPS) for automatic detection of HABs in satellite optical images (Kurekin et al., 2014). It is based on the relationships between water absorption properties and algal pigment composition, and between water backscatter and phytoplankton cell size, as features for HAB discrimination. The features were classified by Linear Discriminant Analysis (LDA) technique to produce HAB risk maps, as shown in **Figure 5**.

The method has been trained to discriminate Karenia mikimotoi and Pseudo-nitzschia sp. in the UK coastal waters, as well as Phaeocistis globosa algal blooms in the Southern North Sea. Measurements on board the RV Cefas Endeavor, provided by CEFAS, were integrated in the assessment of HAB risk. Joint analysis of satellite ocean color and Ferrybox data has been successfully applied for detection of a Karenia mikimotoi bloom off the North East of Scotland in August-September in 2013 and in 2014. The experiment has confirmed a strong correlation between satellite observations of HAB risk (Kurekin et al., 2014) with measurements of CTD profiles (including fluorescence and oxygen profiles) and in-situ samples (algal pigments, chlorophylla, cell count by microscopy and flow cytometry).

This method allows daily estimation of certain HABs over a wide area, depending on cloud cover. However, it is limited to phytoplankton species that produce high biomass blooms with a characteristic surface water coloring, whereas many toxin-producing algae are harmful in low concentrations. HAB risk maps are already operational for early warning of blooms affecting Scottish salmon farms, so it would be practical to extend the method toward further monitoring programs. The method is dependent upon the quality training data available for each HAB type, and so this aspect requires ongoing development.

# Remote Sensing of Shelf-Sea Fronts for Estimating Pelagic Biodiversity

A novel approach to the mapping of pelagic diversity has been implemented for the UK continental shelf, using a long time-series of remotely-sensed SST data to automatically detect thermal ocean fronts and then aggregating observations into climatological seasonal metrics (Miller and Christodoulou, 2014). These metrics have characterized the spatial, seasonal and interannual variability of fronts observed in 30,000 satellite passes over a 10-year period. Many researchers have determined that fronts are related to the abundance and diversity of pelagic vertebrates such as seabirds and cetaceans (reviewed by Scales et al., 2014). The resulting front maps were successfully applied as a proxy of pelagic diversity to the UK Marine Conservation Zone (MCZ) project—a key element of efforts to improve environmental status of European seas, and this influenced the designation of 11 of the recommended MCZs (Miller and Christodoulou, 2014) (**Figure 6**).

Although seasonal locations of frequent fronts were found to be fairly consistent, there are considerable interannual and week-to-week variations in the location and frequency of fronts, with consequential changes in the water column likely to affect species distributions. Hence satellite monitoring of shelf-sea fronts can serve as a proxy for certain mobile pelagic animals and as a physical boundary that structures other components such as zooplankton. Real-time front maps can be compared and integrated with other tools such as Ferrybox to assess aspects of the ecosystem and its biodiversity. Real-time satellite front maps have been applied to a UK project to optimize the MCZ/MSFD monitoring strategy using sea gliders and autonomous underwater vehicles across frontal biodiversity gradients (Suberg et al., 2014).

Hence the key benefits of this technique for marine monitoring are to assist the optimization of sampling strategies and to inform predictions of the abundance of fish and other pelagic animals that are difficult to measure directly.

# Broadscale Seabed Mapping Using Opportunistic, High-Resolution Seafloor Acoustic Data

One of the core requirements of the MSFD is the use of habitat maps at the regional or sub-regional scale (Annex III, **Table 1**). In addition, there is an expectation that the assessment takes account of environmental conditions when deciding assessment boundaries [Article 3(2)] and this involves an understanding of predominant habitat types, including the structure and substrata composition of the seabed. The importance of knowing the changes in seabed conditions in detail are particularly relevant for the directives Habitats (D1), Seabed Integrity (D6), and changes to Hydrographical Conditions (D7) (**Tables 1, 2**). So whilst assessments must be reported on at the regional level the actual scale of assessment is on subdivisions of the subregions (European Commission, 2014). Determining the relevant scale for assessment is especially important when we consider that these must be aggregated and reported at a higher level, so that errors and uncertainties will propagate up from the minimum assessment areas (Dong et al., 2015). So whilst identifying the most appropriate assessment method for indicators is a challenge in itself (Berg et al., 2015), the spatial component fundamentally affects our ability to accurately assess ecosystem components.

For the benthic environment we are severely restricted as to the amount of existing data we have to define ecologically relevant areas. The failure of market-value to adequately represent the societal importance of the marine environment has been widely recognized (Brouwer et al., 2016) and the practical reality is that there is less short-term economic incentive to collect seabed information (compared to terrestrial remote sensing), as a result little of the European seabed has been

of satellite SST data, compared with fronts predicted by a numerical model based on tidal currents and bathymetry (dashed lines where Simpson-Hunter stratification parameter S = 1.5). FF, Flamborough front; UF, Ushant front; CF, Celtic Sea front; DB, Dogger Bank; W, Wash; TE, Thames Estuary. (From Miller and Christodoulou, 2014, UKCS region, 1.2 km resolution, 1999–2008 data).

mapped using modern methods. A direct consequence of such data deficiency is that 76% of seabed habitats are in unknown status (EEA, 2015) and there are no systematic habitat mapping programmes in place at national or pan-European scales.

In the absence of adequate seabed data, the urgent need to define seabed habitats for management has resulted in the construction of modeled seabed data such as UKSeaMap (Connor et al., 2006). These existing broadscale maps will inevitably contain errors due to data deficiencies and generalizations. However, the alternative of using the scattering of existing high-resolution maps, does not address our needs to define biogeographical limits of species or overall habitat distribution at a regional scale. To overcome this difficulty (of high resolution data only existing as a localized patchwork) and make best use of existing resources, the novel strategy of continuously logging high-resolution multibeam data during existing monitoring cruises has been adopted on the RV Cefas Endeavor using the Olex software programme. This allows nonhydrographers to automatically mosaic and navigate around the seafloor data in real time through a simple graphical interface. It is then possible to use the data operationally rather than waiting for it to be processed and made available in an accessible format. As there are no dedicated personnel required and the system has no adverse effect on existing operations, large amounts of high-resolution data are collected with negligible additional cost (continuous operation is not expected to reduce its serviceable life expectancy of sonar systems).

Integrating the high resolution bathymetry and backscatter data with existing broadscale environmental data (such as modeled currents and seabed morphology) using randomforest models (e.g., Hengl et al., 2015), it is then possible to create a complete coverage map of the seabed conditions (**Figure 7**). By using only acoustic data in our study the modeled variables produced (whilst not ground-truthed) are repeatable, provide outputs at a uniform resolution, and allow a consistent assessment of uncertainty to be made across the area (Mascaro et al., 2014). These properties are valuable when addressing questions of map interpretation (Steiniger and Weibel, 2005) and ecosystem status at regional scale (Walz and Syrbe, 2013; Galparsoro et al., 2015a). It is possible to use these data to produce categorical maps. However, there are concerns as to the validity of categorizing continuous environmental variables for habitat delimitation (Wilson et al., 1999; Orpin and Kostylev, 2006; Galparsoro et al., 2015b). Defining a fixed set of conditions which delimit the extent of a single species is conceptually problematic (Randin et al., 2006; Heads, 2015), and, as habitats are taxon and scale-specific (Mairota et al., 2015; Mathewson and Morrison, 2015), the use of existing, readily available, categorical GIS habitat maps for biotope assessments should not be considered as scientifically defensible.

Using the method outlined above to collect large quantities of high-resolution data over a broad extent, we can also directly map highly localized features and impacts, such as the direct mapping of species distribution and condition of biogenic reefs. In this way we have a direct relationship between sonar image and species distribution without the need to go through the process of inferring their distribution from correlations. Models can be used to identify areas where the feature is likely to be present and additional monitoring effort can be deployed as necessary, both to monitor condition, as well as to better define their extent (as required by the relevant indicators).

There is no practical hindrance to the collection of spatiallyextensive, high-resolution data from a wide range of platforms already conducting regular monitoring activities. The challenge is in recognizing the benefits of such data in supporting the spatial assessment of multiple indicators, implementing the necessary routines and then incorporating the outputs into monitoring, assessment, and management strategies.

#### INNOVATIVE SAMPLING METHODS

Here we summarized the experience made on innovative sampling methods, some of which have been applied for the

FIGURE 7 | Random forest model of seabed acoustic intensity, extrapolated from high-resolution multibeam data collected opportunistically during fisheries research cruises (ships tracks as red lines; Source: OceanDTM).

first time in European seas. These include methods to monitor hard-bottom substrata, but also the use of citizen science to obtain massive information.

# Artificial Structures to Monitor Hard-Bottom Benthic Biodiversity ARMS

Small invertebrates, including sessile and encrusting organisms as well as mobile specimens inhabiting ecological niches in hidden spaces, represent most of the benthic biodiversity in rocky areas. Despite its importance for ecosystem functioning, a considerable percentage of benthic biodiversity is untargeted during traditional surveys and thus likely to be unreported (Pearman et al., 2016a). In the current scenario of global change, caused by natural and anthropogenic pressures, species may be pushed to extinction even before their identities and roles in ecosystem functioning can be understood (Costello and Wilson, 2011).

To overcome the difficulty in obtaining standardized and comparable information on benthic biodiversity from different habitats and regions, the Coral Reef Ecosystem Division (CRED) of the United States National Oceanic and Atmospheric Administration (NOAA) developed a standardized biodiversity assessment tool called an "Autonomous Reef Monitoring Structure" (ARMS; **Figure 8A**). This device consists of nine 23 × 23 cm gray, Type I PVC plates stacked in an alternating series of layers that are either open to the current or obstructed, which are intended to mimic the three-dimensional structure of the reef environment. They should be deployed for 1–3 years and colonized by bacteria, algae and sessile and mobile fauna, including cryptic species, of different size ranges (meiofauna, 20–500 µm; macrofauna, >500 µm; large macrofauna, >2000 µm). After recovery, both sides of each plate are photographed, and then surfaces are scraped, homogenized and analyzed using barcoding and metabarcoding techniques. The ARMS processing protocol applies a combination of morphology (for organisms >2000 µm) and molecular-based (all components) identification approaches to assess species richness (Leray and Knowlton, 2015).

The use of a standard sampling unit and the application of homogeneous protocols for morphological and molecular identifications can produce comparable datasets over different geographical areas. Despite some limitations of the metabarcoding technique (Carugati et al., 2015; see metabarcoding section), such as the incompleteness of reference databases, the sequence inventory obtained is already valuable for biodiversity assessment that be further improved in the future without additional laboratory work by rerunning the bioinformatics analyses on updated reference databases. Over a deployment of 1–3 years, colonization and succession patterns could be affected by changes in environmental conditions, making ARMS proper tools for marine monitoring of coastal areas. ARMS can be also re-deployed in the same locations and used to assess biodiversity changes over time. The characterization of the surrounding environment where ARMS units are deployed should be carried out for a comparison with natural assemblages. Temporal variability in key environmental variables, such as temperature, nutrients and chlorophyll a, should be investigated during the deployment period. Combining the use of ARMS with standard surveys, generally targeting fish and conspicuous invertebrates (**Table 2**), it is possible to obtain a comprehensive picture of the biodiversity

the 3D structure of a natural reef environment. (B) Artificial Substrate Unit (ASU) developed to mimic the filamentous algae or kelp holdfasts.

and more accurate information on the health status of the system.

The use of ARMS for routine marine monitoring presents some problems that need to be addressed. Although the costs of sequencing are dropping, and even if the ARMS-based approach is more cost effective than morphological-based one (Hayes et al., 2005), overall costs may still be high. Moreover, protocols for the assessment of biodiversity associated to ARMS rely upon the use of molecular approaches and thus the use of such devices present the same problems described above for metabarcoding. The ARMS protocol of Leray and Knowlton (2015) proposed the use of the mt COI gene. However, the database for this gene is highly biased toward metazoans and may thus be limited in the detection of other groups (such as algae and unicellular eukaryotes). Other genes have been targeted for ecological studies (e.g., 18S rDNA, Logares et al., 2014a, 28S rDNA, Hirai et al., 2015, and the ITS region Tonge et al., 2014) and a combination of these genes and COI may give a more comprehensive assessment of diversity. In the future, molecular studies using ARMS may also investigate the functional ability of the assemblage using shotgun metagenomic techniques.

#### ASUs

Another example of standardized sampling devices for marine biodiversity assessment is represented by Artificial Substrate Units (ASUs; **Figure 8B**). ASUs are nylon pot scrubbers, which have been used to study recruitment and taxonomic composition for over 20 years (Menge et al., 1994, 2002, 2009; Gobin and Warwick, 2006; Underwood and Chapman, 2006; Hale et al., 2011). They are particularly used to mimic filamentous algae or kelp holdfasts (Menge et al., 1994), a preferred habitat for recruits of many species (e.g., mussels, Paine, 1974).

After their recovery, ASUs are traditionally processed to identify species by using their morphological characters (Menge et al., 2002; Underwood and Chapman, 2006; Hale et al., 2011). With the advent of metabarcoding, the diversity associated with ASUs has been assessed by combining morphological and molecular methods.

The advantages and disadvantages of ASUs are similar to those of the ARMS, which are detailed above. Comparing the two structures, ASUs are easier to deploy than ARMS and the materials needed to construct an ASU are less expensive than those used to build ARMS. Moreover, the processing of an ASU takes fewer person-hours per unit (18 person-hours per ARMS vs. 6 per ASU). This makes ASUs more amenable to fine-scale sampling, for instance to measure temporal changes in biodiversity. They would be a valuable contribution to current monitoring programs, which require intensive samplings. The use of ASUs in monitoring programs can be relatively simple (e.g., Hale et al., 2011). Another consequence of simpler processing is that there are fewer risks of deviation from standardized procedures for ASUs than for ARMS during the processing of samples. However, they do not sample the same ecosystem component as the ARMS, since the two devices mimic different habitats. The small size of the ASUs relative to the ARMS imposes a selection for smaller organisms and species, such that large-bodied organisms cannot be collected by using the ASUs.

# High Resolution Sampling

Recent advances in robotic technologies provide new opportunities to conduct high-resolution sampling of patchily distributed organisms (such as zooplankton), by using AUV, carrying bottles for collecting discrete seawater samples and a sensor for gathering contextual environmental data. Environmental Sample Processors have been developed as stationary (moored) devices able to conduct in situ molecular assays (sandwich hybridization assay) by using 18S ribosomal RNA oligonucleotide probes, in order to detect actual plankton diversity (from calanoid and podoplean copepods, to larvae of barnacles, mussels, polychaete worms, brachyuran crabs, and invasive green crabs; Carcinus maenas; Harvey et al., 2012).

The Continuous Automated Litter and Plankton Sampler (CALPS) is a custom-made semiautomatic sampler which collects water using a pump system at a single depth along a predetermined transect as the ship sails. The system consists of a pump system and additional elements fitted onto the research

vessel. The additional elements include a water inlet of 20 cm diameter, a flowmeter, 6 cylinder traps and associated valves and level detectors to prevent overflowing and the system is controlled by computer (**Figure 9**). When activated, the system pumps sea water from a depth of 4 m at rates of between 35 and 45 L per minute, and distributes the water into one or more of the 6 possible traps. Each trap consists of a PVC cylinder (height: 73.3 cm, diameter: 28.0 cm) containing a plankton net (length 66.0 cm and diameter 26.5 cm) of chosen mesh-size. The volume of water filtered is measured with an electronic flowmeter. The performance of the CALPS against traditional vertical net sampling was evaluated in a study by Pitois et al. (2016). The authors concluded that the CALPS is suited to describe broad geographic patterns in zooplankton biodiversity and taxonomic composition; its particular advantage over more traditional vertical sampling methods is that it can be integrated within existing multidisciplinary surveys at little extra cost and without requiring additional survey time. These features make the CALPS a particularly useful tool as part of integrated monitoring of environmental status to underpin policy areas such as the MSFD.

# Ocean Sampling Day

The Ocean Sampling Day (OSD) is a simultaneous sampling campaign of the world's coastal oceans which took place for the first time on the summer solstice (June 21st) in the year 2014 and was repeated in 2015 and 2016 (Kopf et al., 2015). In this way, the collected samples related in time, space and environmental parameters, will provide new insights regarding microbial diversity and function and contribute to the blue economy through the identification of novel, ocean-derived biotechnologies. Micro B3's OSD project aims to generate, in a single day and in a cost-effective way, the largest standardized marine microbial data set, complementary to what obtained by other large-scale sequencing projects. The standardized procedure including a centralized hub for laboratory work and data processing via the Micro B3 Information System, ensures the collection and the processing of sea water samples with a high level of interoperability and consistency between data points worldwide. All OSD data (i.e., sequences and contextual data) are archived and immediately made openly accessible without an embargo period (Ten Hoopen and Cochrane, 2014). OSD sampling sites are typically located in coastal regions within exclusive economic zones (EEZ) and thus the OSD data set provides a unique opportunity to test anthropogenic influences on microbial assemblages. The final aim is to create an OSD time-series indicators to assess environmental vulnerability and resilience of ecosystems and climatic impacts. In the long term such indicators may be incorporated into the Ocean Health Index (OHI) (Halpern et al., 2012), which currently does not include microorganisms due to the lack of reliable data. OSD has the potential to close that gap expanding oceanic monitoring toward

microbes. This could lead to a global system of harmonized observations to inform scientists and policy-makers, but also to raise public awareness for the major, unseen component of world's oceans.

# CONCLUSIONS

There is an urgent need to improve our knowledge of the spatio-temporal variations of marine biodiversity and of the consequences of global changes on marine ecosystems. This should be done quickly, in real time, using harmonized, standardized and low-cost tools (Borja and Elliott, 2013), and extending our ability to monitor the deep-sea ecosystems (Danovaro et al., 2014; Corinaldesi, 2015). Recently developed technologies and instruments should help to determine not only the biodiversity but also the functioning of ecosystems, feeding the needs of the recently enacted Marine Strategy Framework Directive (Cardoso et al., 2010).

Some of the innovative methodologies and technologies described here (e.g., AUVs, high-resolution sampling instruments) are tested and validated in different geographical areas and they can help to achieve in real time information on different ecosystem components (from microbes to megafauna), rapidly and in a rigorous way, at a lower cost than traditional ones. Other tools, especially molecular ones, e.g., metabarcoding, need further evaluation (Bourlat et al., 2013).

In this context, such innovative approaches for marine monitoring need to be further implemented through: (i) defining standardized manuals and protocols for sampling and sample processing; (ii) developing new indicator metrics and indices fitting the new approaches and also useful for policy and decision-making; (iii) integrating, in monitoring surveys, information on biodiversity with other data sources (CTD, remote sensing, multibeam, taxonomy databases) for an holistic marine ecosystem assessment.

# REFERENCES


Innovative methods can improve monitoring and contribute to the definition of criteria for better conservation of marine biodiversity. While the potential of these approaches to work exists, further studies are needed before their complete implementation application in routine marine monitoring programmes.

## AUTHOR CONTRIBUTIONS

RD and LC conceived the paper. All authors have contributed equally to the Introduction. RDan, LC, MB, SCar, AC, CC, AD, EG, JG, JF, IF, JP, AR, NR, and AB contributed to the section of molecular approaches. RDan, LC, ND, VM, SM, KS, ER, SCon, SG, SS contributed to the section of in situ instruments. RDan, LC, SCri, RDav, PG, RF, AK, PM, AN, EG, JG, IF, AR, CW, VS, OM contributed to the section of remote sensing. RDan, LC, SCar, JP, ML, AEC, SP, SG, SC, and AB contributed to the section of innovative sampling methods. All authors have contributed equally to the discussion and conclusions. All authors reviewed the manuscript.

# ACKNOWLEDGMENTS

This manuscript is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392) (http://www.devotes-project.eu). Further financial assistance was provided to VS and ER by the European Union under the ENPI CBC Mediterranean Sea Basin Programme (Sustainability and Tourism in the Mediterranean—S&T Med Strategic Project). The contents of this article are the sole responsibility of the authors and can under no circumstances be regarded as reflecting the position of the European Union or of the Programme's management structures.

marine monitoring and assessments. Front. Mar. Sci. 3:187. doi: 10.3389/fmars. 2016.00187


S. Bourlat (New York, NY: Springer), 237–249. doi: 10.1007/978-1-4939-3774- 5\_16


bacterioplankton community throughout the water column over ten years. ISME J. 9, 563–580. doi: 10.1038/ismej.2014.153


J. Exp. Mar. Biol. Ecol. 330, 234–244. doi: 10.1016/j.jembe.2005. 12.030


system for metagenomic marine science. Ecol. Inform. 12, 10–15. doi: 10.1016/ j.ecoinf.2012.07.003


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declared a collaboration with all the authors and states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Danovaro, Carugati, Berzano, Cahill, Carvalho, Chenuil, Corinaldesi, Cristina, David, Dell'Anno, Dzhembekova, Garcés, Gasol, Goela, Féral, Ferrera, Forster, Kurekin, Rastelli, Marinova, Miller, Moncheva, Newton, Pearman, Pitois, Reñé, Rodríguez-Ezpeleta, Saggiomo, Simis, Stefanova, Wilson, Lo Martire, Greco, Cochrane, Mangoni and Borja. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Evaluation of Alternative High-Throughput Sequencing Methodologies for the Monitoring of Marine Picoplanktonic Biodiversity Based on rRNA Gene Amplicons

Isabel Ferrera\*, Caterina R. Giner, Albert Reñé, Jordi Camp, Ramon Massana, Josep M. Gasol and Esther Garcés

*Biologia Marina i Oceanografia, Institut de Ciències del Mar, CSIC, Barcelona, Spain*

Edited by: *Marianna Mea, Jacobs University of Bremen, Austria*

#### Reviewed by:

*Naiara Rodriguez-Ezpeleta, AZTI, Spain Eugenio Rastelli, Polytechnic University of Marche and Stazione Zoologica Anton Dohrn, Italy*

> \*Correspondence: *Isabel Ferrera iferrera@icm.csic.es*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *14 June 2016* Accepted: *02 August 2016* Published: *22 August 2016*

#### Citation:

*Ferrera I, Giner CR, Reñé A, Camp J, Massana R, Gasol JM and Garcés E (2016) Evaluation of Alternative High-Throughput Sequencing Methodologies for the Monitoring of Marine Picoplanktonic Biodiversity Based on rRNA Gene Amplicons. Front. Mar. Sci. 3:147. doi: 10.3389/fmars.2016.00147* Sequencing of rRNA gene polymerase chain reaction amplicons (rRNA tags) is the most common approach for investigating microbial diversity. The recent development of high-throughput sequencing (HTS) technologies has enabled the exploration of microbial biodiversity at an unprecedented scale, greatly expanding our knowledge on the microbiomes of marine ecosystems. These approaches provide accurate, fast, and cost efficient observations of the marine communities, and thus, may be suitable tools in biodiversity monitoring programs. To reach this goal, consistent and comparable methodologies must be used over time and within sites. Here, we have performed a cross-platform study of the two most common HTS methodologies, i.e., 454-pyrosequencing and Illumina tags to evaluate their usefulness in biodiversity monitoring and assessment of environmental status. Picoplankton biodiversity has been compared through both methodologies by sequencing the 16 and 18 S rRNA genes of a set of samples collected in the coast of Barcelona (NW Mediterranean). The results show that, despite differences observed in the rare OTUs retrieved, both platforms provide a comparable view of the marine picoplankton communities. On a taxonomic level, there was an accurate overlap in the detected phyla between the two methods and the overall estimates of alpha- and betadiversity were comparable. In addition, we explored the concept of "indicator species" and found that certain taxa (i.e., members of the Gammaproteobacteria among others) as well as the ratio between some phylogenetic groups (i.e., the ratio of Alphaproteobacteria/Gammaproteobacteria, *Alteromonas*/SAR11, and *Alteromonas* + Oceanospirillales/SAR11) have potential for being useful indicators of environmental status. The data show that implementing new protocols and identifying indicators of environmental status based on rRNA amplicon sequencing is feasible, and that is worth exploring whether the identified indices are universally applicable.

Keywords: plankton diversity, high-throughput sequencing, marine ecosystems, prokaryotes, picoplankton, monitoring programs, indicators, environmental status

# INTRODUCTION

The oceans are the largest ecosystem on Earth and provide countless ecosystem services to society (Liquete et al., 2013). Oceans regulate our planet's climate and represent one of the largest carbon reservoirs in the globe. Over a third of the world's population live in coastal areas, but virtually all humans depend to some extent on the ocean. Marine ecosystems provide resources for human survival and well-being, from fishing to natural products used in medicine or biotechnology. However, human-impacted marine ecosystems (i.e., coastal areas) are being increasingly threatened by pressures exerted due to changes in land use, overfishing, climate change, pollution, aquaculture, invasive species and other impacts of a rapidly growing human population (Halpern et al., 2007, 2008). Therefore, there is a need to report on the condition of the marine ecosystem in response to these human pressures, which may have an effect on all the components of the marine food web, from microorganisms to top animal predators (Brown et al., 2010; Claudet and Fraschetti, 2010; Hoegh-Guldberg and Bruno, 2010) and in the ocean services.

Legislation regarding the management of human impacts on the marine environment has been implemented worldwide to protect and conserve marine ecosystems. Several international (United Nations Convention on the Law of the Sea), regional [i.e., Marine Strategy Framework Directive (MSFD) in Europe, Oceans Act in the USA (Birk et al., 2012) among others] and local initiatives to protect the oceans exist. These initiatives include a number of criteria and methodological standards for assessing the environmental status of marine waters. The effect of anthropogenic impacts on the marine ecosystem is currently assessed through a variety of approaches (Birk et al., 2012). In any marine environmental assessment carried out for legislative or non-legislative reasons, there is a need to develop and test indicators at the species, habitat and ecosystem level. There is also a need for the cost-effective implementation of these indicators by defining monitoring and assessment strategies that are as simple, fast, and cheap as possible.

Among the different biological components of the marine ecosystem potentially used as indicators, the least known are the microbial communities, which are the major contributors to global marine diversity, and are a dominant component of the whole aquatic biota in terms of biomass and activity. Furthermore, they play a crucial role in its contribution to primary production and processing of organic matter (Kirchman, 2008). Microorganisms are the smallest biotic components and their intrinsic growth rates are the fastest among all biological components of natural aquatic systems. Microbial communities increase cell numbers as response to nutrients inputs, and as a consequence decrease their diversity, which also occurs in response to events of acute contamination (see review by Nogales et al., 2011). Since microorganisms are the fastest biotic responders to environmental changes, their abundances, community composition (i.e., the taxa present and their relative abundances) and relative indications of their activity have the potential for becoming useful indicators of ecosystem condition. Indeed, microbial indicators have been proposed in several legislative directives, such as the MSFD descriptors of biodiversity, food webs, eutrophication, and seafood contaminants. Including microbes in future monitoring programs has already been suggested (Caruso et al., 2015), and an intense research on this direction is being carried out particularly since the introduction of genetic methodologies.

Genetic technologies have the potential to provide accurate, rapid, and cost efficient observations of the marine environment. Molecular methods also represent a reliable taxonomic identification tool especially for organisms lacking conspicuous morphological traits such as microorganisms. Several molecular methods have been proposed for integration into existing monitoring programs (e.g., qPCR, SNP based methods, DNA barcoding, microarrays, metagenomics, metatranscriptomics; see review by Bourlat et al., 2013). Among those, DNA tagging (i.e., DNA barcoding or assigning taxonomy to a specimen/sample by sequencing a short DNA fragment) has a high potential for marine monitoring and assessment because of its relatively low cost and easy standardization once a reference database has been built.

The recent development of high-throughput sequencing (HTS) technologies has enabled the exploration of microbial biodiversity at an unprecedented scale, greatly expanding our knowledge on the microbiomes of different ecosystems (Cho and Blaser, 2012; Gilbert et al., 2014) including the oceans (Ferrera et al., 2015; Moran, 2015). Sequencing of rRNA gene polymerase chain reaction amplicons (rRNA tags) is currently the most common approach for investigating microbial biodiversity. Because this approach provides accurate, fast and cost efficient observations of the marine environment, it may be a suitable tool in biodiversity monitoring programs. While the potential for this method exists, testing and pilot studies are needed to answer relevant questions, for example, their benefits as compared to more traditional methods, and to test their general applicability (Bourlat et al., 2013).

In this study we evaluated two of the most commonly used HTS methodologies, i.e., 454-pyrosequencing (from now on 454) and Illumina, to study marine picoplanktonic biodiversity and explored their use in the assessment of ecosystem health status. The 454 method has been the most popular methodology since the development of HTS as it was the first to become commercially available and offers relatively long read length. The International Census of Marine Microbes program (Huse et al., 2008) used this approach. In contrast, Illumina provides shorter reads but offers significantly greater throughput than 454 at lower cost (Glenn, 2011) and is becoming the most popular deep sequencing platform for diversity applications, including the Earth Microbiome Project (Gilbert et al., 2014). Currently, only a few cross-platform studies are available; these two methodologies have been compared in metagenomic studies (Luo et al., 2012), and other applications such identifying single nucleotide substitutions in whole genome sequences (Ratan et al., 2013). Regarding tag sequencing, comparisons have been performed in lake, soil or human samples (Claesson et al., 2010; Sinclair et al., 2015). The initial results showed that the taxonomic classification of reads from the first Illumina sequencers was worse than 454 due to their shorter length and higher error rates (Claesson et al., 2010). Nonetheless, the improvement in quality and length reads of later Illumina sequencers has shown promising results; Illumina performed in a similar manner than 454 with regards to estimates of alpha- and betadiversity except when estimating evenness in soil and lake samples (Sinclair et al., 2015). Here, a careful comparison of the performance of sequencing 16 and 18 S (for marine planktonic prokaryotes and small Picoeukaryotes, respectively) rRNA gene tags by using 454 and Illumina (pair-ended 2 × 250 bp) has been performed to determine and quantify marine picoplankton biodiversity, and the robustness of the results has been tested. The results show minor differences in the performance of both sequencing methodologies for rare taxa, but overall both methodologies provide a comparable view of marine planktonic biodiverstity. Moreover, we also show that certain taxa as well as the ratio between some phylogenetic groups may be good indicators of ecosystem health status. HTS may thus provide valuable information for the assessment of the environmental status in marine waters.

# MATERIALS AND METHODS

#### Sample Collection and Basic Data

Surface waters were collected on 8th Aug 2013 in a 6 km inshore to offshore transect off the coast of Barcelona, NW Mediterranean. Five stations were sampled along the transect (**Figure 1**). Samples were sieved through a 200-µm mesh and transported to the laboratory within 2 h. Basic physical data was measured in situ with a conductivity, temperature and depth probe and surface salinity was analyzed with an AUTOSAL salinometer. The concentration of inorganic nutrients was determined spectrophotometrically by using an Alliance Evolution II autoanalyzer according to standard procedures (Grasshoff et al., 1983). Chlorophyll a (Chl a) concentration was measured from acetone extracts by fluorometry from the total fraction (<200 µm) and the fractions less than 20 and 3 µm. To collect microbial biomass, about 5 l of surface seawater was sequentially filtered through a 3- and a 0.2-µm pore-size polycarbonate filters (Poretics, GE Osmotics, Delft, Netherlands) using a peristaltic pump. The filters were stored in cryogenic vials containing 1.7 ml of lysis buffer (50 mM Tris-HCl pH 8.3, 40 mM EDTA pH 8.0 and 0.75 M sucrose) at −80◦C until further processing.

# DNA Extraction and Sequencing

The 0.2-µm filters were treated with lysozyme, proteinase K and sodium dodecyl sulfate, and the nucleic acids were extracted with phenol and concentrated in an Amicon 100 (Millipore), as described in Massana et al. (1997). DNA was quantified spectrophotometrically (Nanodrop, Thermo Scientific), and two subsamples from each extraction were sent for sequencing. Sequencing was performed by the Research and Testing

Laboratory (Lubbock, TX, USA; http://www.researchandtesting. com/). Primers 341F (5′ -CCTACGGGNGGCWGCAG-3′ ), and 805R (5′ -GACTACHVGGGTATCTAATCC-3′ ) were used to amplify bacterial 16 S rRNA gene (Herlemann et al., 2011) and primers TAReukFWD1 5′ -CCAGCASCYGCGGTAATTCC-3 ′ and TAReukREV3 5′ -ACTTTCGTTCTTGATYRA-3′ were used to amplify the V4 region of the eukaryotic 18 S rRNA gene (Stoeck et al., 2010). Pyrosequencing was performed using the bTEFAP method by 454 GL FLX technology as described previously (Dowd et al., 2008). Illumina MiSeq 2 × 250 flow cells were used for Illumina sequencing following protocols described elsewhere (Cúcio et al., 2016). Approximately 30,000 raw sequences per sample were obtained.

### Data Analyses

High-Performance computing analyses were run at the Marine Bioinformatics Service of the Institut de Ciències del Mar (ICM-CSIC) in Barcelona. Reads from the two sequencing methodologies underwent method-specific quality filtering before being pooled. Bacterial-454 data was filtered by quality using QIIME (Quantitative Insights Into Microbial Ecology, Caporaso et al., 2010) as described in Sánchez et al. (2013). Briefly, sequences from the 454 run were assigned a sample IDs using a mapping file and the barcode assigned to each sample. After sample IDs were assigned, bacterial sequences were removed from the subsequent analyses if they were shorter than 150 bp or longer than 500 bp, had an average quality score <25 calculated in sliding windows of 50 bp, contained more than two ambiguous characters or had an uncorrectable barcode. Eukaryotic-454 reads were quality checked and demultiplexed with QIIME following the same parameters described in Pernice et al. (2015). Shortly, sequences shorter than 150 bp or longer than 600 bp, with more than three mismatches in the primer, or having homopolymers longer than 8 bp were removed. Phred quality was analyzed in 50 bp running windows. Illumina sequences from bacteria and picoeukaryotes were quality filtered following a custom made pipeline (https://github.com/ramalok). Briefly, BayesHammer error correction of sequence reads was performed with SPAdes software (Nurk et al., 2013). Sequences were assembled with PEAR (http://pear.php.net/) and quality filtered in UPARSE (fastq\_maxee value = 1). Clean bacterial-454 and bacterial-Illumina sequences were pooled and processed together; eukaryotic-454 and eukaryotic-Illumina sequences were also pooled together. Since 454 and Illumina sequences may have different length, bacterial sequences were truncated at equal depth (400 bp). However, for picoeukaryotes we did not truncate sequences, since large natural variability in the length of 18 S rRNA from different taxa occur. Sequences of both datasets were clustered into operational taxonomic units (OTUs) at 97% cutoff using the UPARSE algorithm implemented in USEARCH (Edgar, 2013). Both de novo chimera check and by comparison to reference database (SILVA) were done using the UCHIME algorithm (Edgar et al., 2011). Chimeric sequences and singleton OTUs (those represented by a single sequence) were removed. Taxonomic assignment of bacterial OTUs was performed using the BLAST classifier and the version 119 of the SILVA SSURef non-redundant database. OTUs assigned to chloroplasts were removed for subsequent analyses. For picoeukaryotes, OTUs were taxonomically classified by using BLAST against two reference databases: PR<sup>2</sup> (Guillou et al., 2013) and a marine microeukaryote database (MASS9013, Pernice et al., 2013). After taxonomic assignment, metazoan OTUs were removed. Sequence data has been submitted to the Genbank Sequence Read Archive under accession number SRP079955.

Statistical analyses were performed using the R statistical software (R Developement Core Team, 2015) and the packages vegan, labdsv, venneuler, hmisc, and corrgram. Alpha- and betadiversity analyses were performed using an OTU abundance table that was previously subsampled down to the minimum number of reads in order to avoid artifacts due to an uneven sequencing effort among samples. For alphadiversity analyses, we calculated the Chao1 index as a measure of richness and Shannon and Simpson indices as diversity metrics. Differences in microbial composition (betadiversity) were assessed using hierarchical clustering of Bray-Curtis dissimilarity matrices and the Unweighted Pair Group Method with Arithmetic Mean algorithm (UPGMA). To search for "indicator species" we used the IndVal (INDicator VALues; Dufrêne and Legendre, 1997) analysis, which identifies indicator species based on OTU fidelity and relative abundance. Only OTUs with significant p-values (< 0.05), and >0.3 IndVal values were considered. To assess links between diversity and environmental data we performed linear regressions and pairwise correlations (Pearson's correlation coefficient). The results were thresholded at p < 0.05. Analysis of variance was run to test for differences among diversity data and categories (sequencing method, station) with Tukey-Kramer post hoc comparisons at the 5% significance.

# RESULTS

Five stations were sampled in an inshore-to-offshore transect off the coast of Barcelona. Station 1 was located closest to the shore. The following stations were sampled in increasing depth and distance to the shore (**Figure 1**). Basic physicochemical data is shown in **Table 1**. The sampled area is expected to suffer impacts from human activities due to a large urban development, and putatively receiving pollutants from urban and industrial activities (domestic waste, organic and inorganic nutrient enrichment). A decreasing nutrient concentration was observed as distance to shore increased. Despite this variability can in part be associated to natural processes, it can also reflect the degree of human impact (i.e., nutrient enrichment). Concentration of all nutrients measured showed the lowest values in Station 5 (offshore) and higher values closer to the shore.

# Influence of the Sequencing Platform on Microbial Diversity

Sequencing of all bacterial and most picoeukaryotic samples was successful yet two picoeukaryotic replicates (1a, 454, and Illumina) resulted in a low number of reads and were discarded from further analyses. The bacterial dataset resulted in 277,212 high quality reads that clustered in a total of 658 OTUs at 97% similarity. From those only 34.7% of OTUs were shared


TABLE 1 | Values of physicochemical variables measured along the inshore-to-offshore transect.

between samples sequenced by either 454 or Illumina (**Figure 2**). However, the unique OTUs in each methodology correspond to rare members; the proportion of shared OTUs (350 out of 658) represented 99.4% of the reads. We found a good correlation between the relative abundance of each OTU sequenced by both methodologies (R = 0.87, p < 0.001). Likewise, when grouping OTUs into the main bacterial taxa, a good agreement between contributions obtained by 454 or Illumina was found (R = 0.81, p < 0.001). In both cases, most bacterial sequences were related to the phyla Proteobacteria (average of all bacterial dataset, 72%), Bacteroidetes (20%), and Cyanobacteria (5%). Within the Proteobacteria, the most prevalent classes were the Alpha-(50%) and the Gammaproteobacteria (20%), whereas the Beta-, Delta-, and Epsilon- were present at low relative abundances (grouped as "Other Proteobacteria," **Figure 3**). Within the Alphaproteobacteria, the OTUs showing higher relative abundances were affiliated to the Rhodobacterales, Rhodospirillales, Rickettsiales, and the SAR11 clade. The Bacteroidetes were largely represented by members of the Flavobacteriia. The Actinobacteria represented on average 1% of the total reads. Several other groups such as the Acidobacteria, Firmicutes, Gracilibacteria, Parcubacteria, Planctomycetes, and the Verrucomicrobia were also detected but at low read abundances (<1%) and were grouped as "Other Bacteria" for plotting purposes. Analysis of the variance resulted in no significant differences in the contribution of the major taxa retrieved by each sequencing methodology.

A similar pattern was observed for picoeukaryotes (0.2–3 µm size fraction). The 556,143 clean reads were clustered into an OTU table at 97% similarity that contained 768 OTUs; from those only 37.1% were shared between the two methodologies, but these represented the vast majority of reads (96.4%; **Figure 2**). OTUs recovered with only one of the sequencing methodologies represented very rare members. In fact, as for bacteria, we found very good correlations when comparing the relative abundance of the different taxa both at the OTU level or clustering them at the taxonomic group level (R = 0.84, p < 0.001, and R = 0.91, p < 0.001, respectively; see **Figure 3**). The picoeukaryotic OTUs were classified into 70 class-level groups. The taxonomic affiliation was dominated by four groups that accounted on average for >55% of the total number of reads within the picoeukaryotic dataset: Mamiellophyceae (19% of the reads, dominated by Micromonas OTUs [97% of Mamiellophyceae]), Dinophyceae (17%), MALV-II (10%), and Cryptophyceae (10%). Other less abundant groups included MALV-I, Chlorarachnida, Picozoa, Prasinophyceae, Dictyochophytes, Chlorodendrophyceae,

MAST-3, and Pelagophytes. The remaining 58 taxonomic groups presented very low relative abundances (<1.1%) and were grouped as "Other Eukaryotes." No statistically significant differences in the relative abundance retrieved by 454 or Illumina for the difference groups were found.

In order to further explore whether the sequencing methodology had an influence on the bacterial and

picoeukaryotic diversity, we calculated various widely used indices of alphadiversity: the Chao 1 index for richness, and the Shannon and Simpson indices for diversity estimation (Hill, 1973; Magurran, 1988; Chao and Lee, 1992; **Figure 4**). Analysis of variance showed no significant differences between sequencing platforms for any of the indices tested, neither for Bacteria nor for Picoeukaryotes (P > 0.05). Additionally, to infer the variation of the microbial assemblages along the gradients, that is, beta diversity, the Bray–Curtis dissimilarity index was used on community composition. Dissimilarity matrices were constructed based on the relative abundance of each OTU. The distance between samples and replicates was visualized using hierarchical clustering. The results show that, in general, replication was good within each sequencing platform, but replicates sequenced using the same methodology were more similar among each other, which indicates that the sequencing chemistry has a certain influence on the community composition observed (**Figure 5**). For Bacteria, the samples grouped according to station regardless of the sequencing platform, except for Stations 3 and 4, which grouped by method, indicating that the platform introduces errors and artifacts to a certain extent at the OTU level. A similar trend was observed for the picoeukaryotic dataset, in which samples grouped by station, and in general were more similar among replicates subjected to the same methodology. Yet, in one case, the replicate obtained by Illumina (Illu−4a) was fairly different to the rest of the replicates from the same station. The number of OTUs in Illu-4a sample was much lower than in the other three replicates of Station 4 (one from Illumina and two from 454), indicating some biases in amplification or sequencing of this specific sample.

### Bacterial and Picoeukaryotic Plankton Diversity along a Inshore-to-Offshore Gradient

In order to obtain direct descriptors of the bacterial and picoeukaryotic diversity of plankton assemblages, we compared the diversity retrieved along the inshore-to-offshore gradient (**Figure 1**, **Table 1**). We observed significant differences in alphadiversity between stations (**Figure 6**). In particular, significant differences for Chao1 and Simpson indices were found for the bacterial dataset. The Chao1 showed a clear

inshore-to-offshore decrease, whereas the Simpson index showed higher values in the transition zone from the coastal to the offshore station (**Figure 6**). The Shannon index showed a similar trend to the Simpson index but the differences detected were not significant. Interestingly, eukaryotic picoplankton showed a different trend. Whereas the values of Chao1 were quite constant along the gradient, the Shannon and Simpson diversity indices increased from coast to offshore. In fact, statistical analyses (ANOVA) confirmed significant differences, particularly between stations 1 and 5.

In addition, clear inshore-to-offshore changes in community composition were found both for bacteria and eukaryotic picoplankton. The larger differences were detected between Station 1 (coastal) and Station 5 (offshore), whereas a transition in community composition was observed at intermediate stations (**Figure 3**). In the case of Bacteria, some phylogenetic groups (Phylum, Class, and Order levels) showed a clear increase in their abundance from coast to offshore. These include the phylum Actinobacteria and the orders Rhodospirillales, Rickettsiales, and SAR11 within the class Alphaproteobacteria. An opposite trend was observed for the phylum Bacteroidetes, order Rhodobacterales (Alphaproteobacteria), and orders Alteromonadales and Oceanospirillales of the Gammaproteobacteria. Phylum Cyanobacteria were small contributors to community composition in the coastal station and peaked at Station 3 coinciding with the highest value of Chl a. The greater differences were observed for the order Alteromonadales which represented >25% of the reads in the coastal station and decreased to almost nil in the offshore station. Conversely, the SAR11 clade increased from 1 to >20% of the reads along the transect. Analyses of variance confirmed significant differences between stations for all the above-mentioned groups (details now shown).

The picoeukaryotic community also changed along the gradient being likewise Station 1 the most different from Station 5. The lineage Mamiellophyceae showed similar high relative abundance (>20%) in all stations except in Station 5, where they were virtually absent. The relative abundance of Cryptophyceae increased from Station 1 to Station 3 and then decreased toward offshore stations. Dinophyceae, Dictyochophytes, marine alveolates (MALV-II and MALV-I), and Stramenopiles showed increasing contributions along the transect. Contrarily, Ciliophora were important contributors only in the coastal station. Other groups presented quite constant contributions in all stations (Picozoa, Prasinophyceae, Dictyochophytes; **Figure 3**).

# Potential Indicators of Environmental Status

Potential "indicator species" were explored by calculating the indicator value (IndVal; Dufrêne and Legendre, 1997; Podani and Csányi, 2010) which identifies indicator species based on species (or OTU) fidelity and relative abundance, both for bacterioplankton and eukaryotic picoplankton. The IndVal of a species is a popular measure to express species importance in community ecology. Its potential to measure species explanatory power and to reflect environmental quality has been explored in biodiversity surveys (Gevrey et al., 2010; Keith et al., 2012; Lumbreras et al., 2016). We classified the stations into three categories, i.e., coastal (Station 1), transition (Stations 2, 3, and 4), and offshore (Station 5) and searched for indicator OTUs. We found 114 bacterial OTUs with significant IndVal values, potentially useful as indicator species. However, we considered only those OTUs showing (i) IndVal values >0.3, as this is the value that has been proposed to be a good threshold for habitat specialization (Dufrêne and Legendre, 1997), and (ii) overall relative abundance >1% since the potential as indicator species of rare OTUs is questionable taking into account the differencesfound between sequencing methods for the rare OTUs and the known biases of the PCR-based methodologies (Polz and Cavanaugh, 1998; Acinas et al., 2005). After selection, the list was reduced to 23 bacterial OTUs. We found OTUs with explanatory power for all three categories. The OTU with higher IndVal value was affiliated to a species of Gammaproteobacteria (Marinobacterium) and was indicative of coastal waters. On the contrary, alphaproteobacterial members of the SAR11 clade were explanatory for offshore waters and mainly Bacteroidetes for the transition zone. Within picoeukaryotes, a total of 164 OTUs with significant values were found but after filtering the table using the same criteria only 13 OTUs were retained. Most of them were explanatory for Station 5 in offshore waters. However, the indicator OTU presenting a higher contribution, OTU1, was classified as Micromonas pusilla, and was indicator for coastal waters in agreement with previous reports that have shown the preference of Micromonas species for coastal waters (Not et al., 2005, 2008). Overall, IndVal values for picoeukaryotic OTUs were lower than for bacterial OTUs. In both cases, the highest values were associated to rare species (details not shown) that were discarded based on abundance data. The selected IndVal scores and associated OTUs are listed in **Table 2**.

In addition to exploring potential "indicator species," we explored the microbial profiles as possible descriptors of environmental status. That is, analyzing the relative abundance of the most abundant phylogenetic groups in each sample in relation to the degree of impact. The transect analyzed off the coast of Barcelona reflects a decreasing gradient of human impact from inshore (Station 1) to offshore (Station 5) which is somewhat reflected in the concentration of inorganic nutrients (see **Table 1**). The analysis of changes in community composition along the gradient together with the OTUs showing highest IndVal scores suggested the exploration of the ratios between taxa as potential indices of ecosystem health status. Interestingly, we found strong positive and negative correlations between the relative abundance of different bacterial groups as well as the ratio between taxa and the concentration of nutrients. The strongest correlation detected was a positive correlation between the relative abundance of Alteromonadales and all nutrients measured (phosphate, nitrite, nitrate, ammonium, silicate, R > 0.96, p < 0.0001). Likewise, the ratios Alphaproteobacteria/ Gammaproteobacteria, Alteromonas/SAR11, and Alteromonas + Oceanospirillales/SAR11 were strongly correlated to nutrient concentration (R > 0.90, p < 0.0001). For picoeukaryotes, we found significant correlations between the relative abundance of certain taxa and the nutrient load, yet these correlations were in general weaker than for bacteria. The strongest positive correlations were found for Ciliophora and all nutrients (R = 0.85–0.88, p < 0.0001). Significant negative correlation between Chlorarachnida and nitrite (R = 0.79) and nitrate (R = 0.75) as well as between Dinophyceae and phosphate (R = 0.71) were also observed.

# DISCUSSION

# Do Different Sequencing Methodologies Provide Comparable Views of Microbial Biodiversity in Marine Ecosystems?

Up to date, several studies have investigated the potential biases on the estimations of richness and evenness in microbial communities associated with the primer selection and the PCR


TABLE 2 | Potential indicator OTUs identified with IndVal.

*of the OTU in the dataset (abundance),*

 *the GenBank accession number, the taxonomy and the similarity to the reference database used (see Section Material and Methods).*

step in amplicon-based studies (Acinas et al., 2005; Hong et al., 2009; Engelbrektson et al., 2010; Parada et al., 2015). However, since cross-platforms studies are rare, currently it is unclear whether the inherent differences in chemistry and sequencing protocols will affect the quality of the sequences and the estimates of genetic diversity and community structure. Furthermore, despite variability is known to be introduced during sample manipulation, PCR amplification and sequencing, the numerous studies on microbial diversity using HTS lack analysis of replicates (Prosser, 2010). For these reasons, we compared the two most frequently used HTS platforms, the Roche 454 FLX Titanium, and the Illumina MiSeq, on a set of DNA samples obtained from an inshore-to-offshore transect in the coast of Barcelona. Additionally, we explored the reproducibility of the results by sequencing replicates. Overall, the platforms provided a comparable view of the marine picoplankton communities but some differences were found when comparing the datasets at the OTU level.

Different HTS platforms produce millions of short sequence reads, which vary in length. It is known that sequence length can impact diversity estimates (Claesson et al., 2010). Nowadays, pair-end Illumina can produce up to 300 bp nucleotide reads, and thus is feasible to do a careful compassion with 454 using the same primer set, providing the same amplicon length, and thus distinguish the performance of both methodologies based only in potential differences in the chemistry of the sequencing. Here, we found that the sequencing methodology does not significantly influence estimates of alphadiversity. No significant differences in Chao1, Shannon, and Simpson indices were found between platforms. A recent study comparing the Illumina and 454 platforms to study bacterial diversity via 16 S rRNA gene amplicons in sediments and soda lakes also found that both methodologies performed in a similar manner and that the general trends in alphadiversity were conserved with the exception of evenness estimates where correspondence between methods was low (Sinclair et al., 2015). It is known that the OTU clustering method can influence the estimates of diversity (Edgar, 2013; Flynn et al., 2015; Sinclair et al., 2015). We used the UPARSE algorithm, which offers an improved accuracy compared to other methods, resulting in fewer OTUs likely closer to the expected number of species in a community (Edgar et al., 2011). Using this methodology may have reduced the influence of sequencing and amplification artifacts and resulted in comparable estimates of diversity by the two sequencing methodologies.

We did observe some differences for betadiversity, that is the variation of the microbial assemblages along the transect, despite the trends identified were in general similar for both methodologies. Replication was good within each sequencing platform but in general replicates were more similar among each other depending on methodology, revealing thus a certain influence of its chemistry. We found that the bacterial communities in Stations 3 and 4 were more similar depending on the method indicating that the platform introduces biases. Oceanographic conditions were quite similar between these two stations (**Table 1**), and therefore microbial communities could be expected to be fairly similar. Sampling artifacts associated with random sampling (Zhou et al., 2008), PCR biases (Polz and Cavanaugh, 1998; Acinas et al., 2005) or errors directly related to the performance of the technology per se (Berry et al., 2011; Schirmer et al., 2015) can occur at any time, but when comparing samples, the impact of these artifacts will depend on the similarity among those samples. In this case, it is feasible to assume that the potential artifacts associated to the methodology overwhelmed the natural differences between the communities in these closer stations. For picoeukaryotes, in general samples grouped by station as expected, indicating that the sequencing biases, if any, were minor. However, there is one replicate from Station 4 that differs substantially from the other replicates. Problems during PCR amplification or degradation of the DNA could explain this difference.

Venn diagrams showed that less than half of the total OTUs were equally retrieved by both methodologies. However, the non-shared OTUs correspond to very rare contributors of these microbial communities. The concept of the rare biosphere has attracted a lot of attention in the last years (Pedrós-Alió, 2012; see reviews by Lynch and Neufeld, 2015). Microbial communities are dominated by a small number of species that account for most of the biomass and a large number of species that are represented by only a few individuals (rare members). The development of HTS has allowed accessing at least some of these rare microbial species. However, it is known that some of the rare OTUs retrieved in microbial diversity surveys correspond to sequencing errors (Kunin et al., 2010). We discarded the singletons (OTUs represented by a single sequence in the whole dataset) to avoid potential artifacts in diversity estimates. Nevertheless, still over half of the OTUs were only retrieved by one methodology. Part of it can be explained because rare OTUs may or may not appear in a dataset only by random chance but we cannot discard that part of this diversity is due to sequencing errors. For that reason, for the purpose of finding indicator species, we decided to focus only on those OTUs that represented >1% of the total relative abundance. Regardless of the differences in the rare OTUs, the two sequencing technologies revealed very similar profiles when grouping OTUs at the class and family levels (**Figure 3**). Relative taxa abundances were consistent across technologies and thus, the view of the community composition was fairly comparable. The results show that, due to the improvement in the length of Illumina sequence reads, Illumina tags offer similar classification efficiencies than 454 tags at a much lower cost (Glenn, 2011), being therefore a cost efficient approach for biodiversity monitoring.

# Does Plankton Diversity Have Informative Potential for Environmental Status Assessment?

Diversity and trophic state are two quality descriptors for evaluating ecosystem function in the MSFD. Despite the main goal of this work was to compare HTS methodologies for biodiversity monitoring, we further explored whether picoplankton biodiversity can be used as an alternative indicator of environmental status. The Mediterranean Sea is a valuable paradigm to assess anthropic pressure, because of the contrasting nature of its offshore and coastal areas. The offshore waters of the Mediterranean Sea are among the most oligotrophic areas of the world. In these waters, nutrient availability is low and inorganic phosphorus concentrations limit primary production. On the contrary, coastal areas are nutrient rich, as they receive river discharges, runoff from populated areas, and submarine groundwater, but they are also influenced by offshore oceanographic conditions. The coastal marine zone is therefore a transitional area characterized by strong physical, chemical, and biological gradients that extend from land to sea. Here, biological production is closely coupled to processes that deliver nutrients to surface waters. Anthropogenic forcing clearly influences the absolute availability of these nutrients and their stoichiometry, both of which impact phytoplankton productivity and species composition (Camp et al., 2015). The studied transect is expected to have a decreasing degree of anthropogenic pressure as the distance from the coast increases (from Station 1 to 5). Concentration of inorganic nutrients, as indication of eutrophication, showed indeed a decreasing concentration. We determined common alphadiversity indices as possible descriptors of the environmental status since pressures can lead to changes in microbial composition (Torsvik et al., 2002; Smith and Schindler, 2009) and those could reflect variations in biodiversity. For bacteria, Shannon and Simpson indices showed a similar trend with higher values at intermediate stations of the transect. The observed trend could be explained by the "intermediate disturbance hypothesis" (Connell, 1978), which suggests that intermediate intensity of disturbance maximizes diversity, and therefore systems with low and high disturbance, such Stations 5 and 1 in terms of nutrient load, can harbor similar levels of diversity. In any case, as previously observed in other systems (Garrido et al., 2014) these indices do not seem promising as indicators to asses environmental status. Contrarily, a clear decrease in richness was observed from coast to offshore. A sharp decrease of richness from coastal to offshore locations in the NW Mediterranean has been previously documented (Pommier et al., 2010). On the other hand, an increase in Shannon and Simpson indices was observed along the transect for picoeukaryotes, indicating a higher diversity in more oligotrophic stations (Cheung et al., 2010). Furthermore, the most abundant OTU in all stations but Station 5, Micromonas, is known to be more common in coastal areas than open ocean (Not et al., 2005), possibly related to higher nutrient load in coastal waters. The results found here suggest that it may be worth exploring the links between bacterial and picoeukaryotic diversity and environmental status on coastal waters over time and space covering a wide range of impacts.

Traditionally, several species of plants and animals have been and still are being used as indicator species for different types of pollution in monitoring programs (Borja et al., 2000, 2008; Ferrat et al., 2003; Montefalcone, 2009; Marbà et al., 2013). Likewise, plankton indicators have been proposed for diagnoses of ecosystem state (Beaugrand, 2005). Most studies have focused on species of zooplankton (i.e., Calanus finmarchicus) or some phytoplankton bloom-forming species. For example, Phaeocystis sp. produces spring blooms in the North Sea which magnitude might indicate an excess of available N or P in relation to dissolved silica and thus, is considered and indicator for eutrophication (Tett et al., 2007). However, several flaws in the usefulness of using large phytoplankton to reflect significant pressure-impact relationships have been identified (Cloern and Jassby, 2008, 2010; Camp et al., 2015). Bacterial and eukaryotic picoplankton constitute the smallest but most abundant organisms of plankton and are key players in ecosystem functioning. Since disturbances can affect community structure and ecosystem functioning, the smallest members of marine plankton may be crucial in understanding the magnitude of these disturbances particularly because of their fast response to environmental change. In fact, microorganisms have been already proposed as indicators of marine environmental quality, and not only the presence of pathogens such as E. coli, commonly used as indicator of fecal contamination, but in relation to biodiversity and ecosystem functioning (Caruso et al., 2015). Here we tested the Indicator Species Value from Dufrêne and Legendre (1997) in the different sampled stations. The IndVal identifies indicator species based on OTU fidelity and relative abundance. Different bacterial and picoeukaryotic OTUs showed high scores for Stations 1 and 5, as well as for intermediate stations and could represent potential "indicator species." Alternatively to "indicator species," we explored the potential of using the abundance of certain taxa and the ratio between different groups of microorganisms as an alternative indicator of environmental status. These indices may also offer ecological information (i.e., species relative composition). In fact, this approach has been explored in other ecosystems; for example in reclaimed waters, the ratio between the Bacteroidetes, Gammaproteobacteria, and Nitrospira/Betaproteobacteria (BGN:β) seems a possible alternative indicator of water quality (Garrido et al., 2014). We tested the correlation of different taxa and the degree of eutrophication (i.e., nutrient concentration) and found significant correlations between certain picoeukaryotic taxa e.g., Ciliophora, and nutrient load; this taxa has been found previously in high abundances in eutrophic waters (Romari and Vaulot, 2004). Yet, the strongest correlations were with the ratio of Alphaproteobacteria/Gammaproteobacteria, Alteromonas/SAR11, and Alteromonas + Oceanospirillales/SAR11. Whether these "indicator species" and indices can be used as robust alternative indicators of environmental status remains to be explored in different locations subjected to contrasting pressures and over time. The challenge is to discriminate between antropogenic-induced changes and the confounding effects of the natural variability of the marine environment.

# CONCLUSIONS

HTS methods are commonly used to determine the diversity of complex marine microbial communities and have been proposed as a suitable tool in biodiversity monitoring programs. However, validating their usefulness is crucial for conducting rigorous analyses. Comparison of 454 and Illumina methodologies showed minor differences in the performance of both sequencing methodologies that can in part be attributed to inherent differences in chemistry and sequencing protocols, which may affect the quality of the sequences. Nevertheless, these differences were assigned to very rare OTUs and overall, both platforms provided a comparable view of the marine picoplankton communities. On a taxonomic level, there was very good overlap in the detected phyla between the two methods. The comparative analyses performed suggest that 454 and Illumina data can be combined if the same bioinformatic workflow for describing overall patterns of diversity and taxonomic composition is used. On the other hand, we found that plankton biodiversity surveys have the potential to be used as alternative indicators of environmental status. In particular, using bacterioplankton biodiversity (bacterial richness as well as the ratio between certain bacterial taxa) as an alternative indicator of water quality deserves further investigation. However, these preliminary results have to be further investigated by performing intensive surveys covering wide spatial and temporal scales in order to discriminate between changes resulting from human activities and the natural variability of the marine environment and test whether the identified indices are universally applicable.

# AUTHOR CONTRIBUTIONS

Conceived and designed the study: IF, AR, JC, JG, EG; Performed the experiments: IF, CG, AR. Contributed materials: JC, RM, JG, EG; Analyzed the data: IF, CG. Interpreted the data and wrote the paper: IF, CG, AR, JC, RM, JG, EG. All authors reviewed the manuscript.

### ACKNOWLEDGMENTS

This manuscript is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing GEnS) project, funded by the European Union (grant agreement no. 308392), and a MINECO Grant GRADIENTS Fine-scale structure of cross-shore GRADIENTS along the Mediterranean coast (CTM2012-39476-C02). We thank the Coastal Ocean Observatory (http://coo.icm.csic.es/) from the ICM for making possible the sampling and providing ancillary data, and the Marine Bioinformatics Service from the ICM, particularly Drs. Pablo Sánchez and Ramiro Logares for help with computing analyses. We also thank Dr. Eva Ortega-Retuerta for assistance using Ocean Data View and Laura Arin for chlorophyll analyses.

# REFERENCES


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declared a collaboration with the authors and states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Ferrera, Giner, Reñé, Camp, Massana, Gasol and Garcés. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Marine Sediment Sample Pre-processing for Macroinvertebrates Metabarcoding: Mechanical Enrichment and Homogenization

Eva Aylagas \*, Iñaki Mendibil, Ángel Borja and Naiara Rodríguez-Ezpeleta

*AZTI, Marine Research Division, Sukarrieta, Spain*

Metabarcoding is an accurate and cost-effective technique that allows for simultaneous taxonomic identification of multiple environmental samples. Application of this technique to marine benthic macroinvertebrate biodiversity assessment for biomonitoring purposes requires standardization of laboratory and data analysis procedures. In this context, protocols for creation and sequencing of amplicon libraries and their related bioinformatics analysis have been recently published. However, a standardized protocol describing all previous steps (i.e., processing and manipulation of environmental samples for macroinvertebrate community characterization) is lacking. Here, we provide detailed procedures for benthic environmental sample collection, processing, enrichment for macroinvertebrates, homogenization, and subsequent DNA extraction for metabarcoding analysis. Since this is the first protocol of this kind, it should be of use to any researcher in this field, having the potential for improvement.

#### Edited by:

*Marianna Mea, Ecoreach s.r.l, Italy; Jacobs University of Bremen, Germany*

#### Reviewed by:

*Franck Lejzerowicz, University of Geneva, Switzerland Erik Michael Pilgrim, U.S. Environmental Protection Agency, USA*

\*Correspondence:

*Eva Aylagas eaylagas@azti.es*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *15 June 2016* Accepted: *30 September 2016* Published: *18 October 2016*

#### Citation:

*Aylagas E, Mendibil I, Borja Á and Rodríguez-Ezpeleta N (2016) Marine Sediment Sample Pre-processing for Macroinvertebrates Metabarcoding: Mechanical Enrichment and Homogenization. Front. Mar. Sci. 3:203. doi: 10.3389/fmars.2016.00203* Keywords: environmental samples, laboratory procedures, sample manipulation, DNA, biomonitoring

# INTRODUCTION

Biomonitoring has become essential to address changes in the quality of the environment as a response to the several pressures that are threatening marine ecosystems (Halpern et al., 2008). The rapid response of benthic organisms to a range of natural and anthropogenic pressures makes this community a suitable ecological component for marine biomonitoring (Johnston and Roberts, 2009). Above all, macroinvertebrates are widely used to assess environmental quality through the calculation of benthic indices (Diaz et al., 2004; Borja et al., 2015). Yet, the fast environmental degradation and the necessity of cost-effective methods for biodiversity assessment urge the need of new tools that allow species identification in a much faster way compared to morphological methodologies (Bourlat et al., 2013). The advent of high-throughput sequencing (HTS) technologies has favored the application of DNA-based biodiversity assessment methods (Creer et al., 2016) and in particular, DNA metabarcoding has become a promising technique for rapid, accurate, and cost-effective taxonomic identification of the benthic macroinvertebrate community in environmental samples (Elbrecht and Leese, 2015; Aylagas et al., 2016).

DNA metabarcoding involves the amplification of a particular DNA region (barcode) to resolve the total genomic DNA extracted from an environmental sample into distinct taxa, typically species, by using universal primers (Taberlet et al., 2012). Coupled with HTS, the technique enables the simultaneous identification of the taxonomic composition of several independent samples by matching the unknown amplified DNA barcode to a DNA reference database (ideally, every organism within a sample can be detected). Metabarcoding has been proven useful in the identification of metazoan community composition from a wide variety of aquatic environments (Chariton et al., 2010; Cowart et al., 2015; Dowle et al., 2015; Elbrecht and Leese, 2015; Lejzerowicz et al., 2015; Leray and Knowlton, 2015; Zaiko et al., 2015), and recent studies have proved that the ecological ecosystem condition addressed through the calculation of DNA-based biotic indices is comparable to that inferred using morphological identification (Dowle et al., 2015; Lejzerowicz et al., 2015; Aylagas et al., 2016). However, metabarcoding is not a fully established methodology for marine monitoring. Therefore, standardization of procedures is necessary, which requires of optimized protocols that allow the reliability and reproducibility of the approach. In this sense, significant efforts have been made to standardize different steps of the metabarcoding workflow by addressing the issues regarding to PCR amplification (Aylagas et al., 2016), barcode region (Carew et al., 2013), primer selection (Leray et al., 2013), library preparation (Bourlat et al., 2016), and bioinformatics analysis for data interpretation (Aylagas and Rodríguez-Ezpeleta, 2016).

A major limitation for environmental DNA metabarcoding studies of benthic macroinvertebrate communities that has not been properly addressed is the manipulation of the sample to be analyzed. Usually, sediment and organic matter carried over using marine benthic community sampling methods result in large sample volume, which needs to be correctly processed so that DNA representing the whole community can be extracted. However, the amount of collected material, the nature of the sample (e.g., mud sediments require different processing than coarse sands) and the size of the target organisms make, in some cases, DNA extraction of the entire sample unfeasible. The requisite of an adequate metabarcoding study is that the sample must be representative of the whole community. Thus, because each sample is different, the pre-processing strategy must be carefully considered in order to retrieve a reliable representation of the macroinvertebrate community. Additionally, routine application of metabarcoding for biomonitoring requires each step of sample collection, handling, pre-processing, DNA extraction, and DNA library preparation and sequencing be standardized so that results from different laboratories can be compared and combined (Deiner et al., 2015).

Different approaches can be used to recover DNA from sediment samples. Generally, the size range of the target organisms determines the amount of sediment to be processed and the protocol used (Creer et al., 2016). For studies targeting small size metazoans (e.g., meiofauna), the procedures can rely on extracting DNA from small sediment samples (i.e., 5 gr of sediment) without any pre-processing step (Lejzerowicz et al., 2015), targeting extracellular DNA (Guardiola et al., 2015; Pearman et al., 2016), or performing some separation via decantation/flotation (Creer et al., 2010). However, when the fraction to be investigated is larger (e.g., macroinvertebrates) samples need first be processed via decantation protocols so that the macroinvertebrate community is separated from the sediment. Recently, Aylagas et al. (2016) showed that following protocols to target the extracellular DNA from sediment samples, only a small proportion of the macroinvertebrate taxa is retrieved, whilst the isolation of organisms followed by homogenization and DNA extraction allows a reliable characterization of the macroinvertebrate community through DNA metabarcoding.

The objective of the present protocol is to extract good quality and integrity DNA from complex environmental samples which is representative of the whole macroinvertebrate community. For that purpose, we present guidelines for the processing of benthic sediment samples collected for metabarcoding-based biomonitoring. We detail the steps necessary to (i) preserve the benthic sample to ensure DNA integrity, (ii) isolate organic fraction from the sediment by decantation, (iii) homogenize the sample in order to achieve a good community representation, and (iv) extract DNA of good quality and integrity. The efficiency of sediment decantation and homogenization steps detailed in this protocol have previously shown to help providing accurate metabarcoding taxonomic inferences that are comparable to those inferred from morphology (Leray and Knowlton, 2015). Thus, followed by the well-established metabarcoding procedures for library preparation (Bourlat et al., 2016) and bioinformatics analysis (Aylagas and Rodríguez-Ezpeleta, 2016) this protocol represents the first steps of the procedure to gather the taxonomic list of several benthic samples simultaneously. This information can be ultimately used for a variety of applications that rely on the macroinvertebrate community characterization of the samples such as the calculation of benthic indices for ecological status assessment (Aylagas and Rodríguez-Ezpeleta, 2016), the detection of non-indigenous species (Zaiko et al., 2015), or large-scale spatio-temporal biodiversity assessments (Leray and Knowlton, 2015; Chain et al., 2016). Finally, a Notes section is dedicated to discuss various artifacts and pitfalls to consider throughout the description of the protocol.

### MATERIALS AND EQUIPMENT

#### Sample Collection and Preservation


# Sample Processing

#### Decantation


#### Homogenization and DNA Extraction


#### DNA Overall Quality Assessment, Purification and Normalization


# PROCEDURES

# Sample Collection and Preservation

DNA-free materials thoroughly cleaned between locations must be used to avoid cross-contamination (see **Note 1**), and samples should be preserved under appropriate conditions to guarantee DNA integrity.


# Sample Processing

#### Decantation (0.5 h)

Humic substances, co-extracted with DNA, inhibit enzymes such as the Taq Polymerase used in PCR reactions to amplify DNA, representing the primary inhibitory compound associated with sediment samples (Matheson et al., 2010). This inhibition represents a potential bias for DNA metabarcoding studies performed on sediment samples and, if not properly addressed, can lead to generation of false negative results (Thomsen and Willerslev, 2015). At the same time, the heterogenic composition of the benthic macroinvertebrate community would require extracting all DNA within a sample in order to detect all species present. As this step is logistically unfeasible, the homogenization of the sample is required, so that a subsample is representative of the whole community. The volume of sediment processed may significantly vary among samples, which could imply a great impact on the sample representativeness. In this sense, low amounts of sediment in the sample allow for more representative homogenized subsamples. For these reasons, it is recommended to separate the organic fraction from the sediment before proceeding with DNA extraction. Depending on sediment type (**Figure 1**), this separation can be totally or partially performed through a decantation process. Medium to coarse grain sediments can often be completely removed through decantation but muddy or fine sediments may decant with the organic matter and impede the complete sediment removal. The sample processing workflow is shown in **Figure 2**.


#### Homogenization and DNA Extraction (2 h, Overnight and 3 h)

The biomass of the decanted organic material may greatly differ among samples, which predetermines subsequent sample pre-processing and DNA extraction procedures. Large amounts of organic material recovered (i.e., the recovered material contains macroinvertebrates and lots of organic matter or bigsized organisms) are followed by Blender homogenization and DNA extraction using the PowerMax Soil DNA Isolation Kit; conversely, samples with a range of recovered biomass from 10 to 200 mg (i.e., the recovered material contains animals for the most part) are processed using Mortar homogenization followed by DNA extraction using the PowerSoil DNA Isolation Kit (see **Figure 2** for schematic representation of the workflow).

#### Blender Homogenization


#### Mortar Homogenization


the power bead solution and incubating samples in a shaking incubator overnight at 56 ◦C (Leray and Knowlton, 2015).

#### DNA Overall Quality Assessment, Purification and Normalization (3 h)


#### ANTICIPATED RESULTS

The protocol described here provides guidelines to resolve the first steps needed for metabarcoding-based benthic macroinvertebrate community assessment: sample collection and preservation, processing, and extraction of representative. DNA of good quality and integrity. The standardization of these three steps is crucial to further obtaining accurate taxonomic inferences from metabarcoding data.

FIGURE 3 | DNA integrity of 8 environmental samples processed as described in the present protocol. DNA extraction was performed using the PowerMax Soil DNA Isolation Kit. HyperLadderTM1 kbp.

Macroinvertebrate samples used for benthic monitoring can occur in different types of sediment (coarse, medium and fine sands, and muds), and contain organisms of heterogeneous size (from 1 mm to several cm) and nature (soft or containing hard, shell, or spiny calcium carbonate exoskeleton, gelatinous, etc.), which implies that DNA extraction may not be equally efficient for all types of sediment or organismal types. Our protocol is based on large sediment volumes (>100 ml) to ensure that all organisms are present, preserved in appropriate conditions to prevent DNA degradation, that are mortar or blender beaten to ensure breaking of hard exoskeletons.

DNA extracted from complex environmental samples need to be representative and of good quality and integrity. The steps presented here ensure both (i) macroinvertebrate community representation by homogenizing samples from which subsamples are taken before DNA extraction and (ii) good quality and integrity DNA by utilizing kits-based extraction protocols specifically designed for isolating high-quality environmental DNA from soil or sediment. The procedures described in the present protocol for decantation, homogenization, and DNA extraction have been recently applied to sediment samples from estuarine and coastal locations with different level of anthropogenic pressures. The DNA extracted from each environmental sample was amplified following the protocol for amplicon library preparation and sequencing (Bourlat et al., 2016) and the resulting reads analyzed using the pipeline for bioinformatics analysis of metabarcoding data (Aylagas and Rodríguez-Ezpeleta, 2016). Using the retrieved macroinvertebrate taxonomic list from each sample, the marine biotic index AMBI (Borja et al., 2000) was calculated, showing comparable results to that inferred using morphological species identification from samples of the same locations (Aylagas et al., in preparation). Thus, the promising results obtained using the present protocol for environmental biomonitoring contributes to accelerating the implementation of metabarcoding for environmental status assessment.

Finally, in response to the necessity of more costeffective approaches than the traditional morphological species identification, the present protocol followed by DNA amplification coupled with HTS proves to be a suitable cheaper alternative for biodiversity assessment. Although several procedures involving less sample manipulation prior DNA extraction are well-established for small metazoans metabarcoding studies (Guardiola et al., 2015; Lejzerowicz et al., 2015; Pearman et al., 2016), these approaches cannot be accommodated for macroinvertebrates. In this context, the standardization of the sample pre-processing through mechanical enrichment and homogenization before DNA extraction will ensure the reproducibility of the results and may help to the establishment of macroinvertebrates metabarcoding for environmental biomonitoring.

# NOTES

# Note 1. Recommendations to Prevent Cross-Contamination

DNA-based approach to characterize metazoan communities is very sensitive to contamination. Avoiding cross-contamination is essential to ensure the success of DNA metabarcoding-based biodiversity studies. During sample collection, decantation and homogenization steps, material (sieves, graduated cylinders, blender jar, mortar, and tweezers) must be cleaned between samples by soaking in 10% bleach for a minimum of 5 min and gently rinsing with deionized water. Finally, these recommendations must be followed:


# Note 2. Environmental Sample Preservation for DNA-Based Studies

DNA degradation is critical for metabarcoding marine benthic community assessment. In this sense, the detection of some of the species present in an environmental sample may be reduced if DNA integrity has been altered. The process of DNA degradation starts at the moment an organism dies, when cell membranes break and allow entrance of bacteria and other threats with the subsequent release of DNAses that degrade DNA. Thus, avoiding DNA degradation requires storing the sample as soon as collected in appropriate preserving agents (ethanol or other reagents such as RNA later) that prevent DNAse activity (Rodriguez-Ezpeleta et al., 2013). Although formalin has traditionally been used to store marine benthic organism samples, as it preserves morphological structure and allows visual identification, it is toxic and degrades DNA (Serth et al., 2000); thus, ethanol 96% is recommended to preserve samples for molecular studies (Stein et al., 2013).

#### Note 3. Safe Stopping Points


#### Note 4. Subsample Representativeness

Homogenization is performed in order to solve the problem of representativeness issues in large volume samples from which the whole macroinvertebrate community is aimed to be characterized. The best community characterization using DNAbased approaches would require the DNA extraction of the total sample; yet, this cannot be achieved in a reasonable time and commercial kits are not designed for samples up to 10 g. Therefore, a good homogenization step is crucial to ensure the representativeness of the whole community in a subsample. However, we recommend performing two DNA extractions on two subsamples from the homogenized sample to further guarantee a reliable representation of the whole community. In order to ease following steps of the protocol, the DNA replicates are pooled and purified prior amplicon library preparation. Finally, one of the issues related with metabarcoding of different size organisms (from 1 mm to several cm) is the homogenization of exceptionally large specimens with the remaining sample. The DNA of large organisms may mask the presence of other biota in the sample, which may lead to false negative results. In this case,

#### REFERENCES


body parts from large specimens can be subsampled or set aside for standard DNA barcoding.

# Note 5. Recommendation to Avoid Inhibition Issues Related to Humic Substances

Even though DNA extraction kits used in this protocol are appropriate to remove humic substances, applying cleaning columns further removes other potential PCR inhibitors such as calcium carbonates, silicates, proteins, and algal polysaccharides.

#### AUTHOR CONTRIBUTIONS

Conceived and designed the protocol: EA and NR. Developed and performed the protocol: EA and IM. Wrote the first draft of the protocol: EA and NR. All authors contributed equally in writing the last version of the protocol.

### FUNDING

This work was funded by the European Union (7th Framework Program "The Ocean of Tomorrow" Theme, grant agreement no. 308392) through the DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status—http://www.devotes-project.eu) project and by the Basque Water Agency (URA) through a Convention with AZTI. EA is supported by a doctoral grant from Fundación Centros Tecnológicos—IG.

#### ACKNOWLEDGMENTS

The authors would like to thank Dr. Maria C. Uyarra for kindly advising us on some details on the manuscript. This paper is contribution number 776 from AZTI (Marine Research Division).

MiSeq: the dual-PCR method," in Marine Genomics Methods and Protocols, Methods in Molecular Biology, ed S. J. Bourlat (New York, NY: Springer), 1452.


pitfalls and promises. Mol. Ecol. 19(Suppl.1), 4–20. doi: 10.1111/j.1365- 294X.2009.04473.x


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declared a collaboration with the authors and states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Aylagas, Mendibil, Borja and Rodríguez-Ezpeleta. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Benchmarking DNA Metabarcoding for Biodiversity-Based Monitoring and Assessment

#### Eva Aylagas <sup>1</sup> , Ángel Borja<sup>1</sup> , Xabier Irigoien<sup>2</sup> and Naiara Rodríguez-Ezpeleta<sup>1</sup> \*

*<sup>1</sup> Marine Research Division, AZTI-Tecnalia, Sukarrieta, Spain, <sup>2</sup> Red Sea Research Center, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia*

Characterization of biodiversity has been extensively used to confidently monitor and assess environmental status. Yet, visual morphology, traditionally and widely used for species identification in coastal and marine ecosystem communities, is tedious and entails limitations. Metabarcoding coupled with high-throughput sequencing (HTS) represents an alternative to rapidly, accurately, and cost-effectively analyze thousands of environmental samples simultaneously, and this method is increasingly used to characterize the metazoan taxonomic composition of a wide variety of environments. However, a comprehensive study benchmarking visual and metabarcoding-based taxonomic inferences that validates this technique for environmental monitoring is still lacking. Here, we compare taxonomic inferences of benthic macroinvertebrate samples of known taxonomic composition obtained using alternative metabarcoding protocols based on a combination of different DNA sources, barcodes of the mitochondrial cytochrome oxidase I gene and amplification conditions. Our results highlight the influence of the metabarcoding protocol in the obtained taxonomic composition and suggest the better performance of an alternative 313 bp length barcode to the traditionally 658 bp length one used for metazoan metabarcoding. Additionally, we show that a biotic index inferred from the list of macroinvertebrate taxa obtained using DNA-based taxonomic assignments is comparable to that inferred using morphological identification. Thus, our analyses prove metabarcoding valid for environmental status assessment and will contribute to accelerating the implementation of this technique to regular monitoring programs.

Edited by: *Michael Elliott, University of Hull, UK*

#### Reviewed by:

*Katherine Dafforn, University of New South Wales, Australia José Lino Vieira De Oliveira Costa, Centro de Oceanografia, Portugal*

> \*Correspondence: *Naiara Rodríguez-Ezpeleta nrodriguez@azti.es*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *15 April 2016* Accepted: *30 May 2016* Published: *10 June 2016*

#### Citation:

*Aylagas E, Borja A, Irigoien X and Rodriguez-Ezpeleta N (2016) Benchmarking DNA Metabarcoding for Biodiversity-Based Monitoring and Assessment. Front. Mar. Sci. 3:96. doi: 10.3389/fmars.2016.00096* Keywords: Illumina MiSeq, COI barcodes, extracellular DNA, AMBI, biotic indices, macroinvertebrates

# INTRODUCTION

Environmental biomonitoring in coastal and marine ecosystems often relies on comprehensively, accurately, and repeatedly characterizing the benthic macroinvertebrate community (Yu et al., 2012). These organisms are considered a good indicator of ecosystem health and have demonstrated a rapid response to a range of natural and anthropogenic pressures (Johnston and Roberts, 2009). As a result, the macroinvertebrate community has been largely used to develop biotic indices (Diaz et al., 2004; Pinto et al., 2009; Borja et al., 2015), such as the AZTI's Marine Biotic Index (AMBI; Borja et al., 2000), used worldwide to assess the marine benthic status (Borja et al., 2015). Nevertheless, biomonitoring based upon benthic organisms has limitations because species identification requires extensive taxonomic expertise and it is time-consuming, expensive, and laborious (Yu et al., 2012; Wood et al., 2013; Aylagas et al., 2014). The rapid development of high-throughput sequencing (HTS) technologies represents a promising opportunity for easing the implementation of molecular approaches for biomonitoring programs (Bourlat et al., 2013; Dowle et al., 2015). In particular, DNA metabarcoding (Taberlet et al., 2012a) allows the rapid and cost-effective identification of the entire taxonomic composition of thousands of samples simultaneously (Zepeda Mendoza et al., 2015) and the ability to provide a more comprehensive community analysis than traditional assessments (Dafforn et al., 2014), which can enable the calculation of benthic indices in a much faster and accurate way compared to morphological methodologies.

Metabarcoding consists of simultaneously amplifying a standardized DNA fragment specific for a species (barcode) from the total DNA extracted from an environmental sample using conserved short DNA sequences flanking the barcode (primers; Hajibabaei, 2012; Cristescu, 2014). The obtained barcodes are then high-throughput sequenced and compared to a previously generated DNA sequence reference database from well-characterized species for taxonomic assignment (Taberlet et al., 2012a). In the case of animals, different barcodes such as portions of the small and large subunits of the nuclear ribosomal RNA (18S and 28S rRNA) genes (Machida and Knowlton, 2012) and of the mitochondrial cytochrome oxidase I (COI; Meusnier et al., 2008) and 16S rRNA genes (Sarri et al., 2014) have been proposed for metabarcoding. The COI gene is by far the most commonly used marker for metazoan metabarcoding (Ratnasingham and Hebert, 2013), for which thousands of reference sequences are available in public databases [the Barcode of Life Database (BOLD) contains >1,000,000 COI sequences belonging to animal species] and several amplification primers have been designed [more than 400 COI primers are published in the Consortium for the Barcode of Life (CBOL) primer database].

Several studies have used metabarcoding to characterize the metazoan taxonomic composition of aquatic environments (Porazinska et al., 2009; Chariton et al., 2010; Fonseca et al., 2014; Dell'Anno et al., 2015; Leray and Knowlton, 2015; Chain et al., 2016), and an increasing number of studies have directly applied the approach for environmental biomonitoring purposes (Ji et al., 2013; Dafforn et al., 2014; Pawlowski et al., 2014; Chariton et al., 2015; Gibson et al., 2015; Pochon et al., 2015; Zaiko et al., 2015). Initial studies inferring biotic indices from molecular data show the potential of metabarcoding for evaluating aquatic ecosystem quality (Lejzerowicz et al., 2015; Visco et al., 2015). However, before implementation of metabarcoding in regular biomonitoring programs, this approach needs to be benchmarked against morphological identification so that accurate taxonomic inferences and derived biotic indices can be ensured (Aylagas et al., 2014; Carugati et al., 2015). The accuracy of metabarcoding-based taxonomic inferences relies on the retrieval of a wide range of taxonomic groups from a given environmental sample using the appropriate barcode, primers, and amplification conditions (Deagle et al., 2014; Kress et al., 2015), and on the completeness of the reference database (Zepeda Mendoza et al., 2015). Some attempts have been performed to compare morphological vs. metabarcodingbased taxonomic inferences; yet, results are inconclusive as some studies do not apply both approaches to the same sample and/or have focused on a particular taxonomic group (Hajibabaei et al., 2012; Carew et al., 2013; Zhou et al., 2013; Gibson et al., 2014; Cowart et al., 2015; Zimmermann et al., 2015). A recent study (Gibson et al., 2015) has performed morphological and metabarcoding-based taxonomic identification on the same freshwater aquatic invertebrate samples, but limited their visual identifications to family level. Only two studies (Dowle et al., 2015; Elbrecht and Leese, 2015) have performed a robust benchmarking of metabarcoding using freshwater invertebrates and showed that this technique can be successfully applied to biodiversity assessment. In marine metazoans, all studies have focused only on plankton samples (Brown et al., 2015; Mohrbeck et al., 2015; Albaina et al., 2016). Thus, an exhaustive evaluation of metabarcoding for marine benthic metazoan taxonomic inferences is still lacking.

The use of extracellular DNA (the DNA released from cell lysis; Taberlet et al., 2012b) for biodiversity monitoring is increasingly applied to water (e.g., Ficetola et al., 2008; Foote et al., 2012; Thomsen et al., 2012; Kelly et al., 2014; Davy et al., 2015; Valentini et al., 2016), soil (Taberlet et al., 2012b), and sediment samples (Guardiola et al., 2015; Turner et al., 2015; Pearman et al., 2016). Constituting a significant fraction of the total DNA (Dell'Anno and Danovaro, 2005; Pietramellara et al., 2009; Torti et al., 2015), it is assumed that the taxonomic composition of the free DNA present in the environment reflects the biodiversity of the sample (Ficetola et al., 2008), which would simplify DNA extraction protocols (Pearman et al., 2016) and allow the detection of organisms that are even larger than the sample itself (Foote et al., 2012; Thomsen et al., 2012; Kelly et al., 2014; Davy et al., 2015). Thus, this method appears as a promising costeffective alternative for macroinvertebrate diversity monitoring, but no robust evidence that the entire macroinvertebrate community can be detected using extracellular DNA exists so far.

The lack of a thorough comparison between morphological and metabarcoding-based taxonomic inferences of marine metazoa and of an evaluation of the use of metabarcoding for marine biotic index estimations prevents the application of metabarcoding in routine biomonitoring programs. Here, we benchmark alternative metabarcoding protocols based on a combination of different DNA sources (extracellular DNA and DNA extracted from previously isolated organisms), barcodes (short and long COI regions), and amplification conditions against benthic macroinvertebrate samples of known taxonomic composition. Additionally, we test the effect of the discrepancies between morphological and DNA-based taxonomic inferences in marine biomonitoring through the evaluation of the molecular based taxonomies performance when incorporated for the calculation of the AMBI and prove the suitability of molecular data based biotic indices to assess marine environmental status.

# METHODS

The experimental design followed to compare the performance of molecular and morphological based taxonomic inferences is summarized in **Figure 1**.

# Sample Collection and Processing

Benthic samples were collected from 11 littoral stations (sampling depth ranging from 100 to 740 m) along the Basque Coast, Bay of Biscay (Supplementary Figure 1), during March 2013, using a van Veen grab (0.07–0.1 m<sup>2</sup> ). At each location, after sediment homogenization, one subsample of sediment was taken from the surficial layer of the grab and stored in a sterile 15 ml falcon tube at −80 ◦C until extracellular DNA extraction (see below). In order to collect the benthic macroinvertebrate community (organism size >1 mm) present in each sample, the remaining sediment was sieved on site through a 1 mm size mesh, and the retained material preserved in 96% ethanol at 4 ◦C until processing (<6 months). Macroinvertebrate specimens were sorted and identified to the lowest possible taxonomic level based on morphology. Following taxonomic classification, each sample was divided into two identical subsamples by taking equal amount of tissue per taxa for each subsample. Tissues from one subsample were pooled and used for bulk DNA extraction. Each tissue of the second subsample was used for individual DNA extraction (see below).

# Extracellular, Individual, and Bulk DNA Extraction

Extracellular DNA was extracted following an optimized protocol (Taberlet et al., 2012b). Briefly, 5 g of each sediment sample were mixed with 7.5 ml of saturated phosphate buffer and an equal volume of chloroform:isoamyl alcohol (IAA). After centrifugation for 5 min at 4,000 g, the aqueous phase was passed through a second round of chloroform:IAA purification and ethanol precipitated before elution of resulting DNA pellet in 100 µl Milli-Q water. For individual and bulk processing, total genomic DNA from each tissue and from the mix of tissues composing each sample, respectively, were extracted using the Wizard <sup>R</sup> Genomic DNA Purification kit (Promega, WI, USA) in a 125 µl of Milli-Q water final elution. The possible presence of PCR inhibitors in the bulk and extracellular DNA were removed using the Mobio PowerClean <sup>R</sup> DNA Clean-Up Kit. Genomic DNA integrity was assessed by electrophoresis, migrating about 100 ng of GelRedTM-stained DNA on an agarose 1.0% gel, DNA purity was assessed using the Nanodrop <sup>R</sup> ND-1000 (Thermo Scientific) system and DNA concentration was determined with the Quant-iT dsDNA HS assay kit using a Qubit <sup>R</sup> 2.0 Fluorometer (Life Technologies). About 20 ng of each individually extracted DNA were used for DNA barcoding of single species (see details below). Subsequently, 5 µl of each individually extracted DNA at original concentration were pooled (hereafter referred as "pooled DNA"). Extracellular, bulk, and pooled DNA were used for PCR amplification and sequencing (see below).

# Individual PCR Amplification and Sanger Sequencing

Individual DNA barcoding was performed for the species for which no COI barcode was available in public databases (see **Table 1**, Supplementary Material). The standard 658 bp COI barcode (folCOI) was targeted using the dgLCO1490 × dgHCO2198 primer pair (Meyer, 2003). Each individual DNA sample was amplified in a total volume reaction of 20 µl using 10 µl of Phusion <sup>R</sup> High-Fidelity PCR Master Mix (Thermo Scientific), 0.2 µl of each primer (10 µM), and 20 ng of genomic DNA. The thermocycling profile consisted of an initial 30 s denaturation step at 98 ◦C, followed by up to 35 cycles of 10 s at 98 ◦C, 30 s at 48 ◦C, and 45 s at 72 ◦C, and a final 5 min extension step at 72 ◦C. PCR products were considered positive when a clear single band of expected size was visualized on a 1.7% agarose gel. Samples with negative product were further amplified with the mlCOIintF × dgHCO2198 primer pair (Leray et al., 2013) targeting a 313 bp fragment of the COI gene (mlCOI). Negative samples were included with each PCR run as external control. PCR products were purified with ExoSAP-IT (Affymetrix) and Sanger sequenced.

# PCR Amplification for Library Preparation and Illumina Miseq Sequencing

Indexed paired-end libraries of pooled amplicons were prepared using two nested PCRs from the extracellular, bulk and pooled (mix of 5 µl of individually extracted DNA at original concentration) DNA obtained from each of the 11 collected samples. In parallel, three of the samples were processed per triplicate and considered independently in downstream analysis. For the first PCR, two universal primer pairs with overhang Illumina adapters were used to amplify two different length COI barcodes (the mlCOI and the folCOI). Three different PCR profiles were used to amplify each COI barcode from the bulk and pooled DNAs (46 and 50 ◦C annealing temperatures and a touchdown profile), whilst the extracellular DNA COI barcodes were amplified with 46 ◦C annealing temperature. PCRs were performed in a total volume of 20 µl using 10 µl of Phusion <sup>R</sup> High-Fidelity PCR Master Mix (Thermo Scientific), 0.5 µl of each primer (10 µM), and 2 µl of genomic DNA (5 ng/µl). The PCR conditions for the two different annealing temperatures consisted on an initial 30 s denaturation step at 98 ◦C, 27 cycles of 10 s at 98 ◦C, 30 s at 46 or 50 ◦C, and 45 s at 72 ◦C, and a final 5 min extension at 72 ◦C. For the touchdown profile the PCR conditions consisted on an initial 30 s denaturation step at 98 ◦C, 16 cycles of 10 s at 98 ◦C, 30 s at 62 ◦C (−1 ◦C per cycle), and 60 s at 72 ◦C, followed by 17 cycles at 46 ◦C annealing temperature, and a final 5 min extension at 72 ◦C (Leray et al., 2013). Negative controls were included with each PCR. Generated amplicons were purified with AMPure XP beads (Beckman Coulter), eluted in 50 µL MilliQ water and used as templates for the generation of the dual-indexed amplicons in the second PCR round following the "16S Metagenomic Sequencing Library Preparation" protocol (Illumina). Purified PCR products were quantified using the Quant-iT dsDNA HS assay kit using a Qubit <sup>R</sup> 2.0 Fluorometer (Life Technologies) and further normalized for all samples. Pools

TABLE 1 | Results from the regression model between traditional and molecularly inferred pa-AMBI values.


\**Significant correlations (P* < *0.05), TD: touchdown PCR profile.*

of 96 equal concentration amplicons were sequenced using the 2 × 300 paired-end on a MiSeq (Illumina).

#### DNA Barcode Reference Database

Trace files of Sanger sequences obtained from individual PCR amplifications were edited and trimmed to remove low quality bases (Q < 30) using SeqTrace 0.9.0 (Stucky, 2012) and checked for frame shifts using EXPASY (Gasteiger et al., 2003). COI sequences are available in "BCAS project" at BOLD (http://www.boldsystems.org) and in GenBank (accession numbers KT307619–KT307707). To generate our DNA reference database, we retrieved a total of 1,123,601 public COI aligned sequences from 96,641 different taxa from BOLD (October 2014), including the sequences generated in this study (COI RefSeq). After removing duplicates, a total of 505,033 sequences were kept and trimmed to the 658 bp Folmer COI fragment to generate the "BOLD database." A smaller customized DNA reference database was generated using the 4231 sequences corresponding to species included in the AMBI list (see below; available at http://ambi.azti.es) extracted from the "BOLD database" to build the "AMBI database." For the analyses of the folCOI reads, the 249 bp not sequenced internal fragment (see below) was removed from these two databases to construct the "BOLD gapped database" and the "AMBI gapped database." The four resulting databases were formatted according to mothur (Schloss, 2009) standards.

#### Amplicon Sequence Analysis

Demultiplexed reads were quality checked using FastQC (Andrews, 2010) and primer sequences removed using Trimmomatic 0.33 (Bolger et al., 2014). Since the mlCOI paired-end reads overlap in 237 bp and the folCOI paired-end reads do not overlap, different preprocessing steps are needed for each COI fragment. Forward and reverse mlCOI reads were merged using FLASH (Magoc and Salzberg, 2011 ˇ ) with a minimum and maximum overlap of, respectively, 20 bases below and above the expected overlapping region, and the resulting reads were trimmed using Trimmomatic at the first sliding window of 50 bp with an average quality score below 30. The folCOI forward and reverse reads were trimmed at 260 and 200 bp, respectively, based on the quality decrease after these positions observed on FastQC plots. Each pair of forward and reverse-complemented reverse read was pasted to create a 409 bp read that corresponds to the folCOI barcode without a 249 bp internal fragment. Further details on this new pipeline developed to analyze the universal 658 bp COI barcode which is too long for most HTS applications such as the Illumina MiSeq are detailed elsewhere (Aylagas and Rodríguez-Ezpeleta, 2016). Preprocessed reads from both barcodes were independently analyzed with mothur following the MiSeq standard operating procedure (Kozich et al., 2013). Briefly, sequences with ambiguous bases were discarded and the rest, aligned to the corresponding BOLD and AMBI reference databases. Only those mlCOI and folCOI reads aligning inside the barcode region and longer than 200 and 300 bp, respectively, were kept. After chimera removal using the de novo mode of UCHIME (Edgar et al., 2011), sequences were grouped into phylotypes according to the taxonomic assignments made based on the Wang method (Wang et al., 2007) using a bootstrap value of 90. The sequences that did not return any taxonomic assignment against the BOLD database were blasted against the NCBI non redundant database. Sequences have been deposited in the Dryad Digital Repository (http://dx.doi.org/10. 5061/dryad.0sc0s).

### Comparison of Morphological and Metabarcoding-Based Taxonomic Compositions

Only taxa representing at least 0.01% of the reads in one station were considered present in the taxonomic composition inferred from molecular data. An in-house script (Supplementary Figure 2) was used to calculate the degree of match between the molecular and morphologically inferred taxonomic compositions of each station. The detection success was normalized for each sample and transformed to percentage of matches (100% of matches means all taxa identified based on morphology have been detected using DNA-based approaches). Differences in mean values of the taxa detection percentages between DNA extraction methods, primers and PCR conditions were examined using a t-test at alpha = 0.05. Patterns of sample dissimilarity were visualized using non-metric multidimensional scaling (nMDS) based on taxa presence/absence and abundance using the Jaccard and Bray-Curtis indices, respectively, obtained using molecular approaches.

# Comparison of Morphological and Metabarcoding-Based Biotic Indices

In order to compare morphological and metabarcoding-based biotic indices, we used AMBI, which is a status assessment index based on the pollution tolerances of the taxa present in a sample, with tolerance being expressed categorically into ecological groups (EGI, sensitive to pressure; EGII, indifferent; EGIII, tolerant; EGIV, opportunist of second order; and EGV, opportunist of first order). We calculated the presence/absence morphology-based AMBI (pa-AMBI) and the presence/absence genetics-based AMBI (pa-gAMBI; Aylagas et al., 2014) inferred through DNA metabarcoding of each sample, using the AMBI 5.0 software (http://ambi.azti.es). The relationships among pa-AMBI and pa-gAMBI values were examined using standardized major axis (SMA) estimation (Warton et al., 2006) using the software SMATR (Falster et al., 2003). In order to evaluate the performance of pa-gAMBI for each condition, root-mean-square error (RMSE) and bias were calculated (Walther and Moore, 2005).

# RESULTS

#### Morphological and Molecular Analysis

In total, 138 macroinvertebrate taxa belonging to nine different phyla were morphologically identified in the 11 stations. Representatives of two main phyla, Annelida, and Arthropoda, are present at all stations, with 94 and 21 taxa, respectively, whereas less represented phyla (Mollusca, Chaetognata, Cnidaria, Echinodermata, Nemertea, Nematoda, and Sipuncula) are absent from some stations and include less number of taxa (Supplementary Table 1). Individual DNA barcoding was successful on 61 and 24 of the 106 identified species with no COI barcode in public databases, for which new folCOI and mlCOI barcodes were generated, respectively, and included in the reference database. Despite this effort to increase the reference database, 21 species remain without barcode because amplification of both barcodes failed.

For each station, two condition combinations were tested for the extracellular DNA (two different barcodes) and six for the bulk and pooled DNAs (two different barcodes and three different PCR profiles). From the 238 samples analyzed, including triplicates performed on three of the stations, 14 had no PCR amplification (see Supplementary Table 2 for clarification on the number of samples produced for molecular analysis). The 224 remaining samples resulted in 16 million reads, from which about 56% passed quality filters and were used for taxonomic analysis (Supplementary Table 2). Of the total reads obtained from extracellular DNA, 71.5 and 73.4% could not be assigned to any metazoan phylum using the customized BOLD database and 24.9 and 25.6% were not assigned to Metazoa for mlCOI and folCOI, respectively. When blasted against NCBI, the reads obtained using mlCOI matched with bacteria (0.6%), non-metazoan eukaryotes (84%), metazoans (12.2%), or did not provide any match (3%), and the reads obtained using folCOI matched with bacteria (66.6%), nonmetazoan eukaryotes (6%), metazoans (4.2%), archaea (0.05%), or did not provide any match (23.2%). The percentages of nonmetazoan reads are much lower for bulk (0.03 and 0.04%) and pooled DNA (0.1 and 0.3%), and the proportion of Metazoa reads with no phylum assigned are lower for mlCOI (23.2 and 10.6% for bulk and pooled DNA, respectively) than for folCOI (29.94 and 31.6% for bulk and pooled DNA, respectively).

# Comparison of Morphological and Molecular-Based Taxonomic Compositions

From the taxonomic inferences obtained using molecular approaches, only macroinvertebrates were considered for sample comparison (e.g., Chordata records were excluded for downstream analysis). The average percentage of recovered taxa (molecular taxonomy matches visual taxonomy) over all stations using different conditions is shown in **Figure 2** (see Supplementary Figure 3 for percentage of recovered taxa considering only species level identification). Matches for taxonomic inferences based on metabarcoding of extracellular DNA are very low (3.4 and 3.1% for folCOI and mlCOI respectively), with only taxa from three phyla (Mollusca, Annelida, and Nemertea) retrieved (Supplementary Table 3). Results obtained between replicates from the same sample reveal similar taxonomic inferences. No significant differences were observed between the percentage of matches obtained using bulk and pooled DNA (p > 0.05). Interestingly, the mlCOI barcode outperforms the folCOI barcode (p < 0.05 for bulk and pooled DNA) and, within the mlCOI, the 46 and 50 ◦C annealing temperatures outperform the touchdown profile both for bulk and pooled DNA (p < 0.05). Overall, the best performing condition is the mlCOI barcode amplified using 46 ◦C annealing temperature, which results in a percentage of recovered taxa of 62.4% for all matches and of 76.3% for only matches at species level.

Using molecular approaches we were able to retrieve taxa that had not been morphologically identified. Representatives of Annelida (e.g., Tubificoides amplivasatus, Chloeia parva, and Mugga wahrbergi), Arthropoda (e.g., Scyllarus arctus and Limnoria sp.), Mollusca (e.g., Nucula nucleus, Galeomma turtoni, Thyasira ferruginea, and Entalina tetragona), and Echinodermata (e.g., Ophiura albida and Macrophiothrix sp.) were solely identified using DNA-based approaches. Moreover, we were able to find taxa belonging to two phyla that were not morphologically identified even at phylum level: two families (Triaenophoridae and Echinobothriidae) and one order (Acoeala) of Platyhelminthes and one family (Hemiasterellidae) of Porifera. As illustrated by the nMDS ordination plot of beta diversity (**Figure 3**), the greatest disparity in macroinvertebrate composition inferred using molecular taxonomy of each station was shown by the extracellular DNA approach.

# Comparison of Morphological and Metabarcoding-Based Biotic Indices

The correlation between pa-AMBI and pa-gAMBI values obtained from the taxonomic composition inferences using the AMBI database is shown in **Figure 4**. The pa-AMBI values that best correlate with pa-gAMBI values are those obtained using bulk and pooled DNA approaches at 46 or 50 ◦C annealing temperatures obtained with mlCOI (**Table 1**). Generally, pagAMBI values tend to score lower than pa-AMBI values (negative

bias over all stations). This tendency can be also observed in the variation of the percentage of taxa found belonging to each ecological group obtained using morphological and molecular taxonomic identifications (Supplementary Figure 4). The nondetection of taxa belonging to tolerant and opportunistic ecological groups (III, IV, and V) when using folCOI, especially for pooled DNA method, leads to poor correlations between pa-AMBI and pa-gAMBI values.

# DISCUSSION

### Effect of PCR-Based Analysis Biases on Taxonomic Inferences

Finding the primer pair and PCR conditions that most accurately recover the organisms present in an environmental sample is crucial for a successful application of metabarcoding to biomonitoring. Several studies analyzing the same samples with morphological and molecular taxonomy have been performed so far to benchmark COI based metabarcoding in animals, all focusing exclusively on freshwater or terrestrial macroinvertebrates (Hajibabaei et al., 2012; Carew et al., 2013; Gibson et al., 2014; Dowle et al., 2015; Elbrecht and Leese, 2015) or carried out under morphological identifications limited to high taxonomic levels (Gibson et al., 2015). Thus, studies on marine benthic communities that prove the suitability of

DNA-based approaches for environmental biomonitoring are lacking. Using samples of known taxonomic composition, we show that an alternative barcode that targets a shorter region of the COI gene outperforms the 658 bp region that is commonly used for metabarcoding metazoans (Carew et al., 2013; Ji et al., 2013; Dowle et al., 2015; Elbrecht and Leese, 2015; Zaiko et al., 2015). Our data corroborate previous studies unveiling the lack of universality in the COI primers, which is translated to biases during PCR step (Pochon et al., 2013; Deagle et al., 2014). However, the increased performance of the short region, previously demonstrated for individual barcoding on marine metazoans (Leray et al., 2013) and metabarcoding in insects (Brandon-Mong et al., 2015) proves that the mlCOI barcode retrieves a high proportion of the morphologically identified taxa. This fact also corroborates the preferred use of small barcodes for metabarcoding, which provide pair-end overlaps on Illumina sequencing and good taxonomic resolution for species identification (Meusnier et al., 2008). Additionally, the folCOI barcode returns more reads with no match and metazoan reads not assigned to any specific phylum, which could be attributed to the fact that longer barcodes can accumulate more errors

and *folCOI*).

or 50 ◦C annealing temperatures or Touchdown profile) displayed separately for each barcode—*mlCOI* (top 3 rows) and *folCOI* (bottom 3 rows). Each dot shows the relationship between the pa-AMBI (x-axis) and pa-gAMBI value (y-axis) for each station. The dotted lines represent the results of model II regression and the diagonal showing perfect correlation between the two observations is depicted.

during the PCR and sequencing processes (Schirmer et al., 2015).

The effect of the PCR annealing temperature has been shown to affect retrieved taxonomic composition in bacterial and archaeal metabarcoding using the 16S rRNA gene (Sipos et al., 2007; Lee et al., 2012; Pinto and Raskin, 2012). Here, we show that the use of inappropriate PCR conditions can also affect the final taxonomic assignment in metazoan metabarcoding analyses. Our results show that a constant low annealing temperature (46 or 50 ◦C) provides more accurate taxonomic inferences compared to the touchdown profile, which contrasts with previous studies (Hansen et al., 1998; Simpson et al., 2000; Leray et al., 2013). Moreover, it is well-established that the more PCR cycles, the more spurious sequences and chimera are formed during PCR (Haas et al., 2011), which could explain the lower taxa detection rate when using the touchdown profile (which includes five more cycles). Further, the nature of the organisms and their size may bias DNA extraction (i.e., hard shells or chitin exoskeleton can prevent cell lysis and DNA from small organisms can be less effectively extracted). Here, we have ensured that DNA from all organisms is present in the pooled sample by pooling individually extracted DNAs, and show that the results of the pooled DNA and bulk extracted DNA are comparable.

# The Use of Extracellular DNA for Biodiversity Estimations

The extracellular DNA-based metabarcoding for biodiversity assessments has the potential of detecting big-size organisms in small samples, which facilitates sampling strategies and could resulting in a more cost-effective approach for environmental biomonitoring (Taberlet et al., 2012b; Thomsen et al., 2012; Thomsen and Willerslev, 2015). Several studies have used extracellular DNA from the water column to detect vertebrates (Ficetola et al., 2008; Thomsen et al., 2012; Valentini et al., 2016) freshwater macroinvertebrates (Goldberg et al., 2013; Mächler et al., 2014) and benthic eukaryotes (Guardiola et al., 2015; Pearman et al., 2016). Yet, so far, this approach has not been proved valid for biodiversity assessment as no comparison with samples of known taxonomic composition has been performed. To our knowledge, only one attempt exists to detect the whole freshwater benthic macroinvertebrate community from extracellular DNA extracted from samples of known composition (Hajibabaei et al., 2012), but the authors used the preservative ethanol as controlled environment containing the free DNA rather than natural scenarios. In our analyses, only a small proportion of the taxa identified using morphological methods are retrieved using extracellular DNA present in the sediment. Indeed, even considering the taxa not identified through morphological taxonomy, the extracellular DNA-based analyses only identify 30 macroinvertebrate taxa over all stations, which is much lower than the total diversity inferred from morphology and from DNA extracted from the isolated organisms. Therefore, the striking differences obtained between morphological and extracellular DNA metabarcoding based taxonomic inferences suggest that further studies are needed before using sediment extracellular DNA as a suitable source for macroinvertebrate biodiversity assessment; yet, more experiments testing the effect of sediment sample size, DNA degradation scenarios, or DNA extraction protocols are required, as it is possible that sampling more deeply in the sediment, or using the water column provides better results, and/or that the optimal DNA extraction procedure has not been employed (Corinaldesi et al., 2005).

# Effect Misinterpreting Community Composition in Environmental Biomonitoring

Environmental biomonitoring programs rely on the detection of a wide range of taxonomic groups, which are usually amplified using universal primers (Leray et al., 2013). The abovementioned biases inherent to PCR-based analyses can lead to greater recovery of sequences of some species and the exclusion of others (Elbrecht and Leese, 2015; Piñol et al., 2015). Thus, it is important to see whether in samples containing species from numerous phyla, metabarcoding is also able to retrieve a high proportion of taxa that suffices for environmental monitoring. In general, we show a high percentage of recovery using bulk DNA among the nine different phyla identified using morphological approach. However, in our metabarcoding analyses, some taxa identified using morphological methodologies remain undetected using both short and long COI barcodes, whereas others appear only using metabarcoding. The species exclusively detected using metabarcoding represent potential cryptic species (e.g., Tyasira flexuosa/Thyasira ferruginea and Ophiura texturata/Ophiura albida) or unable to be classified based on morphological characters. Further, some additional identified taxa [i.e., two phyla detected from extracellular DNA (Platyhelminthes and Porifera)] may either represent organisms which had been missed by taxonomy based on morphology and metabarcoding from previously isolated organisms due to their small size (<1 mm) or detected due to the fact that the free DNA has been transported from other localities (Roussel et al., 2015).

Consequences of the misinterpretation of the taxonomic composition could result in erroneous biodiversity assessment, which may impede the implementation of DNA metabarcoding in regular biomonitoring programs (Chariton et al., 2015; Cowart et al., 2015; Lejzerowicz et al., 2015; Zaiko et al., 2015). In particular, calculation of biotic indices based on pollution tolerances assigned to the taxa retrieved from the sample (Maurer et al., 1999; Borja et al., 2000) may be affected by the approach used for taxonomic assignment. We show that, despite using the metabarcoding conditions that most accurately detect the morphologically identified taxa, some differences between both approaches are observed. Yet, in general, pagAMBI values obtained from metabarcoding analyses provide significant presence-absence community estimations and can be used for calculating biotic indices.

# CONCLUSIONS

Representing a promising opportunity to overcome the time-consuming and high cost of traditional methodologies for species identification, it is anticipated that DNA metabarcoding will be routinely used in biomonitoring programs in the near future. Yet, the application of this technique to regular biomonitoring programs requires benchmarking and standardization. Here, we demonstrate through an exhaustive study design that, using the appropriate conditions, metabarcoding presents a great potential to characterize biodiversity and to provide accurate biotic indices. Thus, our findings will contribute to accelerating the implementation of metabarcoding for environmental status assessment.

### ADDITIONAL INFORMATION

Accession codes: All Sanger and Illumina generated sequences have been deposited in GenBank (accession numbers KT307619– KT307707) and DRYAD (http://dx.doi.org/10.5061/dryad.0sc0s).

# AUTHOR CONTRIBUTIONS

Conceived and designed the study: EA, AB, and NRE. Performed the experiments: EA. Contributed reagents/materials: XI. Analyzed the data: EA and NRE. Interpreted the data and wrote the paper: EA, AB, and NRE. All authors reviewed the manuscript.

## REFERENCES


#### FUNDING

This manuscript is a result of the DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status http://www.devotes-project.eu) project funded by the European Union (7th Framework Program "The Ocean of Tomorrow" Theme, grant agreement no. 308392) and the Basque Water Agency (URA) through a Convention with AZTI. EA is supported by the "Fundación Centros Tecnológicos" through an "Iñaki Goenaga" doctoral grant.

# ACKNOWLEDGMENTS

We thank Iñaki Mendibil and Craig T. Michell for technical assistance, Iñigo Muxika, Jon Corell, and Germán Rodríguez for discussions and Vega Asensio (www.norarte.es) for preparing **Figure 1**. The specimen taxonomic identification was done by experts from the Cultural Society INSUB. This paper is contribution number 770 from AZTI (Marine Research Division).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00096

and pipelines. Bull. Entomol. Res. 105, 717–727. doi: 10.1017/S0007485315 000681


Sanger and 454-pyrosequenced PCR amplicons. Genome Res. 21, 494–504. doi: 10.1101/gr.112730.110


metabarcoding: assessing the impact of fish farming on benthic foraminifera communities. Mol. Ecol. Resour. 14, 1129–1140. doi: 10.1111/1755-0998. 12261


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Aylagas, Borja, Irigoien and Rodríguez-Ezpeleta. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Historical Data Reveal 30-Year Persistence of Benthic Fauna Associations in Heavily Modified Waterbody

#### Ruth Callaway \*

*College of Science, Department of Biosciences, Swansea University, Wales, UK*

#### Edited by:

*Michael Elliott, University of Hull, UK*

#### Reviewed by:

*Lech Kotwicki, Polish Academy of Sciences, Poland Rodrigo Riera, Atlantic Environmental Marine Center (CIMA SL), Spain Laura Uusitalo, University of Helsinki, Finland Raquel Vaquer-Sunyer, Universitat de les Illes Balears, Spain*

> \*Correspondence: *Ruth Callaway r.m.callaway@swansea.ac.uk*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *12 May 2016* Accepted: *27 July 2016* Published: *12 August 2016*

#### Citation:

*Callaway R (2016) Historical Data Reveal 30-Year Persistence of Benthic Fauna Associations in Heavily Modified Waterbody. Front. Mar. Sci. 3:141. doi: 10.3389/fmars.2016.00141* Baseline surveys form the cornerstone of coastal impact studies where altered conditions, for example through new infrastructure development, are assessed against a temporal reference state. They are snapshots taken before construction. Due to scarcity of relevant data prior to baseline surveys long-term trends can often not be taken into account. Particularly in heavily modified waterbodies this would however be desirable to control for changes in anthropogenic use over time as well as natural ecological variation. Here, the benthic environment of an industrialized embayment was investigated (Swansea Bay, Wales, UK) where it is proposed to build a tidal lagoon that would generate marine renewable energy from the tidal range. Since robust long-term baseline data was not available, the value of unpublished historical benthos information from 1984 by a regional water company was assessed with the aim to improve certainty about the persistence of current benthic community patterns. A survey of 101 positions in 2014 identified spatially discrete benthic communities with areas of high and low diversity. Habitat characteristics including sediment properties and the proximity to a sewage outfall explained 17–35% of the variation in the community structure. Comparing the historical information from 1984 with 2014 revealed striking similarity in the benthic communities between those years, not just in their spatial distribution but also to a large extent in the species composition. The 30-year-old information confirmed spatial boundaries of discrete species associations and pinpointed a similar diversity hotspot. A group of five common species was found to be particularly persistent over time (*Nucula nitidosa*, *Spisula elliptica*, *Spiophanes bombyx*, *Nephtys hombergii*, *Diastylis rathkei*). According to the Infauna Quality Index (IQI) linked to the EU Water Framework Directive (WFD) the average ecological status for 2014 was "moderate," but 11 samples showing "poor" and "bad" status indicated possible negative impacts of dredge spoil disposal. Generally the study demonstrated the value of historical information for assessing the persistence of benthic community characteristics, while also highlighting shortcomings if raw data is lost and if the historical baseline does not reflect pristine ecological conditions.

Keywords: Swansea Bay, tidal lagoon, baseline, reference conditions, WFD

# INTRODUCTION

Coastal infrastructure such as seawalls, breakwaters, or jetties impact marine ecosystems (Bulleri and Chapman, 2010; Firth et al., 2013). In recent years there has been growing demand to build marine renewable energy devices, contributing to even more infrastructure (Wilson et al., 2010; Binnie, 2016). In order to assess its environmental impacts developers have to carry out baseline surveys of the diversity and composition of benthic communities in the affected area (Franco et al., 2015). However, these are just snapshots of the situation immediately before construction and questionable long-term reference states. Baseline surveys may be affected by short-term natural impacts such as severe storms, extreme temperatures or unusual riverine freshwater input due to heavy rainfall, or anthropogenic impacts such as maintenance dredging (Kröncke et al., 1998; Bolam et al., 2010; Rangel-Buitrago et al., 2016; Robins et al., 2016). For the baseline to be a critical reference point it ought to establish the long-term condition and patterns of the benthic communities. Ideally, environmental monitoring data from statutory bodies or scientific research can be consulted, but longterm data is often scarce or non-existent or the spatial resolution is insufficient to serve as a suitable baseline. In those cases, where present, historical information may provide a valuable source of information. Marine historical ecology contributes profoundly to our understanding of the coastal environment and is increasingly applied in long-term management and policy (Robinson and Frid, 2008; Engelhard et al., 2015). Generally, historic data provides information on past baselines of biological and environmental parameters and enhances our understanding of the effects of anthropogenic disturbances on marine ecosystems and the role played by humans in shaping our coastal habitats (Lotze et al., 2006).

Currently the construction of a tidal lagoon is proposed for Swansea Bay (Wales, UK), a coastal area considered to be a "heavily modified waterbody" under the EU Water Framework Directive (WFD; 2000/60/EC) as a result of coastal protection measures (**Figure 1**). Under the WFD Swansea Bay's predicted ecological quality is classified as "Bad Potential." The proposed lagoon would exploit the tidal range to generate 320 MW using bulb turbines and power 155,000 homes (Waters and Aggidis, 2016). The wall would be 9.5 km long enclosing 11.5 km<sup>2</sup> of foreshore and seabed. As part of the environmental impact assessment (EIA) to inform the planning consent the developer had to carry out a baseline survey, but long-term information about the benthic communities in the area over recent decades was sparse (Harkantra, 1982; Shackley and Collins, 1984). Suitable historical information for a long-term comparison of benthic communities was located in an unpublished report (Conneely, 1988). In 1984 the regional water authority took 172 benthos samples throughout Swansea Bay to fulfill legislative requirements to protect the natural environment from adverse activities and the need to prepare a discharge policy; the area bordering the bay was, and still is, heavily industrialized and there were a number of domestic sewage discharges (Chubb et al., 1980). The survey was a data treasure chest that could be used to assess long-term changes of benthic community patterns. Since 1984 Swansea Bay experienced changes in its anthropogenic use, for example, a major sewage discharge was closed and relocated in 1999 and some areas were used for mussel cultivation, and these activities had had measurable, localized impacts on benthic communities (Smith and Shackley, 2004, 2006). Over the past decades the shipping channels to Swansea and Port Talbot ports as well as the River Neath were regularly dredged, and the material was discarded at a spoil ground in the outer Swansea Bay area (**Figure 1**).

In 2014 the wider Swansea Bay area was surveyed with a sampling design similar to 1984, and the data was analyzed in the same way as reported by Conneely (1988) for the historical study. The raw data from 1984 had been lost, and hence the comparison between 1984 and 2014 was limited to the figures and tables shown in the historic report. Fortunately the author had applied several techniques which are still valid today, including multivariate community analyses, but for the 2014 data additional analytical tools were applied.

The objectives of this study were to


# MATERIALS AND METHODS

#### Study Area and Habitat

Swansea Bay is a shallow embayment on the northern coastline of the Bristol Channel (Wales, UK) with depth generally < −20 m Ordnance Datum (OD; Pye and Blott, 2014). It is exposed to severe hydrodynamic forces due to strong winds and tides generated in the Bristol Channel, as well as North Atlantic swells (Allan et al., 2009). Swansea Bay is characterized by a complex patchwork of bottom substrata (Collins et al., 1980). It consists of depositions of poorly sorted, consolidated glacial boulder clay (glacial till), pebbles and cobbles, sometimes mixed with unconsolidated mud and silt as well as mixed sand, silts and clays with associated peats (Culver and Bull, 1980). Marine sediments in the eastern Swansea Bay area are mixed with redistributed dredge spoils from the Swansea and Port Talbot docks (Culver and Bull, 1980; Shackley and Collins, 1984). Generally, surface sediments are highly temporarily variable depending on storminess, with an increase in the proportion of sand and the exposure of relic gravel deposits after periods of wave exposure and deposition of mud following calm weather (Shackley and Collins, 1984). Water quality is largely influenced by the hydrology of three river catchments that serve Swansea Bay. It is also influenced by the historical and current industrial activity and associated diffuse and point pollution (surface water run-off) toward the eastern side of the bay. The main sewage outfall for the wider Swansea area is located in the center of the inner bay (**Figure 1**).

#### Benthos and Sediments

In 1984 benthos samples had been collected in a 1 km<sup>2</sup> sampling grid in the wider Swansea Bay area, and the same design was adopted in 2014; the reported figure of sampling points in 1984 suggests that the 1 km<sup>2</sup> grid design was only partly realized (**Figure 1**). Altogether 272 sites were visited in 1984, but glacial till limited the number of successful faunal samples to 176. At each position a single sample was taken. In 2014, 129 positions were visited and more samples were taken closer to shore compared with 1984, but fewer samples further offshore due to logistical limitations. Bed rock, boulders, or large shells prohibited the jaws of the benthos grab to close at some positions and here sampling was unsuccessful. Successful benthos samples were retrieved at 101 positions.

Benthos samples were taken with the same 0.1 m<sup>2</sup> Day grab in 1984 and 2014. About 200 g of surface sediment were removed for particle size analysis, and the remaining sediment was washed through a 1 mm sieve. The amount of sediment in each grab sample varied between 3 and 10l. Benthic community parameters were correlated with the amount of sediment per grab sample to test for possible effects of sediment volume on benthic community results, but it had no statistically significant impact (DistLM, p > 0.05).

The sieve residue was fixed in 4% formaldehyde and stained with Rose Bengal. All benthic species were sorted from the samples, identified to species level and counted. Sediment samples were air dried and passed through a series of sieves from 2 mm to 63µm according to the Wentworth–Udden classification scale to determine particle-size distribution. The sediment parameters mean grain size (x), sorting (σ), skewness (Sk), and kurtosis (K) were calculated with GRADISTAT (Blott and Pye, 2001).

Taking a single sample at each position without replication allowed using available resources to achieve a high resolution in the spatial spread of sampling positions, but it carried the risk that individual sites were not sampled representatively. It was therefore important not to place too much importance on results of individual sampling points, but rather consider groups of sampling points and broader spatial patterns.

#### Data Analysis

All information of the 1984 study was taken from Conneely (1988). Scanned copies of key figures and tables in the report are provided as Supplementary Material S1.

#### Cluster Analysis

In 1984 as well as 2014 groups of samples with a similar benthic community were identified by cluster analysis based on Bray– Curtis similarities and group average (Clarke and Warwick, 2001). In order to down-weigh numerically dominant species, the data was ln(x+1) transformed in both studies. For 1984 clusters were identified from the dendrogram shown in Conneely (1988) (Supplementary Material S1). For 2014 clusters were additionally analyzed by the "similarity profile" (SIMPROF) permutation test in PRIMER (Clarke et al., 2009). This explores the evidence of statistically significant clusters in samples which are a priori unstructured.

#### Indicator Species of Cluster

For the 1984 survey Conneely (1988) determined indicator species of the sample clusters by a set of "pseudo F-tests" (Mirza and Gray, 1981), and the method was replicated for the 2014 data. The test is "pseudo F" because it is applied to groups of samples determined by cluster analysis and not to pre-defined, independent sets of samples. Species that are significantly different between clusters in terms of abundance are considered useful discriminators between communities. There are potential pitfalls of such an approach, such as violation of the underlying assumption of normality and multiple comparisons problems, and therefore an increased chance of type I and II errors. The pseudo-F-test is rarely used these days to identify species that discriminate groups of samples. Instead, one of the most widely used methods in benthic community studies is SIMPER, which examines the contribution of each species to the average resemblance between sample groups (Clarke and Warwick, 2001). SIMPER additionally determines the contributions of species to the average similarity within a group of samples and hence identifies the species that typify a group; this analysis does, however, not identify discriminator species.

For the 2014 data both methods, SIMPER and pseudo Ftest, were applied. Identifying indicator species by pseudo-F tests and SIMPER was not directly comparable since SIMPER contrasts pairs of clusters, while pseudo-F compares all clusters simultaneously. It was however possible to broadly assess the resemblance of the species identified by the methods as discriminator species.

#### Inverse Classification

Associations of species with similar spatial distribution were identified by inverse classification for 1984 and 2014. Two species are thought of as similar if their numbers tend to fluctuate in parallel across sites. Conneely (1988) had performed an inverse cluster analysis based on Bray-Curtis species similarities and reported the species associations for 1984, and the same analysis was carried out for the 2014 data. The Sørenson Index was calculated between species association identified for 1984 and 2014 to assess the similarity between them over time.

#### Link with Environmental Variables

The extent to which habitat characteristics could explain the multivariate community structure found in 2014 was explored by distance-based linear models (DistLM). The routine allows analyzing and modeling the relationship between a multivariate data cloud, as described by a resemblance matrix, and one or more predictor variables (Anderson et al., 2008; PERMANOVA+ for PRIMER software). DistLM provides quantitative measures and tests of the variation explained by the predictor variables. The sediment properties mean grain size (x), sorting (σ), skewness (Sk), kurtosis (K), % coarse sand and silt/clay were included as variables as well as depth. Distance of each sampling position to the mouth of the rivers Tawe, Neath, and Afan was entered as a proxy for exposure to freshwater. This was calculated as a cumulative factor weighed according to the size of the catchment: River Tawe 49%, River Neath 32%, River Afan 19%. The distance to the sewage outfall was entered to quantify the impact of nutrient enrichment and point-source pollution. Before DistLM regression was carried out a Draftsman plot was evaluated for multi-collinearity and skewness of data. The Draftsman plot indicated strong correlations between some of the sediment parameters, but r was always below the usual cut-off point of 0.95. Hence, all variables were entered into the model, but it was kept in mind that inter-correlations may render some sediment characteristics redundant as explanatory factors. AIC was used as selection criterion since, unlike R 2 , it will not necessarily continue to get better with increasing numbers of predictor variables in the model; a "penalty" term is included in AIC for increases in the number of predictor variables (Anderson et al., 2008). Results of the DistLM were visualized by distance-based redundancy analysis (dbRDA). A vector overlay was added to the ordination diagram of the dbRDA, with one vector for each predictor variable.

#### Infaunal Quality Index (IQI): Water Framework Directive (WFD) Classification

The IQI was developed to assess the ecological status of the macrobenthic invertebrate infaunal assemblages of sediment habitats in UK coastal and transitional water bodies for the WFD (Phillips et al., 2014). It is a multi-metric index that expresses the ecological health of benthic assemblages as an Ecological Quality Ratio (EQR). It is composed of three individual components: AZTI Marine Biotic Index (AMBI), Simpson's Evenness (1–λ'), and number of taxa (S). To fulfill the requirements of the WFD, the IQIv.IV incorporates each metric as a ratio of the observed value to that expected under reference conditions. For reference conditions sediment properties were entered for each sample. Salinity was standardized to 28 for positions closest to rivers, 31 for other positions in the inner Bay and 32 in the outer bay south of Mumbles Head; salinities were averaged from data provided by Natural Resources Wales.

The IQI was calculated as

$$IQI\_{vIV} = \left( \left( 0.38 \times \left( \frac{1 - \left( \frac{AMBI}{7} \right)}{1 - \left( \frac{AMBI\_{Ref}}{7} \right)} \right) \right)$$

$$+ \left( 0.08 \times \left( \frac{1 - \lambda}{1 - \lambda \left( \_{Ref} \right)} \right) \right)$$

$$+ \left( 0.54 \times \left( \frac{S}{S\_{Ref}} \right)^{0.1} \right) - 0.4 \right) / 0.6$$

The resulting EQR ranges from an ecological status "High" (no or very minor disturbance) to "Bad" (severe disturbance; Phillips et al., 2014). It was calculated with the IQI Calculation Workbook

UKTAG v.1: update 11/03/2014, which is freely available from the WDF UKTAG webpage.

# RESULTS

#### Benthic Communities

For the 2014 study 188 benthic species were identified from 101 infauna grab samples. The multivariate benthic community analysis comparing all samples classified 21 clusters, and these were grouped into six broader clusters of samples (SIMPROF test based on Bray-Curtis resemblance matrix). The similarity within each of the six clusters was 23–38% (SIMPER).

Cluster 1 covered most of the eastern side of Swansea Bay (**Figure 2**). It was characterized by typical fine-sand species such as the bivalves N. nitidosa and S. elliptica, the polychaetes S. bombyx, and N. hombergii, as well as the cumacea D. rathkei.

Cluster 2 at the western side of Swansea Bay off Mumbles Head was the most biodiverse cluster with almost four times as many species and five times the number individuals compared



with cluster 1 (**Table 1**). While clusters 1 and 2 had several species in common these were still discriminating the clusters since most species were more abundant in cluster 2, except N. nitidosa and D. rathkei. Additionally encrusting, sessile, tube-dwelling polychaetes, sipunculids, and phoronids, as well as fully marine species such as the brittle star Ophiura ophiura colonized the area grouped as cluster 2 (SIMPER; **Table 2**).

The third cluster was located inshore, characterized by typical lower intertidal to shallow subtidal species such as amphipods of the genus Bathyporeia and Nephtys caeca (**Table 2**). All other clusters did not have a discrete spatial identity but were interspersed within the other clusters. They were characterized by low numbers of species, which were sub-sets of the three other clusters; cluster 4 additionally contained Nephtys cirrosa, cluster 5 the polychaetes Magelona mirabilis and Owenia fusiformis.

The spatial identity and distribution of the clusters was similar to 1984 (**Figure 2**): cluster 1 in 2014 and cluster D in 1984 both covered the eastern side of Swansea Bay; cluster 2 in 2014 and cluster C in 1984 were located off Mumbles Head; cluster 4 in 2014 was found in similar areas to cluster A in 1984. In 1984 samples had not been taken as far inshore as in 2014 and there were hence no sample positions that could be compared with cluster 3 in 2014. In 2014 too few samples were taken in off-shore areas to make a meaningful comparison with the area of the 1984 cluster B.

There was considerable resemblance in the species composition of individual clusters between 1984 and 2014. Based on the pseudo F-test table published in Conneely (1988) the two studies had 26 species in common (**Table 3**); Conneely (1988) reported F-tests for 40 species but the full species list was not published for the 1984 study. It was therefore not possible to identify the exact number of common, missing and additional species between studies. Of the 26 species recorded in 1984 as well as 2014, 16 had significant F-values in both studies, and 12 species were found in highest numbers in matching clusters.

Similar to 2014, in 1984 the cluster at the eastern side of Swansea Bay was characterized by N. hombergii, N. nitidosa, D. rathkei, and S. elliptica. Also, the largest number of species was reported for the cluster C off Mumbles Head; 32 of the 40 indicator species had highest abundances in cluster C (**Table 3**, approx. cluster 2 in 2014). These were mostly polychaetes, tubedwelling species or sessile species such as Mytilus edulis. Nephtys TABLE 2 | Discriminating species between the main groups of samples of the 2014 benthos survey in Swansea Bay (SIMPER).


*Groups were delineated by hierarchical clustering based on Bray-Curtis sample similarities. Cluster 1 (n* = *46) eastern Swansea Bay; Cluster 2(n* = *9) western side of Swansea Bay off Mumbles Head; Cluster 3 (n* = *6) inshore areas; Cluster 4 (n* = *18);* Figure 2*; mean densities per 0.1 m<sup>2</sup> are shown for clusters.*

cirrosa was the indicator species in the species-poor cluster A in 1984, which matched cluster 4 in 2014.

In comparison with 1984 the mean abundance of N. hombergii was lower on the eastern side of Swansea Bay in 2014. Conversely, the average density of N. nitidosa was higher in areas off Mumbles Head in 2014 compared with 1984; however, the species' distribution was generally patchy and standard deviations were high (**Table 2**, **Figure 3**). The difference in F-values for individual species in 1984 and 2014 supports that mean abundances in clusters differed between the studies. Further, some species were relatively abundant in the 2014 survey but were not reported for 1984, such as the polychaete Aphelochaeta marioni or O. ophiura.

#### Species Associations

Inverse cluster analysis identified 5 species associations in 1984 (Conneely, 1988) and 10 associations in 2014. Several associations had common species in 1984 and 2014 (**Table 4**). The greatest similarity (Sørensen Index) was found for the Nucula-association (N. nitidosa, S. elliptica, D. rathkei, N. hombergii, and S. bombyx). The species were mostly found in the eastern half of Swansea Bay and off Mumbles Head, broadly coinciding with sample clusters 1 and 2 (**Figures 2**, **3**).

#### Link between Environment and Benthos

Distance-based linear models (DistLM) allowed quantification of the degree to which one or more environmental parameters explained the benthic community structure in 2014; this analysis could not be carried out for the 1984 survey. The overall best model explained 35% of the resemblance in species richness and contained five variables (S: depth, mean grainsize, sorting, % coarse sediment and distance to sewage outfall). Of all entered variables "distance to the sewage outfall" explained most of the variation (6.3%, n = 101, p = 0.0032) in the data. The best model for the Nucula-association explained 22% of the variation and consisted of six factors: depth, mean grainsize, sorting, % coarse sand, % silt/clay, and distance sewage outfall; individually sediment sorting explained most of the variation (5.5%, n = 79, p = 0.0018).

For the entire multivariate benthic community matrix containing all species, each of the entered explanatory variables was individually a statistically significant predictor of the multivariate community structure (n = 101, p < 0.005 for each variables), each explaining 2.6–5.6% of the variation in the benthic community (Supplementary Material S2). The overall best model explained 17% of the variation and contained five factors: mean grainsize, sorting, % coarse sand, distance to sewage outfall, and distance to rivers. Individually sediment sorting explained most of the variation (5.6%, n = 101, p = 0.0001); distance to the sewage outfall explained 2.5% of the variation (n = 101, p = 0.001), and distance to rivers 2.4% (n = 101, p = 0.0023). The model is illustrated in **Figure 4**, where the dbRDA ordination of the benthic community is superimposed by explanatory variables. The dbRDA plot broadly groups the samples similar to the cluster analysis, at least for cluster 1. It ought to be noted that the dbRDA shows just 63% of the fitted variation and therefore captures only part of the model.

# Infaunal Quality Index (IQI): Water Framework Directive (WFD) Classification

The majority of samples indicated "moderate" or "good" environmental status according to the WFD classification (**Figure 5**). The IQI in the inner bay was 0.61 ± 0.08 (mean ± sd, n = 45) and 0.56 ± 0.15 in the outer bay (mean ± sd, n = 56). Both, the outer and inner Swansea Bay fell into the ecological status category "moderate." In the inner bay one sampling location in the vicinity to the sewage outfall was classified as "poor." Eleven samples classified as "poor" or "bad" according to the WFD were in proximity to the spoil disposal site in the outer Swansea Bay area (**Figure 1**).

FIGURE 3 | Distribution of species recorded in 1984 and 2014 which were identified by inverse classification as an association with similar spatial trends in both studies. The size of symbols represents the relative abundance of species in the 2014 study.

# DISCUSSION

Over the past decades some benthic communities along European coasts changed markedly in response to sea-level rise, invasive species or eutrophication, while others remained relatively unchanged (Hinz et al., 2011; Schumacher et al., 2014; Singer et al., 2016). The benthos of the urbanized Swansea Bay in South Wales (UK) showed strong resemblance in 1984 and 2014, despite changes in anthropogenic use during the past decades. This study provided evidence of striking similarities in the species composition and spatial mosaic of the benthic fauna. Since the two surveys were 30-years apart it is possible that the communities experienced changes during the intervening years. However, published records from before 1984 tie in well with the surveys described here, which suggests that the results may not reflect ephemeral conditions but relatively persistent community patterns (Warwick and Davies, 1977; Harkantra, 1982; Shackley and Collins, 1984). Still, given the uncertainty regarding the nature of the community during the intervening 30 years the term "persistence" is used sensu Grimm and Wissel (1997). According to their definition "persistence" is a stability property that allows for temporal variation in an ecological system which remains essentially the same over time; in contrast, the term "constancy" describes a system that stays unchanged.

In 1984 as well as 2014 a biodiversity hotspot was identified in an area further off-shore in mixed sediment and rocky grounds off Mumbles Head, a carboniferous limestone headland (**Figure 2**, **Table 1**). Reasons for the diversity-promoting conditions in that area were not obvious. Tidal flow velocities are exceptionally high around Mumbles Head where they are enhanced by an anticyclonic gyre, and generally tidal current speed and species richness are negatively correlated in sedimentary habitats (Warwick and Uncles, 1980; Rees et al., 1999; Pye and Blott, 2014). However, the current speed may not be the direct challenge for benthic species but rather the associated sediment movement (Warwick and Uncles, 1980). It is possible that off Mumbles Head high current velocities coincide with relatively stable substratum due to its mixed nature of glacial deposits and marine sediments with low sedimentation rates (Pye and Blott, 2014). This provides hard substratum and environmental conditions suitable for sessile species vulnerable to sedimentation and erosion. Further, tidal currents transport



*All tests significant (p* < *0.05) except those marked with*(*a*) *. F-values and means of ln(x+1) transformed abundances are shown for species in clusters; only clusters 1,2, and 4 of the 2014 survey had indicator species in common with 1984 and are shown in this table. Clusters with similar spatial identities were A/4, C/2, and D/1.*

large quantities of plankton from the inner bay area. This favors suspension and filter feeders and would explain the diverse sessile polychaete fauna, including several tube-dwelling species as well as sipunculids and phoronids, leading to higher diversity and abundance than elsewhere in the bay. While overall community patterns persisted over time, there was evidence of changes in density of individual species. These ought to be interpreted with caution. In Swansea Bay densities of individual species change dramatically not just seasonally, but from month to month and annually, and these are therefore unlikely indicators for long-term changes (Shackley and Collins, 1984; Conneely, 1988; Smith and Shackley, 2004, 2006).

In the 1984 and 2014 surveys a group of five species was prevalent: the bivalves N. nitidosa and S. elliptica, the polychaetes N. hombergii and S. bombyx, and the cumacean D. rathkei were grouped as species that showed overlap in their distribution (**Figure 3**). The species were found in both main benthic clusters and occurred in large parts of Swansea Bay. They are adapted to living in mobile sediments and coping with erosion and sedimentation, and it seems plausible that this group of species persisted over time because they can tolerate the rigor of the environment (Valentin and Anger, 1976; MarLIN, 2016). Their distribution was significantly linked to sediment properties. Generally, benthic monitoring can be onerous because of the taxonomic expertise necessary to identify large numbers of invertebrate species, and it may be possible to speed up the process by focusing on this group of species. This could provide a rapid indication of spatial change in the benthic community; it would though preclude conclusions about changes in biodiversity.

Distance-based linear models indicated that the composition of the Swansea Bay benthic fauna was significantly linked with sediment properties, the proximity to rivers and the sewage outfall. The close relationship of benthic organisms with sediment characteristics has long been established (Gray, 1974), and in the Bristol Channel and Swansea Bay area faunal associations were shown to be directly related to tidally-averaged bed shear stress, which provided evidence for the physical control of the benthic communities (Warwick and Uncles, 1980). The broad spatial pattern of sediment distribution remained identical over the past 30 years (Harkantra, 1982; Pye and Blott, 2014). However, the 1984 and 2014 studies also highlighted that the traditional method of grab sampling to assess benthic fauna and substratum may not be entirely appropriate for an area

#### TABLE 4 | Similarity of species associations in 1984 and 2014 (Sørensen Index).


*Associations were determined for each year by inverse cluster analysis based on Bray-Curtis similarities of ln(x*+*1) transformed abundances of species. Species jointly found in 1984 and 2014 are listed here; the total number of species in each cluster is shown above and in front of the 2014 and 1984 clusters of each species association. The darker the shading the greater the similarity between associations.*

with considerable glacial deposits, because the grab fails in rocky grounds. The coarse glacial material was not sampled representatively and some of the unexplained variation in the data is likely to stem from ignoring the impact of rocky substratum. Our understanding of the benthic ecology in areas with a mosaic of marine sediments and glacial till would improve by applying additional methods such as dredging and underwater video or stills. Further, the benthic models could be improved by more detailed, high resolution information about salinities in Swansea Bay. Distance to three rivers was a significant factor in explaining variation in benthic community characteristics, and it is possible that areas close to the rivers are at least temporarily subjected to full estuarine conditions (Heathershaw and Hammond, 1980).

The sewage outfall was also significantly linked with the benthic community composition, suggesting that this pointsource pollution affected the fauna. Invertebrates in heavily modified waterbodies in the vicinity of urban centers can be significantly impacted by an altered food chain, caused by higher nutrient concentrations from domestic and industrial sewage (Puccinelli et al., 2016). This can translate into a compromised ecological status, particularly if it is linked to oxygen depletion (Borja et al., 2009). However, while the distance to the sewage outfall was a statistically significant factor explaining 2–6% of the variation, it was generally part of a group of 5–6 habitat characteristics that best explained larger portions of the overall variation of the benthic community structure.

### Infauna Quality Index (IQI) and Ecological Status

The EU WFD water body classification suggests that Swansea Bay has "bad potential," partly because of possible constraints in the distribution of invertebrates due to coastal defense infrastructure and diffuse source pollution. This study showed that in 2014 the average ecological status of Swansea Bay fell into the category "moderate" in terms of its invertebrate fauna. The inner bay in particular was predominantly classified as "moderate" or "good" ecological status (**Figure 5**). A single location in the inner bay was categorized as "poor." The site was in close proximity to the current sewage outfall, which would be a plausible explanation for the poor ecological status (Borja et al., 2006). However, since this was just a single sample the result needs to be viewed with caution, and more replicate samples would be needed to

verify the finding. Similar to the benthic community models, the WFD classification for the area would also benefit from more accurate salinity data. The impact of salinity on multimetric parameters is recognized and following from this the importance of geographical separation of areas according to environmental conditions when implementing the WFD (Fleischer and Zettler, 2009). This is however particularly challenging in a relatively small area such as Swansea Bay with spatially and temporarily widely fluctuating salinities.

Eleven samples from the outer bay indicated "poor" or "bad" ecological status. A possible explanation is the vicinity of the dredge spoil ground, used to discard material from maintenance dredging of three shipping channels in Swansea Bay (**Figure 1**). The spoils may either directly impact the benthic community at the disposal site, or sediments may be transported over a wider area, explaining the west-to-east spread of sites with poor ecological status. Dredging and spoil disposal generally increases turbidity, changes sediment composition and mobilizes toxic materials such as heavy metals (Marmin et al., 2014). The nature of the impact of disposing dredge spoils on the benthic fauna varies with site specific environmental factors such as wave exposure and sediment dynamics (Roberts and Forrest, 1999; Bolam et al., 2010). The management of dredging and disposal of spoils would also be of relevance for new infrastructure projects, including the proposed Tidal Lagoon Swansea Bay, since maintenance dredging may be necessary after operation commences. With improving discharge management the risks decrease, but there is considerable uncertainty about the behavior of dredge spoils, and the impact on the ecology of affected areas merits closer investigation.

There was little resemblance between the pattern of the WFD ecological status classifications and the benthic community patterns identified by multivariate community analysis. For example, the biodiversity hotspot off Mumble Head was not

Biotic Index (AMBI), Simpson's Evenness (1-λ ′ ), and number of taxa (S).

generally categorized as having high ecological status. This precluded extrapolation of the 2014 ecological status assessment to 1984: while the broad benthic community patterns may have been similar in 1984 and 2014 it does not necessarily follow that the ecological status was similar too. Long-term studies of the sensitivity and robustness of benthic indicators to natural variability suggested that multimetric parameters such as the IQI will not just respond to anthropogenic impacts which they were designed for, but also to natural variation and disturbance, for example cold winter events and gradual changes in the climate regime (Kröncke and Reiss, 2010). They are however more robust against seasonal and interannual variability than univariate diversity indices. In the 1984 and 2014 comparison of Swansea Bay both natural long-term change as well as changed anthropogenic pressure was intertwined. Had there been significant differences in the benthic community structure, it would have been challenging to disentangle natural and anthropogenic effects. This highlights the importance of not only temporal reference conditions, but also spatial reference states (Borja et al., 2012).

Although this study precluded analyzing temporal differences in ecological conditions, it seems plausible that the ecological status may have changed over time in Swansea Bay. For example, Conneely (1988) suggested that the 1984 benthic fauna was affected by high concentrations of heavy metals in sediments from industrial and sewage discharge (Chubb et al., 1980). In 1999 the sewage treatment and discharge location was moved from the western bay to its current central position (**Figure 1**), triggering a shift in benthic diversity from filter-feeders to deposit feeders in the immediate vicinity of the old sewage pipe (Smith and Shackley, 2006). Further, the start of a commercial mussel lay in the western bay led to a localized increase in carnivores and deposit feeders, but also to an overall decrease in species richness within the mussel bank (Smith and Shackley, 2004). While these changes in anthropogenic use seem not to have altered the broad community patterns over the past 30 years, they are likely to have changed the ecological status in pockets of Swansea Bay.

The historical use of Swansea Bay highlights that the 1984 baseline did not represent a pristine state at which anthropogenic effects could be considered to be negligible (Collins et al., 1980). We know, for example, that about a century ago the area had thriving oyster beds (Shackley et al., 1980). Similar to other European stocks populations declined through overfishing, untreated sewage discharge, heavy metal contamination and shellfish disease (Laing et al., 2006). Despite improvements in water quality and the absence of commercial oyster dredging for decades, the stocks have not recovered, and hence, the anthropogenic activities a century ago may have changed the system beyond natural recovery. Further, coastal defense and infrastructure development in the bay severely modified the bay for over a century. Port Talbot Harbor or the Swansea Dockland/Tawe dredge channel create surrogate headlands which affect localized sediment movement (Thomas et al., 2015). This needs to be taken into account for environmental management, particularly when determining targets and reference conditions.

# CONCLUSIONS

This study provides further evidence of the value historical data can add to marine and coastal management, particularly if the repeat surveys are standardized to the historic methodology and complemented with contemporary techniques. On balance this approach maximizes the power of the comparison, although it may not capitalize on all currently available survey techniques. While this study emphasized the opportunities of historical data, it also grappled with limitations of using sub-standard information. The aspiration remains to determine meaningful reference conditions or baselines that can be repeated to track change (Borja et al., 2012). Generally this study highlights the importance to store raw survey data and make them available for future research in public archives. For the development of an infrastructure project such as the tidal lagoon in Swansea Bay this study offers a baseline of spatial benthic diversity patterns and provides information about key species and their relationship with the habitat. While significant environmental variables affecting the benthic community composition were identified, much of the spatial variation in the fauna remained unexplained. In order to improve models, more accurate and detailed information about freshwater input and salinities needs to be generated. The impact of glacial till on the benthic community needs attention, and this study suggested that for areas with a mosaic of marine sediments and glacial deposits traditional sediment property measures may be poorer indicators of the hydrodynamic regime than elsewhere. Direct values of current speed as well as wave exposure and sediment transport could improve the benthic models. Since the proximity of the sewage outfall was a significant contributing factor in explaining benthic characteristics, it would be advisable to measure oxygen concentration in sediments more accurately as a possible explanatory factor linked with nutrient enrichment. Importantly, this study suggests that dredge spoil disposal may affect the current ecological status of the benthic community, and future studies ought to focus on the behavior of dredge spoil disposals in the outer Swansea Bay in order to understand processes that may affect the benthic fauna.

The comparison with a 30 year old baseline removed some uncertainty about the temporal variability of the benthic communities and confirmed that current associations are unlikely to be ephemeral but instead reflect persistent patterns. The severe natural environmental conditions in this heavily modified waterbody appear to have overshadowed localized changes linked with anthropogenic use over the past decades when assessing the area on a larger spatial scale. However, the 1984 baseline portrayed an already highly anthropogenically impacted situation as a result of industrial activities for over a century and did not represent a pristine ecological state. Historical data are therefore not necessarily suitable for setting future targets regarding the environmental status and for assessing if an area is as expected under prevailing conditions, as required, for example, by the Marine Strategy Framework Directive (MSFD, 2008/56/EC; European Commission, 2008). Such a task may be particularly challenging an areas such as Swansea Bay, which have been subjected to century-long anthropogenic impact.

# DATA ACCESSIBILITY

This paper highlights the importance of data accessibility and the author strongly supports public availability of raw data. The data of this study will be made publically available through an appropriate public archive once the paper is accepted for publication.

# AUTHOR CONTRIBUTIONS

The author confirms being the sole contributor of this work and approved it for publication.

# FUNDING

The study was part-funded by the EU ERDF project SEACAMS.

#### ACKNOWLEDGMENTS

I am indebted to everyone from the Swansea SEACAMS team who helped with the boat work on RV Noctiluca, in particular Keith Naylor, Chris Lowe, Ian Tew, Hanna Nuuttila, and Christine Gray. Chiara Bertelli identified the invertebrates in the benthos samples and Anouska Mendzil processed the sediment samples. Many thanks to Gill Lock from TLSB for her continuous cooperation and the

#### REFERENCES


discussions about the subject. Four reviewers made constructive suggestions during the revision process and greatly improved the paper.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00141

the history of fish and fisheries in current policy context. ICES J. Mar. Sci. 73, 1386–1403. doi: 10.1093/icesjms/fsv219


**Conflict of Interest Statement:** The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declared a past collaboration with the author and states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Callaway. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Application of Long-Lived Bivalve Sclerochronology in Environmental Baseline Monitoring

Juliane Steinhardt <sup>1</sup> \*, Paul G. Butler <sup>2</sup> , Michael L. Carroll <sup>3</sup> and John Hartley <sup>1</sup>

*<sup>1</sup> Hartley Anderson Ltd., Aberdeen, UK, <sup>2</sup> School of Ocean Science, Bangor University, Anglesey, UK, <sup>3</sup> Akvaplan-niva, FRAM–High North Centre for Climate and the Environment, Tromsø, Norway*

Assessments of the impact of construction, operation, and removal of large infrastructures and other human activities on the marine environment are limited because they do not fully quantify the background baseline conditions and relevant scales of natural variability. Baselines as defined in Environmental Impact Assessments typically reflect the status of the environment and its variability drawn from published literature and augmented with some short term site specific characterization. Consequently, it can be difficult to determine whether a change in the environment subsequent to industrial activity is within or outside the range of natural background variability representative of an area over decades or centuries. An innovative approach that shows some promise in overcoming the limitations of traditional baseline monitoring methodology involves the analysis of shell material (sclerochronology) from molluscs living upon or within the seabed in potentially affected areas. Bivalves especially can be effective biomonitors of their environment over a wide range of spatial and temporal scales. A rapidly expanding body of research has established that numerous characteristics of the environment can be reflected in morphological and geochemical properties of the carbonate material in bivalve shells, as well as in functional responses such as growth rates. In addition, the annual banding pattern in shells can provide an absolute chronometer of environmental variability and/or industrial effects. Further, some species of very long-lived bivalves can be crossdated back in time, like trees, by comparing these annual banding patterns in their shells. It is therefore feasible to develop extended timeseries of certain marine environmental variables that can provide important insights into long temporal scales of baseline variability. We review recent innovative work on the shell structure, morphology, and geochemistry of bivalves and conclude that they have substantial potential for use as monitors of environmental variability and the effects of pollutants and disturbance.

Keywords: bivalve, shell, environmental monitoring, baseline, sclerochronology

# INTRODUCTION

Effective monitoring plays a key role in protecting the environment and limiting anthropogenic impacts by providing evidence of the efficacy or otherwise of mitigation measures. Monitoring programs are carried out to minimise uncertainties and ensure that in situ effects do not go beyond the modeled predictions and remain within defined limits.

Edited by: *Angel Borja, AZTI, Spain*

#### Reviewed by:

*Jose M. Riascos, Universidad del Valle, Colombia Eugenio Alberto Aragon-Noriega, Centro de Investigaciones Biológicas del Noroeste S C, Mexico*

> \*Correspondence: *Juliane Steinhardt juliane.steinhardt@gmail.com*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *26 May 2016* Accepted: *31 August 2016* Published: *23 September 2016*

#### Citation:

*Steinhardt J, Butler PG, Carroll ML and Hartley J (2016) The Application of Long-Lived Bivalve Sclerochronology in Environmental Baseline Monitoring. Front. Mar. Sci. 3:176. doi: 10.3389/fmars.2016.00176*

While standard environmental monitoring strategies assess the impacts of marine infrastructure in great detail, up to and including decommissioning, a full description of the baseline against which the results of such monitoring are assessed is precluded because the period between planning and the start of construction is too short to allow the range of natural background variability to be defined. Pre-existing long term instrumental records are sparse in the marine environment, often situated very close to the coast, and limited in the range of variables measured. In addition, new industrial developments and associated infrastructure can be situated in remote areas where long term records of environmental conditions are unavailable. Nevertheless, in the context of long period cycles and regime shifts in the marine environment that may be of the order of several decades (Drinkwater, 2006), and of even longer trends resulting from natural cycles and anthropogenic climate change (Kaufman et al., 2009), assessment of the full environmental impact attributable to an operation requires a realistic understanding of impacts that may be attributable to other sources (**Figure 1**).

One approach to this issue involves measurement of the responses of organisms in an area of interest to environmental change. In particular, the use of the hard parts of bivalve molluscs—the field known as sclerochronology—is now established in marine palaeoclimate studies and presents an exciting option for extended baseline marine monitoring.

Sclerochronology is the study of physical and chemical variations in the microgrowth of hard tissues (Davenport, 1938; Buddemeier et al., 1974; Jones, 1980; Oschmann, 2008), which can be placed into a temporal context by periodic banding. Advances in bivalve-based sclerochronology over recent years have generated a range of new possibilities for environmental monitoring, including in particular the ability to obtain long (multi-decadal) baseline environmental information. However, this potential has not yet attained a wide recognition in the commercial and regulatory sectors, and as a result sclerochronological techniques have not become a standard aspect of monitoring programs. Here we review the state of the art of bivalve sclerochronology and assess its potential to address needs and gaps in current marine environmental monitoring practices.

### MONITORING THE ENVIRONMENT

Environmental monitoring is defined as "the regular collection, generally under regulatory mandate, of biological, chemical,

assessment programs.

or physical data from predetermined locations such that the present status and any ecological changes attributable to industry operations can be quantified" (GESAMP, 1999). Only with sufficiently good knowledge of the current status, extent, changes, and trends in the condition of the marine environment can stakeholders and policy makers make informed and reasonable decisions, set rational priorities, calculate costs and benefits, and evaluate risks of proceeding with or refraining from the proposed projects. A tight feedback response within the environmental management framework reinforces successful environmental management and its ability to protect the environment through quick responses and continuous improvements in prediction, observation, regulation, and operation. Furthermore, there is a particular need for monitoring procedures that can identify and quantify variability and trends through time and geographical space. Monitoring programs aim to provide an overview of general environmental conditions, variability and trends with the goal of robustly estimating and attributing any changes associated with marine developments.

Objectively interpreting results and evaluating impacts upon the environment is dependent upon the standardization of monitoring approaches (e.g., standardized taxonomic analysis and geochemical measurements), which in turn allows the intercalibration of data sets and comparison of relative impacts. While the contaminants of concern and the biological species available for monitoring will vary between marine regions, monitoring methods and assessment criteria should be harmonized as appropriate. In recent years, substantial effort has been put into developing biological effect tools (Beliaeff and Burgeot, 2002; Broeg et al., 2005; Dagnino et al., 2007; ICES, 2007; Thain et al., 2008; Edge et al., 2014). Moreover, new developments and improved technologies allow for measurements that were previously challenging and expensive to be made rapidly, in large numbers, and by technical staff with only basic skill requirements (Bowen and Depledge, 2006).

Baseline characterizations aim to establish pre-operational background environmental conditions, including variability and long-term trends, allowing site-specific operational planning to minimize environmental impact. Baseline survey data is used to define natural background levels and variability of physical, chemical, and biological parameters, which can subsequently be used as a point of comparison for later monitoring surveys. On a broader scale, baseline studies can also contribute to determining anthropogenic impacts and potentially distinguishing between natural and anthropogenic climate variations.

A further target of environmental monitoring is determination of "good environmental status" (GES) as defined by the European Union's Marine Strategy Framework Directive (MSFD). A number of descriptors have been put forward for the MSFD, intended to specify what is meant by GES (2010/477/EU, document C(2010) 5956, EC and MSFD, 2010).

The implementation of MSFD aims to protect and conserve the marine environment, prevent its decline, and, where necessary, restore marine ecosystems in areas where they have been negatively affected.

# Biological Techniques in Environmental Monitoring

A biological effect is defined as the response of an organism, a population, or a community to changes in its environment. The usefulness and applicability of any biological-effect method will depend on how well it is able to separate anthropogenic stressors from the influence of other environmental factors or internal biological processes. Biological effects techniques include tools to indicate whether concentrations of contaminants are at safe levels, specifically in relation to the pollution effects of naturally occurring and synthetic chemicals in the marine environment. Applying these techniques enables the identification of pathways between contaminant and ecological receptor, as well as the detection of the impact of substances (or any combination of substances) that may not be analyzed as part of routine chemical monitoring programmes in the sense of "what you don't measure you don't find" (van der Oost et al., 2003; Thain et al., 2008). In general, biological indicators carry information about the status and health of the environment, and may reflect biological, chemical and/or physical conditions.

There have been clear examples in the aquatic environment where biological effect techniques have been used to identify problems and subsequently to monitor the efficacy of management interventions. For example, the Mussel Watch programs have been applied worldwide to assess pollution levels within coastal zones (Goldberg et al., 1978; Goldberg and Bertine, 2000). Primarily, indicators are used to characterize the current status and to track or predict significant changes; they often track effects with multiple causes in a more simplified and useful manner than direct contaminant measurements. In addition, quantitative measurements can be taken in the form of biomarkers; these are measurable responses in the biological system to exposure to doses of substances and are potential tools to detect exposure to or effects of contaminants. Biomarkers can measure responses at different biological levels: biochemical, physiological, organism, and population (Lam and Gray, 2003). Most of these biological effect techniques make use of markers in the tissue of the living animal.

# Bivalves in Monitoring Studies—Current Approaches

In general, the term "biomarker" is applied to any change that can be detected in an individual living organism as a consequence of exposure to a harmful chemical (or chemicals). Depledge (1994) defines biomarkers more specifically as: "biochemical, cellular, physiological, or behavioral variations that can be measured in the tissue, body fluid samples, geochemistry, and morphology of the shell/exoskeleton or at the level of the whole organism, which provide evidence of exposure to and/or effects of, one or more chemical pollutants (radiation)." The selection of a biomarker depends on the nature of the environmental study and the information already available about the physical and chemical nature of the study area.

Biomarkers are being used in different countries as part of various marine monitoring programs e.g., the joint biological monitoring program for the North Sea (JAMP, 1998a,b), the UK National Marine Monitoring Programme (NMMP UK, CEFAS, 2004) and in Spain (Franco et al., 2002; Borja et al., 2004). One of the major concerns regarding the application of biomarkers as a regional monitoring tool has been the lack of standardization of methodologies. In response, a number of programs have developed standard operating procedures (SOPs) and undertaken inter-laboratory comparisons, including the standardization programs QUASIMEME, BEQUALM (Moore et al., 2004), and the SOPs of ICES. The use of biomarkers based on organic tissue or behavior only allows sampling of a "snapshot" in time, and longer time-series requires periodic sampling efforts. In this paper we argue that sclerochronological techniques present a viable and effective approach to long baseline environmental monitoring.

Shell material carries an imprint, in its chemistry, crystal structure and morphology, of the state of the ambient environment at the time of deposition (**Figure 2**). The periodically banded shells of long-lived bivalves can therefore be thought of as an archive of data reflecting pre- and postimpact conditions, with specimens from an unimpacted control site being used to define the pre-impact baseline. Data from extended multi-shell chronologies, where the precise timing of geochemical data points is determined by the periodic banding in the shells, can be used as a basis to study the background effects of seasonality and decadal (or even centennial) scale climate variability within the area of interest.

# Monitoring Targets

#### Coastal Monitoring

The increase in "competing users" of coastal and marine areas means that marine planning is becoming increasingly important for resolving conflicts between users and for managing environmental impacts on the sea (Douvere and Ehler, 2009). Marine spatial planning is a bottom-up process intended to improve collaboration and coordination among all coastal and ocean stakeholders, and to better inform and guide decision-making where it is perceived to affect their economic, environmental, security, and social and cultural interests (e.g., CMSP NOAA, 2015). In particular, marine coastal monitoring focuses on three general categories: major spill events (e.g., oil spills), chronic point-source pollution (e.g., outfalls, oil/gas platforms, coastal industry waste discharge), and multiple inputs (e.g., rivers), and all of these strategies have both a temporal and spatial aspect. During major spill events both the temporal and spatial components are complicated by rapidly changing conditions. In such cases pre-spill data is highly desirable for subsequent evaluation of the impacted site, but is most often not available.

#### Renewable Energy and Near-Shore Activities

Many technologies are currently being developed to harvest low carbon marine energy; these include offshore wind, wave and tidal power, salinity gradients, and thermal gradients. In 2007, European leaders agreed to source 20% of their energy needs from renewable sources by 2020 (Energy Policy for Europe 7224/07, the so-called Renewable Energy Roadmap; European Parliament, 2007). However, the large scale of planned offshore renewable energy developments (OREDs), will add to the existing pressures on coastal ecosystems, increasing the need for environmental and ecological monitoring. Based on Nedwell and Howell (2004) the lifecycle of a wind farm (∼30 years) can be divided into four general phases: (1) the pre-construction phase

the typical transect for cross-section. (B) Umbo-ventral margin cross-section as indicated in (A); this exposed surface is the typical site for high resolution geochemical sampling. Microstructure of *Cerastodernum edule* (right) showing (C) periodic banding with tidal resolution and (D,E) characteristic microstructural and crystallographic features, which show a high correlation with seawater temperatures. (A,B adapted from Schöne, 2003 C–E adapted from Milano et al., 2016).

(lasting 1–5 years), (2) the construction phase (1 year), (3) the operational phase (20–25 years), and (4) the decommissioning phase (1 year). Any significant environmental effects (**Figure 3**) are likely to depend on the natural disturbance regime and the stability and resilience of the communities (Gill, 2005). In terms of long-term ecological effects (**Figure 3**), the lifecycle time span and its individual phases need to be considered in ecological monitoring studies. A number of potential impacts of offshore wind farms on ecosystems have been identified, including underwater noise effects, bird collisions, the barrier effect, electromagnetic fields, visual/physical intrusion, and seabed disturbance during construction and operation (**Figure 3**; ETNWE, 2003; CEFAS, 2004; Carter et al., 2008).

Offshore renewable energy installations are likely to impact coastal ecosystems, since single developments have ecological footprints extending over several square kilometers of nearshore waters (e.g., Boehlert and Gill, 2010). Nevertheless, data on the environmental impacts of OREDs are limited (EC, 2015) and there is an urgent need to develop specific long-term observation methods to monitor their effects on the environment (Shumchenia et al., 2012; Lindeboom et al., 2015). Furthermore, various terrestrial and near-shore activities (Mason, 2002; Matthiessen and Law, 2002) have led to disturbance of habitat or local habitat loss. Other environmental aspects that need to be considered include changes in nutrient availability and

cycling, sediment erosion or reduced sediment supply, changes in sea level and any change in vulnerability to natural and anthropogenic disturbances (McLusky et al., 1992; Schekkerman et al., 1994; Rogers and McCarty, 2000). Sclerochronology-based indicators, such as indicators of changes in food availability and/or foodweb structure [these indicators include shell growth, δ <sup>13</sup>Cshell, Ba/Cashell, etc.; see Section Ecosystem Variables (Trophic Levels, Food Supply, Metagenome)], can potentially be used for the establishment of baselines related to the pre-impact environment and for the in situ monitoring of any environmental change during the construction and operation of the installation.

#### Oil and Gas

While the total amount of oil and gas produced within the OSPAR area has decreased since 2001, the number of offshore installations has increased (OSPAR, 2010 Quality Status report). OSPAR is the legislative instrument by which 15 governments and the EU cooperate to protect the marine environment of the North-East Atlantic. In addition, rising global temperatures, particularly in the high northern latitudes, have coincided with a rapid decline in sea ice cover, making hydrocarbon resources in the Arctic increasingly accessible. The use of standard monitoring procedures to assess the impact of existing operations, decommissioning, and the opening up of new areas is enhanced and relevant where long term baseline conditions can be established for comparison.

#### **Drilling discharge—a potential contamination source**

Routine operations of production platforms can lead to the release of oil, chemicals, and naturally occurring radioactive materials; these occur especially through discharges of produced water, drainage water (water from platform decks etc.), drilling fluids, and from drill cuttings. Furthermore, accidental oil spills can occur during exploration and production operation. A main source of input are drill cuttings (drilling mud and fragments of overlying and reservoir rock) deposited onto the seafloor during the exploration phase. Drilling fluids, used to lubricate the drilling string, are divided into three types: water-based (WBM), oil-based (OBM), and synthetic-based (SBM). Waterbased drilling fluids, consisting of inorganic components, are regarded as less harmful than the other two, and these are allowed to be discharged without treatment. The main constituents of WBM (bentonite clay and barite) are non-toxic, but they can smother sessile benthic animals.

In the North Sea alone the volume of drill cuttings is estimated at 12 million m<sup>3</sup> (OLF, 2000). A study by Breuer et al. (2008) of a drill cuttings pile resulting from drilling oil based mud (such discharges were effectively banned in the OSPAR area in 1996) showed that microbially mediated diagenetic reactions in organic-rich cuttings can result in rapid removal of O<sup>2</sup> within the top few millimeters resulting in anoxia within the cuttings pile. As a result, the rate of degradation of hydrocarbons is slowed and elevated concentrations of trace and heavy metals occur. Trace metals released into the porewater of the cutting pile can potentially diffuse into the overlying water by adsorbing onto Mn and Fe oxyhydroxides at the sediment water interface. Other metals (Cr, Cu, and Pb) can diffuse downward, becoming

nature of the environmental effects.

incorporated into Fe monosulfides. If the Fe sulfides are exposed to O2, e.g., by bioturbation, advection and/or pile resuspension during the decommission process, this may lead to the release of the associated metals into the water column (Huerta-Diaz et al., 1998; Saulnier and Mucci, 2000). Sclerochronological approaches that can be used to track some of these pollutants are described in Section Metal Pollution.

Another major source of oil discharge from routine production is produced water (PW), which is extracted from the reservoir along with the oil. PW can represent up to 80% of the waste and residual discharge to sea from natural oil production operations (Tibbetts et al., 1992; Carroll et al., 2001; McCormack et al., 2001) and contains hazardous substances including residues of chemicals used in the production process such as corrosion inhibitors and demulsifiers (chemicals that enhance the separation of oil from water) and inorganic compounds including heavy metals and radionuclides. The composition and amount of PW varies between fields. The ratio of PW to other fluids depends on the type of reservoir and its geochemical characteristics, as well as the production/processing techniques. The maximum permitted concentration of oil in discharged PW in the OSPAR area is 30 mg/l (e.g., Lee et al., 2011).

Other pressures from oil and gas activities include chemicals that can leak, e.g., from hydraulic valves, and leach from coatings and anodes of pipelines and other subsea structures. The seabed is physically disturbed not only during installation of pipelines, cables, subsea structures, and platforms, but also when they are decommissioned (an increasingly important issue as installations come to the end of their lifecycles, see below). Additionally, the risk of accidents such as leaks and spills may increase as the infrastructure ages. In recent years the oil industry has begun to decommission redundant installations and pipelines. The removal of the installations and associated infrastructure can cause sediment disturbance and subsequent localized impacts, such as turbidity. If there is a cutting pile at the base of the platform this may be disturbed and contaminated cuttings re-suspended. However, evidence indicates that these re-suspended particles do not disperse and settle back in the same area.

#### **Decommissioning**

Decommissioning is the process of removing or otherwise making safe oil or gas exploitation structures at the end of their life cycle. Decommissioning may be carried out by any one of four methods: complete removal, tow-and-place, partial removal (i.e., "topping"), or toppling (laying the structure on its side; Schroeder and Love, 2004; Macreadie et al., 2011; Fowler et al., 2014). Globally, more than 7000 oil and gas platforms distributed over 53 countries (Parente et al., 2006) will need to be decommissioned in the coming decades.

Of national and international bodies that regulate decommissioning in the NE Atlantic, OSPAR is particularly significant in terms of influencing the adopted approach. OSPAR Decision 98/3 prohibits leaving offshore installations wholly or partly in place but provides certain derogations to the legislation including the exception of concrete structures and the footing of large steel jackets with a weight of more than 10,000 tonnes.

Decommissioning assessment reports should take into account the effects of all decommissioning options, including energy budgets, biological and technological impact of discharges, secondary emissions, physical and habitat issues, fisheries, waste management, littering and drill cutting deposits (Parente et al., 2006).

Decommissioning alternatives to complete removal may include the creation of artificial reefs (so called "rig to reef " approach), which provide substrates for marine organisms (as for example the Gulf of Mexico). The objective of the "rig to reef " approach is to use the decommissioned structures for fisheries yield and production, for recreational activities, to prevent trawling, to repair degraded marine habitats, and for overall economic and social benefit. Ultimately a reefbased food chain may develop providing food sources for larger organisms such as fish (e.g., Claisse et al., 2014). Such a rig-reef community may vary considerably from the naturally occurring species composition and also affect local nutrient recycling within the water column and settling of nutrients to the seafloor, thereby affecting benthic organisms surrounding the reef area. One criterion for consideration of partial removal of decommissioned offshore oil platforms is its potential for conversion to a man-made reef that would provide a "net benefit" to the environment compared with complete removal of the structure. However, it has to be kept in mind that the removal of the structure may affect the environment negatively as well.

The decommissioning programme incorporates monitoring of the seabed after the asset has been decommissioned. The alternative disposal options will usually have different environmental effects and economic consequences, and with marine ecosystems expected to change rapidly in response to increasing anthropogenic influences and climate change, there is a strong need to assess and understand the long-term spatiotemporal variation in the environment and the marine ecosystem in the context of the physical presence of offshore structures. This would include the assessment of variability before, during and after the operational and decommissioning phases. Standard monitoring procedures reflect only the changes associated with the presence of the structure and compare them with a relatively short period before construction commences. Long baseline measurements of the kind described later in this paper could allow the combined effects of the construction, operation and decommissioning cycle to be evaluated in the context of longterm natural environmental variability.

# Climate Change

In response to rapidly increasing greenhouse gas concentrations in the atmosphere, the oceans act as a buffer, taking up around 66% of the excess CO<sup>2</sup> and around 93% of the excess heat content (IPCC et al., 2014). Consequences of increasing emissions from fossil fuels include rising seawater temperatures, deoxygenation, and ocean acidification, all of which can affect the benthos in various ways (**Figure 4**; ICES, 2008; Birchenough et al., 2015). In the context of the techniques discussed in this paper, it is important to understand and predict how these changes affect the benthic species used as environmental monitors, and how climate

effects can be separated from natural background variability and effects attributable to an installation.

#### Temperature and Hydrodynamics

Responses to changing temperature include distributional shifts, phenological changes, life history effects (reproduction and recruitment) and physiological responses (stress). The hydrodynamic regime can affect transport, dispersal, and settlement of larvae, consequently affecting species population dynamics (Levin, 2006).

#### Hypoxia

Changes in hydrodynamics may result in oxygen depletion as a result of stratification or eutrophication. Because pelagic and benthic processes can be tightly coupled, especially in coastal and selected shelf locations, the benthic ecosystem may be affected by changes in primary production and the transport pathways of benthic food sources. The quality and quantity of organic matter settling vertically through the water column is a vital factor affecting benthic abundance, biomass, growth and health (e.g., Dauwe et al., 1998).

Hypoxic zones (<2 mg l−<sup>1</sup> dissolved oxygen) are projected to expand because of (a) increased water column stratification, (b) temperature-related increase in respiration and (c) changes in precipitation that cause amplified terrestrial fresh water discharges and an elevated volume of nutrients (incl. agricultural fertilizers). Mass mortality and decreased diversity in benthic species have been observed (e.g., Diaz and Rosenberg, 2008; Levin et al., 2009; Seitz et al., 2009). Furthermore, bottom water oxygen depletion is likely to alter biogeochemical processes and affect nutrient supply at the sediment-water interface (e.g., phosphorus release, denitrification). In contrast, climate change is also projected to induce more storm activity in some regions; a stormier environment could increase vertical mixing of the water column and decrease stratification, reducing the potential for oxygen depletion (Rabalais et al., 2007).

# STATE OF THE ART OF BIVALVE SCLEROCHRONOLOGY FOR ENVIRONMENTAL MONITORING

#### Long-Lived Bivalves: A Brief Introduction

While the shells of a number of species have been used for environmental monitoring, the implementation of long baseline monitoring ideally depends on the availability of long-lived animals whose shells contain periodic growth increments which grow synchronously within populations and can therefore be used to generate multidecadal and multicentennial chronologies. These chronologies define a timeline of precisely dated shell material that can be used for geochemical analysis and to link patterns in functional responses (i.e., growth) to variability in environmental drivers. The two species that currently lend themselves best to these techniques are Arctica islandica (Linnaeus, 1767) and Glycymeris glycymeris (Linnaeus, 1758). Both occur over a relatively large geographic range in the temperate North Atlantic. A. islandica is the longest-lived (up to 507 years; Butler et al., 2013) and the most extensively researched species. In addition to its very significant presence in paleoclimatology research, this remarkable animal has contributed to studies in ecology, biology, pollution monitoring, gerontology and genetics. As well as being the longest living non-colonial animal whose actual age can be ascertained, A. islandica is known to grow synchronously within populations, allowing long chronologies to be constructed by crossdating living specimens with fossil shells (Butler et al., 2010, 2013). In addition to its potential for environmental monitoring, it can also be used for the effective reconstruction of marine climate on multicentennial timescales and at annual- and sub-annual resolution (Witbaard, 1997; Schöne et al., 2002, 2003; Goodwin et al., 2003; Butler et al., 2013; Mette et al., 2016).

Arctica islandica lives buried in surficial sediments with relatively short siphons which open at the sediment-water boundary. It is not attached to the substrate, and shows some ability to move vertically through the sediments (Abele, 2002; Morton, 2011), with regular reports of A. islandica burrowing several centimeters into the sediment (e.g., Taylor, 1976; Strahl et al., 2011) and remaining there for periods of several weeks. In respect of food, A. islandica is thought to be very selective, feeding only on fresh organic matter, and discarding older organic material lying at the sediment surface (Erlenkeuser, 1976). During prolonged periods (>60 days) of unfavorable conditions (e.g., anoxia) it can switch to anaerobic respiration and a reduced metabolism (Oeschger, 1990; Strahl et al., 2011).

G. glycymeris is found in surface areas of coarse-grained subtidal sediments, thus complementing A. islandica in terms of habitat and usefully extending the area for which shells suitable for long baseline monitoring are available. A study using stable oxygen isotopes in the shell (Berthou et al., 1986) demonstrated that the growth lines of G. glycymeris are formed annually, corresponding to winter shell growth checks. While not quite as long-lived as A. islandica, some individuals can live up to two centuries (Reynolds et al., 2013). Like A. islandica, G. glycymeris grows synchronously within populations.

Because bivalves such as A. islandica and G. glycymeris live in the boundary layer between the sediment and the water column, they are directly exposed to heavy metal pollution (Szefer and Szefer, 1990), and several studies have shown that heavy metal levels in both shells and soft tissue of A. islandica are elevated at polluted sites (Supplementary Table I). Measurements of trace and heavy metals have been carried out on transects through the annual growth increments in shells of A. islandica, showing its suitability as a monitor of contamination through time (Liehr et al., 2005; Dunca et al., 2008). This potential has been enhanced more recently with advances in analytical methods and improved knowledge of biomineralization (e.g., Holland et al., 2014; Shirai et al., 2014; Poulain et al., 2015).

### Biomineralization: Environmental vs. Biological Effects

Molluscs construct their shells of calcite, aragonite, or both (mostly aragonite in A. islandica and G. glycymeris). The bulk of the shell is composed of calcium carbonate (CaCO3), with some trace elements that can substitute for calcium (e.g., Mg, Sr) and some organic substances (e.g., proteins, lipids, polysaccharides) that tend to concentrate in the growth lines that separate the wider increments.

The shell formation process is initiated at the early stage of larval development (trochophore) and sequential carbonate deposition continues to contribute to the shell growth after metamorphosis and throughout the entire life of the animal (Marin et al., 2007). The oxygen of the bicarbonate ions (HCO3<sup>−</sup> ) is in isotopic equilibrium with that of the ambient water, enabling seawater temperature reconstructions to be based on stable oxygen isotopes in the carbonate shell [see Section Physical Variables (Temperature, Salinity)]. Most essential elements can diffuse through the mantle epithelia, gills and digestive gland and can be absorbed from ingested food and water. Because of isotopic fractionation, vital effects, and detoxification processes in the various transport pathways, elements other than oxygen are not precipitated in equilibrium with the ambient environment (see Sections Shell Growth—Measurement and Interpretation and Deciphering the Environmental Information in the Shell).

Organic materials from surrounding waters are incorporated in the soft tissue and shell of bivalves. Since soft tissues are continuously added and turned-over by metabolic processes, the stable isotope signature of bivalve soft tissue integrates environmental and dietary signals over the entire water column on relatively short timescales (Ellis et al., 2014). In contrast, material isolated within the shell mineral matrix is deposited in discrete annual increments, is not affected by subsequent metabolic processes (Bayne and Newell, 1983; Serban et al., 1988; Rosenberg and Hughes, 1991; Risk et al., 1996; Quitmyer et al., 1997) and can provide permanent proxy records of the depositional environment for the entire life span of the bivalve (Ellis et al., 2014). The proportions of metabolic and environmental carbon in the shell change with ontogeny during early growth and are likely species-specific (Lorrain et al., 2004), and this should be taken into account when using the stable carbon isotope as an environmental monitor.

Elements are typically reported as a molar ratio to calcium. The partitioning between the water and shell is expressed as a partition coefficient (DElement):

DElement = (Element/Ca)carbonate/(Element/Ca)water

Mg fractionation in inorganic CaCO<sup>3</sup> has been attributed to organic molecules within the calcifying solutions regulating the mineralization process (Orme et al., 2001; Elhadj et al., 2006), although it has also been reported that Mg incorporation into the crystal lattice is enhanced during inorganic precipitation (Stephenson et al., 2008). Other divalent cations, e.g., Sr, might also be affected by this fractionation preceding the mineralization process. Negatively charged carboxyl and sulfate groups, which influence electronic charging at the calcification site, are thought to be involved in regulating the biomineralization process and hence also elemental composition within biogenic CaCO3(e.g., Addadi et al., 2006; Marin et al., 2012).

# Shell Growth—Measurement and Interpretation

Several studies have shown that annual growth increments in long-lived bivalve shells can be used to establish absolute chronologies for the marine environment, similar to the annually resolved terrestrial records based on tree rings (Witbaard, 1997; Marchitto et al., 2000; Schöne et al., 2002; Schöne, 2003; Scourse et al., 2006; Butler et al., 2010, 2013). By measuring the distance between consecutive growth lines, time series of increment widths (growth increment series; GIS) can be defined for each individual shell. Because the patterns of growth within populations are synchronous, multi-shell chronologies can be constructed by crossdating individual shell GIS. Where the date of death of any individual in the chronology is known, precise calendar dates can be assigned to the whole chronology, which can therefore extend for many hundreds of years before the lifetime of any living specimen (e.g., Butler et al., 2010, 2013). As well as providing the basis for chronology construction, synchronous growth within a population also constitutes prima facie evidence that shell growth is responding to a common environmental driver, allowing environmental changes to be inferred from variation in the growth increment width (Witbaard, 1997; Ambrose et al., 2006; Carroll et al., 2011). Further, environmental indicators can be derived from the geochemical and structural properties of the shell, and these indicators are also precisely dated.

Measurement of growth increments is carried out on the polished surface of a shell that has been sectioned perpendicular to the growth lines along the axis of maximum shell growth (umbo to ventral margin; see **Figure 2A**). The increments may be measured either in the umbo region or along the outer shell margin (**Figure 2B**).

Growth in A. islandica does not occur throughout the year, but is restricted to a growth season, which is itself affected by water depth and the depth of the thermocline (Schöne et al., 2005b,c). The duration of the growing season remains more or less the same during ontogeny. The start of growth line formation is thought to occur shortly after the seasonal temperature maximum (Weidman et al., 1994). In addition to the prominent annual growth lines, the existence of daily growth lines in A. islandica has been supported by successfully linking the micro-growth pattern to stable isotope data (Schöne et al., 2005b). With knowledge of daily/seasonal growth rates (Schöne et al., 2005b) it is possible to position geochemical analyses very precisely in time and thus establish very precise proxy records for environmental monitoring.

While food availability is likely the main driver of growth in A. islandica, analysis of A. islandica chronologies (e.g., Butler et al., 2010; Mette et al., 2016), as well as laboratory experiments (Witbaard, 1997) have also shown a link between seawater temperature and shell growth, although this is rather weak in natural settings (Witbaard, 1997; Witbaard et al., 2001; Schöne et al., 2005b,c).

Brocas et al. (2013) constructed cross-dated chronologies from two populations of G. glycymeris, from the east and south coasts of the Isle of Man. They identified a common growth signal in the two populations, indicating that a common environmental driver controls growth across the two sites. The positive correlation between the chronology and SST (sea surface temperature) was found to be much stronger for G. glycymeris than for A. islandica from the same region (Butler et al., 2010). In addition, G. glycymeris chronologies have been shown to reflect synoptic scale signals originating in the North Atlantic (Reynolds et al., 2013). Since the habitat preference of G. glycymeris complements that of A. islandica, it is possible to crossdate G. glycymeris chronologies with those from adjacent populations of A. islandica, allowing longer multi-species chronologies to be constructed.

# Deciphering the Environmental Information in the Shell

In this section the existing and potential applications of molluscan sclerochronology to environmental monitoring will be discussed. These applications may make use of information encoded in the shell in various forms, including the growth increment pattern, the shell geochemistry, the shell crystal structure and the metagenome. Advantages and constraints associated with each of these archives will be discussed in the context of the relevant monitoring target.

#### Metal Pollution

Shell geochemistry has long been investigated as a potential means to record ambient seawater chemistry. The elemental composition of the shell can potentially be used not only to track long-term climate variability (e.g., Lazareth et al., 2003, 2006; Schöne et al., 2005a; Welsh et al., 2011; Butler et al., 2013) and improve climate predications but also to record environmental pollution events (e.g., Raith et al., 1996; Liehr et al., 2005).

Geochemical properties (trace element and heavy metal incorporation) offer a number of possibilities for the reconstruction of environmental variables. The elemental composition of the shell is strongly controlled by element availability and partitioning. The relationship between the chemistry of ambient water and shell biogeochemistry is complicated by multiple confounding factors (Carroll and Romanek, 2008; see refs. in Zuykov et al., 2013). Typically Element/Calcium ratios deviate from thermodynamically predicted values because of changes in growth rate (i.e., ontogeny) and crystal fabric (Swan, 1957; Gillikin et al., 2005a; Schöne, 2013; Schöne et al., 2013; Shirai et al., 2014).

More recently, significant advances in micro-scale analytical techniques have improved understanding of incorporation mechanisms. Calcium carbonate samples (tens of µg) from the outer shell layer are usually obtained by microdrilling/micromilling (**Figure 5**) for stable isotope mass spectrometry or ICP-OES element analysis, which can achieve a resolution on the order of weeks. Higher resolutions (down to days) can be achieved using laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP MS) or secondary ion mass spectrometry (SIMS). Electron microprobe analysis (EPMA) allows for even higher resolution and provides the means to study the element distribution in bivalve shells with respect to their microstructure (Shirai et al., 2014).

Metals occur naturally in the environment, but their concentrations have increased since the beginning of mining activities in the Iron Age (800 BC–AD 100 in western Europe; Hooke, 2000; Breitenlechner et al., 2010; Wells, 2011). Since the industrial period, anthropogenic activities (mining, agriculture, combustion, and the use of metal containing products) have significantly increased pollution locally and globally (Larsen et al., 2011). Non-essential metal pollutants, such as mercury (Hg), cadmium (Cd), and lead (Pb), can affect the central nervous system and cause kidney damage in mammals. The European Union has regulated these metals by adopting the Water Framework Directive (2013/39/EU), in which environmental quality standards (EQSs) of 45 prioritized substances have been defined for water and biota samples. Other metals (e.g., Cu, Zn, Fe, Co, Mo, Se, Mg, Ca, and Mn), while they are an essential part of metabolic and biochemical processes (e.g., enzymes), can cause toxic effects at higher concentrations. Bivalves take up metals both through food and directly from the seawater and incorporate them in their shell and soft tissue. Assuming that metal accumulation is more concentrated in the soft tissue of bivalves (Brown and Depledge, 1998), most research efforts have focussed on soft tissue metal accumulation. However, significant advantages of using the shell to monitor metal contamination include less variability (Bourgoin, 1990; Lingard et al., 1992), serial incorporation of elements over the entire period of shell formation, higher preservation potential even after the organism's death, and relatively cheap and easy storage (Protasowicki et al., 2007).

The preference in monitoring programmes for the use of soft tissue in bivalves to record and measure pollutant concentrations in water (e.g., Goldberg et al., 1978; Larsen et al., 2011) likely reflects generally greater concentrations of metals in some organs; as a result there is little information in the literature about the quantitative relationship between metal concentrations in soft tissue and in the shell. For example, Pb is bioaccumulated by living organisms and more particularly by marine invertebrates (Neff, 2002). Chow et al. (1976) found that both the soft tissue and the shell of mussels reflected Pb concentrations in the environment, with higher concentrations in more urban areas

indicating anthropogenic pollution. If such relationships can be reliably modeled, some monitoring studies could be simplified, since shells are easier to store than soft tissue (Koide et al., 1982).

Koide et al. (1982) found that Zn, Pb, and Cd levels were higher in the tissues of Mytilus edulis and M. californianus than in their shells, while conversely Cu concentrations were higher in the shells. Bourgoin (1990) also found higher Pb levels in the tissues of M. edulis than in the shell (specifically the nacre). In a trace metal study using M. edulis from the Baltic Sea, Protasowicki et al. (2007) emphasized the importance of the actual metal bioavailability at the sampling location. For the cockle Chione (Austrovenus) stutchburyi, Pb concentrations in the shell were similar to those in the soft tissues (based on dry weight; Purchase and Fergusson, 1986). Moreover, Gillikin et al. (2005b) found that there is large inter- and intra-annual Pb variability within shells of the clam Mercenaria mercenaria, suggesting that year to year, as well as intra-annual variations in Pb/Ca ratios, could potentially be interpreted (**Figure 6**). Even though it appears that concentrations of metals are generally lower in the shell compared to tissue, the feasibility of using bivalve shells for investigating metal contamination in the marine environment has nevertheless been demonstrated, and it may be that the practical advantages of using the chemically stable and easily storable shells outweigh the analytical advantages of the higher concentrations in the tissue, increasingly so as more sensitive techniques for measuring the elemental makeup of shells are developed (please see Supplementary Table I).

Arctica islandica has been shown to be an effective bioindicator for contaminated sediments (Steimle et al., 1986), and its longevity (e.g., Ropes et al., 1984) makes it particularly suitable as a record of historical contamination events and of long term trends. By carrying out measurements in successive internal growth increments it is possible to obtain a chronological reconstruction of the metal content. Raith et al. (1996) showed using LA-ICP-MS that the Pb concentration in the shell of A. islandica reflects changes in seawater metal pollution. This study indicated that high-resolution methods such as LA-ICP-MS offer an accurate method of determining pollution levels in the environment, and can potentially be used in tracing the source of the pollution. This was confirmed by Richardson (2001) for Pb and Zn using shells of the horse mussel Modiolus modiolus. In a detailed comparative study, Liehr et al. (2005) investigated the potential use of A. islandica for pollutant biomonitoring, analysing the heavy metal concentrations in shells and soft body tissue of specimens from the western Baltic Sea (**Figure 7**). Samples were taken from an historic dump site in the inner Mecklenburg Bight and from an adjacent, less contaminated site (representing background contamination of the western Baltic Sea). The soft tissue from the dumping site had significantly higher Pb and Cu concentrations than those from the reference site. No difference in Zn concentrations was found. For the shell, Liehr et al. (2005) used LA-ICP-MS to analyze Cu, Pb and Zn concentrations and found that these were higher at the dumping site, indicative of contamination (**Figure 7**). The chronological reconstruction (profile) of the measured metals in the shell showed no significant trend, most likely because the analyzed specimens only reached an age of ∼35 years and did not record the initial contamination (40–45 years). Liehr et al. (2005) concluded that processes such as bioturbation and other physio-chemical processes at the sediment water interface might affect metal incorporation into the shell. Nevertheless, this study showed the potential for pollutant biomonitoring using sclerochronological methods.

Uranium is less well-studied in bivalve shells; examples include a study of U/Ca in aragonite of the shells of the marine cockle, Cerastoderma edule (Price and Pearce, 1997) and a study of U/Ca ratios in aragonite of freshwater mussel shells as an indicator of uranium pollution through time from a copper– uranium mine in Australia (Markich et al., 2002).

Barite (BaSO4) is a naturally occurring mineral in many sediments and is a major component of almost all drilling muds (Hartley, 1996). Although barite in drilling fluids and PW is usually relatively inert (Neff, 2002), under reducing conditions it can dissolve slowly, releasing Ba into the overlying water column (Boothe and Presley, 1989). It may be possible to use the Ba/Ca ratio in bivalve shells to track the release over time of Ba in the proximity of drill cutting disposal. Marali et al. (2016) have measured Ba/Ca ratios in multicentennial A. islandica chronologies from four sites (Gulf of Maine, Iceland, Faroe Steinhardt et al. Bivalve Sclerochronology in Environmental Monitoring

Islands, and Isle of Man) to show that Ba/Cashell spikes are synchronous within chronologies, and that this synchroneity is independent of ontogenetic age.

Furthermore, Faubel et al. (2008), Lopes-Lima et al. (2012), Nuñez et al. (2012), and Zuykov et al. (2013) demonstrated that several metals, fabricated metal nanoparticles, and organic contaminants, can bioaccumulate in low concentrations in the extrapallial fluid, and are able to disturb the calcification of shell by forming distinct structures on the internal shell surface.

# Physical Variables (Temperature, Salinity)

#### **Temperature and Salinity**

Oxygen isotopes. The stable oxygen isotope ratio (δ <sup>18</sup>Oshell) of carbonate structures reflects both temperature-driven fractionation and the δ <sup>18</sup>O of the ambient seawater (δ <sup>18</sup>Osw) from which it is precipitated. The relationship between δ <sup>18</sup>Oshell and seawater salinity is complicated because of the varying isotopic composition of freshwater and the meridional evaporation/precipitation pattern (e.g., Schmidt et al., 2007). Commonly, in order to calculate temperatures from δ <sup>18</sup>Oshell, a slightly modified (Sharp, 2007) version of the (Grossman and Ku, 1986) equation is used:

$$\mathrm{T\_{\mathbb{\delta^{18}O}}\left(^{\circ}C\right)} = 20.6 - 4.34^\* \left(\mathbb{\delta^{18}O\_{\mathrm{shell}}} - \left(\mathbb{\delta^{18}O\_{\mathrm{sw}} - 0.27\right)\right)$$

with δ <sup>18</sup>Oshell calibrated to the Vienna PDB scale and δ <sup>18</sup>Osw calibrated to the V-SMOW scale. A δ <sup>18</sup>Oshell change of 1‰ is equivalent to a seawater temperature change of ∼4.34◦C, assuming constant δ <sup>18</sup>Osw (and hence constant salinity). As an example, Marsh et al. (1999) used δ <sup>18</sup>Oshell analysis on annual bands from A. islandica, together with empirical modeling, as proxy evidence for a cold episode in the Gulf of Maine in the 1880s. In a later study, a subseasonal temperature record was established, based on a single A. islandica specimen from the North Sea, which indicated a 1◦C warming of SST over a 100 year period (Schöne et al., 2004; **Figure 8**). However, note that the reconstruction of temperature based on shell δ <sup>18</sup>O is complicated by likely variability in δ <sup>18</sup>Osw, particularly in coastal waters. It is therefore important to isolate a reliable and independent proxy for either temperature or salinity so that both variables can be reconstructed without ambiguity.

Royer et al. (2013) demonstrated that seawater temperatures can be accurately estimated through sclerochronological and δ <sup>18</sup>O analyses using shells of G. glycymeris. Comparing δ <sup>18</sup>Oshellderived temperatures with water temperatures measured at several monitoring stations in the Bay of Brest, the study found that the seasonal pattern of δ <sup>18</sup>Oshell data indicated a May– October growing season. Furthermore, based on reconstructed temperatures from two different sampling sites, Royer et al. (2013) inferred that shell growth is constrained by a lower thermal threshold of ∼13◦C.

#### **Potential methods for independent temperature calibration.**

Other potential methods for the independent reconstruction of temperatures from calcium carbonate shells include ∆<sup>47</sup> clumped isotope thermometry (Ghosh et al., 2006; Henkes et al., 2013)

Individual δ <sup>18</sup>Oaragonite-derived temperatures (Tδ18*O*) are plotted as dots. Mean temperatures (open circles) during February through September [Tδ18*<sup>O</sup>* (Feb–Sep)] were calculated as weighted averages from individual Tδ18*<sup>O</sup>* values. Exceptional Tδ18*<sup>O</sup>* and Tδ18*<sup>O</sup>* (Feb–Sep) values are highlighted. Long-term mean as indicated by the dashed line is 10.16◦C (Figure from Schöne et al., 2004).

and δ <sup>44</sup>/40Ca paleothermometry (e.g., Nagler et al., 2006; Hippler et al., 2013). While these techniques show promise, they are limited at present because of the required sample amount and the large analytical uncertainty.

Sr/Ca and Mg/Ca. The partitioning of Sr/Ca and Mg/Ca is strongly affected by a complicated interplay between biological, physio-chemical and kinetic processes during biomineralization and has so far only yielded controversial results in respect of thermometry applications (Toland et al., 2000; Gaetani and Cohen, 2006; Foster et al., 2008, 2009; Schöne et al., 2011b, 2013). Future ultra-high-resolution analyses are needed to test whether particular portions of the aragonitic shell (e.g., near the growth lines) of A. islandica or other bivalves can consistently be used for Sr/Ca or Mg/Ca (paleo)thermometry.

The crystal fabric in A. islandica has been shown to be influenced by environmental variables (Schöne et al., 2013; Stemmer et al., 2013). Heterogeneities in the crystal fabric coincide with heterogeneous distribution of Sr/Ca, Mg/Ca, and Ba/Ca (Schöne et al., 2010, 2013; Shirai et al., 2014). During the formation of the annual growth lines, extremely slow growth occurs and Sr/Ca and Mg/Ca are seemingly incorporated in equilibrium with the ambientseawater. However, during the rest of the growing season, ratios of Sr/Ca and Mg/Ca remained below thermodynamic equilibrium values. The incorporation of Ba does not seem to be related to changes in crystal fabric and appears to be less influenced by vital effects, as Ba/Ca peaks occur erratically at different times of the year.

Evidence so far suggests that Sr/Ca variations are minimally influenced by temperature (e.g., Schöne et al., 2013; Shirai et al., 2014). Schöne et al. (2013) concluded, from results consistent with those of Foster et al. (2009), that crystal size, shape, and orientation are very influential in trace element variability. Similarly, Shirai et al. (2014), using very high analytical resolution showed that the Sr/Ca difference between the outer and middle shell layers were associated with microstructural differences (**Figure 9**). They concluded that crystal morphology cannot entirely explain the compositional changes of Sr within the shell, since while different Sr/Ca ratios were observed between the irregular prismatic crystals and the adjacent acicular crystals, the Sr/Ca ratio was not correlated with the length and shape of the crystals. It seems that organic composition at the site of calcification is an additional control on Sr/Cashell.

Na/Ca. Sodium (Na) has been proposed both as an indicator of post depositional alteration and as a proxy for salinity. As one of the major constituents of sea salt, Na has clear potential as a salinity proxy (the other constituent, like chlorine, occurs at far lower concentrations Kitano et al., 1975). Early research on modern marine mollusc shells suggested that seawater salinity had the strongest control on shell Na content (Rucker and Valentine, 1961; Pilkey and Harris, 1966; Gordon et al., 1970; Kitano et al., 1975), although at the time it was impossible to tell whether Na was bound in the crystal lattice or in microscopic seawater inclusions. However, the very low chlorine concentrations suggest that Na is structurally bound in the crystal lattice (Ishikawa and Ichikuni, 1984). Sodium ions cannot directly substitute for Ca2<sup>+</sup> (because of the difference in ion radius and charge) during calcite precipitation (Ishikawa and Ichikuni, 1984). Primarily, Na<sup>+</sup> incorporation depends on the activity of Na in seawater, which is a function of its concentration and, to a lesser extent, its activity coefficient (Ishikawa and Ichikuni, 1984). Increasing salinity (and hence [Na+]) increases the activity of Na whereas seawater temperature, over the relevant range, has only a minor effect on the activity coefficient of Na. Hence the effect of temperature on Na incorporation is negligible (Ishikawa and Ichikuni, 1984; Delaney et al., 1985; Lea et al., 1999; Zeebe and Wolf-Gladrow, 2001). Findlater et al. (2014) found a clear relationship between Na concentration and inferred salinity in fossil and modern bivalve shells. It still remains to be tested whether Na incorporation is species-specific, whether the same relationship between Na and salinity occurs in both aragonite and calcite, and how sensitive the potential salinity proxy is.

Crystallography. A recent study investigated the use of the crystal structure of the common cockle, Cerastoderma edule, as a proxy for seawater temperature and salinity (Milano et al., 2016). Living specimens, exhibiting shell growth increments with tidal resolution (**Figures 2C–E**), were collected after a year of continuous temperature and salinity measurements. Using scanning microscopy, shell microstructures were analyzed, showing that size and shape of the mesocrystals strongly correlated with water temperature during the growing season (May–September). The results suggest that shell microstructures of C. edule may serve as a new, independent proxy for water temperature. Future research will be needed to test whether the relationship between calcite microstructures and temperature can also be observed in other species.

FIGURE 9 | Electron probe microanalyses (EPMA) showing the micrometer-scale distribution of (A) Sr/Ca, (B) S/Ca, (C) Mg/Ca in the shell of Arctica islandica. Signal intensity (cumulative count) ratio is illustrated as a map using a color scale (ct/ct: count/count) shown at lower right of each elemental map. (D) Sr/Ca and Mg/Ca maps are partly overlapped on S/Ca map for the comparison among each elemental map. Position of annual and sub-annual growth lines are clearly recognized and indicated by the blue and pink lines, respectively, at the top of shell signal images and on the yellow lines (Figure from Shirai et al., 2014).

#### Ecosystem Variables (Trophic Levels, Food Supply, Metagenome)

#### **Food web analysis using stable isotopes in the organic shell material**

Stable isotope analysis is frequently used to investigate patterns of productivity and (based on isotopic fractionation) degradation and metabolic transfer of organic matter through the water column (Peterson and Fry, 1987). Carbon and nitrogen isotope ratios record information about the types of primary producers and further alterations as carbon is transferred through the food web. Therefore, carbon and nitrogen isotopes can be used to assess different classes of primary producers or benthic vs. pelagic marine production (DeNiro and Epstein, 1978; Rounick and Winterbourn, 1986; Farquhar et al., 1989).

O'Donnell et al. (2007) showed that proteins in the organic matrix of modern and fossil Mercenaria shells can resolve spatial and temporal changes in dietary carbon sources. However, the organic constituents in shell mineral material do not faithfully record whole-diet carbon isotopic ratios. Controlled feeding experiments have shown that an offset of between 1 and 6‰ can exist between diet and bone collagen bulk δ <sup>13</sup>C values, depending on the composition of the diet (DeNiro and Epstein, 1978; Howland et al., 2003). Nevertheless, measurements of compound specific δ <sup>13</sup>C should account for variation in the amino acid makeup of the organic matrix (Howland et al., 2003) and yield results directly reflective of dietary sources (Sykes et al., 1995; Ellis et al., 2014).

Combining stable carbon isotopes with nitrogen isotopes could refine modern and historical trophic assessments and distinguish natural from anthropogenic influences on coastal ecosystems, e.g., coastal nitrogen input, and can ultimately be used to define effects of eutrophication on an ecosystem-level (Carmichael R. H. et al., 2004; Carmichael R. et al., 2004; Valiela, 2009). Delong and Thorp (2006) have used stable carbon and nitrogen isotopes in the shell periostracum to study food web dynamics and changes in trophic complexity over time. Using very high resolution techniques, it may also be possible to analyse stable nitrogen isotopes in the organic fraction of the shell matrix.

#### **Food web analysis using stable isotopes in the inorganic shell material**

Studies of long-term (multidecadal to century-long) δ <sup>13</sup>Cshell timeseries in A. islandica have indicated that age-related vital effects are limited to the shell portions close to the umbo of individuals that have exhibited very rapid growth during early ontogeny (Butler et al., 2011). Using an extensive set of >3500 individual δ <sup>13</sup>Cshell values, Schöne et al. (2011a) found that shell carbonate is secreted with a constant offset from expected equilibrium (by −1.54 to −2.7 ± 0.2‰), which would correspond, assuming δ <sup>13</sup>Cfood ∼ −25‰, (Erlenkeuser, 1976), to a contribution of metabolic carbon of between 6.2 and 10.8 ± 0.8%. Using data from laboratory and in situ experiments, Beirne et al. (2012) later confirmed these findings. Further, they supported the conclusion that δ <sup>13</sup>Cshell values of A. islandica provide a robust proxy for seawater dissolved inorganic carbon (DIC) values, since they did not observe an ontogenetic effect or an impact of growth rates on the measured δ <sup>13</sup>Cshell values. They found the following relationship between δ <sup>13</sup>Cshell and δ <sup>13</sup>CDIC:

$$8^{13} \text{C}\_{\text{DIC}} = 8^{13} \text{C}\_{\text{shell}} - 1.0 \text{\%o} (\pm 0.3 \text{\%o})$$

Microchemical analysis of fish otoliths has recently been suggested as a method of food web analysis, for example by determining the residency of the fish (how long and where they reside e.g., on oil structures; Fowler et al., 2015). Trace elements are incorporated into the calcium carbonate of the otoliths, leaving a record within the otoliths of residency seawater conditions (Campana et al., 2000; Gillanders and Kingsford, 2000; Fowler et al., 2015). It may be possible to observe distinct geochemical signatures in the otolith, characteristic of particular structures, that can potentially be used to evaluate residence times at individual sites and help assess habitat value and contribute to decommissioning decisions.

Although the potential of otolith microchemistry for longterm reconstructions is limited by the relatively short lifespans of fish (usually <25 years), it may be possible to construct otolith chronologies using otoliths of known date held in national fisheries archives. Ultimately, the combination of fish otolith chronologies with centennial scale bivalve chronologies may enable researchers to link pelagic and benthic processes over extended periods.

#### **Mn/Ca**

Manganese (Mn2+) has been suggested as a proxy for increased riverine discharge events (Lazareth et al., 2003) and productivity (e.g., Vander Putten et al., 2000; Lazareth et al., 2003). Intra-shell variation in Mn/Ca may reflect the seasonal variation of seawater Mn2<sup>+</sup> concentrations (Freitas et al., 2006; Bougeois et al., 2014), indicating that redox processes control its concentration in seawater. One source of marine manganese (Mn2+) is terrestrial input via rivers. Normally it is slowly removed from solution by oxidation to (Mn4+), but reducing conditions in sediments and oxygen depletion within the water column can recycle Mn2<sup>+</sup> back into solution. Hence oxygen depletion potentially supports Mn incorporation into the shell. Markich and Jeffree (1994) showed that Mn2<sup>+</sup> can be substituted for Ca2<sup>+</sup> during uptake from the external medium. Temperature increase may also support incorporation of manganese as a result of increased Mn2<sup>+</sup> uptake and transfer to the calcification site (Wada and Fujinuki, 1976).

Freitas et al. (2006) suggested that the intra-annual variation of Mn/Cashell reflects the seasonal variation of seawater Mn2<sup>+</sup> concentrations, but they also point out that further research is necessary to test the effect of other variables controlling seawater Mn2<sup>+</sup> concentrations (e.g., primary productivity, oxygen concentrations, temperature). Manganese has also been used as a fast chemical marking technique using cathodoluminescence; this might be an additional method to estimate shell growth rate (e.g., Langlet et al., 2006; Barbin et al., 2008; Lartaud et al., 2010). Potentially, Mn/Ca concentrations could be used in monitoring and baseline studies concerned with eutrophication, hypoxia and deep sea mineral exploration (e.g., manganese crusts).

#### **Ba/Ca**

Barium is a minor constituent of seawater, found as Ba2<sup>+</sup> at very low concentrations (∼34 nmol/L). Planktonic organic matter is thought to deliver Ba to the benthos in particulate form, where it is digested by filter feeders. The incorporation of Ba into the shell is therefore linked to primary productivity and metabolic activity (Lazareth et al., 2003; Carré et al., 2006). Nevertheless, a recent study using A. islandica indicated that Ba/Ca peaks (>40 µmol/mol) can occur at different times during the growing season, raising further questions about the factors that influence Ba/Ca concentrations in A. islandica shells (Schöne et al., 2013) and showing the need to better understand the controls and mechanisms of Ba incorporation. Future studies could use multiple shell proxies and culturing experiments to identify the major factors controlling Ba incorporation and to refine proxy calibrations. For example, following a suggestion by Thébault et al. (2009) that the Li/Ca ratio in A. islandica may be associated with Li-rich silicate particles and terrestrial weathering and could ultimately be used as a proxy for river discharge, it may be possible to combine Li/Ca and Ba/Ca measurements to determine the extent to which terrigenous sediment input by river discharge is contributing to the Ba distribution in shell material. There are other studies which suggest that Ba/Ca ratios in bivalve shells track intraannual variability in river discharge (Carroll et al., 2009) or salinity (Gillikin et al., 2006, 2008; Poulain et al., 2015). Poulain et al. (2015) confirmed the conclusion of Barats et al. (2009), working with only the background Ba/Ca signal, that there is a strong inverse correlation between salinity and Ba/Cashell, which suggests that Ba/Cashell could offer a highresolution proxy for the reconstruction of salinity fluctuations within estuarine and nearshore waters and could also be used to distinguish the salinity and temperature signals in the stable oxygen isotope ratio. They also found a positive correlation between Ba/Cashell and Ba/Casw.

Poulain et al. (2015), using the Manila clam Ruditapes philippinarum, demonstrated that the seawater Ba2<sup>+</sup> concentration was reflected at almost daily resolution in Ba/Cashell and might be used as a proxy with very high temporal resolution. The same study also found variable Mg/Cashell ratios, suggesting that the incorporation of magnesium into shell carbonate is strongly regulated by the organism and not by environmental conditions.

#### **Mo/Ca**

Molybdenum (Mo) is one of the most abundant transition group metals in seawater, mainly present as the oxyanion MoO4<sup>−</sup> 2 in oxygenated environments (Collier, 1985). Mo is considered to be a conservative element in seawater, with low concentrations around 110 nmol/L and little apparent influence of biogeochemical processes on its concentration (Collier, 1985). Coastal Mo distribution is also influenced by freshwater-seawater mixing (Dalai et al., 2005). Some studies, however, have indicated non-conservative behavior in coastal waters, both at the sediment water interface (SWI; Crusius et al., 1996; Adelson et al., 2001; Chaillou et al., 2002; Elbaz-Poulichet et al., 2005) and in the water column (Tuit and Ravizza, 2003; Dellwig et al., 2007). When phytoplankton is decomposed by bacteria, organic compounds are released and Mo-enriched aggregates are subsequently formed which settle on the sediment water interface where they are rapidly decomposed by microbial activity, contributing to a substantial release of Mo in bottom waters (Dellwig et al., 2007). Barats et al. (2010) evaluated ([Mo]/[Ca])shell profiles of Pecten maximus as a potential record of specific biogeochemical processes occurring at the SWI. They found that Mo incorporation is promoted by the significant spring pelagic productivity, although they could not directly link ([Mo]/[Ca])shell maxima with specific phytoplankton species. Barats et al. (2010) found that the background partition coefficient (DMo) indicates anionic Mo precipitation pathways within the calcium carbonate shell but also consider a particulate phase uptake of Mo into the shell to play a role. In order to explain ([Mo]/[Ca])shell maxima in the tropical bivalve species Comptopallium radula, Thébault et al. (2009) suggested the ingestion of N2-fixing cyanobacteria. This seems to be confirmed by the 7-year Mo/Cashell record with late spring maxima, which were not directly related to the spring bloom biomass maximum but rather to a post-bloom period characterized by nutrient depletion (silicic acid and nitrate depletion) and Pseudonitschia spp. dominance. Mo inputs at the SWI can be induced by a diatom biogenic material downward flux and might therefore enrich Mo/Ca ratios of scallop shells at the SWI. Hence, ([Mo]/[Ca])shell records in Pecten maximus may serve as a new proxy for biomonitoring studies of primary production, including eutrophication, in temperate coastal environments.

#### Ocean Acidification

#### **Crystallography**

Increased pCO<sup>2</sup> can affect physiology, acid-base homeostasis and biomineralization in bivalves, including decreased growth rates and altered shell structure, composition and mechanical properties (Michaelidis et al., 2005; Gazeau et al., 2007, 2013; Kurihara et al., 2007; Beesley et al., 2008; Kurihara, 2008; Ellis et al., 2009; Dickinson et al., 2012; Ivanina et al., 2013; Fitzer et al., 2014a,b, 2015, 2016).

Milano et al. (2016) (see Biomineralization: Environmental vs. Biological Effects) showed that crystallography strongly correlates with water temperature. However, changes in the carbonate chemistry of ambient seawater (i.e., pH), related to ocean acidification, are also likely to affect calcification and cause changes in the microstructure of shell calcite. A number of studies have recently shown effects of ocean acidification on shell material properties and crystallography (Fitzer et al., 2014a,b, 2015, 2016). Fitzer et al. (2014a) found that M. edulis appears to improve its ability to continue biomineralization, although this occurs at some cost to the structural integrity of the mussel shell. At pCO<sup>2</sup> (>1000 µatm) various mechanisms that support biomineralization were observed, including increased protein metabolism, chitinase mRNA and tyrosinase gene expression. Later, Fitzer et al. (2016) confirmed that OA reduces the crystallographic control of shell formation. The study suggests that in order to combat shell damage, more amorphous calcium carbonate formation is induced, lowering the crystallographic control in mussels.

#### **U/Ca**

A study by Gillikin and Dehairs (2013) investigated the potential of shell U/Ca as a proxy for ocean acidification (OA). They measured U/Cashell in the aragonitic clam Saxidomus giganteus. Since the elemental analyses were perfectly aligned with δ <sup>18</sup>O analyses (Gillikin et al., 2005a), it was possible to assign interand intra-annual dates using the δ <sup>18</sup>O profiles, allowing a direct comparison of elemental profiles between two shells. The analyzed shells exhibited seasonal cycles in U/Cashell up to ∼6 years of growth, after which the concentration decreased below the detection limit. The authors concluded that U/Ca ratios in S. giganteus shells are not controlled by pH (or [CO2<sup>−</sup> 3 ]). However, it has to be pointed out that the U/Ca data was only compared to pH and not to other variables in the carbonate system.

#### Future Application Approaches Valve-Movement Behavior in Bivalves—A Tool to Calibrate Shell Signatures and Potential Applications for Water Quality Control

Valve movement behavior of bivalves can be used to indicate physiological rate functions (e.g., feeding rate). Studies on A. islandica have shown that the frequency of siphon opening and growth of the shell and tissue were strongly related (Witbaard, 1997). Generally, filter-feeding bivalves reduce their filtration rate by reducing or completely closing their valve-gape (e.g., Riisgård and Larsen, 2015). Video recordings of valve-gape responses of M. edulis to absence or presence of algal cells in the ambient water have revealed that the critical algal concentration (at which concentration the bivalve closes its valves) is between 0.5 and 0.9 mg chl a l−<sup>1</sup> (Riisgård et al., 2006; Pascoe et al., 2009).

The principle of a valve position monitor is based on the measurement of electromagnetic field strength between two electronic sensors permanently fixed on the outside of each of the valves and facing each other perpendicularly. Both sensors are connected to a data logger and an energy source (battery). On a pre-set frequency, one of the sensors receives an electric pulse, producing an electro-magnetic field that is detected by the sensor on the opposite valve. What is recorded is the strength of the electro-magnetic field, which depends on the distance between the two sensors (i.e., the valve opening). The distance between the sensors is expressed as the percentage of the signal at maximum valve opening. The time-stamped signals are transferred to the data logger and saved, enabling valve opening behavior to be monitored over time, and allowing it to be related to different a(biotic) parameters (e.g., algal concentrations, sediment turbidity, metal/toxin concentrations in seawater).

High frequency measurements have been carried out to address fine-scale bivalve behavioral physiology (e.g., Wilson et al., 2005) in detail. Such techniques may involve: assessment of valve gape, siphon movements (changes in aperture), filtration and pumping behavior in relation to associated environmental parameters such as depth, light, temperature, particulate matter, food availability, and predator interactions (e.g. Ropert-Coudert and Wilson, 2004).

Several studies (e.g., Manley and Davenport, 1979; Kramer et al., 1989; Huebner and Pynnönen, 1992; Markich et al., 1996; Fdil et al., 2006; Schwartzmann et al., 2011) have confirmed that valve movement behavior can be used to sensitively quantify biological reactions in real-time (**Figure 10**) for assessing the toxicological effects of metal exposures. The observations found that upon exposure to toxic concentrations of metals, bivalves have the ability to reduce the exposure of their soft tissues for extended periods by closing their valve (Manley and Davenport, 1979; Kramer et al., 1989; Salánki and Balogh, 1989; Huebner and Pynnönen, 1992).

Hence, bivalves are potentially useful as biological early warning systems of water quality (Matthias and Römpp, 1994; de Zwart et al., 1995). Furthermore, observations of bivalve gape and the siphon area might find limited application in areas where turbidity is high or for monitoring burrowing behavior in bivalves (Robson et al., 2009). Although monitoring of valve gape behavior in bivalves has been reported in several studies, no published research has yet focused on connecting shell growth and growth increment patterns with valve gape behavior. If such a connection can be made (Ballesta-Artero et al., 2016, **Figure 10**), there is potential to use shells to indicate historic changes in levels of nutrient supply, or the presence of threshold levels of toxins/harmful substances in the ambient environment.

# CONCLUSIONS

We find significant potential in the use of proxy archives in bivalve shells to establish long baseline conditions for environmental monitoring. For very long baselines, this is linked to the longevity and capacity for crossdating of certain species (in particular Arctica islandica and Glycymeris glycymeris), but other, less long-lived species can still be very useful, especially in remote areas such as parts of the Arctic that lack consistent instrumental records (Ambrose et al., 2006; Carroll et al., 2014).

Although the soft tissues in bivalves have long been used in biomonitoring, their use for contamination rate and recovery trend assessment relies on repeated measurements, which are labor intensive and expensive. Furthermore, toxins and heavy metals which are not excreted accumulate in the body tissue, so that concentration changes in the environment are overprinted over time by biological processes, making the recovery of historical changes in these contaminants difficult. Signals in the shell (**Figure 11**), on the other hand, constitute a stable and temporally sequential archive whose value and significance will likely increase as more sensitive measurement techniques allow analyses to be carried out using smaller amounts of material, and as more linkages are made between shell geochemistry and the ambient environment.

Specifically, the incorporation of elements into the shell matrix can be linked to environmental changes (e.g., variability of elements in seawater, primary productivity, eutrophication; Barats et al., 2008, 2009, 2010; Gillikin et al., 2006; Holland et al., 2014), so that trace element and heavy metal concentrations in bivalve shells can provide suitable bioindicators (**Figure 11**) for some descriptors of the Marine Framework Strategy Directive. It remains the case that more needs to be known about

the relationship between bioavailability and incorporation of elements into the shell calcite, so that the shell concentrations can be robustly interpreted. Detailed investigations of the mechanisms of incorporation into the shell material will be an essential part of future research. For example, by undertaking continuous sampling and analysis of the composition of the extra pallial fluid (EPF) during controlled growth experiments in culture, shell geochemistry could be related to preset culture parameters (e.g., seawater composition), and the composition of the EPF. Another valuable direction of research would be the analysis of heavy metal isotope ratios in order to assign specific sources to metal contamination.

Further insight into the incorporation of trace elements, heavy metals or other contaminants into the shell can be obtained from studies of bivalve behavior in highly fluctuating environments (e.g., Jørgensen, 1988; Wilson et al., 2005, 2008; Robson et al., 2007, 2009; Riisgård and Larsen, 2015). Ultimately, this information will help researchers to understand in detail the physico-chemical interactions between the environment and the animal, and in particular the impact of biological effects and environmental variables on shell geochemistry and growth. For example, techniques are now available for studying valve gape behavior and filtration activity and connecting them to shell growth. Combining observations of long-term valve gape behavior with continuous measurement of seawater chemistry and subsequent measurement of shell geochemistry could help to better understand the processes by which trace elements, heavy metals and other harmful substances are incorporated into the shell and the mechanisms that control the calcification processes.

Other environmental proxies in the shell include the stable isotopes of oxygen, nitrogen and carbon and the shell microstructure (e.g., Raith et al., 1996; Liehr et al., 2005; Lazareth et al., 2006; Welsh et al., 2011), as well as the increment widths themselves which can be an effective indication of nutrients in the environment.

Reduced shell growth can be indicative of less favorable conditions, for example those induced by anthropogenic disturbance (Stott et al., 2010). Increased shell growth can indicate higher nutrient supply (including eutrophication) which can be linked to aquaculture (Stott et al., 2010). Physical damage to larger shells (scars) often reflects impact with fishing gear (Gilkinson et al., 1998; Ramsay et al., 2000), while in smaller shells it is more likely to be a result of factors other than fishing (predator attacks or reburrowing).

The range of applications based on sclerochronology now offers a wide and increasing repertoire of techniques for monitoring natural and anthropogenic environmental variability and distinguishing between them, with applications to a broad range of commercial and regulatory users (**Figure 11**). Future research should be targeted at understanding the processes of incorporation into the shell, and at developing regionally specific and species specific calibrations to enable the robust interpretation of the shell geochemistry.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

The Leading author, JS was responsible for the conception of the work, information collection, analysis and interpretation, as well as drafting the article. PB and MC were responsible for the conception of the work, the critical revision of the article and the final approval of the version to be published. JH was involved in the conception of the work, the critical revision of the article.

#### FUNDING

Funding for this study was kindly provided by the EU within the framework (FP7) of the Marie Curie International Training Network ARAMACC (604802). Additional funding for MLC was provided by the Research Council of Norway (Project # 227046).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00176


Mytilus edulis shells. J. R. Soc. Interface 12:20141227. doi: 10.1098/rsif.2 014.1227


Geochim. Cosmochim. Acta 70, 395–407. doi: 10.1016/j.gca.2005. 09.015


virginica and Mercenaria mercenaria. Comp. Biochem. Physiol. Part A Mol. Integr. Physiol. 166, 101–111. doi: 10.1016/j.cbpa.2013.05.016


the north-eastern United States? Fish. Oceanogr. 8, 39–49. doi: 10.1046/j.1365- 2419.1999.00092.x


Bivalvia), from NW Europe as marine environmental archives. in Bulletin de l'Institut océanographique (Musée océanographique), 105–111. Available at: http://cat.inist.fr/?aModele=afficheN&cpsidt=2505512 (Accessed July 2, 2015).


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Steinhardt, Butler, Carroll and Hartley. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# High Frequency Non-invasive (HFNI) Bio-Sensors As a Potential Tool for Marine Monitoring and Assessments

Hector Andrade<sup>1</sup> \*, Jean-Charles Massabuau2, 3, Sabine Cochrane1, 4, Pierre Ciret 2, 3 , Damien Tran2, 3, Mohamedou Sow2, 3 and Lionel Camus <sup>1</sup>

*<sup>1</sup> Akvaplan-niva AS, Tromsø, Norway, <sup>2</sup> Environnements et Paléoenvironnements Océaniques et Continentaux, UMR 5805, University of Bordeaux, Arcachon, France, <sup>3</sup> Centre National de la Recherche Scientifique, Environnements et Paléoenvironnements Océaniques et Continentaux, UMR 5805, Arcachon, France, <sup>4</sup> SALT Lofoten AS, Svolvær, Norway*

Marine ecosystems all over the globe are facing multiple simultaneous stressors including rapid climatic change and increased resource exploitation, such as fishing, petroleum exploration and shipping. The EU-funded DEVOTES project (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) aims to better understand the relationships between pressures from human activities and climatic influences and their effects on marine ecosystems. To achieve these goals, it is necessary among others, to test and validate innovative monitoring tools to improve our understanding of ecosystem and biodiversity changes. This paper outlines the application of a high frequency non-invasive (HFNI) valvometer as a potential tool for long-term marine monitoring and assessments. The principle of the method is based on the regular gaping behavior (closing and opening of the valves) of bivalve molluscs and the fact that physical or chemical stressors disrupt that gaping reference pattern. Bivalve gaping behavior is monitored in the natural environment, remotely, continuously over a time period of years, requirements that must be fulfilled if bivalve behavior is to be a useful biomonitoring tool. Here, we review the literature and highlight potential uses of the HFNI valvometry as a biosensor, to monitor and provide early-warning alerts of changes in water quality, such as global temperature increase, releases of contaminants and toxic algal blooms. Finally, potential relevant applications for monitoring and assessing environmental status of marine waters in the context of the Marine Strategy Framework Directive are identified. Relevant descriptors, criteria, and indicators of Good Environmental Status that might be monitored using the HFNI valvometer are discussed for monitoring bathing beaches and harbors, petroleum installations and aquaculture sites.

Keywords: chronobiology, environmental monitoring, valvometry, rhythmicity, real time data

# INTRODUCTION

Marine ecosystems all over the globe are facing multiple simultaneous stressors including rapid climatic change and increased resource exploitation, such as overfishing, petroleum exploration and shipping. To protect more effectively the marine environment across Europe, Member States of the European Union committed to adopting an ecosystem approach to marine management.

Edited by: *Angel Borja, AZTI, Spain*

#### Reviewed by:

*Joana Patrício, Executive Agency for Small and Medium-Sized Enterprises, Belgium Jose Rafael Garcia March, Universidad Católica de Valencia San Vicente Màrtir, Spain*

\*Correspondence: *Hector Andrade hector.andrade@akvaplan.niva.no*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *02 June 2016* Accepted: *13 September 2016* Published: *04 October 2016*

#### Citation:

*Andrade H, Massabuau J-C, Cochrane S, Ciret P, Tran D, Sow M and Camus L (2016) High Frequency Non-invasive (HFNI) Bio-Sensors As a Potential Tool for Marine Monitoring and Assessments. Front. Mar. Sci. 3:187. doi: 10.3389/fmars.2016.00187* The EU Marine Strategy Framework Directive (MSFD 2008) mandated Member States to assess the environmental status of their territorial waters by July 2014, and to develop strategies to achieve "good environmental status" within 2020 (European Commission, 2008). The DEVOTES project (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) aims to better understand the relationships between pressures from human activities and climatic influences and their effects on marine ecosystems, including biological diversity, in order to support the ecosystem based management and fully achieve the Good Environmental Status (GES) of marine waters. Among the main objectives is to develop, test and validate innovative integrative modeling and monitoring tools to improve our understanding of ecosystem and biodiversity changes (www.devotes-project.eu/).

This paper outlines the application of high frequency non-invasive (HFNI) valvometers (http://molluscan-eye.epoc. u-bordeaux1.fr/index.php?lang=en&page=enregis&wid)<sup>1</sup> , as a potential tool for marine monitoring and assessments. The aim of this article is to provide a brief description of how the HFNI valvometer works, review some results achieved so far by studying bivalve gaping behavior with this method under both natural and laboratory conditions and discuss how the HFNI can be employed in the context of the Marine Strategy Framework Directive as a biosensor. A literature review is carried out to provide an overview of the geographic locations where the technology has been deployed, the species tested and the effect of different environmental and anthropogenic drivers and stressors upon valve behavior, growth and reproduction.

# THE HFNI VALVOMETER AND THE PRINCIPLES OF THE METHOD

The HFNI valvometer is a high frequency (10 Hertz), noninvasive (HFNI) biosensor employed to monitor gaping behavior (closing and opening of the valves) of bivalve molluscs. It is a new-generation remote technique enabling the online study of the behavior of bivalve molluscs living in their natural habitat, without interfering with normal behavior. Gaping activity of many species is closely related to physiological processes such as breathing and nutrition and waste elimination which respond to environmental conditions following rhythmic cycles (García-March et al., 2008; Sow et al., 2011; Tran et al., 2011). Bivalves alter their normal gaping behavior in the presence of stressors, indicating perturbations in the environment (e.g., Tran et al., 2003; Fournier et al., 2004; Sow et al., 2011) and this can be employed for marine monitoring and assessments. In a typical HFNI field deployment, a pair of electrodes with 1–1.5 m flexible cables are glued on to each half shell of 16 bivalves (**Figure 1**). The electrodes, designed to minimize disturbance to bivalve's behavior, are made up of two resin-coated electromagnets (56 mg each). An electromagnetic current between the electrodes is generated allowing measurement of the amount of valve opening and closing (Sow et al., 2011). At minimal distance, the electronic noise contribution to the signal is minimal (≤ 1µm).

In a classic HFNI valvometer deployment, each pair of electrodes is coupled to a waterproof box next to the animals. This box contains a first stage analogical electronic card that manages the electrodes. The first stage card is connected to a second held on the sea surface, or located on land by an umbilical cable. The whole system constitutes a Linux embedded system that acquires, saves and digitizes the data for transfer to the laboratory workstation (**Figure 2**). The system is built to sample data at 10 Hz from 16 animals in a sequential order. Every 0.1 s, three packets of information are produced: distance between valves at the electrode level, sampling time, and animal number. Thus, as a whole, a total of 3 × 864,000 pieces of information/day describes the behavior of a whole group of 16 animals (2,592,000 data points/day). At the individual level, it means that today the system performs a measurement of the opening status every 1.6 s, and that a total of 54,000 data points characterizes the gaping/closing behavior of any individual every day. If more than 16 animals are needed in a deployment, up to four systems can be installed. The raw data is transferred from the field to the laboratory using either an Ethernet network socket or a cellular telephone network (GPRS; General Packet Radio Service), with a mobile phone module embedded in the second electronic card. In both cases, data are transferred daily at 00:00 GMT local time (a standard configuration that can be modified) to a laboratory workstation where analyses are performed using both Bash Linux and mathematical codes written in R, (https://cran.r-project.org/). Thus, basically, the system architecture is composed of multiple robust slave-modules in the field (low to very low energy consumption, 0.5–1 watt; high-precision) and a single powerful master workstation in Arcachon, France, to handle megabytes of daily data. The latter is designed to capture, process and distribute information derived from the original data on the web (Sow et al., 2011).

The data generated are processed and analyzed on a daily basis, and easy-to-read graphs are automatically produced. In this manner daily, weekly or monthly trends can be identified quickly, offering a real-time monitoring framework to investigate the environmental status of multiple marine ecosystems. The metadata also is stored, providing a background allowing for gaping behavior comparisons when any change, subtle, overt or violent occurs over time. Records for all the sites where the HFNI has been deployed remain available since their original launch (the first one was done in February 2006), thanks to a data stocking policy and two back-ups located in two geographically different sites. Registered users have access to all the records from the individuals deployed at their site using the PRO website version. Here dynamic graphs deploy simultaneously gaping behavior of 1, 4, or 16 bivalves. The PRO version also allows inter-site comparisons.

In the pages of MolluSCAN eye PRO, professionals can integrate various types of information and derive graphs describing the various aspect of the bivalve's ethology, chronobiology or physiology. The graphs allow the relevance of single observations to be placed in a broader context

<sup>1</sup>MolluSCAN eye, the website. http://molluscan-eye.epoc.u-bordeaux1.fr/index.php ?rubrique=accueil&lang=en&site=EYRAC.

of comprehensive behavior, especially when the system is coupled with a multi-parameter probe. Five key examples of the types of data produced are shown on the website (http://molluscan-eye.epoc.u-bordeaux1.fr/index.php?rubrique= contenu\_sitePro&lang=en): biological rhythms, growth rates, spawning and death records and disturbance by toxic algae.

To model bivalve behavior, a non-parametric modeling approach based on kernel estimations is employed. This method has the advantage of summarizing complex data into a simple density profile obtained from each animal at every 24-h period, to ultimately make inference about time effect and external conditions on this profile (Sow et al., 2011; Tran et al., 2011). Hypotheses can be formulated to study bivalve biological rhythms and how local environmental drivers affect valve activity, e.g., tide amplitude, light regime, temperature, chlorophyll a, turbidity, etc. (Schwartzmann et al., 2011; Sow et al., 2011; Mat et al., 2012, 2013; Tran et al., 2016). The rationale behind the principle is that reference behavior and biological rhythms are basically synchronized by an endogenous molecular clock and external environmental factors (Tran et al., 2011).

The effects of external stressors such as pollution and climate change upon bivalve activity can be studied by comparing experimentally (under laboratory and/or field conditions) whether deviations from normal expected gaping patterns occur. The main goal in such experiments has been to test whether the HFNI valvometer can be effectively employed as a viable biosensor for water-quality assessment. Comparisons are made by either contrasting valvometry records prior to and after the introduction of a stressor, or by carrying out exposure experiments between treatment (exposed) and control (unexposed) units. Behavioral parameters such as valve opening duration, valve amplitudes, etc. are recorded and compared (**Figure 3**). In the case of bivalve exposures to toxic substances for example, a minimal sensitivity threshold can be calculated as the trace element concentration inducing a valve closure on 50% of the exposed organisms. By recording the time when alterations occur, it is possible to derive dose-response curves as well as timeresponse curves (Tran et al., 2003, 2004, 2007; Fournier et al., 2004).

Growth rates have been measured using the HFNI valvometer based on the fact that calcification in bivalves occurs in the mantle cavity, all over the shell's internal structure (**Figure 4**). When daily growth layers are deposited, the minimum distance between the electrodes glued to the shells increases providing a good proxy of growth (Schwartzmann et al., 2011; Berge et al., 2015). In the same manner, the maximum daily valve opening can be used to trace the overall health condition of the bivalves. Maximal opening status in bivalves is an index of a clam's well-being, because decreasing the valve opening is employed by bivalves as a primary strategy to protect the soft tissues when under threat (Schwartzmann et al., 2011). By plotting these values of minimal and maximum daily valve opening, mortality events can be easily distinguished as when a bivalve dies, its valves become widely open and inert. Comparisons to typical records allow to establish the exact time, up to the minute, at which the adductor muscle ceases to contract.

FIGURE 3 | Typical records of valve activity behavior of Crassostrea gigas during a 1-day period feed with the non-toxic algae (A) Isochrysis galbana clone tahiti (T-ISO), (B) Heterocapsa triquetra and the toxic dinoflagellate (C) Alexandrium minutum. Three parameters were used to characterize the behavior: daily valve-opening duration; daily valve micro-closure; valve-opening amplitude. Behavior comparisons are drawn under simplified but well controlled laboratory conditions to study the effects of environmental (e.g., temperature) and anthropogenic (e.g., toxic algae exposure) stressors upon bivalves. Reprinted from Aquaculture 298 (2010) 338–315 with permission. Minor text modifications performed on the original figure text.

# CASE STUDIES OF APPLICATION OF HFNI

Gaping behavior in bivalves has been studied using the HFNI valvometer both in the field and under laboratory conditions. Gaping behavior in the field has been recorded from tropical to arctic locations, for up to nearly four years continuously without human intervention (**Table 1**). At these locations, the biological rhythms of several native species have been studied as well as how extreme environmental conditions (e.g., increased water temperature, storms) affect their gaping behavior. Growth and reproduction events have also been recorded remotely, over multiple-year cycles allowing to study life history aspects of bivalves. Investigations have been conducted to assess the effects of toxins upon the behavior of several species using the HFNI valvometer in the laboratory. Ecotoxicology experiments are allowing to develop the HFNI technology as a biosensor for anthropogenic impacts in marine and freshwater.

## Valve Behavior, Growth and Reproduction of Bivalves

Patterns of valve behavior, growth rates and/or reproduction activity have been studied thoroughly in the giant clam Hippopus

the Bay of Arcachon, France. The index corresponds to the minimum daily opening value measured between HFNI electrodes as shown in Figure 1. White arrows, 2 oysters with continuous growth; black arrows, 3 oysters exhibiting a growth arrest which started during the period 11 October – 4 November 2011 (dashed lines), at the beginning of the winter period. Note the growth rate acceleration that started in early August.

hippopus, the Pacific oyster Crassostrea gigas (**Figure 3**) and the Icelandic scallop Chlamys islandica at their natural locations.

In New Caledonia, it has been shown that patterns of daily behavior and growth rate of the giant clam were related to light availability and changes in water temperature (Schwartzmann et al., 2011). Growth rate, as measured by HFNI valvometry, was demonstrated to be continuous throughout the year, but periods of both zero and altered daily growth were recorded. Typically, giant clam valves were open during the day and partially closed during the evening. This pattern became erratic during stressful environmental conditions brought about by a cyclone, and during increased irradiance periods and maximum temperatures (>27◦C) in the summer months. The later indicated that the species might be living beyond its upper thermal comfort limits during the summer at this location (Schwartzmann et al., 2011).

Growth patterns and daily behavior also have been studied in the Icelandic scallop C. islandica using HFNI valvometry. Above the polar circle (at 79◦ North, Ny-Ålesund) it was shown that despite what one might expect, growth rates of this bivalve can be similar during the polar night compared to the rest of the year (Berge et al., 2015). Behavior records showed that the valves of scallops remained opened most of the time and showed a steady biological rhythm suggesting that metabolism is kept active without any marked resting periods. The results showed that despite the seasonal polar night/day cycles, Icelandic scallops maintain a circadian rhythm for the majority of the year much like bivalves at other latitudes (Mat et al., 2012; Tran et al., 2016).

Reproduction and spawning behavior have been studied for the Pacific oysters Crassostrea gigas in the Bay of Arcachon and the Bay of Marennes-Oléron, France using the HFNI valvometer


(Bernard et al., 2016). Spawning behavior in female Pacific oysters is characterized by rapid contractions of the adductor muscle. This gaping signal is recorded by the HFNI valvometry as several rapid and ample movements (large openings) of the valves and allows for studying in detail the exact timing and possible environmental drivers of spawning activity. Spawning events between 2003 and 2014 consistently occurred during spring high tides at both locations, when the moon is closest to the earth (perigee). Peaks in water current were proposed as the final spawning trigger (Bernard et al., 2016). On the whole, the above studies demonstrate that the HFNI valvometer has been successful to investigate various aspects of bivalve life history and how these are affected by the prevailing environmental drivers.

#### Trace Metal Detection

Initial try-outs with the HFNI system were carried out on the freshwater mussel Corbicula fluminea to test the potential and limitations of using bivalves as a rapid response and/or sensitive biosensor for different contaminants (Tran et al., 2003). Under laboratory conditions, C. fluminea was exposed to increased levels of cadmium, copper, uranium and inorganic mercury in independent experiments to test whether gaping behavior differed between exposed and unexposed organisms. The experiments effectively showed that changes of valve closure patterns occurred in organisms exposed to increasing trace metal concentrations. Importantly, time was taken in consideration and an inverse relationship between concentration and response velocity was systematically demonstrated in all conditions. Minimal sensitivity threshold, i.e., the trace element concentration inducing a valve closure on 50% of the exposed organisms were calculated as well as the time needed to achieve such closures (Tran et al., 2003, 2004, 2007; Fournier et al., 2004). Cadmium concentrations above 50µg/l could be detected within less than 1 h. The lowest cadmium concentration detected was 16µg/l and required 5 h of exposure (Tran et al., 2003). Copper concentrations as low as 4µg/l were detected within 5 h (Tran et al., 2004). For uranium, the minimal sensitivity threshold varied depending on the pH. At pH 5.5, minimum detection levels at 0.05µmol/l were achieved after 5 h (Fournier et al., 2004). In a latter experiment using inorganic mercury, minimum detection occurred at 3µg/l at the same exposure time (Tran et al., 2007). Interestingly, the inorganic mercury experiment showed that stressed valve behavior of C. fluminea exposed was different from those exposed to the other trace metals indicating that pollutants might produce a contaminantspecific gaping signal. In general, these studies demonstrated that the HFNI valvometer has potential as a biosensor for monitoring anthropogenically induced trace metals in the water column.

# Algal Toxicity

Bivalves are filter-feeders that can accumulate paralytic shellfish toxins which are harmful to human health (Bricelj and Shumway, 1998). Experiments using the HFNI valvometry tested whether increased concentrations of these toxins could modify the valve behavior of Pacific Oysters C. gigas (**Figure 3**). Under laboratory conditions, oysters were exposed to various simulated algal blooms of the toxic dinoflagellate Alexandrium minutum and the non-toxic dinoflagellate Heterocapsa triquetra or the Isochrysis galbana clone Tahiti. Gaping behavior of oysters differed between toxic and non-toxic treatments and were detected after ≈ 1 h. Organisms exposed to A. minutum increased both micro-closure activity and daily valve-opening duration while valve-opening amplitude decreased (Tran et al., 2010; Haberkorn et al., 2011; Mat et al., 2013). In a later study it was shown that daily gaping rhythmicity completely vanished in oysters exposed to the harmful algae (Tran et al., 2015). These results demonstrate that the HFNI vavometer have the potential to be employed to monitor toxic algal blooms.

In general, the heavy metal and algae toxicity experiments have shown that the HFNI technology has been effective to detect toxic substances in the water under laboratory conditions. The methodological changes developed with the HFNI valvometer allowed to better define the optimal response capacities of various bivalves in simplified, although perfectly controlled, conditions. In this regard, the HFNI technology has a clear potential as a biosensor to monitor water quality.

#### HFNI—AN INNOVATIVE TECHNOLOGY

As demonstrated by the papers reviewed here, the HFNI valvometry has been employed successfully to study multiple life history traits (biological rhythms, growth rate, spawning events, death) of several bivalve species in relation to their natural environment and in ecotoxicological studies as a biosensor for various toxic substances and contaminants. The use of various technical designs to record molluscan gaping behavior, for the purposes of water quality assessment is not new in principle (see for example http://www.mosselmonitor.nl/, although other systems also exist; Kramer et al., 1989; Borcherding, 2006; Kramer, 2009; Chen et al., 2010). However, the HFNI valvometer differs from others in a number of significant ways:


and ecosystems, this attribute may have an application in commercial mussel farm areas.

# HFNI AS A POTENTIAL TOOL FOR MSFD MONITORING

# General Applications

The current set-up of the non-invasive sensor system is applicable to the monitoring phase of the MSFD (Marine Strategy Framework Directive), assess the environmental status across the European seas. Within the MSFD, Good Environmental Status (GES) is defined in terms of 11 qualitative descriptors, within which a total of 29 associated criteria and 56 indicators have been identified, which include biological, physico-chemical state indicators as well as pressure indicators (EU Commission Decision of 1 September 2010 on criteria and methodological standards on good environmental status on marine waters (European Commission, 2010). In short, the descriptors as listed by Borja et al. (2013) comprise Biological diversity (D1), Non-indigenous species (D2), Exploited fish and shellfish (D3), Food webs (D4), Human-induced eutrophication (D5), Seafloor integrity (D6), Hydrographic conditions (D7), Contaminants (D8), Fish and seafood contaminants (D9), Marine litter (D10) and Energy including underwater noise (D11). The HFNI can be directly employed to monitor the descriptors Human-induced eutrophication (D5), Contaminants (D8, D9) and Noise (D10), but also indirectly Sea-floor integrity (D6) and Food webs (D4). However, the method is most suitable for long-term, 24/7, insitu monitoring of changes in water quality—not in terms of actual values measuring directly an indicator (e.g., nutrient concentration in the water column), but in terms of arising disturbances, either acute or gradual (e.g., abnormal gapping behavior due to increased concentration of nutrients in water). If the continuous data feeds do not show any abnormalities, then one may assume the water quality is as usual (according to previously measured levels). If there is a change, either abrupt or progressive (in case of silent pollution), then it will provide an early-warning that a change has occurred, and a more detailed water quality measurement can be done, should this be the appropriate action. An alarm system would make monitoring of the deployed systems efficient in terms of human effort.

Three important points have to be considered when using HFNI valvometry for detection of toxic substances in the water: first, the valve behavior due to stress of the experimental set-up must be minimized; second, the natural valve closing/opening rhythm (equivalent to background noise) has to be defined so that optimal comparisons can be made between stressed and unstressed organisms; and third, mathematical descriptions using analysis of the dose–response-type curves that integrates time of any detection mechanism(s) must be developed (e.g., Tran et al., 2003). Some relevant monitoring applications are described below (in alphabetical order):

#### Aquaculture Sites

At aquaculture sites, where it is not possible to locate the cage groups in exposed or deep water, a form of online monitoring alerting to changes in water quality can contribute to maintaining efficient fish health and thus growth. Especially in areas where seasonal upwelling is prominent, an early-warning of deterioration in water quality, usually as a result of over-enriched sea-floor conditions, could in the most extreme case prevent mass mortalities due to oxygen depletion. If the clam sensors indicate stressful conditions, then quantitative water measurements can immediately be carried out. Direct and indirect effects of nutrient enrichment, increased contaminant concentrations and organic matter over a threshold level can be detected as abnormal behavior in bivalves prompting an immediate monitoring response where more exhaustive sampling is required. Such sampling would include among others, nutrient and contaminant concentrations in the water column, turbidity and oxygen levels. Temporary transfer of cages to a less exposed area could be a remedial action from such a warning.

Additionally, as mentioned above, the sensors allow the detection and possibly also prediction of spawning behavior, which may have considerable application to bivalve farms (using the same species been farmed as a biosensor). The collection of larvae at the appropriate timing is of considerable value in bivalve farming, thus knowledge of when animals are spawning is a key piece of information. Another application might be to infer upon causes of decreases in bivalve growth, as shown for oysters infected with parasites (Chambon et al., 2007).

From a management perspective, the main potential environmental impacts of aquaculture come from the introduction of non-indigenous species, nutrients, organic matter, contaminants including pesticides and litter, the disturbance to wildlife, and the possibility for escape of farmed fish (European Commission, 2016). The based on the papers reviewed here, the HFNI valvometer could potentially detect evidence of increased eutrophication in enclosed areas (D5), declining sea-floor integrity due to siltation (D6) and the presence of contaminants in the form of hazardous substances and microbial pathogens (D8).

#### Bathing Beaches and Harbors

EU Member States monitor the quality of their bathing sites according to the provisions of the EU's revised Bathing Water Directive (2006/7/EC). This directive requires Members States to monitor and assess the bathing water for at least two parameters of (fecal) bacteria and prepare bathing water profiles containing information about the kind of pollution and sources that affect the quality of the bathing water (http://ec.europa.eu/ environment/water/water-bathing/index\_en.html). Recordings of molluscan gaping behavior at bathing beaches or harbors will allow detection of change in water quality assisting in monitoring the MSFD descriptors Human-induced eutrophication (D5) and Contaminants (D8) (European Commission, 2010). A working hypothesis would be that if the clams behave as normal, we may assume no adverse change has occurred. If a sudden change is episodic, it could be linked with a single event, but if the aberrant behavior persists, then quantitative monitoring of water quality should be implemented. Coupled with the actual measurements of toxic algal blooms, bacterial content and other contaminants, this would provide an efficient system to safeguard human safety, even during periods where daily physical measurements are not being carried out by the municipality (see Sections Contaminant Detection and Toxic Algae Alerts). The system would have a further public appeal, because the information would be made available in a user-friendly way on an openly accessible web site. Television reporters have covered a story about these ideas on the Franco-German TV station, ARTE in 2009 (X:enius, 15/7/2009), which illustrates the public interest for such questions (Oberhauser, 2009).

# Climate-Related or Other Changes in Hydrographic Properties

The HFNI valvometer allows the user to relate changes in bivalve behavior and growth rate to climate-induced stress. As mentioned above, the valve activity and growth of the giant clam H. hippopus becomes erratic at increased temperatures and solar irradiance (Schwartzmann et al., 2011). Moreover, bivalve populations at the edge of their thermal maxima (temperature above which most individuals respond with unorganized locomotion, subjecting the animal to likely death) can present massive mortalities due to effects of increased temperature (Jónasson et al., 2004) and can be recorded and dated with the HFNI system as dead bivalves remain with the shells open and motionless. To investigate further valve activity behavior and life history of bivalves near their thermal limits, 16 individuals of the blue mussel M. edulis have been deployed in the high Arctic archipelago of Svalbard in April 2016. Blue mussels were absent from this location for about 1000 years but new settlements have recently re-colonized the area due to increased sea surface temperatures along the west coast of Svalbard (Berge et al., 2005). In general, these investigations show the HFNI technology potential to study effects of climate variability upon bivalves in tropical and arctic environments.

Relationships between primary productivity and bivalve growth established with the HFNI can also provide information relevant to the Descriptor 4 "Food Webs" (European Commission, 2010). Filter-feeding bivalves use their gills to catch particulate food such as phytoplankton. The latest HFNI deployments include a multi-parameter probe equipped with a fluorometer that estimates chlorophyll production, a widely used index of phytoplankton biomass. Effects of increased phytoplankton biomass on bivalve growth rates can then be easily studied with this approach, providing an indicator of trophic level productivity and trophic interactions.

#### Contaminant Detection

The HFNI valvometer can be used to detect acute pollution in case of accidental contamination, but it has a specific added-value for the detection of chronic pollutions and "silent" and/or transitory pollutions which are difficult or impossible to detect otherwise, due to its 24/7 monitoring capabilities over very long periods of times, the very high sensitivity of bivalve molluscs and a policy of data stocking. Cumulative effects of toxicity can also be assessed based on changes in expected behavior. A transient effect to a single dose exposure can have dramatic impact. HFNI can be a remote witness of these events. Interestingly, the bivalve behavior before death is very typical. Identifying the last moments during which it behaved normal and rhythmic can tell when the animal started to become disturbed. If a whole group starts to be disturbed in a similar timeframe, one can speculate that a common driver exists. Such a driver could be a drastic, or subtle but deadly, change in water quality. Preventing and reducing anthropogenic inputs to the marine environment is one of the main objectives of the Marine Directive and the aim of Descriptor 8 is to ensure that levels of contaminants in the marine environment do not give rise to pollution effects (European Commission, 2010) (http://ec.europa.eu/environment/marine/good-environmentalstatus/descriptor-8/index\_en.htm). The HFNI potential as a biosensor for heavy metal detection even at low concentration levels has been extensively demonstrated in the laboratory as discussed in several of the papers reviewed here (e.g., Tran et al., 2003).

### Petroleum Installations

Produced water (the water which is produced as a byproduct of oil and gas extraction) is notoriously difficult to monitor, primarily due to its rapid dilution in water currents. Deploying the HFNI system at appropriate locations around oil and gas production units will allow the continuous detection of water quality at biologically-relevant levels. This would cover both intentional discharges but would also give early-warning of unplanned leakages to the water column. A parallel system set up at the sea floor would also have an application to detection of unintentional discharges at the sea floor, for example from pipelines or sub-sea production templates. In this regard, the HFNI valvometer has potential as a monitoring tool for the MSFD-related indicators 8.2.1. "Levels of pollution effects on the ecosystem components concerned, having regard to the selected biological processes and taxonomic groups where a cause/effect relationship has been established" and "8.2.2. Occurrence, origin, extent of significant acute pollution events and their impact on biota physically affected by this pollution" (European Commission, 2010).

# Toxic Algae Alerts

In areas where toxic algae may present a risk to human or culture organism health, remote biosensors could function as an early warning system through the documented changes in the gaping behavior, in reaction to toxic alga, to alert when conventional water sampling is needed. This has been demonstrated in the laboratory as mentioned above, by studying behavior of oysters exposed experimentally to a mimic bloom of harmful algae (e.g., Tran et al., 2010). This would help to reduce the total costs of algal bloom monitoring schemes as conventional sampling (e.g., sample collection by a person, laboratory analysis, etc.) will be required only in the case of a bloom event; or supplement with data from areas where routine sampling is not carried out. This application is relevant to monitor the MSFD indicator 5.2.4 "Species shift in floristic composition such as diatom to flagellate ratio, benthic to pelagic shifts, as well as bloom events of nuisance/toxic algal blooms (e.g., cyanobacteria) caused by human activities" (European Commission, 2010) (see also under aquaculture and bathing beaches).

# CONCLUSIONS

The HFNI valvometry has been effective in recording valve activity behavior of bivalves across several locations, from the tropics to the high Arctic. Its deployment has been performed in harsh weather and sea conditions, and/or in daylight or darkness down to −30◦C. Studies have been performed to understand biological rhythms, various life history traits and effects of natural and anthropogenic stressors upon bivalve behavior in their natural environment or under laboratory conditions. Relationships between pressures from human activities and climatic influences and their effects on bivalve species have been established using the HFNI valvomenter during the last 15 years. In this regard, this innovative tool holds promise for marine monitoring, allowing managers to assess environmental status of marine and freshwater ecosystems remotely and in near-real time. In the MSFD context, the HFNI valvometer has direct applications for monitoring several Indicators of the MSFD Descriptors "D5. Eutrophication," "D8. Contaminants" and "D11. Energy and Noise"; and indirectly the Descriptors "D4. Food webs" and "D6. Seafloor Integrity" (European Commission, 2010). While bivalve growth rates can provide an indirect measurement of production of key species or trophic groups, bivalve behavior provides an indicator of stress that can be related to worsening water quality conditions in benthic environments as well as in the water column. For this purpose, it is necessary to first, establish the baseline normal gaping behavior of the species been studied. Then, behavioral responses to stressors must be characterized. Comparisons can then be made between stressed and unstressed organisms and infer upon the causes of such stress. Abnormal gaping behavior can reveal episodic incidences of disturbance but there will be times when it is not possible to pinpoint directly the exact cause of that disturbance. When this occurs, the technology reveals the problem becoming the starting point of an enquiry for which complementary sampling in the field and analyses are required. How are we able to distinguish the causes of change in gap behavior? By building a repertoire of behavior and analysing in parallel all available environmental data (local ephemerids, tide table, water temperature, noise). In this view the knowledge obtained by multi-parameter probes is evidently quite helpful: chlorophyll a, turbidity, oxygen, etc. Further research should focus on evaluating the HFNI technology performance directly in the field, as a biosensor for different anthropogenic stressors such as increased sedimentation, mining, oil production etc.

# AUTHOR CONTRIBUTIONS

HA: Main writer, literature review, data analysis and interpretation, article concept; JM, Project proponent, technology development, data analysis, interpretation, writer; Data analysis and interpretation, writer, figures; SC: Writer, data analysis and interpretation, article concept; DT: Data analysis and interpretation, technology development, writer; PC: Data analysis, technology development; MS: Data analysis and interpretation, technology development; LC: Critical revisions of the document.

## ACKNOWLEDGMENTS

This paper is part of an ongoing collaboration between Akvaplan-niva AS and the UMR CNRS 5805 EPOC (University of Bordeaux and CNRS) aiming to develop the HFNI technology as a reliable biosensor in arctic conditions. This manuscript is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and

#### REFERENCES


assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), www.devotes-project.eu. Financial support has also been received through the CNRS, the University of Bordeaux, the Région Aquitaine, the Norwegian Regional Research Fund in the North (Project number 208974), The French National Research Agency (ANR-15-CE04-0002-02), The Fram Centre Flagship "Effects of climate change on sea and coastal ecology in the north," the Svalbard Environmental Protection Fund (Project 15/133) and by the Russian Foundation for Basic Research and the Research Council of Norway (Project 233635/H30 "Environmental management of petroleum activities in the Barents Sea: Norwegian-Russian collaboration."

(Corbicula fluminea): quantification of the influence of pH. Environ. Toxicol. Chem. 23, 1108–1114. doi: 10.1897/02-604


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Andrade, Massabuau, Cochrane, Ciret, Tran, Sow and Camus. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Microplastics in Seawater: Recommendations from the Marine Strategy Framework Directive Implementation Process

Jesus Gago<sup>1</sup> \*, Francois Galgani <sup>2</sup> , Thomas Maes <sup>3</sup> and Richard C. Thompson<sup>4</sup>

*1 Instituto Español de Oceanografía, Vigo, Spain, <sup>2</sup> Institut Français de Recherche pour l'Exploitation de la Mer, Bastia, France, <sup>3</sup> Lowestoft Laboratory, Centre for Environment, Fisheries and Aquaculture Science, Lowestoft Suffolk, UK, <sup>4</sup> School of Marine Science and Engineering, Plymouth University, Plymouth, UK*

Microplastic litter is a pervasive pollutant present in marine systems across the globe. The legacy of microplastics pollution in the marine environment today may remain for years to come due to the persistence of these materials. Microplastics are emerging contaminants of potential concern and as yet there are few recognized approaches for monitoring. In 2008, the EU Marine Strategy Framework Directive (MSFD, 2008/56/EC) included microplastics as an aspect to be measured. Here we outline the approach as discussed by the European Union expert group on marine litter, the technical Subgroup on Marine litter (TSG-ML), with a focus on the implementation of monitoring microplastics in seawater in European seas. It is concluded that harmonization and coherence is needed to achieve reliable monitoring.

#### Edited by:

*Maria C. Uyarra, AZTI Tecnalia, Spain*

#### Reviewed by:

*Mario Barletta, Federal University of Pernambuco, Brazil Stefano Aliani, National Research Council, Italy*

> \*Correspondence: *Jesus Gago jesus.gago@vi.ieo.es*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *13 June 2016* Accepted: *24 October 2016* Published: *08 November 2016*

#### Citation:

*Gago J, Galgani F, Maes T and Thompson RC (2016) Microplastics in Seawater: Recommendations from the Marine Strategy Framework Directive Implementation Process. Front. Mar. Sci. 3:219. doi: 10.3389/fmars.2016.00219*

#### Keywords: marine debris, plastics, microplastics, monitoring

# INTRODUCTION

The ubiquity of plastics in the marine environment and in biota from across the globe has highlighted the prevalence of this contaminant within our oceans. The global mass-production of plastics which started mid last century has been followed by the accumulation of plastic litter in the marine environment (Rochman et al., 2013).

The term "microplastics" (referred to as MPs from hereon) first entered the published literature in 2004 (Thompson et al., 2004), but is now used extensively to describe small fragments of plastic. There is no widely accepted "lower boundary" in size as the limit of detection is dependent on the sensitivity of the sampling technique used (e.g., mesh size of the net or size of the filter).

Microplastics are widely dispersed in the marine environment and are present in the water column, on beaches and on the seabed (Barnes et al., 2009; Law et al., 2010; Browne et al., 2011). Microplastics are a newly recognized type of marine pollution and as such there are few regulations in terms of production, use or emissions.

In the EU, the Marine Strategy Framework Directive (hereinafter MSFD) adopted in 2008 (European Commission, 2008), aims to establish a good environmental status (GES) of the European seas by 2020. The MSFD represents the first instance, worldwide, that MPs in the marine environment have been included in a legislative proposal. In this sense is important to mention that MPs were not included in the Water Framework Directive (WFD), the main EU directive dealing with pollution of river basins.

The main findings of the MSFD marine litter expert group in relation to MPs in seawater are described here. This information may help researchers and governments of EU member states and also other countries to establish legislative tools and to implement programs aimed to study abundance and the impacts of microplastics in the marine environment.

# MICROPLASTICS IN THE MARINE ENVIRONMENT

Microplastics can enter the marine environment directly as primary MPs (e.g., pre-production pellets and/or granules used as abrasives in cleaning products) or indirectly, as secondary MPs, i.e., the result of progressively fragmentation in the environment of larger items. The relative importance of primary and secondary sources of microplastics to the marine environment is not known (Andrady, 2011).

One of the main threats emanating from MPs is their potential to be taken up by marine organisms. Potentially affected species include primary producers at the base of the food chain through zooplankton, and all the way up to macro invertebrates, fish, and mammals (CBD, 2012). There is limited information on the extent to which microplastics might cause harm in the marine environment. Cell damage, infections, tumor formation, death are just some of the reported toxicity effects by MPs (CBD, 2012).

# THE MSFD: AN INTEGRATED ENVIRONMENTAL POLICY FOR THE MARINE ENVIRONMENT

The European Directive 2008/56/EC (MSFD) is a key element in Europe's actions to protect seas and oceans. The Directive calls for all of the EU's marine regions and sub-regions to achieve or maintain "Good Environmental Status" (GES) by 2020. GES is defined by means of 11 qualitative "descriptors." The relevant criteria and indicators applicable to those descriptors are defined in the Commission Decision 2010/477/EU (European Commission, 2010).

One of the most important strengths of the MSFD is the aim to provide a holistic, functional approach; it separates the ecosystem into a set of process-related (functional) objectives, and then recombines these, to ensure the integrity of the ecosystem.

Descriptor 10 relating to marine litter, and their formulation according to the MSFD is that "Properties and quantities of marine litter do not cause harm to the coastal and marine environment." It is the first time that marine litter is addressed, in an integrated way for the protection of the marine environment, in a European directive (Galgani et al., 2013a).

A Technical Subgroup on Marine Litter (TSG-ML) was established in 2010 to support Member States in harmonizing monitoring protocols and streamlining monitoring strategies in the framework of the MSFD (Galgani et al., 2013a,b).

#### Microplastics in the Context of the MSFD

Microplastics are considered specifically in descriptor 10 of the MSFD [10.1.3 "Trends in the amount, distribution, and where possible, composition of micro-particles (in particular microplastics)"], and not directly but implicitly in the indicator related with impacts of litter on marine life. The descriptor will establish baseline quantities, properties, and potential impacts of MPs. It must be noted however that the decision was reviewed recently for changes in order to make it simpler and clearer, to introduce minimum standards and to be coherent with other EU legislation.

Within the process, the TSG-ML suggested that micro-litter be considered as a size fraction integrating micro-litter along with other litter fractions in the matrix related indicators. Not all of the experts support this view, arguing that micro litter is different from other litter types (meso/macro) and that micro-litter may have considerably different effects to those caused by larger items of litter. The idea of merging indicators 10.1.2 (litter at sea, floating and on the sea floor) with indicator 10.1.3 (microplastics) aimed to avoid treating microparticles as a separate issue while measures to combat marine litter need to be formulated covering all size classes.

Finally, the revised decision (article 9/3 and 11/4) kept (the review has been done but not published yet) criteria separated for macro litter (10DC1) and microplastics (D10C3), now defined as "The composition, amount, and spatial distribution of micro-litter in the surface layer of the water column, in sea-floor sediment, and possibly on coastlines, is at a level that does not cause harm to the coastal and marine environment."

MPs should be categorized according to their physical characteristics including size, shape, and color (see **Table 1**). It is also important to obtain information on polymer type.

The size definition of MPs according to the TSG-ML (Galgani et al., 2013b) is in line with the NOAA definition. We strongly suggest using this size (<5 mm) as an international standard. One aspect that should be refined is the definition of the lower size boundary for MPs in the MSFD. The lower size has not been defined strictly and nanoparticles have not been considered as a category despite their potential relevance (Galgani et al., 2013a).

Sampling of MPs in the different marine compartments (sea water, sediment, and biota) requires different approaches: samples can be selective, bulk, or volume-reduced (see e.g., Hidalgo-Ruz et al., 2012). Selective sampling in the field involves visual identification and manual sorting of fragments from different matrices and is not very effective for MPs due to difficulties in handling small size items. The subsequent identification of plastic particles in the matrix follows similar procedures (section Quantification and nature of MPs).

Bulk samples refer to samples where the entire volume of the sample is taken without reducing it during the sampling process. Bulk samples are most appropriate when MPs cannot be easily identified visually because in the field because (i) they are covered by sediment particles, (ii) their abundance is small requiring sorting/filtering of large volumes of sediment/water, or (iii) they are too small to be identified with the naked eye (Hidalgo-Ruz et al., 2012).

Volume-reduced samples, in seawater, refers to sampling where the bulk volume of the sample is reduced during sampling, preserving only that portion of the sample that is of interest for further processing. While on board a vessel seawater samples can be volume-reduced by filtering water through nets or screens.


#### TABLE 1 | Categories used to describe microplastic appearance in the MSFD.

### A Need for Standardization: The Exemplary Case of Sampling Seawater

In the last years studies determining the global quantity of plastic particles in the ocean have been published (Eriksen et al., 2014; Cózar et al., 2014, 2015). In order to ensure inter-comparability between these studies to evaluate when (seasonality) and where (space) contamination is taking place, harmonization is urgently needed.

Seawater samples for MPs are mostly taken by nets. The main advantage of using a net is that large volumes of water can be sampled quickly, only retaining the volume-reduced sample. Most studies have been from surface water using neuston nets (Hidalgo-Ruz et al., 2012); manta and bongo nets have also been used at the sea surface. Since most plastics are buoyant they are likely to accumulate at the sea surface. Another instrument, that is widely deployed on a global scale and that has also been used for MPs sampling is the Continuous Plankton Recorder (CPR) (Thompson et al., 2004). Some instruments, including bongo and the CPR, are used sub surface making direct comparison rather difficult (Hidalgo-Ruz et al., 2012; Frias et al., 2014).

The most relevant characteristics of the sampling nets used are the mesh size and the net opening. Mesh sizes used for microparticle sampling range from 0.053 to 3 mm, with a majority of the studies (rather than individuals samples collected) ranging from 0.30 to 0.39 mm (Hidalgo-Ruz et al., 2012). The net aperture for rectangular openings of neuston nets (sea surface) ranged from 0.03 to 2.0 m<sup>2</sup> .

Techniques using apparatus to collect surface seawater and pass it through a filter on-board ship are being developed for example by CEFAS, UK (T. Maes; personal communication). They use the ships water inlet, collecting seawater from the side at specified depths, mostly ranging between 4 and 1 m depth. The seawater is being passed along a set of sieves or nets after which the sieves or nets can be removed and analyzed for MPs in the laboratory (Pitois et al., 2016).

The advantage of such systems is that it can collect marine litter samples from the water column while steaming and thus long transects over several kilometers can be collected autonomous in connection with in-line analytical systems for other parameters like nutrients or oxygen. The development of filtration systems for the quantification of MPs appears promising (Lusher et al., 2014).

The recommendation from the TSG-ML is to obtain samples from sea water wherever possible, and to ensure the following details are recorded to accompany each sample: type of net (preferably Manta net), aperture (usually 60 cm), and mesh size (preferably 333 µm). It is also important to record the following parameters: depth (preferably either at the sea surface or within surface 10 m, for greatest inter-comparability among sampling programmes) distance towed, location of tow (in/out of water) and volume of water filtered (with a current meter).

Also prevailing weather conditions and sea state, together with any relevant information on the volume of plankton or other particulates sampled, for example if there is concern that the net may have become clogged due to high concentration of plankton, must be recorded. Samples should be stored in glass jars. MPs are determined as the total quantity of items per volume of seawater captured by the net during the period it is deployed.

Samples in seawater can be passed through a 500 µm sieve, and liquid passing through the sieve then filtered through a filter paper using a Buckner funnel. Filter papers can then be examined under a dissecting microscope to quantify microplastics below 5 mm. Sample on CPR silk filter screens can be examined directly under the microscope.

At present and from the experience in the implementation of the MSFD discussed in the TSG-ML, it is not appropriate to recommend one approach over all others. As an example, in **Table 2** are shown MPs values available for the Mediterranean with sampling details (mesh size, net). Each approach has advantages and disadvantages and may be preferable according to local availability/sampling opportunities, the characteristics of the area to be sampled and other factors. The mesh size and water volume are important if one wants to compare different surveys and thus harmonization between these parameters is recommended.

# QUANTIFICATION AND NATURE OF MPs

Once MPs have been separated from their environmental matrix (seawater, sediments or biota) they must be quantified and identified.

# Identification of MPs

The identification of MPs polymers is achieved by comparing the spectra from the unknown sample against that of a known standard polymer in a database. We encourage consulting Hummel (2002) for more details on this methodology. It should be noted that this method is only definitive where a good match is obtained and this is not always possible. Due to biofouling and degradation processes of microplastics in the environment, their


spectra are not totally similar to spectra from the virgin material in the library.

If formal identification of particles using Fourier Transformed- Infra Red (FT-IR) or Raman Spectroscopy is applied then polymer type should also be recorded. Spectroscopy is not critical for routine monitoring of larger fragments > 500 µm. However, it should be considered essential for fragments > 50 µm and a proportion (5–10%) of all samples should be routinely checked to confirm the relative accuracy of any visual examination.

A suitable approach proposed by the TSG-ML would be to automatically accept any match >70% similarity (Frias et al., 2016), to individually examine matches between 60 to 70% similarity rejecting any samples which do not show clear evidence of peaks corresponding to known synthetic materials and to routinely reject (as synthetic) any samples which produce spectra with a match < 60%).

It is advocated that when analyzing particles in the range 1–100 µm to subject them to further spectroscopic analysis to confirm polymer identity (e.g., using FT-IR). For particles in the size range 101 µm–4.99 mm we recommend that a proportion (10% of the material in each size class, up to a maximum of 50 items per year or sampling occasion whichever is the least frequent) of the items considered to be MPs is subjected to further spectroscopic analysis to confirm identity (e.g., using FT-IR). This step is important in order to; (1) ensure quality control of visual identification and (2) gain information on the relative abundance of different polymer types which can inform on sources.

One important issue is to mitigate contamination of samples, as plastics are present in our daily lives (in clothes, scrubbers) and in labs (labware). People undertaking the sampling and working in the lab should minimize any synthetic clothing. As procedural controls to check ambient cleanliness, place unused clean filter papers in Petri dishes. Remove the lid and leave the Petri-dish open for a fixed time period relevant to the time period for which samples might be exposed to the air during examination. Procedural contamination should be <10% of the average values determined form the samples themselves.

#### Required Reporting Units

For MPs in seawater items/m<sup>3</sup> seawater, average size of particles, relative abundance of main colors and shape are suggested as units. Relating quantities of MPs to volume is relevant when considering the sampling of water column through filtration. Expressing quantities by volume also allow to link field studies directly with exposure experiments in the laboratory. The estimation of volumes is however impossible when using neuston/manta nets as the trawl frames are permanently moving vertically at the surface of the sea, complicating correct calculation of the sampled water height covered during tows. For this reason, the sampling of the surface density most often rely on items/m<sup>2</sup> , a more relevant estimation of the sample covered. It should be stressed that when possible, more info should be recorded to facilitate reporting in several units in order to ensure comparison with other studies. If FT-IR or Raman is used then polymer type should also be recorded together with shape and color.

#### FINAL REMARKS

When comparing reported abundances of MPs in the water column it is important to keep in mind that even though most surveys are conducted using a neuston net, the mesh size of these nets often differ. In addition, despite recommendations for the definition of MPs as particles smaller of 5 mm (Arthur et al., 2009; Galgani et al., 2013a), many authors worldwide are using other size limits e.g., 1 mm (Costa et al., 2010; Van Cauwenberghe et al., 2013). Furthermore, sometimes it is not possible to compare density values due to different methodologies used for sampling (items/km<sup>2</sup> vs. items/km<sup>3</sup> ). Hence, comparison between studies is quite complex.

There is a need for research to develop and subsequently validate new methods to rapidly and inexpensively identify and quantify MPs. These methods could include image recognition equipment to facilitate rapid identification as is currently used for plankton and particulate characterisation (Sieracki et al., 2009) and separation. Development of bulk chemical approaches to provide either an absolute value or an index of extent to which a water sample is contaminated with MPs and to indicate the type of particles (as e.g., polymer type) could also prove useful. It is also important to note that methods for detecting nanoparticles in the marine environment should be developed in the coming years.

As this is an emerging field and our understanding of the rates of accumulation and the extent to which MPs might cause harm in the environment is very scarce. Therefore, the experts of TSG-ML advocate a precautionary approach and recommends the development and calibration of methods and initiation of wider scale monitoring should commence straight away.

In our view one of the most important long term needs for the MSFD beyond 2020 are to gain a holistic understanding of marine litter by integrating MPs data collected from waters, sediments, and biota with other litter data and by integrating knowledge of temporal and spatial trends across types and sizes of marine litter (Van Franeker and Law, 2015).

Some of the monitoring approaches for the MSFD are still under development, so the implementation and improvement of monitoring will require continuous collaborative efforts. To achieve the greatest efficiency, MPs in seawater should be sampled alongside other routine sampling programmes. Similarly sampling of sea water column could also be incorporated into other monitoring programmes. A key consideration in collecting seawater samples is the cost of ship time. Hence the potential to sample during existing cruises or programmes is well worth considering.

The comparable quantification of MPs, by the use of common methodologies, is also important for identification of the sources,

# REFERENCES


planning of measures against marine litter and for checking the efficiency of these counter-measures under the umbrella of the MSFD.

# AUTHOR CONTRIBUTIONS

JG write the paper and took part in discussions. FG, TM, and RT contribute to the paper discussions.

## FUNDING

Participation of JG was financed by Spanish minister of Environment under Project 3-ESMARAC.

#### ACKNOWLEDGMENTS

This article is based on the activities of the GES-Technical subgroup on Marine litter (2012–2015), specifically on microplastics. We want to express our gratitude to all members of this group. We are very great full to the three reviewers that have made a number of good suggestions to improve this paper.

Field of Marine Environmental Policy (Marine Strategy Framework Directive). Brussels: Official Journal of the European Union L164, 19–40.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Gago, Galgani, Maes and Thompson. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# **Indicators to assess the status**

# Biodiversity in Marine Ecosystems—European Developments toward Robust Assessments

*Fisheries and Aquaculture Science, Lowestoft, UK, <sup>6</sup> Marine Research Division, AZTI, Pasaia, Spain*

Anna-Stiina Heiskanen<sup>1</sup> \*, Torsten Berg<sup>2</sup> , Laura Uusitalo<sup>1</sup> , Heliana Teixeira3 † , Annette Bruhn<sup>4</sup> , Dorte Krause-Jensen<sup>4</sup> , Christopher P. Lynam<sup>5</sup> , Axel G. Rossberg5 † , Samuli Korpinen<sup>1</sup> , Maria C. Uyarra<sup>6</sup> and Angel Borja<sup>6</sup>

*<sup>1</sup> Marine Research Centre, Finnish Environment Institute (SYKE), Helsinki, Finland, <sup>2</sup> MariLim Aquatic Research GmbH, Schönkirchen, Germany, <sup>3</sup> European Commission, Joint Research Centre (JRC), Directorate for Sustainable Resources, D.2 Water and Marine Resources Unit, Ispra, Italy, <sup>4</sup> Bioscience, Aarhus University, Silkeborg, Denmark, <sup>5</sup> Centre for Environment,*

#### Edited by:

*Michael Elliott, University of Hull, UK*

#### Reviewed by:

*Iñigo Muxika, AZTI-Tecnalia, Spain Ursula Scharler, University of KwaZulu-Natal, South Africa*

\*Correspondence:

*Anna-Stiina Heiskanen anna-stiina.heiskanen@ymparisto.fi*

#### † Present Address:

*Heliana Teixeira, Departamento de Biologia and CESAM, Universidade de Aveiro, Aveiro, Portugal; Axel G. Rossberg, School of Biological and Chemical Sciences, Queen Mary University of London, London, UK*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *20 June 2016* Accepted: *08 September 2016* Published: *23 September 2016*

#### Citation:

*Heiskanen A-S, Berg T, Uusitalo L, Teixeira H, Bruhn A, Krause-Jensen D, Lynam CP, Rossberg AG, Korpinen S, Uyarra MC and Borja A (2016) Biodiversity in Marine Ecosystems—European Developments toward Robust Assessments. Front. Mar. Sci. 3:184. doi: 10.3389/fmars.2016.00184* Sustainability of marine ecosystems and their services are dependent on marine biodiversity, which is threatened worldwide. Biodiversity protection is a major target of the EU Marine Strategy Framework Directive, requiring assessment of the status of biodiversity on the level of species, habitats, and ecosystems including genetic diversity and the role of biodiversity in food web functioning and structure. This paper provides a summary of the development of new indicators and refinement of existing ones in order to address some of the observed gaps in indicator availability for marine biodiversity assessments considering genetic, species, habitat, and ecosystem levels. Promising new indicators are available addressing genetic diversity of microbial and benthic communities. Novel indicators to assess biodiversity and food webs associated with habitats formed by keystone species (such as macroalgae) as well as to map benthic habitats (such as biogenic reefs) using high resolution habitat characterization were developed. We also discuss the advances made on indicators for detecting impacts of non-native invasive species and assessing the structure and functioning of marine foodwebs. The latter are based on indicators showing the effects of fishing on trophic level and size distribution of fish and elasmobranch communities well as phytoplankton and zooplankton community structure as food web indicators. New and refined indicators are ranked based on quality criteria. Their applicability for various EU and global biodiversity assessments and the need for further development of new indicators and refinement of the existing ones is discussed.

Keywords: indicators, food web, good environmental status, invasive species, pelagic ecosystem, benthic ecosystem, marine strategy framework directive

# INTRODUCTION

Sustainability of marine ecosystems and their services are dependent on marine biodiversity, which is threatened worldwide (Narayanaswamy et al., 2013; Bennett et al., 2015). Biodiversity is fundamental to sustain marine ecosystem services, such as food, maintenance of water quality, and recovery from perturbations (Beaumont et al., 2007; Liquete et al., 2016). Despite its important role and contribution to human wellbeing, its lost has been reported world-wide. The main threats to marine biodiversity include habitat loss, overexploitation, pollution by hazardous substances, eutrophication, and invasions by nonindigenous species (Kappel, 2005; Venter et al., 2006). Efforts to reduce these pressures for halting the biodiversity loss, a commitment of the signatory countries of the Convention on Biological Diversity (CBD, 1992), is therefore essential for global food security, coastal water quality, ecosystem stability, and buffering the resistance and recovery of ecosystem services, thus enabling different types of future economic valuations and management (Reker et al., 2015). Restoring marine biodiversity through sustainable fisheries management, pollution control, maintenance of essential habitats, and the creation of marine reserves, are some of the opportunities for investments that can support the productivity and reliability of goods and services that the ocean provides to humanity (Worm et al., 2006; Palumbi et al., 2009; Cressey, 2016). Marine management should ensure sustaining all of an ecosystem's biological parts at functioning levels—via conservation of biodiversity at all different levels (from genetic to ecosystems)—in order to maintain ecosystem integrity and stability (Palumbi et al., 2009). The objective of ecosystem-based management of marine environment is to ensure healthy, functional and diverse ecosystems by managing the key drivers of adverse impacts. Biodiversity indicators need to measure variables that are documented to respond to pressures, using methods that can distinguish the anthropogenic impact from natural variability (Borja et al., 2016).

In order to understand the current biodiversity status and its conservation needs (including restoration and prevention), it is imperative to monitor fundamental parameters of biodiversity, both structural and functional (Strong et al., 2015). As biodiversity is such a multifaceted concept, monitoring may need covering genetic variability and physiological or life history diversity within species, surrogate taxa such as habitat-forming species (e.g., seagrasses, kelps), pollutant-recycling species (e.g., marsh grasses, macroalgae), and species diversity all the way from megafauna to microbes, planktonic prokaryotes and microeukaryotes, and energy flow hubs (Strong et al., 2015). In addition to skilful taxonomists, the monitoring process might strongly benefit from information-based tools designed for quicker assessments of taxonomy (Pittman et al., 2007), longterm monitoring sites, new tools for remote and continuous measurement of different biological components (e.g., microbial diversity, oceanic, and coastal phytoplankton and zooplankton, and meio- and mega-fauna in the benthos). Such tools include (i) genomics (Bourlat et al., 2013), (ii) robust marine biosensors (e.g., automated aerial, surface, and underwater drones equipped with sonar or acoustic monitoring), (iii) underwater cameras for detection of ocean fauna, and (iv) improved mathematical models to chart energy flow within food webs amounting to creation of marine life observatories (Palumbi et al., 2009).

The EU Marine Strategy Framework Directive (MSFD; 2008/56/EC), one of the major legal frameworks for the protection of marine biodiversity together with the EU Biodiversity Strategy 2020 (COM/2011/0244) and the Convention on Biological Diversity (CBD, 1992), highlights setting programs for monitoring and assessing the environmental status of the marine waters. According to MSFD, the status of the marine environment is evaluated using 11 descriptors, that comprise both biodiversity related descriptors (D1, biological diversity; D4, food-webs; and D6, seafloor integrity) and pressure descriptors (D2, non-indigenous species; D3, fisheries; D5, eutrophication; D7, hydrological conditions; D8 and D9, contaminants in the environment and in seafood; D10, litter; D11, energy and noise). These are further detailed in the EU Commission Decision 2010/477/EU providing 29 criteria and 56 associated "indicators" that should be monitored for the assessment of the environmental status.

The MSFD puts biodiversity in the center of the assessment of marine environmental status (Borja et al., 2010). The descriptor (D1) on biodiversity has the following target to contribute to the achievement of the Good Environmental Status (GES): "Biological diversity is maintained. The quality and occurrence of habitats and the distribution and abundance of species are in line with prevailing physiographic, geographic, and climatic conditions." The background and definitions, the key attributes (biological components, predominant habitat type, and ecotypes for mobile species) as well as the suggested indicator classes for the descriptor (D1) criteria of the attributes are presented by Cochrane et al. (2010). While this descriptor directly targets biodiversity, MSFD Descriptor 4 (Marine food-webs), which calls for maintenance of the normal functioning of marine food-webs, and some aspects of MSFD Descriptor 6 (Seafloor integrity) are also closely related to the assessment of biodiversity (Borja et al., 2010).

The assessment of the ecological status of coastal waters is also required by the EU Water Framework Directive (WFD; 2000/60/EC) which does not specifically address biodiversity. Nevertheless some of the indicators for biological quality elements under WFD, such as phytoplankton, macrophytes, zoobenthos, and additionally indicators for fish community structure for transitional waters (e.g., Heiskanen et al., 2004), can be also applied to the MSFD (Borja et al., 2010). Some of these indicators include parameters for species composition, community structure and abundance, and are thus also applicable for assessing biodiversity at the community level. For example, zoobenthos indicators such as AMBI (AZTI Marine Biotic Index; Borja et al., 2000) and BBI (Brackish waters Benthic Index; Perus et al., 2007) and macrophyte indicators such as eelgrass depth limit, all describe aspects of those communities (or biogenic habitats) and can be used in the biodiversity assessment for the MSFD (Rice et al., 2012) and have been included in the initial assessments of the marine environment carried out by EU member states in 2012.

The analysis of indicators and assessments applied in the MSFD initial assessments during the first phase of the MSFD implementation revealed some problems regarding the degree of development and operationality of the biodiversity assessment within the EU member states compared to the requirements of the MSFD (Berg et al., 2015; Hummel et al., 2015). Data availability and regional specificities influenced the number of methodologies used and reported by member states (Palialexis et al., 2014). Some ambiguity in the EU Commission Decision (Berg et al., 2015) introduced discrepancies and increased the potential for non-harmonized approaches in the assessment of the marine environment even between member states of the same regional sea.

In addition to the assessments performed by EU member states, marine biodiversity assessments have also been carried out at regional level by the Regional Sea Conventions (RSC), which have identified a number of biodiversity indicators representing the different trophic levels of the marine ecosystem as well as the relevant habitats and ecosystems. OSPAR and HELCOM, the relevant RSC for the North East Atlantic and Baltic Sea regions, respectively, have agreed to develop common indicators for the major elements and species groups of marine biodiversity: benthic and pelagic habitats, seabirds, marine mammals, fish and food webs (HELCOM, 2016; OSPAR, 2016). Existing monitoring programmes in these regions were originally set up to assess pollution effects in the marine system and commercially exploited fish stocks. Therefore, these did not originally cover biodiversity assessment needs as specified by the MSFD. Many of the common indicators currently being defined under OSPAR and HELCOM are new indicators to the regions, specifically developed for the forth-coming environmental assessment in 2018 and many of them have not been properly validated yet.

In order to complement the on-going work for biodiversity indicator development for the MSFD environmental status assessments in the EU, the EU FP7 project DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status; www.devotes-project.eu) carried out a comprehensive overview of existing biodiversity-related indicators used in previous marine assessments carried out under different policy frameworks, in the MSFD initial assessment, and elsewhere (Teixeira et al., 2014). The DEVOTES inventory of the existing biodiversity indicators was compiled in the form of a catalog, which includes over 600 entries. In this paper, we present an overview of the indicator development and refinements in DEVOTES, which were aimed to address some of the identified gaps and further development needs for the next phase of the MSFD assessment (by 2018).

## MATERIALS AND METHODS

DEVOTES biodiversity indicator catalog is available via a database software (DEVOTool), which allows navigating the metadata (http://www.devotes-project.eu/devotool) and to make specific queries to find existing indicators depending on the needs of the user. The catalog currently includes over 600 indicators of marine biodiversity, food web status, sea floor integrity, and alien species, used and proposed to be used in marine assessments. We queried the database to find out how many indicators there are for each ecosystem component, and how they relate to the MSFD Descriptors and criteria (**Table 1**). The exact DEVOTool query used here is presented in the Supplementary Material. **Table 1** presents an overview of the number of indicators included in the DEVOTool Indicator Catalog and representing the different biodiversity components that these indicators cover (i.e., Microbes, Phytoplankton, Zooplankton, etc.), and their applicability to the MSFD Descriptors and criteria.

At the early stages of the project, after the initial indicator catalog compilation, similar data was used to analyse the gaps in the available indicator coverage, and to prioritize the indicator development taking place within the project (Teixeira et al., 2014). The gap analysis was carried out comparing how well the indicators in the catalog cover the requirement of the MSFD Commission Decision (2010/477/EU) criteria and indicators for the biodiversity related descriptors D1, D2, D4, and D6. Also it was evaluated how well the indicators in the catalog covered the biodiversity components, and habitats, as listed in **Table 1** of Annex III to the Directive (MSFD; 2008/56/EC) and specified by Cochrane et al. (2010). The number of indicators refined and new indicators developed for various biodiversity components are also included in **Table 1**. Overview of the methodologies for development of new indicators and refinements of existing indicators are compiled in Berg et al. (2016) and referred in the text.

# RESULTS

### Gap Analysis of Biodiversity Indicators

Despite the high number of marine biodiversity indicators available, there were important gaps in terms of the MSFD requirements (**Table 2**; Berg et al., 2015). Additionally, there is insufficient information regarding the quality and confidence of the indicators (Queiros et al., 2016). Most indicators lack regional targets or quality threshold values, and few have a measure of confidence and a demonstrated link to pressures (Teixeira et al., 2014). Thus, although the indicators were operational in the sense that they were used in previous marine assessments, their applicability to fulfill the criteria of MSFD is less evident (Berg et al., 2015).

Major gaps observed for the Descriptor 1 were indicators to assess the biodiversity at ecosystem level and the genetic composition of populations, indicators for microbes, pelagic invertebrates (Cephalopods), and reptiles (**Table 2**). Abyssal and bathyal zones totally lack indicators addressing those depths (Hummel et al., 2015, Teixeira et al., 2014). Moreover, habitats with restricted distribution in the regional seas (like ice-associated species and communities) had a low overall number of indicators (Teixeira et al., 2014). Indicators for the MSFD biodiversity criterion 1.7 "Ecosystem structure" and the associated processes and functions are relatively scarce (Berg et al., 2015), and those few addressing this criterion focused essentially on communities, often in isolated components. Only few biodiversity indicators related to ecosystem processes and function were reported (**Table 1**), all of them for the North-Eastern Atlantic (Teixeira et al., 2014).

MSFD Descriptor 2 addresses the threat on natural biological diversity caused by the non-indigenous species (NIS), which are the taxa introduced outside their natural range and natural dispersal potential as result of human activities. Particularly the invasive NIS are recognized as a global threat to biodiversity (Olenin et al., 2010). Despite the requirements of MSFD Descriptor 2, there were no reported indicators in Europe



*(http://www.devotes-project.eu/devotool).*

*component.*

 *See Supplementary*

 *Material for the explanation how the data was obtained from the DEVOTool Indicator Catalog*  TABLE 2 | Overview of the identified gaps in the suite of existing indicators (as included in the DEVOTool Catalog v. 7) for MSFD environmental status concerning Descriptors 1, 2, 4, and 6 (Annex 1) with respect to MSFD criteria (as identified in Commission Decision 2010/477/EU) and development of new and refined indicators in the DEVOTES project, as well as the identified future research needs.


specifically assessing the "Impacts of non-indigenous invasive species at the level of species, habitats and ecosystem, where feasible" (MSFD criteria 2.2., indicator 2.2.2; **Tables 1, 2**; Teixeira et al., 2014). In most cases, the MSFD initial assessments of the Member States did not include reporting of the adverse effects in biodiversity or the magnitude of impacts caused by NIS (Palialexis et al., 2014; Berg et al., 2015). RSCs (e.g., Barcelona and Bucharest Conventions, HELCOM, and OSPAR) have not included indicators for NIS impacts within their agreed set of indicators. Only recently there have been research activities to develop practical proposals to help managers assess these requirements in standardized ways (e.g., Zaiko et al., 2011; Katsanevakis et al., 2016; Nentwig et al., 2016; Rabitsch et al., 2016).

Most of the MSFD food web indicators (Descriptor 4) are related to "Abundance or distribution of key trophic groups or species" (MSFD criteria 4.3), and are thus simultaneously D1 (MSFD criteria 1.1) indicators (Berg et al., 2015). There is a need to develop indicators related to MSFD criterion 4.1 ("Productivity (production per unit biomass) of key species on trophic groups") especially focusing on primary and secondary producers, as well as for the criterion 4.2 ("Proportion of selected species at the top of food-webs"; **Table 2**).

Finally, the remaining biological descriptor of the MSFD (Descriptor 6 on sea-floor integrity) is rather well developed in terms of indicators, having a high number of them (≥20 indicators), both for criterion 6.1. ("Substrate characteristics physical damage") and 6.2. ("Condition of benthic community") for angiosperms, macroalgae, and benthic fauna. However, there is a need for developing indicators especially for cephalopods, but also for other biological components. Also there is a gap in the development of targets at the level of benthic habitats and ecosystems.

### Biodiversity Indicator Refinement and Development in DEVOTES

In order to address the identified gaps in the indicator availability for marine assessments, 16 new indicators were developed and 13 indicators were refined (**Table 3**) to better fulfill the MSFD requirements (Berg et al., 2016). The indicators were scored according to the eight indicator quality criteria listed by Queiros et al. (2016) in order to evaluate their fitness as potential indicators for the MSFD assessments (**Figure 1**). The quality criteria are: (1) Scientific basis, (2) Ecosystem relevance, (3) Responsiveness to pressure, (4) Possibility to set targets, (5) precautionary capacity, (6) Quality of sampling method, (7) Costeffectiveness, and (8) Existing and ongoing monitoring data. In general, the newly developed indicators covered both pelagic and benthic ecosystems and addressed several biological components and habitats that were identified as lacking indicators or as being under-represented in several marine regions (Teixeira et al., 2014). In addition, 13 pre-existing indicators were further refined in order to improve their performance and confidence. For example, they were tested for responsiveness to pressures, or using new data sets to validate their applicability in wider sea regions.

#### Indicators for Genetic Diversity of Microbial and Benthic Communities

The genetic structure of a population (MSFD criteria 1.3, and indicator 1.3.2) was the least covered of the MSFD D1—biological diversity indicator requirements (**Table 1**; Teixeira et al., 2014). This highlights the need for new developments addressing the genetic component of biodiversity and introducing such new methods to marine monitoring programmes. Emerging assessment tools based on molecular techniques have received strong attention from the scientific community (Bourlat et al., 2013), which might increase their potential for contributing to assessments of GES.

Genetic diversity is an aspect of biodiversity that has recently gained increased attention, but operational indicators addressing genetic structure of the populations are still scarce. Instead of using traditional sampling, taxonomic data can be obtained using a DNA metabarcoding technique. Microbial indicators were identified as one of the gaps and, consequently, further work to advance bacterial community indicators using nucleic acid microarrays was initiated in DEVOTES. Microbial abundance and biodiversity variables are relevant for several MSFD descriptors i.e., D1 Biodiversity, D4 Food webs, and also D5 Eutrophication (Caruso et al., 2015). Specifically, MSFD indicators 1.2.1 (Population abundance and/or biomass, as appropriate), and 1.3.1 (Population demographic characteristics (e.g., body size or age class structure, sex ratio, fecundity rates, survival/mortality rates), and 1.3.2 (Population genetic structure, where appropriate) were targeted in the development work.

Micro-organisms present in the sediments were also included in the analysis addressing the MSFD indicator 6.2.1. Presence of particularly sensitive and/or tolerant species; in addition to D1 indicators. The microbial sediment indicator followed the approach of gAMBI (genetic AMBI; Aylagas et al., 2014, 2016), and scored quite high in the fitness as an operational indicator, particularly on ecosystem relevance, concreteness, early warning capacity, and cost efficiency (**Figure 1**).

The AMBI indicator for benthic invertebrates (Borja et al., 2000) was modified by applying simultaneous amplification of a standardized DNA fragment from the total DNA extracted from an environmental sample (gAMBI). This allows the rapid, accurate and cost-effective identification of the entire taxonomic composition of thousands of samples simultaneously (Aylagas et al., 2014, 2016). Such DNA data do not provide accurate estimations about the abundance of the taxa in a certain sample, so current use is restricted to presence/absence estimations. Nonetheless, a high proportion of the taxa visually identified can be detected using the DNA metabarcoding technique. The benthic gAMBI indicator scored also high in the evaluation of the indicator fitness (**Figure 1**).

#### Habitats of Key Species as Indicators of Biodiversity on Ecosystem Level Functions

Vegetated marine areas, such as seagrass meadows and kelp forests, are important habitats for a wide diversity of algae, invertebrates and fish (Steneck et al., 2002, Schmidt et al., 2011, Boström et al., 2014, Sheaves et al., 2015, Thormar et al., 2016) including economically important fish such as cod (Lilley and


 3 | New and refined indicators developed during the DEVOTES project.

TABLE


TABLE

3


Continued


TABLE

3


Continued

*spatial coverage of the indicator; (4) if the indicator was relevant to MSFD assessment*

*developed for the indicator; "Trends" means that indicator status is dependent on the trends direction, and "NA" means that target setting is not part of the indicator methodology),*

*i.e., used in established management*

*in detail.*

**241**

 *frameworks*

 *such as the OSPAR system of Ecological Quality Objectives (EcoQOs) for the North Sea. The last column refers to the latest publication,*

 *purposes ("Targets" means that the indicator methodology*

 *includes setting of quantitative targets, while "No targets" means that those are not yet*

 *and (5) whether the indicator is already established,*

 *where the indicator development*

 *is described*

Unsworth, 2014). The vegetation stimulates biodiversity by being a habitat-forming ecosystem component, hugely increasing the colonizable area while also providing shelter and food for a wealth of organisms.

In turn, the organisms of the habitat exert a feed-back effect on the meadows and their functioning. Hence, the presence of fish exerts an import role in contributing to maintaining healthy vegetated ecosystems via top-down control of nuisance algae (Baden et al., 2012). The extent and biomass of sea grass meadows may also couple with the abundance of waterfowl in terms of bottom-up as well as top-down effects, thereby showing an additional link to biodiversity. Accordingly, Berg et al. (2016) propose that "the idea of an indicator is that extended eelgrass cover/biomass in combination with large populations of foraging birds reflects good environmental status" which "requires a suitable balance between bottom up control of eelgrass meadows on bird populations and top-down control of the birds on eelgrass meadows." Testing of the relationships between eelgrass cover/biomass and the abundance of herbivorous waterfowl was carried out in order to develop a new indicator "Distribution of herbivorous waterfowl in relation to eelgrass biomass distribution" (Berg et al., 2016). This indicator is relevant for several GES criteria, D1 (biodiversity) and D4 (food webs), 1.1. Species and 1.4. Habitat distribution; and 4.3. Abundance/distribution of key trophic groups/species. The indicator scored relatively high (**Figure 1**), and thus it is considered as a promising indicator for MSFD. On this basis we suggest that indicators describing habitat extent and biomass of key stone species may serve as indicators of biodiversity both on habitat as well as on species level and concurrently enabling coupling with other biological components, and providing proxies also for ecosystem level of structural and functional biodiversity indicators.

Moreover, one of the commonly used macrophyte indicators is the Lower Depth distribution Limit of Macrophytes (LDLM), which indicates the distribution and abundance of habitat forming macrophyte species. This indicator informs about the following MSFD criteria 1.1. Species, and 1.4. Habitat distribution, as well as 5.3. Indirect effects of nutrient enrichment, particularly the indicator 5.3.2. abundance of perennial seaweeds and seagrasses adversely impacted by decrease in water transparency. In DEVOTES, the target setting of the LDLM indicator for a perennial red alga species, Furcellaria lumbricalis was refined and harmonized (**Table 3**).

A promising tool to characterize and map marine habitats is to use multibeam echosounders on vessels (systematic high resolution habitat characterization) which provide high resolution and georeferenced technology and allow continuous and direct mapping of biogenic reef-forming species (e.g., Harris and Baker, 2011). This methodology can be used to derive operational indicators for many MSFD biodiversity indicators under the criteria 1.5 Habitat extent, and (1.6) condition, as well as 1.7 Ecosystem structure, 1.7.1. Composition and relative proportions of ecosystem components (habitats and species) (Berg et al., 2016).

#### Non-indigenous Species Indicators (D2)

One of the identified gaps in all regional seas was the lack of indicators for measuring the ecological impact of non-indigenous species in the marine ecosystems. The early detection of invasive species, using eDNA and metabarcoding was addressed by Ardura et al. (2015) and Zaiko et al. (2015). The abundance and distribution range (ADR), a semi-quantitative characteristic of the extension of a non-indigenous species population within the biopollution assessment framework (Olenin et al., 2007), can be used as measure of the bioinvasion impact. It was tested on the zebra mussel studying ecosystem-level impacts (Minchin and Zaiko, 2013; Zaiko et al., 2014) and the results showed that ADR of zebra mussels generally corresponded with the overall impact score and might be indicative of the particular invasion phase (establishment, expansion, outbreak, accommodation). ADR could thus serve as a proxy for the overall magnitude of impact of the species. Since data on species abundance and distribution can be retrieved from the regular biodiversity monitoring records, delivering ADR is a cost-effective solution for environmental status assessment. Determining the cumulative impact of invasive non-indigenous species is another approach toward assessing their role in biodiversity. The recently developed Cumulative IMPact index of Invasive ALien Species (CIMPAL) uses a spatially explicit conservative additive model based on the distributions of invasive species and ecosystems, including the reported magnitude of ecological impacts and the strength of such evidence (Katsanevakis et al., 2016).

#### Indicators for Food Webs (D4)

#### **Productivity of key species on trophic groups**

Indicators for phytoplankton primary production (PP), provide information on the vigor of an ecosystem (energy fluxes and ability to recover from disturbance) and thus health of the pelagic ecosystem (Tett et al., 2007). Phytoplankton photosynthesis produces organic matter which is then utilized by organisms at higher trophic levels and provides the base on the pelagic food web. There are currently different methods adopted for measuring PP (e.g., oxygen evolution, <sup>14</sup>C method, PAM fluorometers, models, remote sensing). Traditional methods for measuring PP (e.g., <sup>14</sup>C method) are reliable but timeconsuming, expensive and localized. Other methods (e.g., remote sensing, models) can investigate a wider area but require validation or may have limited applicability to certain water types. Time series of annual gross primary production (AGPP) at different ecohydrodynamic regions of the North Sea (based on van Leeuwen et al., 2015) were calculated using an empirical model (Cloern, 1987) from measurements of chlorophyll (proxy of phytoplankton biomass), light attenuation coefficient (Kd), and surface irradiance. Currently there is no ongoing monitoring that would provide data for AGPP indicator calculation and target setting, thus the indicator had a medium score (**Figure 1**), and it requires further development to be operational.

Phytoplankton blooms in coastal and open marine waters are characterized by high temporal and spatial fluctuations. Therefore remote sensing and continuous fluorometric measurements are promising tools to detect and measure phytoplankton phenomena in the surface layers of marine waters (e.g., Kutser, 2009; Kahru and Elmgren, 2014; Cristina et al., 2015). The phytoplankton biomass assessment method based on surface chlorophyll-a concentration measurements using cost-effective remote sensing data (Gohin, 2011a,b; Novoa et al., 2011; Cristina et al., 2015), builds on the indicator assessment approach developed for in situ samples (Revilla et al., 2009) and gained a high score in the indicator evaluation matrix (**Figure 1**).

The joint use of remote sensing biomass observations and ship-of-opportunity fluorescence measurements are a powerful combination to detect changes both in the phytoplankton biomass and composition. Remote sensing, bio-optics, microscopy, and CHEMTAX results provide a combination of analytical tools that can be used to develop a phytoplankton biomass (chlorophyll a) index for marine and coastal waters off the Iberian peninsula, in Portugal (Gohin, 2011b; Cristina et al., 2015, 2016a,b; Goela et al., 2015).

#### **Proportion of selected species at the top of food-webs: fish and elasmobranch indicators**

A number of promising indicators that capture the effects of fishing on marine biodiversity has been tested (Fu et al., 2015; Lynam and Mackinson, 2015; Coll et al., 2016). Fishing targets particular species and through the inherent selectivity of the fishing gears, often regulated by mesh size restrictions, larger individuals of populations are preferentially caught. So community level indicators that focus on changes in abundance (Kleisner et al., 2015), species composition (such as the mean maximum length of fish and elasmobranchs), trophic level (Shannon et al., 2014), or the relative biomass across a size spectrum (Engelhard et al., 2015; Thorpe et al., 2016) of fish and elasmobranch assemblages can respond strongly to direct fishing pressure. DEVOTES also developed an indicator of size composition in fish and elasmobranch communities (Berg et al., 2016), the biomass weighted geometric mean length of fish, known as Typical Length (ICES, 2014) since changes in size structure has been shown to represent change in trophic level (Jennings et al., 2007).

Shannon et al. (2014) made a comparison of the performance of trophic level (TL) indicators to demonstrate fishing impacts in 9 marine ecosystems and showed that the information content of these indicators differed depending on the data source, the previous changes in state and the historical development of fishing in the system. Catch-based TL indicators calculated using landings statistics represent the pressure on the system, while survey based TL indicators show unbiased state changes at the surveyed community level. In order to gain a complete picture of the wider effects of fishing on food web structure, complementary model derived TL indicators can be examined. While a metaanalysis of the 9 ecosystems revealed a significant pattern of low TL (for either catch-, survey,- or model-based indicators) under high fishing mortality, the relationship varied greatly at the level of the single ecosystem (some positive and negative relationships in addition to non-significant relationships). To understand why a particular trajectory in TL occurred in any single ecosystem, a good knowledge of changes in fisheries management and environmental change specific to the system is necessary.

Kleisner et al. (2015) tested the "Non-Declining Exploited Species" (NDES) indicator across 22 marine ecosystems, where the indicator is the proportion of species with survey catch-rates that have a positive monotonic temporal trend assessed with a significance test determined using a distributional test for the community given the length of time series data available (see Lynam et al., 2010). The authors conclude that the indicator can provide a valuable and relatively easy-to-understand measure of change in the ecosystem. The authors compared their evaluation of changes within ecosystems by the NDES to similar evaluations using community indicators derived from survey data (i.e., the proportion of predatory fish, mean trophic level, and mean life span; see Coll et al., 2016). In many, but not all cases, a decline in NDES was mirrored by the community metrics. In other cases, fishing pressure was found to be impacting only part of

the community and this was not reflected well in the overall community metrics. Thus, the scale of fishing impacts on the ecosystem and thus the responsiveness of community metrics to pressure are dependent on the level of fishing pressure relative to other drivers including natural environmental change as shown by Fu et al. (2015). Similarly, spatial patterns in indicators and their response to drivers are often evident. Engelhard et al. (2015) demonstrated that the Large Fish Indicator of the demersal fish and elasmobranch community has responded to decreases in fishing pressure in those parts of the North Sea where demersal fishing effort was once high but the response was not uniform across the area. This study along with Marshall et al. (2016) demonstrates that that the strength of different drivers of fish community structure varies across the North Sea. As a result, the outcomes of management measures are likely to vary in different localities.

#### **Abundance/distribution of key trophic groups/species**

Phytoplankton community composition. Phytoplankton community composition can be used as an indicator for food web structure as well as being an early warning indicator for subsequent effects on the food web. Food web indicators are an important part of biodiversity assessment because the food web delivers energy to all trophic levels thus sustaining the biodiversity components, and chl-a alone is not an applicable indicator for the dynamic processes in food webs. To this end, a phytoplankton community composition index was developed (Suikkanen et al., 2013). The application to areas in the Baltic Sea revealed that late summer communities in the Gulf of Finland, the Åland Sea, and the northern Baltic proper have shifted toward more microbial, less energy-efficient food webs consisting of more mixotrophic and lower food-quality phytoplankton. This may lead to a decreased availability of energy for herbivorous zooplankton and planktivorous fish, despite an observed increase in chl-a and phytoplankton biodiversity. The food web indicator "Phytoplankton community composition as a food web indicator" scored high in the indicator quality evaluation and is currently a candidate indicator for HELCOM holistic ecosystem assessment (Lehtinen et al., 2015).

Novel indicators focusing on the role and impact of N2 fixing cyanobacteria in the pelagic food web were also considered. Cyanobacterial nitrogen is efficiently assimilated and transferred in Baltic food webs (Karlson et al., 2015). On the other hand, high abundance of cyanobacteria may harm the copepod reproduction and exert negatively on the food web (Engström-Öst et al., 2015). However, the tested indicators did not show a clear and coherent response to pressures, and scored low in ranking of the potential indicators (**Figure 1**).

Also phytoplankton community composition based on food quality traits could potentially be used as an early warning indicator for food web effects on higher trophic levels, as the quality of different phytoplankton taxa as food source for higher trophic levels varies (e.g., Danielsdottir et al., 2007). The same idea was also behind the indicator based on diatom/dinoflagellate ratio that has implications for zooplankton community composition and further in the food web. The different functional properties of diatoms and dinoflagellates have an influence on the fate of the organic matter produced and thus have consequences for the overall biogeochemical cycles (e.g., Klais et al., 2011). Diatoms and dinoflagellates are proposed as a life form indicator in the OSPAR area (Gowen et al., 2011), considered as a supplementary indicator in the Baltic Sea (HELCOM, 2012a,b; Klais et al., 2011), and in the Black Sea (Sahin et al., 2007).

Revision of other existing indicators for phytoplankton diversity were also considered and developed further where feasible. The Shannon95 indicator (Uusitalo et al., 2013) and phytoplankton taxonomic evenness, that has been shown to correlate with the resource use efficiency and stability of the community (Ptacnik et al., 2008), were tested but did not show a clear and coherent response to pressures. The indicator on seasonal succession patterns of phytoplankton groups (Devlin et al., 2009), describes the normal or established seasonal succession patterns of phytoplankton groups and suggests that major deviations from this pattern indicate impairment of environmental status. However, sufficiently frequent sampling is seldom available through monitoring programmes.

#### **Zooplankton community composition**

Zooplankton has a crucial role in the pelagic food web, as it transfers energy from phytoplankton to higher trophic levels (Checkley et al., 2009), and changes in the zooplankton community's abundance and composition are related to the functioning of the aquatic ecosystem (Jeppesen et al., 2011). Changes in the eutrophication status of aquatic systems impacts composition of zooplankton community (Gliwiz, 1969; Pace, 1986), and the growth of planktivorous fish is regulated by composition and biomass of the mesozooplankton that they feed upon (Cardinale et al., 2002; Rajasilta et al., 2014). A number of zooplankton indicators have been proposed to assess the status and functioning of marine food webs (Teixeira et al., 2014; Berg et al., 2016).

Long-term monitoring of mesozooplankton composition has been conducted in the Baltic Sea as well as in the Black sea, and based on these data, zooplankton metrics have been proposed as indicators of environmental status both in the Baltic (HELCOM, 2012b) and in the Black Sea (Bulgarian Initial Assessment report, 2013, and Black Sea Commission zooplankton expertise group).

Biomass of mesozooplankton includes information of major key groups, forming the structure of the planktonic fauna, particularly the groups Copepoda, Cladocera, Meroplankton, and the species Oikopleura dioica and Parasagitta setosa. Copepods are present all year round and distributed within the coastal, shelf, and open sea habitats. They are a key group of mesozooplankton that reflects the food availability for zooplanktivorous fish (particularly sprat and anchovy, partly horse mackerel). Mesozooplankton community composition is indirectly impacted by eutrophication (via changes in primary productivity and phytoplankton community composition), whereas climatic changes, predation, introduction of synthetic compounds (from point sources), and predation of invasive species, result in direct impacts. Relatively high copepod biomass implies food availability for fish and consequently is considered to represent good status of the food web structure (D4

#### criterion 4.3.1 abundance trends of functionally important selected groups/species).

The response of the zooplankton indicators with respect environmental variables was tested using Signal Detection Theory (SDT; Murtaugh, 1996). Phytoplankton total biomass and chl-a values were used as "Golden Standard" meaning that the indicator was considered to represent good status when these metrics were in good status. Zooplankton indicators received relatively low grading using the quality criteria ranking. Zooplankton is patchily distributed and seasonal variations in biomass and species structure result in a large variation in the data. Therefore the failure to detect a response to pressure patterns might partly be a problem in the different spatial and temporal scales of zooplankton and pressure data.

#### Approaches for Setting Thresholds/Targets for the Biodiversity Indicators

Indicator boundaries (thresholds) or target values are necessary to decide whether management action is required. A numeric definition for GES, i.e., the GES boundaries, can be defined by several alternative approaches (HELCOM, 2012a): (a) as an"acceptable" deviation from a reference condition (i.e., reference conditions representing natural conditions with minimal impact of anthropogenic pressures, that is the EU WFD approach), (b) as an "acceptable" deviation from a fixed reference point (i.e., fixed or depending on other variables), (c) as an "acceptable" deviation from a desired hypothetical condition (e.g., based on models), (d) as a threshold derived from ecological or physiological models (e.g., carrying capacity of a system, critical depth for photosynthesis, etc.), (e) as temporal trends, or tipping points (e.g., an analysis for changes in status), and (f) as biological effects on the condition of an organism (e.g., thresholds for contamination effects).

In the Baltic Sea, HELCOM has coordinated development of core indicators, which also included setting up the GES boundaries. In practice, the core indicators' GES boundary is not only a single threshold but can be a range (with a lower and upper thresholds), a direction of a trend, or based on a class-scale. Also it appeared to be practical to apply several approaches in parallel, when setting GES boundaries for the HELCOM core indicators (HELCOM, 2013). The trend-based targets are heavily debated, as they do not address whether the status is GES or not, but only show the direction. An alternative to this could be a class-scale assessment, which could be given under high uncertainty of more definite GES threshold. This has not, however, yet been applied to any of the core indicators in practice.

SDT was tested for setting the threshold values for indicators (Chuševe et al., 2016 ˙ ). SDT was applied to the "Benthic quality index" (BQI; Rosenberg et al., 2004) in order to check its accuracy, sensitivity and specificity. In general, the SDT was found to be a robust and scientifically sound approach to set boundaries for indicator values, and to be helpful for planning environmental monitoring.

Finally, a new approach to address the target setting of the indicators in relation to ecosystem resilience (i.e., the ability to recover rapidly and predictably from pressures) and to select indicators and their target ranges has been introduced (Rossberg et al., 2017). The idea is to simply choose the target range for any ecosystem state indicator as the range of values from where, when all pressures were hypothetically removed, the mean time to reach the indicator's natural range of variation was no longer than the "acceptable recovery time R," which is a societal choice. Based on examples, an acceptable recovery time was settled to 30 years. Where this criterion was applied, Rossberg et al. (2017) showed that this definition naturally leads to (1) related criteria for pressure indicators, and (2) selection criteria for important indicators among a range of candidates and for suites of indicators.

This approach implies that it is not always necessary that the targets of MSFD indicators aim at restoring natural or near-natural ecosystem states. Deviations from natural states are acceptable if recovery to natural states is not too slow. It is acknowledged that some ecosystem components are naturally much less resilient than others; and therefore the focus is paced on indicator-based assessments of these low-resilience components that recover slowly after pressures have been removed or decreased. Rossberg et al. (2017) then argued that state indicators and pressure indicators should always be used jointly in assessments of sustainable use, because due to the slow recovery of low-resilience ecosystem components, there is no immediate relationship between states and pressures. If an ecosystem component recovers quickly after the relaxation of pressures, there is little concern that it might be used unsustainably. The approach by Rossberg et al. (2017) aims to define status boundaries that ensures sustainable use of ecosystem services. It is therefore focused primarily on protecting the interests of future generations. Moreover, the status assessments (within these boundaries) should be complemented by considerations of their suitability for current societal needs. They recommend that these two kinds of assessment, and management decisions based on these, should be carried out by separate management bodies to avoid potential conflicts of interest.

# DISCUSSION

# Indicator Development and Gaps Addressed

The new developments and refined indicators by DEVOTES addressed some of the gaps identified with respect to the MSFD criteria and indicators in different marine regions where possible. In addition, DEVOTES indicator development focused on those biological components and habitats where monitoring data and the expertise in the DEVOTES research consortia were available.

Some of the indicators developed and refined were already established indicators, meaning that those are applied in the management frameworks such as EcoQO indicators (i.e., the OSPAR System of Ecological Quality Objectives—EcoQO—for the North Sea), or in the national MSFD monitoring of the EU member states (**Table 3**). However, many of the new indicators (such as new genetic indicators for microbes) are not yet used in the marine assessments meaning that those need to be tested in different marine regions and approved by the national and regional managers to be part of the MSFD monitoring and assessment programmes. Many of the NIS indicators were evaluated as relevant for MSFD, but those are not yet included in the marine monitoring programmes. Most of the fish and macrozoobenthic indicators developed and refined are used for food web and benthic integrity and biodiversity assessments being more mature for assessment purposes, while many of the phytoplankton and zooplankton indicators were judged not to be particularly useful for MSFD purposes, besides the new indicator on Phytoplankton food quality traits, that is currently considered as a candidate for a HELCOM core indicator.

Two of the identified gaps were the absence of indicators for biodiversity on a genetic level (Descriptor 1, criteria 1.3. Genetic structure of populations) as well as the lack of indicators for microbial communities. A DNA metabarcoding technique was applied to develop indicators to assess both microbial communities (Berg et al., 2016) and benthic invertebrate biodiversity (Aylagas et al., 2014, 2016). Genetic methods such as nucleic acid microarrays were considered a suitable methodology to quickly determine diversity and abundance of microbial communities (DeSantis et al., 2007). Marine prokaryotes respond rapidly to environmental changes and anthropogenic pressures and are thus considered as useful components for the assessment of microbial biodiversity, and impacts of eutrophication and toxic substances (Caruso et al., 2015). When the genetic methods were applied to identify microbial and benthic organisms in already established and tested indicator methodologies such AMBI (Borja et al., 2000), they appeared to provide a costeffective and robust methodology for biodiversity assessment. Development of genetic tools is expanding rapidly and more effort on benchmarking and standardization will be needed to enable the use of genetic tools in biodiversity assessments in the future (Aylagas et al., 2016).

We also developed indicators to assess habitats of key species as indicators of biodiversity on ecosystem functions. Keystone species, such as species of seagrasses, kelps, and intertidal algae are recognized as effective ecosystem engineers forming vegetated habitats with multiple ecosystem functions including the stimulation of biodiversity (Gutiérrez et al., 2011) and the mitigation of climate change (Duarte et al., 2013). As increasing human pressures on coastal ecosystems threaten the continued supply of essential functions and services, the protection of marine vegetated habitats should be a management priority (Duarte et al., 2013). The high number of indicators available for vegetated habitats (**Table 1**) suggests that this is already well recognized and that these indicators (such as lower depth limit of macrophytes) are used in practice in marine assessments for policy purposes. In a recent review on seagrass indicators globally, Roca et al. (2016) carried out a meta-analysis and compared the applicability of different types of seagrass indicators to detect environmental improvement. They concluded that physiological and biochemical indicators are more suitable due their fast response to changes of environmental stressors than structural indicators (e.g., shoot density or biomass). Moreover, there is global and local evidence that biodiversity and top–down control strongly influences the functioning of threatened vegetated ecosystems, and indication that biodiversity is comparably important to global change stressors (Duffy et al., 2015; Matheson et al., 2016). Therefore key stone species prove well suited as operational indicators of biodiversity at the levels of species, habitat and ecosystems.

One of the major global threats to marine biodiversity is the spreading of invasive NIS (Costello et al., 2010). There are a number of regulations for controlling NIS, including the MSFD that aim to restrict the spreading of new NIS (Ojaveer et al., 2014). However, the intrinsic complexity of NIS for being detected on time implies the need to provide early warning indicators for tracking potential vectors of invasions as well as to assess the impacts of NIS on native species, habitats, and ecosystems. Therefore, DEVOTES developed and refined indicators to assess the impacts of invasive species which can be used to identify vulnerable areas, environmental targets and to prioritize management actions.

The MSFD also calls for indicators that address the deviations from the normal structure and functioning of the marine food webs. Such indicators would reflect distortions in the top-down control or bottom-up regulation of the food webs. The cascading impacts of top-down control reach into all trophic levels of food webs. As an example, removal of large predatory species can cause a relief on lower trophic levels causing increase on zooplanktivorous fish and thus, change composition and biomass zooplankton communities. Likewise, the bottom-up regulation based on the changes in the phytoplankton composition and quality of algae as a food source for zooplankton can change productivity and species composition of zooplankton impacting the fish communities. D4, food web, was identified to have relatively few indicators (**Table 1**), and DEVOTES developed and refined indicators that can be used to assess different aspects of food web structure and functioning, such as the abundance and distribution of fish indicators, productivity of key trophic groups (phytoplankton primary production) and phytoplankton composition as food web indicator providing an indication of the palatability of phytoplankton for zooplankton.

Some of the indicators developed or refined in the DEVOTES project were evaluated to show poor responsiveness to pressures (**Table 3**, **Figure 1**). Nevertheless, we suggest maintaining these ecologically relevant indicators in the assessment suites, particularly if they are collected on the side of other monitoring programmes and the monitoring would not be very costly. They could then be used as surveillance indicators and applied to complement the indicators with clear pressure-state links (ICES, 2014). Some factors used in the assessment of marine ecosystem health do not have a very clear pressure-state relationship (ICES, 2014). Many foodweb components are simultaneously affected by multiple pressures and processes, showing response to cumulative and synergistic effects, and singling out the effect of any one of those may be impossible. These surveillance indicators would be supplementary and provide information of the overall "health status" of the ecosystem even in the absence of clear pressure-state relationships.

Despite the current work to supplement the indicator suite already included in the DEVOTool catalog, important gaps still remain; many of the biological components and habitats are not adequately monitored to allow development and testing of potential indicators. Reptiles, such as the European sea turtles, Caretta caretta, Chelonia mydas, and Lepidochelys kempii, are all either endangered or vulnerable species (IUCN Red List), and thus require conservation measures through marine protected areas. There are specific regional or local programmes for sea turtle protection (such as Sea Turtle Protection Society of Greece ARCHELON) that provide information of the status of local populations and advocate conservation measures. Likewise, Cephalopods represent a group of species with only few existing indicators included in the DEVOTool Indicator Catalog. A recent ICES report on Cephalopod biology and fisheries stated: "[Despite the importance of several species for European fisheries, there is limited management of the fisheries and no routine assessment; data collection is often either not part of routine fishery data collection or the data are inadequate for assessment. Increasingly, however, cephalopods are seen as alternative target species to replace overexploited finfish stocks, and the growing fishing effort means that management will almost certainly be needed within the next few years. Also on the horizon is the development of commercial aquaculture]." (Jereb et al., 2015).

One of the major oceanic ecosystems lacking indicators is the deep-sea habitats and their respective communities (Teixeira et al., 2014). The deep pelagic ocean and deep seabed ecosystems represent the largest biomes of the global biosphere, but still the knowledge of their biodiversity, habitats and processes is scarce (Webb et al., 2010; Danovaro et al., 2014). The deepsea physical, biogeochemical and ecological processes present distinctive features from other marine ecosystems (Danovaro et al., 2014; Thurber et al., 2014), and the specific habitats or ecosystems host very specific communities (Danovaro et al., 2010). Also many deep-sea invertebrates are exceptionally longlived and grow extremely slowly (Clark et al., 2016). Current evidence indicates that cumulative stressors from e.g., fishing and resource exploitation will cause important and largely unpredictable ecological changes in these biotopes (Gramling, 2014; Clark et al., 2016). Removal of habitat-forming species, decline in diversity, change in abundance and biomass, reduction in distribution, change in community structure, namely its composition and food web structure are a few of the expected impacts (Clark et al., 2016). Climate change will further expose these already vulnerable ecosystems to combined stresses of warming, ocean acidification, deoxygenation, and effects of altered food inputs (Levin and Le Bris, 2015; Rogers, 2015). Due to the attributes mentioned, the recovery capacity of these deep-sea ecosystems is highly limited and predicted to take much more time to recover after pressures have ceased (Clark et al., 2016). Specific metrics of sensitivity of deepsea fauna and habitats are therefore urgently needed for assessing the risk stemming from impacts and for identifying vulnerable ecological units (Clark et al., 2016). Baselines need to be established for diversity, abundance, and biomass of deep-sea ecosystems, particularly for the less studied pelagic realm and an understanding of ecological processes needs to be developed (Danovaro et al., 2014; Rogers, 2015). Such indicators will allow prioritizing areas for protection and designing more efficient monitoring programmes for the deepsea realm.

There were not any indicators for the sea ice habitats, as these represent quite marginal habitat in the European regional seas. In the Polar Regions, sea ice habitats are important for the productivity of the sea and harbor rich biological communities and food webs associated with those (see Thomas and Dieckmann, 2010, for further references). Due to the climate change and warming of the Polar Regions, the extension of the seasonal, and permanent ice cover is shrinking with an alarming pace (Dieckmann and Hellmer, 2010). In the Baltic Sea, the seasonal ice cover is mostly restricted to the northern parts, and similarly as in the Arctic regions, it is an important habitat with rich community of ice-associated algae and micro-organisms as well as migrating birds, and as a primary breeding ground for the two seal species (Granskog et al., 2006). Climate change is projected to change further the biology and ecology of the Baltic Sea, including the diminishing duration and extent of the sea ice cover with its consequences to the ice associated biota (Viitasalo, 2012). However, due to the lack of monitoring and indicators for the sea ice habitat and biota, it is not possible to assess their impacts on biodiversity status in the ice-covered sub-basins of the Baltic Sea.

# Linking to Other Indicator Based Assessments of Biodiversity

The conservation initiatives worldwide often share common assessment elements and make use of similar baseline information (e.g., Duffy et al., 2013; Pereira et al., 2013). Versatile use of indictors across environmental policies, geographical regions, and spatial scales is apparent as many of methods and biodiversity indicators are applied in several assessment or monitoring programmes (Teixeira et al., 2014). Approximately 30% of the indicators suggested to be applied for MSFD were already used for assessment needs of the other EU Directives or regulations (e.g., Birds and Habitat Directives, Water Framework Directive or Common Fisheries Policy). The need to economize marine monitoring, but yet maintain and enhance operational monitoring networks (Borja and Elliott, 2013) can benefit from careful planning of interoperable monitoring and assessment where the same indicators could be used for several purposes, and combined depending on the needs of each respective assessment purpose.

There are indicators used for MSFD assessments targeting threatened marine species included in the IUCN Red List, as well as several MSFD indicators miming the candidate metrics to Essential Biodiversity Variables proposed by Pereira et al. (2013, e.g., Abundances and distributions, Taxonomic diversity, Habitat structure, Allelic diversity). Also the nine Essential Ocean Variables (EOVs) on Biology and Ecosystem health of marine ecosystems under discussion can benefit from operational indicators included in the DEVOTool Catalog (Teixeira et al., 2014) or those refined and developed by DEVOTES (GOOS, 2016). These relate to the "Status of functional groups" and the "Health of living ecosystems" (Phytoplankton biomass and productivity, Incidence of harmful algal blooms, Zooplankton diversity, Fish distribution and abundance, Apex predator distribution and abundance, Seagrass cover, Macroalgal cover, Live coral cover, Mangrove cover). The wider use of environmental indicators applicable for various marine EU legislation and international agreements such as the RSC or the CBD (Zampoukas et al., 2012; Pereira et al., 2013) promotes harmonization between the different assessment systems and allows effective use of monitoring data for different reporting purposes, provided that the indicators satisfy specific quality criteria (Tittensor et al., 2014; Queiros et al., 2016) that should be common for all programmes.

European status assessments of marine biodiversity have chosen an ambitious path, where data-driven indicators with numeric thresholds should be used to depict definite status classifications. Moreover, the purpose is to link indicators to anthropogenic pressures and further to the human activities. Comparison with the assessment approaches in the marine and coastal areas of the United States (U.S.) shows that there is a conceptual difference, which is mainly due to the different understanding of the indicator concept and approaches for setting the assessment thresholds. A general trend in the U.S. assessments is to give scores for different indicators (or assessment questions) and the scoring is based on descriptive definitions. For instance, the National Oceanographic and Atmospheric Administration (NOAA) Marine Sanctuary Programme<sup>1</sup> makes status assessments based on expert analysis. The experts make the status interpretation based on 17 questions and the descriptions of status classes which are elaborated on the basis of monitoring data. Similarly, the NOAA fish stock assessments<sup>2</sup> and the Sea Turtle Assessment (National Marine Fisheries Service, 2013) are based on four criteria scored by experts and guided by information from the monitoring data. The National Estuarine Eutrophication Assessment (Bricker et al., 2007) uses quantitative thresholds for some eutrophication symptoms (e.g., percentage change in vegetation coverage) but the class boundaries are defined qualitatively using expert knowledge. The NOAA Marine Mammal Assessment (Carretta et al., 2015) is based on monitoring data but the assessment is descriptive, no specific status class is given, and the state of the populations is determined based on the viability analysis of the population. The U.S. National Park Service<sup>3</sup> assesses the country's intertidal zone within protected areas based on expert interpretation of data variability and trends. The indicators in the U.S. Environmental Protection Agency (EPA) National Coastal Condition Report (US EPA, 2012) are closest to the European indicator concept, containing some biodiversity aspects, similar to EU MSFD, and using numerical indicator thresholds to define status classes. Due to the relatively strong European consensus on the indicator concept, the US indicators would not likely be applicable in the MSFD context. Though being used successfully in the US-wide assessments such as the National Coastal Condition Report, the US indicators do not have similar marine assessment framework as in the EU which aims at covering all marine elements, regions and pressures in a coherent and coordinated way.

#### Further Research Needs, and Way Forward

The EU MSFD depicts a cyclical implementation and the next assessment of the environmental status of the marine environment is planned to be completed in 2018. Thus, the EU member states and RSCs are currently on the way in planning this assessment. Based on the experiences from the previous MSFD initial assessment (completed in 2012), the EU Commission is on its way to revise the earlier Commission Decision (2010/477/EU) to advice the on-going initial (MSFD Article 8) assessment that is due to 2018. Different aspects of biodiversity will be in the focus of the assessment, as the ecosystem services provided by the living part of the marine ecosystems are strongly dependent on structural and functional status of biodiversity. Monitoring and managing the health of the seas and oceans is highly relevant for the sustainable use of the marine resources particularly in the light of the recent Blue Growth initiatives is Europe (EU's Blue Growth Strategy<sup>4</sup> ) and worldwide (e.g., FAO's Blue Growth Initiative<sup>5</sup> ) that emphasize the sustainable use of marine resources, but with the expectation that more seafood, energy, and other living and non-living resources can be extracted from the seas. Full ecosystem approach, with concise cover of the marine ecosystem components, is needed for the managers to evaluate that Blue Growth is carried out sustainably, i.e., not threatening the future potential of delivering marine living resources and ecosystem services.

There is a need to have a comprehensive set of indicators, in order to cover all important multifaceted components of marine biodiversity assessments; ideally including both surveillance indicators and those with a clear pressure-state link. Some of the remaining gaps and further specific research needs are presented in **Table 2**. Further development and validation of marine biodiversity indicators requires improved data with better spatial and temporal coverage based on novel monitoring methods. In addition to the tests carried out in the DEVOTES project, more experimental indicator testing is needed to ensure their ecological relevance, robustness, and responsiveness to pressures, and also to enable incorporation into models in order to extrapolate marine assessments for into larger spatial regimes and temporal scales. With the help of combination of different tools, indicators covering both early warming and long-term assessment needs across different spatial scales can be combined into a holistic assessment of marine environment.

#### CONCLUSIONS

Despite the large number of indicators available for the assessment of marine biodiversity there are needs for further development in order to ensure (1) full ecosystem approach (covering all components of marine ecosystem, and all levels of marine biodiversity), (2) improved indicator confidence and responsiveness to pressures, and (3) consistent approach and methodology for setting thresholds for environmental status assessment. DEVOTool provides a comprehensive state-of-theart compilation of biodiversity related indicators developed for

<sup>1</sup>http://sanctuaries.noaa.gov/science/monitoring/welcome.html

<sup>2</sup>http://www.st.nmfs.noaa.gov/stock-assessment/reports

<sup>3</sup>http://science.nature.nps.gov/im/units/nccn/monitor/intertidal.cfm

<sup>4</sup>http://ec.europa.eu/maritimeaffairs/policy/blue\_growth/

<sup>5</sup>http://www.fao.org/zhc/detail-events/en/c/233765/

assessment of coastal and marine ecosystems in Europe and elsewhere. This tool is publicly available for marine managers, experts, and NGOs to rank, evaluate, and choose biodiversity related indicators and to find those that fit best to the needs of the regional and local environmental assessments in Europe and worldwide. There is a relatively concise set of indicators for the second phase of the MSFD implementation, however, some important areas like the deep sea habitats, and trophic levels of marine food webs (e.g., microbes) or taxonomic groups (i.e., reptiles) have fewer indicators operational. Moreover, an assessment of ecosystem processes and functions, i.e., the overall status of ecosystem functioning is an area that requires attention in order to understand interrelations between various ecosystem components and how those impact each other under changing anthropogenic manageable and non-manageable external pressures.

### AUTHOR CONTRIBUTIONS

Conceived and designed the paper: AH, TB, LU. Data was analyzed by HT, AB, AH. All authors participated in the interpretation of the data, reviewing the literature, and drafting the paper. Final approval of the version to be published: AH.

#### FUNDING

This manuscript is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and

# REFERENCES


assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), www.devotes-project.eu. MCU is partially funded through the Spanish programme for Talent and Employability in R+D+I "Torres Quevedo").

### ACKNOWLEDGMENTS

We wish to thank the participants of the DEVOTES work package on biodiversity indicator development Eva Aylagas (AZTI), T.J.S. Balsby (AU), Martynas Bucas (Kucorpi), Elisa Capuzzo (Cefas), ˇ Jacob Carstensen (AU), Anne Chenuil (CNRS), P Clausen (AU), Valentina Doncheva (IO-BAS), Isabel Ferrera (CSIC), Esther Garcés (CSIC), Heidi Hällfors (SYKE), Pirkko Kauppila (SYKE), Harri Kuosa (SYKE), Sirpa Lehtinen (SYKE), Maiju Lehtiniemi (SYKE), Sergej Olenin (Kucorpi), Snejana Moncheva (IO-BAS), Nadia Papadopoulou (HCMR), Albert Reñé (CSIC), Naiara Rodríguez-Ezpeleta (AZTI), Kremena Stefanova (IO-BAS), Sanna Suikkanen (SYKE), Anastasija Zaiko (Kucorpi), Argyro Zenetos (HCMR). Helpful comments of the two referees are gratefully acknowledged.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00184


of phytoplankton community structure. Hydrobiologia 633, 151–168. doi: 10.1007/s10750-009-9879-5


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer IM declared a shared affiliation, though no other collaboration, with the authors MY and AB to the handling Editor, who ensured that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Heiskanen, Berg, Uusitalo, Teixeira, Bruhn, Krause-Jensen, Lynam, Rossberg, Korpinen, Uyarra and Borja. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A Catalogue of Marine Biodiversity Indicators

Heliana Teixeira<sup>1</sup> \* † , Torsten Berg<sup>2</sup> , Laura Uusitalo<sup>3</sup> , Karin Fürhaupter <sup>2</sup> , Anna-Stiina Heiskanen<sup>3</sup> , Krysia Mazik <sup>4</sup> , Christopher P. Lynam<sup>5</sup> , Suzanna Neville<sup>5</sup> , J. German Rodriguez <sup>6</sup> , Nadia Papadopoulou<sup>7</sup> , Snejana Moncheva<sup>8</sup> , Tanya Churilova9, 10 , Olga Kryvenko9, 10, Dorte Krause-Jensen<sup>11</sup>, Anastasija Zaiko12, 13, Helena Veríssimo<sup>14</sup> , Maria Pantazi <sup>15</sup>, Susana Carvalho<sup>16</sup>, Joana Patrício<sup>1</sup> , Maria C. Uyarra<sup>6</sup> and Àngel Borja<sup>6</sup>

#### Edited by:

*Marianna Mea, Ecoreach srl, Italy and Jacobs University of Bremen, Germany*

#### Reviewed by:

*Antoine Jean Grémare, University of Bordeaux 1, France Simone Libralato, National Institute of Oceanography and Experimental Geophysics, Italy*

> \*Correspondence: *Heliana Teixeira heliana.teixeira@ua.pt*

#### Present Address:

*Heliana Teixeira, Departamento de Biologia and CESAM, Universidade de Aveiro, Aveiro, Portugal*

†

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *24 June 2016* Accepted: *04 October 2016* Published: *04 November 2016*

#### Citation:

*Teixeira H, Berg T, Uusitalo L, Fürhaupter K, Heiskanen A-S, Mazik K, Lynam CP, Neville S, Rodriguez JG, Papadopoulou N, Moncheva S, Churilova T, Kryvenko O, Krause-Jensen D, Zaiko A, Veríssimo H, Pantazi M, Carvalho S, Patrício J, Uyarra MC and Borja À (2016) A Catalogue of Marine Biodiversity Indicators. Front. Mar. Sci. 3:207. doi: 10.3389/fmars.2016.00207* *<sup>1</sup> D.2 Water and Marine Resources Unit, European Commission, Joint Research Centre, Directorate for Sustainable Resources, Ispra, Italy, <sup>2</sup> MariLim Aquatic Research GmbH, Schönkirchen, Germany, <sup>3</sup> Marine Research Centre, Finnish Environment Institute, Helsinki, Finland, <sup>4</sup> Institute of Estuarine and Coastal Studies, University of Hull, Hull, UK, <sup>5</sup> Centre for Environment, Fisheries & Aquaculture Science, Lowestoft, UK, <sup>6</sup> AZTI-Tecnalia, Marine Research Division, Pasaia, Spain, <sup>7</sup> Hellenic Centre for Marine Research, Institute of Marine Biological Resources and Inland Waters, Heraklion Crete, Greece, <sup>8</sup> Department of Marine Biology and Ecology, Institute of Oceanology-Bulgarian Academy of Sciences, Varna, Bulgaria, <sup>9</sup> Department of Biophysical Ecology, Kovalevsky Institute of Marine Biological Research of RAS, Sevastopol, <sup>10</sup> Department of Oceanographic Processes Dynamics, Marine Hydrophysical Institute of RAS, Sevastopol, <sup>11</sup> Department of Bioscience, Aarhus University, Silkeborg, Denmark, <sup>12</sup> Marine Science and Technology Centre, Klaipeda University, Klaip ˙ eda, Lithuania, ˙ <sup>13</sup> Coastal and Freshwater Group, Cawthron Institute, Nelson, New Zealand, <sup>14</sup> Faculty of Sciences and Technology, MARE Marine and Environmental Sciences Centre, University of Coimbra, Coimbra, Portugal, <sup>15</sup> Hellenic Centre for Marine Research, Institute of Marine Biological Resources and Inland Waters, Athens, Greece, <sup>16</sup> Red Sea Research Center, King Abdullah University of Science and Technology, Thuwal, Saudi Arabia*

A Catalogue of Marine Biodiversity Indicators was developed with the aim of providing the basis for assessing the environmental status of the marine ecosystems. Useful for the implementation of the Marine Strategy Framework Directive (MSFD), this catalogue allows the navigation of a database of indicators mostly related to biological diversity, non-indigenous species, food webs, and seafloor integrity. Over 600 indicators were compiled, which were developed and used in the framework of different initiatives (e.g., EU policies, research projects) and in national and international contexts (e.g., Regional Seas Conventions, and assessments in non-European seas). The catalogue reflects the current scientific capability to address environmental assessment needs by providing a broad coverage of the most relevant indicators for marine biodiversity and ecosystem integrity. The available indicators are reviewed according to their typology, data requirements, development status, geographical coverage, relevance to habitats or biodiversity components, and related human pressures. Through this comprehensive overview, we discuss the potential of the current set of indicators in a wide range of contexts, from large-scale to local environmental programs, and we also address shortcomings in light of current needs. Developed by the DEVOTES Project, the catalogue is freely available through the DEVOTool software application, which provides browsing and query options for the associated metadata. The tool allows extraction of ranked indicator lists best fulfilling selected criteria, enabling users to search for suitable indicators to address a particular biodiversity component, ecosystem feature, habitat, or pressure in a marine area of interest. This tool is useful for EU Member States, Regional Sea Conventions, the European Commission, non-governmental organizations, managers, scientists, and any person interested in marine environmental assessment. It allows users to build, complement or adjust monitoring programs and has the potential to improve comparability and foster transfer of knowledge across marine regions.

Keywords: assessment, non-indigenous species, food webs, seafloor integrity, pressures, Marine Strategy Framework Directive

### INTRODUCTION

Taking the pulse of natural ecosystems and tracking progress toward environmental goals requires suitable indicators (e.g., Pereira et al., 2013; Tittensor et al., 2014; Geijzendorffer et al., 2015). Worldwide, there are several marine biodiversity conservation initiatives in place demanding robust and scientifically-based environmental assessments. Among the most comprehensive and with relatively wide geographical scope are the EU Marine Strategy, the US National Ocean Policy, the United Nations (UN) Convention on Biological Diversity (CBD), or the Convention on the Law of the Sea. Also at regional and local scales, environmental objectives have long been set to cope with the impacts of human activities in marine waters (e.g., Regional Sea Conventions; HELCOM, 2009; Long, 2012) and to protect natural capital (e.g., Marine Protected Areas policies; Costanza et al., 1997; Agardy et al., 2011; Liquete et al., 2013). These initiatives are increasingly important now, as the seas are facing a "marine Wild West" rush (Cressey, 2016) steered by blue growth prospects worldwide, which will inevitably increase and diversify anthropogenic pressures in our oceans (Børresen, 2013; Gramling, 2014). Nations must therefore act quickly to prevent the accelerated depletion of natural resources and wildlife (McCauley et al., 2015; Cressey, 2016), especially since there is still a lack of understanding on many aspects of our marine ecosystems (Danovaro et al., 2014; EEA, 2014).

The success of management is partially dependent on the availability of scientific tools to managers (Rist et al., 2013; Knights et al., 2014). Robust indicator selection, transparent use of information, and effective communication of results constitute crucial parts of this process, but the development, calibration and validation stages of new indicators and assessment approaches can compromise timely managerial response (Borja and Dauer, 2008). However, there is still a common practice of developing new indicators for each new assessment initiative put forward as well as for any specific case or policy requirement. Indeed, during the last couple of decades we have witnessed a boom of ecological indicators worldwide (see reviews by e.g., Marques et al., 2009; Cardoso et al., 2010; Birk et al., 2012; Borja et al., 2016a), driven either by environmental policies or research, attempting to cover, for example, the most sensitive habitats and endangered organisms (e.g. Gobert et al., 2009; Waycott et al., 2009; Deter et al., 2012; Gatti et al., 2015), or to detect imminent threats to ecosystems (Halpern et al., 2012; Katsanevakis et al., 2016) and their services (Liquete et al., 2013).

Although there is still a lack of practical indicators regarding many aspects of the marine ecosystem (Berg et al., 2015; Hummel et al., 2015; Piroddi et al., 2015), and dedicated research is still needed (Rudd, 2014), it is also recognized that the cost and delays associated with gathering information, learning and development process are often responsible for failures encountered in the implementation stage of management plans (Lee, 1999 in Rist et al., 2013; Pitcher et al., 2009). All this has inspired recent attempts to take advantage of the existing knowledge and past efforts to develop robust assessment tools and optimize their use in fulfilling stakeholders' environmental commitments (e.g., Cardoso et al., 2010; Fautin et al., 2010; Zampoukas et al., 2012; Liquete et al., 2013; Pereira et al., 2013; Tittensor et al., 2014; Berg et al., 2015).

Efficient adoption of the existing knowledge not only accelerates the developmental process per se (e.g., Fautin et al., 2010; Teixeira et al., 2012; Borja et al., 2015) but also implies that data associated with indicator development and subsequent monitoring should be available to some extent. This can be valuable when baselines and spatio-temporal trends need to be established locally, regionally or even globally (Muxika et al., 2007; Duffy et al., 2013; Probst and Stelzenmüller, 2015; Borja et al., 2016b). A major difficulty in producing a coherent picture of the current status and trends of marine diversity is the lack of standardized and coordinated approaches for monitoring it (Duffy et al., 2013). Recently, many conservation initiatives have recognized this and started building their marine strategies (e.g., Zampoukas et al., 2012) or recommendations (e.g., Duffy et al., 2013; Pereira et al., 2013) upon relevant existing activities to promote comparability within and across regions.

An important obstacle to the adoption and effective use of existing tools is the tedious and time consuming task of searching candidate indicators scattered throughout the vast scientific, technical and often also gray literature, along with the need to compare among methods every time a new management plan needs to be set. The idea of collating scattered indicators in order to establish the integrity and biodiversity trends of marine ecosystems is therefore not only very appealing but also much needed and wise (Duffy et al., 2013). By reducing the time spent searching for indicators and by optimizing the comparison between different approaches, time can be devoted to other crucial aspects. For example, the uncertainty associated with assessments, or the effective communication of results, which are too often neglected when applying indicators in assessment and regulatory contexts (Rees et al., 2008; Queirós et al., 2016).

With the European Marine Strategy Framework Directive (MSFD, 2008/56/EC) as scenario, we evaluated the current potential of existing ecological indicators to support the assessment of marine biodiversity and address environmental targets (Berg et al., 2015). The MSFD is a good test of the capability of current indicators as it adopts an ecosystem-based approach that considers 11 broad range descriptors to describe the environmental status of marine waters. These descriptors encompass both state and pressure features, from biological diversity and food webs to contaminants and marine litter.

The primary aim of our study was to identify indicators capable of supporting the assessment of four descriptors (D) sensu MSFD: biological diversity (D1), non-indigenous species (D2), food webs (D4), and seafloor integrity (D6). We present a catalogue containing numerous indicators and respective metadata available as a database through the DEVOTool free software application. This tool enables users to browse the catalogue and extract lists of fit-for-purpose indicators using various selection criteria and ranking options.

The concepts of indicator and index are often used as synonyms but it is important to clarify that in a regulatory context indicator may be a proxy for something different from what it actually measures (Rees et al., 2008). An indicator is intended to highlight the status of the system and, for e.g., the European Environmental Agency recognizes distinct types of indicators depending on what they address: descriptive indicators, performance indicators, efficiency indicators, and total welfare indicators (Smeets and Weterings, 1999). The term should therefore be distinguished from index, an aggregation of indicators into a single representation (Rees et al., 2008). Indices are considered as one possible measure of the systems status, as they relate to a specific qualitative or quantitative feature of the system (Pinto et al., 2009). The selection of key indicators, effective at capturing the system condition and announcing changes compared to the specified objectives, leads then to the elaboration of an assemblage of relevant indices used as operational tools. However, in this manuscript, we use the term "indicator" to refer to what is commonly called an index or assessment system, i.e., a qualitative or numerical expression or a statistic, reflecting an ecosystem feature or magnitude of anthropogenic pressure (Claussen et al., 2011; Berg et al., 2015). This is to ensure some coherence with the MSFD, where the term is used for the metrics needed for baseline assessments and to monitor whether environmental targets have been met.

The analysis of the marine biodiversity indicator catalogue aimed: (i) to identify the strengths and gaps in the existing sets of indicators in order to direct the further development of indicators toward the most urgent needs; and (ii) to foster transfer of knowledge across countries and marine regions, so that indicators operational in one area could be easily adjusted and adopted elsewhere for the environmental assessments.

This review highlights the main attributes of the indicators contained in the catalogue, namely the biodiversity components they address and habitats they apply to, their geographical coverage and potential for addressing relevant pressures. We also describe the type of data behind the indicators and their status of development. Moreover, we discuss the potential of existing marine biodiversity indicators in the context of global biodiversity observation networks that could form the basis for worldwide monitoring programs (Pereira et al., 2013; GOOS, 2016). Finally, we provide recommendations on research priorities for improving quality of the assessment tools.

# MATERIALS AND METHODS

# Compilation of Indicators: Survey Design and Scope

Our survey (conducted in mid-2013) targeted marine indicators with potential to address biological diversity, trends and impacts of non-indigenous species, food webs' properties, and seafloor integrity. A questionnaire for retrieving indicators and associated metadata was circulated among 20 scientists from 14 institutions, identified in the database by a "contributor code". All contributors were either involved in the implementation of the MSFD or having broad knowledge on indicators' development or application in their respective regions or fields of expertise.

The information on indicators was compiled from very different sources in national and international environmental contexts: EU Directives, Regional Seas Conventions (RSC), assessments from non-EU seas, and other regional research programs. Since the primary goal of building this Catalogue of Indicators (**Supplementary Material S1**) was to assist the implementation of the MSFD in Europe, the first sources of information were programs associated with existing EU legislation. In particular we screened indicator proposals under the Water Framework Directive (WFD, 2000/60/EC), the Nature Directives (Habitats Directive–HD, 92/43/EEC and Birds Directive–BD, 2009/147/EC), and other relevant EU legislation (including the Common Fisheries Policy– CFP: Council Regulation (EC) No 199/2008; Commission Decision 2010/93/EU). These EU programs link directly to the biodiversity related descriptors in the MSFD. For the WFD indicators, the survey was primarily based on the WISER project methods database (Birk et al., 2012) (available at http://www.wiser.eu/results/method-database/). Any updates to indicators, since the WISER database, have been included in the current catalogue; for example, after subsequent WFD Intercalibration results (Carletti and Heiskanen, 2009; Commission Decision 2008/915/EC; Commission Decision 2013/480/EU), or after further revisions of the methods by their authors. Approaches developed in the framework of RSC covering European seas were also taken into account, namely those by HELCOM–Baltic Marine Environment Protection Commission–Helsinki Commission, the OSPAR Convention for the Protection of the Marine Environment of the North-East Atlantic, the Barcelona Convention - UNEP-MAP Mediterranean Action Plan, and to a lesser extent the Bucharest Convention–Black Sea Commission. Effort was also made to include the indicators used by Member States in their MSFD reporting on Initial Assessments, Good Environmental Status, Environmental targets and associated indicators, if available in the EU Eionet Central Data Repository (http://cdr.eionet.europa.eu/recent\_etc?RA\_ID=608) or if provided by national researchers. Indicators developed and used in other contexts, i.e., from research or monitoring programs within Europe, but also further afield (e.g., the Red Sea area or in the USA), as well as information published in broader scientific literature, were also included in the catalogue. The first version of the catalogue has been released in January 2014 (version 6) and a recent update was completed in March 2016 (version 7). As a result of this consultation, over 700 literature references have been compiled and made available through the "Sources" field of the database (**Supplementary Material S1**). New developments since our survey may not be represented in the catalogue. For example, work and advances that ICES working groups have been following on food webs (D4) and seafloor integrity (D6) fields of research are a relevant and complementary source of information (ICES, 2014, 2015b).

#### Catalogue Structure

The catalogue contains three main sections: "Indicators," "Metadata," and "Sources," composed of both open and closed fields for reporting information. A fourth section allows performing "Analyses" such as querying the database. The "Indicators" section has ten fields describing intrinsic properties of the indicator and other related information: indicator name, RSC affiliation, indicator description, data requirements, collection method, associated costs, overall indicator status, unit, confidence or uncertainty of the indicator, observations or remarks.

The "Metadata" section has two main types of fields. There are fields linked to MSFD requirements for reporting descriptor coverage, i.e., to assign the relation to the 11 MSFD descriptors, and to relate indicators with the Commission Decision criteria and indicators (COM Dec 2010/477/EU) specifically for descriptors D1, D2, D4, and D6. Other fields allow specification of targets of the indicator in terms of biodiversity components (e.g., phytoplankton, macroalgae, fish), and a set of predominant ecotypes (e.g., pelagic fish, demersal fish) for mobile components, as well as the option to insert any taxonomical specificity of the indicator (i.e., taxon name). There are also fields for reporting the link to pressures such as physical damage, contamination pressures including organic enrichment, marine litter, introduction of pathogens or non-indigenous species, extraction of living resources, underwater noise, and marine acidification. Finally, other fields specify settings for applying the indicator, ultimately including information on targets and/or reference conditions for the indicator, associating them to particular habitat(s) where it applies (i.e., seabed, water column, and ice-associated), and its geographical coverage such as e.g., the EU MSFD marine region(s) or non-EU seas where it has been used (with further specification of the scale of application within marine sub-regions, subdivisions or ecological areas, and subareas). Also within settings, it is possible to associate indicators to country level or establish correspondence to existing monitoring programs or initiatives, such as International Conventions, RSC, EU Directives, National monitoring or Research program. If there is data availability for an indicator, such details (e.g., time series and GIS data) and a link to source can also be provided. Finally, within all sections, source fields link certain attributes to specific literature, and all references are then made available in the section "Sources." Further details on the fields and definitions of categories can be found in the database (**Supplementary Material S1**) and in Teixeira et al. (2014).

#### Data Analysis

The analysis presented in this review is based on the Catalogue of Indicators database version 7 (**Supplementary Material S1**) available through the DEVOTool 0.64 software application (free download at: http://www.devotes-project.eu/devotool/), allowing navigation of the database. The catalogue was explored using the query functions available in the "Analyses" section of this software (see DEVOTool-manual-0.64).

To provide a better overview and summarize the content of the catalogue, the indicators reported in the survey were classified a posteriori according to four criteria: (i) allocation to a "DPSIR stage" within the DPSIR framework (Drivers-Pressures-State Change-Impacts-Responses, Elliott, 2002; Smith et al., 2016) and following Berg et al. (2015); (ii) in relation to their "Main attribute" or theme addressed; (iii) according to the type of data required to calculate the index i.e., "Underlying variable type"; and (iv) according to its classification or "Algorithm type". These fields were also included in database vs.7 and definitions of categories are provided as **Supplementary Material S2**.

Flow diagrams were built with RAW 1.0.2 developed by DensityDesign http://raw.densitydesign.org.

# RESULTS

# General Overview of Indicators' Characteristics and Scope

The catalogue currently contains 611 indicators, of which about half are operational, i.e., tested and validated, with associated target values or classification boundaries, easily interpretable within a good through bad environmental status continuum to be useful under regulatory contexts (**Figure 1**). A significant proportion of the indicators (36%) are still under development, i.e., the indicator proposal exists but, for example, has not yet been validated with real data or is in the process of calibration for use in new locations or habitats. A small percentage of conceptual indicators, i.e., an indicator idea supported by theoretical grounds, although no practical measure or metric is yet available, were also reported (7%).

Most of the entries in the catalogue are state indicators that report on distinct aspects of the ecosystem. Habitat integrity is the most widely used feature to assess the health of the marine ecosystem (26.4%, **Figure 2**). Indicators in this category focus on the biotope relevant features, considering the physical habitat and associated biological communities. That is, they use abiotic or biotic data, such as hydrological and physical-chemical indicators, abundance or biomass of habitat-forming taxa, and very often their spatial distribution. In many cases, they integrate different information, using more than one variable type, sometimes up to six different categories of data (**Figure 2**). For example, the "COralligenous Assessment by ReefScape Estimate index" uses abiotic data, abundance, physio-/morphological data, spatial distribution, taxonomic, and traits composition (Gatti et al., 2015).

The status of the marine environment is also assessed using the biota, from the sub-individual level, to the species and community levels, and to the ecosystem level. That is achieved by

focusing on aspects of community structure, population ecology, production and trophic relations, using indicator species or target groups, accounting for species life traits, or measuring the physiological condition of individuals (**Figure 2**).

Although the number of indicators differs conspicuously between biodiversity components (**Figure 1**), there is a good coverage of the major taxonomic groups required by the MSFD. With the exception of microbes, all are covered by operational indicators. For microbes, only one indicator, still under development, has been reported ("Abundance of bacterioplankton"). Benthic invertebrates and fish have by far the greatest number of related indicators (>100, **Figure 1**). Pelagic species are the least assessed by the available indicators, with only eight indicators for fish and two for elasmobranchs. Also pelagic macroinvertebrates are much less covered compared to benthic ones. Angiosperms and macroalgae, birds, marine mammals, and phyto- and zooplankton are addressed by a considerable number of indicators, while reptiles and cephalopods have a comparably lower number. Finally, independently of their taxonomic group, fauna from extreme habitats, such as deep sea or ice-associated habitats, are, in general, very poorly covered by the current set of indicators. From the reported indicators, only the "Marine Biological Valuation Methodology" and the "Biopollution Level index" accommodated those organisms (together with all other faunal groups) in a broad environmental assessment of the area. These indicators target larger trends, without the intention of focusing on group specific properties.

Over 400 indicators have been developed specifically for a biodiversity component or subcomponent, and the catalogue makes reference to more than 80 different species for which indicators exist. Six of them are threatened marine species (Clangula hyemalis, Melanitta fusca, Monachus monachus, Polysticta stelleri, Squalus acanthias, Thunnus thynnus) included in the Red List of the International Union for Conservation of Nature and Natural Resources (IUCN). Other indicators may not be specific to a single component (n = 86) but rather, target several groups either simultaneously or interchangeably, resulting overall in more indicators targeting biodiversity components than indicator entries in the catalogue as shown in **Figure 1** (sum of indicators per biodiversity component = 819; >611). Indicators may also address biodiversity components in a broader and more encompassing way by focusing on, for example, the processes between certain levels of the ecosystem (like "Energy flows and transfer efficiencies among trophic levels or functional groups") or groups defined independently of biodiversity components (e.g., "Number of biocenosis/facies"). In these cases no specific link to a biodiversity component has been reported. Ninety seven indicators do not target biodiversity components directly, focusing instead on biotope features beyond the biological characteristics, or addressing

anthropogenic activities, e.g., "Areal extent of protected areas," "Depth of sediment redox potential discontinuity," or "Number of dredging permits and the amount dredged related to them".

Only 9% of the entries in the catalogue are pressure indicators (**Figure 2**), essentially focusing on anthropogenic activities (e.g., "Ballast water treatment indicator," "Seafloor exploitation index," or species removal and by-catch indicators), specific target groups (mainly related with trends in non-indigenous species introduction) or biotope features (e.g., "Light pollution for sea birds").

Habitats have been linked to 446 indicators, about half of which are operational (54.5%). Seabed habitats are represented by a higher number of indicators than water column (298 vs. 178), and no indicator was reported for ice-associated habitats. A great part of the seabed indicators report on issues related to spatial distribution of benthic habitats (e.g., "Areal extent of rocky habitats," "Distributional range of circalittoral and bathyal soft bottom habitats"), or target habitat structuring and forming species (e.g., "Posidonia oceanica Rapid Easy Index," "Population structure of long-lived macrozoobenthic species"), or address benthic communities structural status (e.g., "M-AMBI"). There are also several indicators addressing anthropogenic activities' pressures to the seabed (n = 17). If we distinguish seabed indicators according to the bottom type (hard bottom—rock and biogenic reef; soft bottom—sand, mud and sediment; mixed bottom—mixed and coarse sediment), the overall number of indicators relevant to hard bottom is lower than for soft bottom, but the number of indicators specifically addressing hard bottom is, however, higher. Regarding depth zone [littoral, shallow sublittoral, shelf, bathyal (upper and lower), and abyssal], the number of indicators decreases noticeably from shallow to the deep sea, and there are no indicators exclusively addressing abyssal or bathyal zones. Only four indicators are specific for the shelf zone. Water column habitat is represented by indicators mainly targeting pelagic groups, population ecology and the structure of their communities, production, and biotope features (e.g., "Abundance or biomass of key species in the coastal waters," "Secchi depth," "Trends in populations of large pelagic fish," or "Chl a concentration").

Most of the indicators in the catalogue have simple algorithms and methods of calculation when integrating the data (77.7%), using categorical approaches, simple arithmetic or statistics. Only 3.9% of them demand higher expertise or IT capabilities for calculations.

More than half (62.7%) of the indicators reported as operational fail to report specific targets, boundaries or reference levels associated with their use or even mention the possibility of setting them. The 115 indicators that report such information often associate targets or boundaries to specific regions, habitats, species or even methodological aspects. In a few cases they refer to the existence of targets alongside sources but without presenting them.

A majority of the indicators also lack any measure of confidence or uncertainty associated with their assessment results. When reported (6.7%), uncertainty assessments were essentially taking into account sampling effort variation, sampling error measurement or spatial and temporal variation; only a couple of examples performed sensitivity tests to the index parameters through evaluation of the stochastic variation of those variables; and, in one case, a set of requirements for index application was established to ensure some minimum robustness but without providing any measure of confidence of the final estimates.

#### Capability to Address Pressures

The current indicator set gathers a great diversity of approaches capable of addressing the main pressures listed by the MSFD (**Figure 3**). Most of the indicators address nutrient and organic matter enrichments, which reflect eutrophication that is still the most widespread pressure in marine and coastal waters in Europe (EEA, 2013, 2015). There are also a number of policies targeting eutrophication, and a large number of indicators have been developed to display whether these policies have resulted in improvement of the eutrophication status (Ferreira et al., 2011). Likewise, many of the indicators were sensitive toward organic loading, which reflects the high number the benthic invertebrate indicators that generally reflect the status of benthic habitats with respect to organic loading. A second pressure group that was targeted by a high number of indicators was related to physical loss, interference with hydrographical processes, and physical damage to marine habitats. These reflect the abrasion pressures caused by demersal fishing and aggregate dredging, but also silting, smothering, and increase of turbidity due to coastal and underwater constructions (e.g., Knights et al., 2013; Oesterwind et al., 2016; Smith et al., 2016). A third group of indicators are able to reflect the effects caused by contamination and fishing (i.e., removal) pressures. Pressures that have been identified recently such as marine noise, litter or acidification, and pressures such as extraction of seaweeds and maerl are addressed by the lowest numbers of indicators.

#### Geographical Coverage

Most of the entries in the catalogue are linked to at least one marine region. There are some exceptions regarding conceptual and under development indicators that have not yet been tested with regional data sets, or indicators whose conceptual basis makes them potentially applicable to any region (e.g., "BTA– Biological Traits Analysis" or "Strength of bottom-up cascade in marine size spectrum").

The catalogue contains indicators developed and in use under diverse contexts within Europe but also beyond Europe's geographical area (e.g., in the Red Sea area or in the USA), corresponding to marine areas of 34 different countries (**Figure 4**). Despite this wide coverage, a good description of methods' availability can only be guaranteed for the European regional seas (Baltic Sea, Black Sea, Mediterranean Sea, and North-East Atlantic Ocean) and their respective marine regions. The number of indicators differs markedly between regional seas (**Figure 4**), partly reflecting the size and overall biodiversity pattern of the specific regional seas but also the focus of environmental concern and research tradition. For example, Mediterranean ranked highest of the European Seas for the state-of-knowledge index across taxa (Costello et al., 2010) suggesting that the effort for taxonomic description of species has been in historical focus rather than development of environmental indicators for biodiversity assessment purposes. On the other hand, despite the low biodiversity of the Baltic Sea, in comparison to fully marine areas with higher salinity, and despite the gaps in taxonomical description of certain organism groups (Ojaveer et al., 2010), a considerably high number of indicators have been reported to this region. This reflects that environmental status concerns and the governments' corresponding long-term investment policy in biodiversity research have been considered for a long time.

#### DISCUSSION

One of the aims of this catalogue is to promote the coherent use of data and the adoption of compatible metrics and indicators, in line with several policy requirements (Zampoukas et al., 2012; Pereira et al., 2013; GOOS, 2016). In fact, despite that the focus of conservation initiatives worldwide might differ, they often share common assessment elements and make use of similar baseline information (e.g., Duffy et al., 2013; Pereira et al., 2013). The majority of indicators in the catalogue are already associated with at least one specific assessment system or monitoring program, and in many cases, they are linked to more than one. Approximately 30% of the indicators reported are already used by other EU Directives or regulations (Nature Directives, Water Framework Directive or Common Fisheries Policy). Many are used by national monitoring programs or within international agreements and Regional Sea Conventions. This shows the great potential for their application across policies, spatial scales, and geographic regions.

Essential metrics for monitoring global trends in biodiversity and the integrity of the oceans worldwide (Pereira et al., 2013; GOOS, 2016) such as, for example, "Taxonomic diversity," "Habitat structure," "Allelic diversity," "Phytoplankton biomass and productivity," "Incidence of harmful algal blooms," "Zooplankton diversity," "Fish distribution and abundance," "Apex predator distribution and abundance," "Seagrass cover," "Macroalgal cover," and "Live coral cover," are largely covered by operational indicators in the catalogue. This reinforces the opportunity for incorporation of existing knowledge and tools in new marine conservation programs.

The set of indicators compiled can essentially be used within two stages of the DPSIR cycle: To measure pressures (P) on the natural system and to assess changes in its state (S), i.e., the properties and processes of the ecosystem (Berg et al., 2015). However, the catalogue contains by far more indicators that

primarily measure the response of ecosystems to pressures than pressures themselves. This is explained by the fact that the MSFD descriptors targeted here were essentially biodiversity-related ones, which encompass very few pressure requirements.

The adequacy of the current set of indicators to address the requirements of the MSFD has been exhaustively analyzed by Berg et al. (2015). Here we focus on the capability and current knowledge on marine biodiversity indicators to support ecosystem-based approaches (Borja et al., 2016a) and on how indicator gaps may compromise such endeavors (Hummel et al., 2015). We discuss this at several levels: (i) in relation to biodiversity components and habitats; (ii) with regard to relevant pressures on the marine environment; (iii) considering the survey and coverage of the catalogue; and (v) in relation to the status of development of the indicators and most common weaknesses of these methods.

#### Biodiversity Components and Habitats

The availability of indicators per biodiversity component may reflect the species richness of the group, their economic importance, the conservation status of the component or the level of taxonomic knowledge and expertise available. The fewer species exist in a group, or the more restricted their distribution is (e.g., cephalopods or reptiles), the smaller the number of indicators reported. A higher number of indicators, besides driven by high species richness, wide distribution, and environmental hazards related to those (e.g., phytoplankton nuisance blooms and HABs), may also reflect their important function in the food web and in the nutrient cycling (e.g., benthic invertebrates), as well as the economic importance of the biodiversity component (e.g., commercial fish). The relatively high number of marine mammal and bird indicators may instead reflect the high conservation status of these components and their importance as flagship species (Smith et al., 2014), prompting efforts toward their monitoring and protection internationally. In contrast, angiosperms and macroalgae species are seldom protected as species per se but they are often protected as structuring components of biotopes/habitats, which might also explain the great availability of indicators. Like the benthic invertebrates, also zooplankton is high in species richness, as well as in abundance, and forms an important link in many food webs and nutrient cycles. Nevertheless, the number of indicators reported for zooplankton is relatively low compared to phytoplankton. One of the reasons could be the absence of

this component in the EU Water Framework Directive, not stimulating further development of methods to assess its status.

Marine habitats are also covered differently by the indicators available. The higher proportion of operational indicators for seabed in comparison to water column habitats might be partially explained by the longer tradition in monitoring and status assessments of benthic communities (Díaz et al., 2004) and also because they are easier to conduct and interpret compared to the strongly spatially variable and stochastic water column communities. The lack of ice-associated habitat indicators may result from an unclear or misleading definition and classification for those habitats and their related communities, but also from their restricted temporal-spatial extend within EU seas. Two indicators relevant to ice habitats exist in the Baltic but refer to gray seal pupping: "Number of pups of gray seals" and "Abundance of seals (at haul-out sites and within breeding colonies)" (HELCOM, 2013). Thus, they are not directly assigned to the habitat type, as both are targeting a specific species. The reduced number of indicators applicable to the deep-sea habitats is mainly related to the degree of access, until recently limited (Danovaro et al., 2014). As shallow depth zones are easy to reach, they consequently have a longer tradition of surveillance and more comprehensive datasets are available, allowing indicator development. The lack of indicators specifically addressing the bathyal and abyssal zone, and in particular the pelagic domain, can be regarded as an important gap in the current suite of indicators. These zones host characteristic communities and species, entangled within unique ecological processes that require specific sampling and assessment approaches (Costello et al., 2010; Danovaro et al., 2014; Thurber et al., 2014; Rogers, 2015) and, therefore, specific indicators for assessing their status. However, an understanding of deep-sea processes needs to be further developed along with baselines for several parameters, before sensitivity metrics can be incorporated into indicator approaches. This is indispensable to allow following the multiple pressures and impacts from increasing offshore activities and climate change (Gramling, 2014; Levin and Le Bris, 2015).

Considering the main topics addressed by the indicators, our review highlights the need for further development and validation of indicators that inform: On the ecosystem level (addressing structure, processes, and functions), on the genetic diversity, on the effects of non-indigenous invasive species and quantification of their impacts, along with indicators informing on food webs structure and functioning, particularly encompassing lower trophic levels, which are currently poorly addressed. These findings concur with others (e.g., Geijzendorffer et al., 2015; Hummel et al., 2015) who demonstrate that decision-makers are currently constrained by the lack of data and indicators on changes in genetic composition, species populations, and ecosystem function and structure. At the ecosystem level the current set of indicators can be effectively complemented by modeling approaches and their modelderived indicators, in particular for topics such as food webs, connectivity, and the effects of non-indigenous species on the ecosystems (see Piroddi et al., 2015; Tedesco et al., 2016). Functional aspects lag behind in operational indicators, but the recent insights in biodiversity and ecosystem function (BEF) relationships may soon contribute to the use of BEF relationships within ecosystem functioning monitoring (Mouillot et al., 2013; see review by Strong et al., 2015). Recent developments on emerging molecular-based indicators are expected to evolve rapidly with advancing novel analytical technologies, and might fulfill the current lack of genetic indicators availability (Bourlat et al., 2013).

## Capability to Address Pressures

The marine environment is exposed to a variety of different anthropogenic pressures. Some of them are the focus of specific MSFD descriptors (i.e., D2 non-indigenous species, D5 eutrophication, D7 hydrological conditions, D8 contaminants, D9 contaminants in seafood, D10 marine litter, and D11 energy like underwater noise or light). The catalogue here presented contains very few pressure indicators because the main targets were essentially indicators of biological diversity, food webs and seafloor integrity, which are state descriptors sensu MSFD (D1, D4, and D6, Claussen et al., 2011). The few pressure indicators reported relate to the pressure caused by non-indigenous species, or result from the existence of mixed pressure/state requirements within a few MSFD criteria (Berg et al., 2015).

Measuring some types of pressures is fairly self-explanatory but we cannot directly measure something like abrasion, for example. Particle size or topography can contribute to assess abrasion, but those parameters vary with other pressures too and, in such cases, the activity is measured instead (Knights et al., 2013; Smith et al., 2016). An ecosystem-based approach is, therefore, needed, where an improvement of the environmental status requires a combination of measures to control the whole suite of pressures introduced by the full range of human activities that impact the marine ecosystem (Knights et al., 2013). The availability of state indicators capable of capturing signal from an identified pressure(s) can provide direct statistical evidence for the relationship between the activity (e.g., trawling effort, which can be managed) that induces the pressure (e.g., fish removal) and an indicator response (e.g., "Large Fish Indicator," Engelhard et al., 2015).

Therefore, despite focusing on the integrity of the ecosystem, state indicators might still have a more or less evident relationship to anthropogenic pressures (Nõges et al., 2016), even if a direct relationship to one or several pressures is sometimes difficult to prove. This is due to the diversity of pressures, and their cumulative and synergistic effects, that may affect specific ecological characteristics of the ecosystem, and also to the complexity and variability of relationships and feedbacks within the ecosystem itself (Knights et al., 2013; Oesterwind et al., 2016; Smith et al., 2016). For example, the cross-linkages and dependencies between trophic levels and competitors for food and space are too numerous and variable to clearly track the path of chain events (Knights et al., 2013). Nevertheless, several state indicators in the catalogue are sensitive to one or more pressures and can provide powerful insight within specific ecosystem-based management frameworks.

As also expected, most of the state indicators were sensitive to pressures that are predominant across coastal and marine regions, such as "nutrient and organic matter enrichment." Primary and secondary eutrophication impacts cascade through the whole ecosystem and have consequences on biodiversity, on species and habitats, as well as at the food web and ecosystem level, which is reflected by the number of sensitive indicators. Likewise, the several EU policies such as the Nitrates and the Urban Waste Water Treatment Directives (91/676/EEC; 91/271/EEC) and the WFD have specifically imposed obligations to assess the impacts of the implementation of these regulations, and to demonstrate if there are improvements. For this reason, a number of indicators have been developed to reflect the impacts on various compartments of the ecosystem and many of the existing indicators have been also suggested as suitable for the assessment of the (D5) eutrophication status (Ferreira et al., 2011; Berg et al., 2015). Considering the most predominant sectors of activity in most marine ecosystems (Knights et al., 2013), the five pressures most likely to affect biological diversity and food webs were "interference with hydrologic processes," "introduction of non-synthetic compounds," "changes in siltation," "introduction of synthetic compounds," and "marine litter." Seafloor integrity is mostly menaced by pressures causing "physical loss" and "habitat damage" (e.g., causing fragmentation and changes in connectivity), but also nutrients and other contaminants input will strongly impact benthic communities (e.g., with homogenizing effect) (ICES, 2015b). With an exception for marine litter, the catalogue includes many indicators particularly sensitive and responsive to these pressures.

Recently, Joppa et al. (2016) have found that no global datasets are available for addressing IUCN listed pressures most affecting threatened marine species: "transportation and service corridors" and "human intrusions and disturbance". The catalogue contains indicators that tackle both issues (e.g., "Ballast water treatment indicator," "Trends in pathways of introduction NIS"). This shows that, at least regionally, some data and indicators are available, and that these threats are under the eye of monitoring programs. The remaining IUCN listed threats ("residential and commercial development," "biological and resource use," "invasive and other problematic species") are also covered, to some extent, by indicators in the catalogue that focus, for example, on fisheries and extraction of living resources, impacts on the seabed, trends in NIS and toxic species. "Pollution" issues are poorly covered by specific pressure indicators of the catalogue, since it relates more to contaminants (MSFD D8 and D9), marine litter (MSFD D10), and energy /noise (MSFD D11), which were not the target of our survey.

Conceptual models of the pathways of state change (Smith et al., 2016) suggest that the components of the DPSIR are not mutually exclusive and that biological change can be direct or can follow a series of physical state changes. So, the "P" and the "S" part are a continuum and, therefore, it can be challenging to fit indicators into a single stage within the DPSIR cycle. In a few cases, the information reported in the catalogue was not always enough to clarify whether an indicator was a pressure or state one. The type of data feeding the indicator can also determine its potential role within the framework. For example, "bycatch" indicators provide evidence of damage to non-commercial species (e.g., "by-catch of marine mammals and waterbirds in fishing gears"), and changes in this indicator

indicate low or increasing pressure. Notwithstanding, changes in the total numbers of organisms affected by by-catch may also allow tracking an increasing trend or a decline of certain species, if long-term data is available.

#### Limitations of the Catalogue

Four major limitations of the catalogue were identified and are outlined below together with the implications when submitting queries to the catalogue: (i) heterogeneity in the amount and type of information reported on indicators; (ii) multiple reported indicators; (iii) ambiguity while interpreting fields in the catalogue; and (iv) survey gaps in the catalogue.

Missing (not reported) data, especially if forming a pattern (e.g., limited coverage of a given regional sea or failure to cover a scientific area due to lack of access rather than real lack of available information), might have compromised the robustness of the analyses (e.g., catalogue capabilities and gap analysis), and have some influence on the final recommendations and conclusions drawn out of this catalogue.

When the same indicator was repeatedly reported by more than one contributor or different indicators were reported as unique but are actually based on and conveying essentially the same information, this may lead to some approaches being under- or overrepresented in the catalogue compared to their actual availability (e.g., per geographical area or per biodiversity component). But redundancy in the focus of the indicators, i.e., their scientific basis or ecosystem relevance, does not necessarily mean that the indicators share all their properties.

Due to the great number of contributors to this catalogue, ensuring a common understanding of the fields was not always fully achieved. Heterogeneous information reported compromised the use of several entries in our catalogue, preventing optimum usage of the effort devoted to this compilation and more importantly, reducing the amount of data available for meaningful analysis. This ambiguity in interpreting fields in the catalogue was particularly evident for fields related with, for example, assigning habitat types, establishing links to pressures, and some of the 2010 Commission Decision on criteria and indicators (the latter explored in Berg et al., 2015).

Important gaps identified are in line with those reported by Hummel et al. (2015), namely regarding certain types of indicator approaches (e.g., regarding new molecular-based tools) or specific habitats (e.g., deep-sea habitats) or marine subregions. However, despite of the low prevalence of such indicators in the literature, their poor representation in the database could also be due to failing to engage local or topic-specific experts for the development of the catalogue. As a non-exhaustive catalogue, its content must be taken with caution. Those gaps could have implications on: (a) identifying priorities for the development of new indicators after the gap analysis, and (b) limiting the indicators available for selection and use as the most promising ones under global and local monitoring programs.

# Recommendations for Future Improvements of the Catalogue

The use of the catalogue could be strengthened in the future through further integration with additional quality criteria for indicator selection through a newly developed framework for testing of indicators in a standardized way (Queirós et al., 2016). As much as we filter the database and narrow our choices to indicators most promising within a given context (See example in **Box 1**), only a standardized approach based on quality analysis criteria, as proposed by Queirós et al. (2016), would allow a proper evaluation, comparison, and final selection of indicators. One of the limitations mentioned before was the danger of redundancy of indicators approaches in the catalogue. However, indicators with a similar focus may differ greatly in other characteristics considered also relevant criteria for evaluating the quality of indicators (Queirós et al., 2016 for a recent review). For example, the acceptability or comprehensibility by the wider public, the complexity of its calculation, or its cost of implementation, may be determinant criteria at the time of selecting indicators. In this sense, the catalogue could benefit from additional or better baseline information on important criteria such as "responsiveness to pressure," or the "possibility to set targets within the indicator response," or "information on the cost-effectiveness of their implementation," to name a few (ICES, 2013, 2015a; Queirós et al., 2016). This would allow an objective evaluation of the quality of the indicator, as detected also by Hummel et al. (2015). Many of the indicators listed as operational did not report any quantitative or qualitative targets or even the existence of those (n = 193). It is, therefore, questionable if those indicators are truly operational. Likewise, Hummel et al. (2015) detected that less than half of the indicators selected by EU Member States for biodiversity assessment were operational. This could partially be related to contributors filling the catalogue, who might have refrained from indicating the targets previously used in other policy or environmental contexts. Regardless of the reason behind not reporting targets, it is important to stress that an indicator output should easily be interpreted within a good-bad quality continuum. In a legal and regulatory context, such as, for example, the MSFD, it is crucial to pair indicators with thresholds, although deriving them can often be more challenging than developing the indicators themselves (Rees et al., 2008; Rossberg et al., 2017). Such thresholds are fundamental to observe the accomplishment of legally imposed targets. There are many approaches to setting targets and/or defining reference conditions, whose adequacy will be tightly linked to the context of use of the indicators (Borja et al., 2012). A recent proposal by Rossberg et al. (2017) recognizes that some ecosystems are naturally more resilient than others and proposes an approach where the longest acceptable length of the recovery time is used for setting targets. Regarding the catalogue, and since operational status refers by definition to a fully developed indicator, we expect that information regarding thresholds or targets even if not reported might still be available in the "Sources" cited in the catalogue.

Along these lines, it should also be possible to link the indicator to different components of the DIPSR frameworks or, the more recently proposed, DAPSI(W)R(M) framework (Atkins et al., 2011; Wolanski and Elliott, 2015), to allow understanding on how the indicator reflects policy responses and measures impacting the changes in the status. These linkages can be demonstrated conceptually, qualitative, or quantitatively (e.g., using models). On the other hand, different types of indicators reflecting Responses, Measures, Drivers and Pressures are needed to demonstrate the effects of management efforts and to advice the policy development (Rapport and Hilden, 2013). As stated earlier, management needs also descriptive indicators, performance indicators, efficiency indicators and total welfare indicators (Smeets and Weterings, 1999).

Another pertinent property of a robust indicator, especially in the context of comprehensive and wide-scale environmental assessment initiatives, is whether the indicators and their data requirements are already covered or are integrating "part of an existing or current ongoing monitoring or data" (Queirós et al., 2016). The DEVOTool catalogue of marine biodiversity indicators together with the DEVOTES catalogue on marine biodiversity monitoring networks (Patrício et al., 2016) compiled such information, and although it was evident that data are available, it is, nevertheless, difficult to evaluate its adequacy (e.g., Joppa et al., 2016) or account for this feature without a framework for properly ranking and selecting indicators.

Another example of a fundamental but widely neglected criterion is the capability of an indicator to provide "concrete, measurable, accurate, and precise outputs." Our catalogue shows that for the majority of the indicators reported (over 90%) there was no reference to any measure of confidence or uncertainty associated with their assessment results. Despite the

BOX 1 | DEVOTOOL 0.64 (database version 7) advanced query example: selection of indicators targeting angiosperms, which are particularly responsive to pressures caused by nutrients and organic enrichment. The query includes accessory information on the main attribute of the indicators, their developmental status, and the dpsir stage to which they apply. If monitoring time series are available the respective period is indicated. WFD, Water Framework Directive.


recognized importance that the quantification of uncertainty has within an environmental assessment (e.g., Andersen et al., 2010; Chaalali et al., 2015; Uusitalo et al., 2015; Carstensen and Lindegarth, 2016), such procedures are often disregarded during index development and seldom applied as a standardized and sound routine. Examining the propagation of uncertainty from indicators to overall biodiversity assessment (Andersen et al., 2014; Carstensen and Lindegarth, 2016) is of utmost importance to ensure robust assessments within large initiatives (e.g., the MSFD Commission Decision 56 indicators) (Probst and Lynam, 2016). Among the most relevant sources of uncertainty that affect indicators' estimates (Nardo et al., 2008) there are: the choice of sub-metrics or parameters within an indicator, the quality of the underlying data, the approach chosen to deal with missing data, the normalization, weighting, and aggregation procedures. Through our survey, we could observe that: (i) not only this information is usually not available or reported, which by itself is a sign of how the issue is still poorly integrated as a fundamental step in index development and application; but also that (ii) measures of confidence in the results typically cover only few of the sources of uncertainty mentioned above. In this sense, efforts should be focused on increased coverage and standardization of procedures to evaluate sources of uncertainty, to provide better guidance on indicator development, performance evaluation and selection.

Finally, the catalogue could be expanded to further accommodate other types of indicators as, for example, the remaining descriptors of the MSFD, for which several indicators are already included (Berg et al., 2015). The catalogue shows also potential to support selection of indicators that capture the state changes in the natural system that finally result in impacts to the human well-being and to the way we can use the natural resources, i.e., ecosystem services ("I" in DPSIR or "I" and "W" in DAPSI(W)R(M) frameworks; Atkins et al., 2011; Wolanski and Elliott, 2015). In the absence of indicators or metrics specific to ecosystem services (Liquete et al., 2013) the mapping and assessment of ecosystems services could be based on existing approaches. Several indicators in the catalogue have been recently considered useful alternatives for measuring provisioning and regulating marine ecosystem services (Maes et al., 2016) (e.g., catch per unit effort, nutrient load, oxygen concentration, turbidity, pH, primary production, species distribution, extent of marine protected areas). Indicators for provisioning marine ecosystem services depend strongly on fishery statistics, while for regulating services they are based on sea water quality observations or modeling (Maes et al., 2016). The information in the catalogue might reveal other sources of potentially useful and complementary information.

# Practical Application in Environmental Assessments

Recently developed by the DEVOTES project is the Nested Environmental status Assessment Tool (NEAT, http://www. devotes-project.eu/neat, based on Andersen et al., 2014), for assessing the status of marine waters (Borja et al., 2016a). In the NEAT, the indicators are thematically grouped, assigning them to the corresponding habitats, biodiversity components, spatially defined marine areas and pressures for which they are used (such functionality is ensured through the DEVOTool software presented in this manuscript). The NEAT follows an Ecosystem Approach (Tett et al., 2013), ensuring that all ecosystem features relevant to the assessment are accounted for (Borja et al., 2016a). This NEAT tool, facilitating an indicator-based assessment of marine biological diversity, has been successfully tested across ten case studies in Europe (Uusitalo et al., 2016), and the authors lay down recommendations for best practices while using this customizable NEAT for marine status assessments.

Using multiple lines of evidence during environmental assessments has been common practice for a long time now (e.g., Adams, 2005; Bay et al., 2007). The importance of integrating knowledge from different ecosystem aspects has recently grown with the overall acceptance of the Ecosystem Based Management Approach, and has led to several proposals and recommendations on best practices to integrate multiple indicators and assessment scales (review by Borja et al., 2014). This catalogue will certainly reveal a handy tool for screening and comparing complementary indicators to incorporate into more complex assessments (e.g., Uusitalo et al., 2016).

# CONCLUSIONS

Despite the geographical focus on the European Regional Seas, we consider that this catalogue provides a comprehensive overview of the existing knowledge and advances in the field of ecological assessments, by integrating the major type of indicator approaches currently available and used in regular monitoring and environmental assessment programs, particularly regarding biological diversity, food webs, seafloor integrity and nonindigenous species.

This catalogue supports more effective biodiversity monitoring and further investment in indicators, essential to track and improve the effectiveness of management responses (Butchart et al., 2010). We expect this tool can pave the way to rationalizing the development of indicators, and that weaknesses encountered can set research priorities, promoting a more robust use of indicators within the context of environmental policies and assessment programs.

Moreover the DEVOTool is linked with NEAT that provides a tool to decide upon combination of the different indicators into a holistic assessment of the environmental status. We advocate that these tools linked together will support development toward more coherent assessment of marine ecosystems across the regional seas. Due to their potential to support the use of common indicators and the adoption of standardized approaches across marine conservation initiatives, these tools will certainly facilitate conservation efforts by a wide range of users, such as EU Member States, Regional Sea Conventions, the European Commission, governmental organizations outside the EU, non-governmental organizations,

scientists, and any person interested in marine environmental issues.

# AUTHOR CONTRIBUTIONS

All co-authors contributed in building the catalogue structure, in the writing process and in discussion of the results. HT, TB, LU, KF, KM, CL, SN, JR, NP, SM, TC, OK, DK, AZ, HV, MP, and SC have contributed with indicators and metadata to the catalogue. HT, TB, LU, and AH have planned the article. HT, TB, KF, and LU have processed the metadata and performed the analyses. TB has written the catalogue software code.

#### ACKNOWLEDGMENTS

We thank J. H. Andersen, A. G. Rossberg, who took valuable time to contribute with indicators to this catalogue. This manuscript is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project, funded by the

### REFERENCES


European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), www.devotes-project.eu. MU is partially funded through the Spanish programme for Talent and Employability in R + D + I "Torres Quevedo." AZ is partially funded through the BONUS project BIO-C3 funded jointly by the EU and the Research Council of Lithuania.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00207/full#supplementary-material

Supplementary Material S1 | "SuppMatS1\_Catalogue\_vs7\_2016.db"–this file contains the Catalogue of Indicators database version 7; db file opens with DEVOTool 0.64 free software application, downloadable from http://www.devotes-project.eu/devotool/.

Supplementary Material S2 | "SuppMatS2\_Categories-definitions.xlsx"– this file contains the definitions of indicators' categories adopted in our study within "Main attribute," "Underlying variable type," and "Algorithm type."


Across Regional Seas. Deliverable 6.1, 105 pp. + 1 Annex. Devotes FP7 Project.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Teixeira, Berg, Uusitalo, Fürhaupter, Heiskanen, Mazik, Lynam, Neville, Rodriguez, Papadopoulou, Moncheva, Churilova, Kryvenko, Krause-Jensen, Zaiko, Veríssimo, Pantazi, Carvalho, Patrício, Uyarra and Borja. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# An Objective Framework to Test the Quality of Candidate Indicators of Good Environmental Status

Ana M. Queirós <sup>1</sup> \*, James A. Strong<sup>2</sup> , Krysia Mazik <sup>2</sup> , Jacob Carstensen<sup>3</sup> , John Bruun<sup>1</sup> , Paul J. Somerfield<sup>1</sup> , Annette Bruhn<sup>4</sup> , Stefano Ciavatta1, 5, Eva Flo<sup>6</sup> , Nihayet Bizsel <sup>7</sup> , Murat Özaydinli <sup>7</sup> , Romualda Chuševe˙ 8 , Iñigo Muxika<sup>9</sup> , Henrik Nygård<sup>10</sup> , Nadia Papadopoulou<sup>11</sup> , Maria Pantazi <sup>12</sup> and Dorte Krause-Jensen<sup>4</sup>

*<sup>1</sup> Plymouth Marine Laboratory, Plymouth, UK, <sup>2</sup> University of Hull, Hull, UK, <sup>3</sup> Department of Bioscience, Aarhus University, Roskilde, Denmark, <sup>4</sup> Department of Bioscience, Aarhus University, Silkeborg, Denmark, <sup>5</sup> Plymouth Marine Laboratory, National Centre for Earth Observation, Plymouth, UK, <sup>6</sup> Institut de Ciències del Mar, Consejo Superior de Investigaciones Científicas, Barcelona, Spain, <sup>7</sup> Institute of Marine Science and Technology, DokuzEylul University, Inciralti-Izmir, Turkey, <sup>8</sup> Marine Science and Technology Center, Klaipeda University, Klaip ˙ eda, Lithuania, ˙ <sup>9</sup> AZTI, Gipuzkoa, Spain, <sup>10</sup> Marine Research Centre, Finnish Environment Institute, Helsinki, Finland, <sup>11</sup> Institute of Marine Biological Resources and Inland Waters, Hellenic Centre for Marine Research, Crete, Greece, <sup>12</sup> Institute of Marine Biological Resources and Inland Waters, Hellenic Centre for Marine Research, Athens, Greece*

#### Edited by:

*Christos Dimitrios Arvanitidis, Hellenic Centre for Marine Research, Greece*

#### Reviewed by:

*Melih Ertan Çinar, Ege University, Turkey Matt Terence Frost, Marine Biological Association, UK*

> \*Correspondence: *Ana M. Queirós anqu@pml.ac.uk*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *12 February 2016* Accepted: *28 April 2016* Published: *26 May 2016*

#### Citation:

*Queirós AM, Strong JA, Mazik K, Carstensen J, Bruun J, Somerfield PJ, Bruhn A, Ciavatta S, Flo E, Bizsel N, Özaydinli M, Chuševe R, Muxika I, ˙ Nygård H, Papadopoulou N, Pantazi M and Krause-Jensen D (2016) An Objective Framework to Test the Quality of Candidate Indicators of Good Environmental Status. Front. Mar. Sci. 3:73. doi: 10.3389/fmars.2016.00073* Large efforts are on-going within the EU to prepare the Marine Strategy Framework Directive's (MSFD) assessment of the environmental status of the European seas. This assessment will only be as good as the indicators chosen to monitor the 11 descriptors of good environmental status (GEnS). An objective and transparent framework to determine whether chosen indicators actually support the aims of this policy is, however, not yet in place. Such frameworks are needed to ensure that the limited resources available to this assessment optimize the likelihood of achieving GEnS within collaborating states. Here, we developed a hypothesis-based protocol to evaluate whether candidate indicators meet quality criteria explicit to the MSFD, which the assessment community aspires to. Eight quality criteria are distilled from existing initiatives, and a testing and scoring protocol for each of them is presented. We exemplify its application in three worked examples, covering indicators for three GEnS descriptors (1, 5, and 6), various habitat components (seaweeds, seagrasses, benthic macrofauna, and plankton), and assessment regions (Danish, Lithuanian, and UK waters). We argue that this framework provides a necessary, transparent and standardized structure to support the comparison of candidate indicators, and the decision-making process leading to indicator selection. Its application could help identify potential limitations in currently available candidate metrics and, in such cases, help focus the development of more adequate indicators. Use of such standardized approaches will facilitate the sharing of knowledge gained across the MSFD parties despite context-specificity across assessment regions, and support the evidence-based management of European seas.

Keywords: ecosystems, European union, good environmental status, indicator, marine strategy framework directive, pressure, water framework directive (WFD)

# INTRODUCTION

The current paradigm of marine management in Europe determines that decisions should be weighed on their impacts on whole ecosystems rather than on individual ecosystem components (United Nations, 1992; MEA, 2005). This "ecosystem approach" is enshrined in the EU Marine Strategy Framework Directive (the MSFD, EC, 2008; EU, 2014) and associated Maritime Spatial Planning Directive (EU, 2014). Component parts to this approach are the aims to attain and preserve "good environmental status" in EU waters ("GEnS," EC, 2008), the definition of which has been summarized across 11 descriptors. Various initiatives have consequently proposed metrics that could serve as indicators for these descriptors to support their monitoring (hereafter "indicators," e.g., Rice et al., 2012; Borja et al., 2013), and efforts are being made to review a wealth of available and new metrics (hereafter, "candidate" indicators, or "candidate" metrics, Borja et al., 2014; Teixeira et al., 2014). As the assessment of GEnS is the fundamental aim of the MSFD, the credibility of this policy depends on the choice of adequate GEnS indicators for its descriptors. Various indicator quality criteria have since been suggested as the desirable characteristics of GEnS indicators that are fit for purpose, and discussions regarding their assessment are being undertaken (Borja et al., 2013; ICES, 2013b, 2015; Rossberg et al., 2013; Hummel et al., 2015). Additionally, scoring systems for the assessment of candidate indicators have been proposed by ICES (2013b, 2015) using a set of 16 quality criteria. However, a stringent framework for assessing whether these candidate indicators actually meet this or other sets of desired quality criteria, that is both comprehensive and applicable across the 11 descriptors of GEnS, has not been described. Though the desirable traits of a GEnS indicator may be intuitive, it is difficult to define objectively whether a candidate metric actually possesses such traits. Judgments or values thus need to be objectively laid out to enable the comparison of candidate metrics, so that an informed selection can be made across descriptors, and a smaller list of indicators ultimately suggested for implementation of the MSFD. This study aimed to provide a standardized procedure to evaluate the quality of candidate indicators across the descriptors, through objective analysis and testing. This framework lays out a transparent and repeatable methodology to test the fulfillment of quality criteria that can be used to define indicator quality, and to rank candidate indicators to facilitate indicator selection within the MSFD assessment.

From a wide range of published alternatives (**Table 1**, adapted from Krause-Jensen et al., 2015) the ICES quality criteria for selecting MSFD GEnS indicators for the North Sea (ICES, 2013a,b, 2015) were chosen as a basis for the present study because this list already resulted from previous exercises to synthesize published efforts, reflecting common aspirations within the community. This ICES quality criteria list has already been applied for selecting common OSPAR (the Convention for the Protection of the Marine Environment of the North-East Atlantic) indicators for the MSFD (ICES, 2015). The list describes 16 quality criteria which were here further distilled to eight Indicator Quality criteria [henceforth, "IQ(s)," **Figure 1** and Table S1]. This simplification was deemed necessary to facilitate the operationalization of indicators by reducing perceived overlap within that list and keeping the focus on state indicators and key performance criteria for these (Table S1 for justification, from Krause-Jensen et al., 2015). Based on these eight IQs, a framework for the analysis of candidate GEnS indicators is presented here which: (1) formulates objective, transparent and repeatable tests of indicator quality; (2) constructs a ranking system to enable the comparison of alternative candidate indicators and thus facilitate indicator selection; and (3) quantitatively displays indicator strengths and weaknesses, and hence the potential need for additional indicator development. Within a wide range of available candidate metrics, four falling within the remit of expertise of the authors, were chosen to investigate and demonstrate the application of this framework as worked examples.

# MATERIALS AND METHODS

The proposed indicator quality testing framework is detailed below, followed by three worked examples detailing its application to four candidate metrics. These metrics currently exist at different stages of operationalization as candidate indicators for the MSFD. Presentation of these worked examples was thus not primarily aimed to serve as actual tests of their quality as actual indicators for the MSFD (although this text could potentially come to support that aim). Rather, they are detailed here with the specific aims of investigating and demonstrating the application of the proposed testing framework across a variety of GEnS descriptors, indicator and ecosystem types, to help build the case for, and support, its further uses by the community. Specifically: the quality of presence of keystone kelp species and the depth limit of eelgrass as candidate metrics for descriptors 1 (biodiversity) and 5 (eutrophication) in the Danish coast is evaluated in worked example I; the quality of the temporal trend of N:P in coastal waters as a potential indicator for the occurrence of harmful algal blooms under descriptor 5 (eutrophication) in the UK is evaluated in worked example II; and the quality of the Benthic Quality Index (BQI, Fleischer et al., 2007) as a potential indicator for descriptors 1 (biodiversity) and 6 (seafloor integrity) in the Lithuanian coast is evaluated in worked example III. With regard to their current status of operationalization: seagrass depth limits are already considered in Denmark and other European countries as indicators for ecological status under the Water Framework Directive ("WFD"), and are being considered within the MSFD (Marbà et al., 2013); presence of kelps is being considered by ICES and specific European countries as a potential indicator for descriptor 1 of the MSFD, though not yet in Denmark (Burrows et al., 2014; Hummel et al., 2015); the trend of N:P is not yet being considered by the MSFD, although the data required for its estimation is collected routinely as part of WFD monitoring efforts around Europe; the BQI is already extensively in use by Baltic countries to assess ecological status for the WFD, including by the Lithuanian Environment Ministry (Šiaulys et al., 2011), and it is under consideration for the

#### TABLE 1 | Literature survey of the use of indicator quality criteria (IQ).


*In the present study, we incorporated the eight criteria marked in bold, IQ1–IQ8. A number of additional criteria were considered implicit in IQ1–IQ8, some criteria were considered of secondary importance, and some criteria (regarding large-scale applicability and commonly accepted status) were excluded from our framework on the basis that indicators fulfilling the key criteria IQ1–IQ8 are also potentially relevant for large-scale application and acceptance. Adapted from Krause-Jensen et al. (2015). Further justification for distilling the 16 ICES criteria to 8 key criteria are provided in Table S1.*

<sup>∧</sup>*The ICES (2013a,b); ICES (2015) criterion "state or pressure indicator" is not included here as our test is focused on state indicators.*

∧∧*Based on Schomaker (1997), OECD (2001), NRC (2000), Dale and Beyeler (2001), CBD (1999), Pannell and Glenn (2000), Kurtz et al. (2001), EEA (2005).*

∧∧∧*Based on a total of nineteen evaluation criteria gleaned from the literature (O'Connor and Dewling, 1986; Landres et al., 1988; Noss, 1990; Harwell et al., 1999; Jackson et al., 2000; Kurtz et al., 2001; Rice, 2003; Jennings, 2005; Rice and Rochet, 2005; Niemeijer and de Groot, 2008; Doren et al., 2009; Jørgensen et al., 2010).*

\**Includes also: metrics should fit indicator function (ICES criterion #14); biologically important (Elliott, 2011), representable (OSPAR), integrative and general importance (Niemeijer and de Groot, 2008).*

\*\**Includes also: space-bound (sensitive to changes in space, Niemeijer and de Groot, 2008).*

\*\*\**Includes also: practicable.*

\*\*\*\**Includes also: confidence evaluation; uncertainty about level (Niemeijer and de Groot, 2008), and "limitations defined" JCN/HBDSEG (2012).*

\*\*\*\*\**Includes also: suitability w. assessment tools (HELCOM, 2012).*

\*\*\*\*\*\**Includes also: reliability (Niemeijer and de Groot, 2008).*


MSFD; it is already being implemented in Sweden under this directive.

#### Quality Testing: The IQ-ES Framework

The indicator evaluation framework is detailed in the next section. For a given candidate indicator (A) or a pair of candidate indicators (A and B) of the same descriptor of GEnS being compared, a sequence of five Evaluation Steps (henceforth "ES") was defined for each of eight IQs to determine whether each is met (**Figure 1**). In summary, ES1 states the null hypothesis associated with the IQ tested; ES2 defines which assessment approach should be employed to test the hypothesis, i.e., qualitative or quantitative, and is conditional to its nature; ES3 states the type of evidence required to undertake the assessment; ES4 defines the methodology (e.g., type of statistical analysis or otherwise) undertaken to test the hypothesis considered and its outcome; ES5 states the quality score for the particular IQ tested given ES4. If the test is successful (within the assessment of each of the eight IQs), the indicator scores 1 in the final step (**Figure 1**, ES5) and 0 otherwise. Once IQs 1–8 have been assessed through these steps individually, all scores are summed in a final step (**Figure 1**, ES6) and a total quality score for the candidate indicator is calculated, which can be compared to that of other candidate indicators for the same descriptor.

At the core of this assessment structure is the expression of each IQ into a testable null hypothesis (ES1). In keeping with a statistical testing background, the hypothesis is stated as a negative that is rejected if the indicator meets the IQ tested for, and accepted otherwise (ES5). Without this first step, there is no clarity about what attribute of quality is being assessed. For example, in IQ1 (**Figure 1**, "scientific basis") ES1 (the null hypothesis) states that "there is no scientific basis for the indicator." Based on the review of associated literature, an informed judgment can be made: the analysis that tests this hypothesis is therefore qualitative and the outcome is categorical (yes or no). Examples of qualitative approaches may therefore include expert judgment, by which e.g., a review of literature may be sufficient to establish whether the indicator satisfies a particular criterion of quality. Conversely, in IQ3 (**Figure 1**, "responsiveness to pressure"), ES1 is only truly testable under a quantitative approach, requiring that a minimum pressure change of interest induces a measureable and consistent indicator response, for the system analyzed. Quantitative approaches could include statistical analyses, graphical exploration of data, or any type of numerical modeling to define a quantitative relationship. The nature of the hypothesis defined by ES1 therefore dictates which type of approach should be preferred in ES2 (qualitative c.f. quantitative). The preferred type of approach (**Figure 1**, ES2) in turn helps identify which type of evidence, resources (**Figure 1**, ES3), and analyses (**Figure 1**, ES4) need to be considered for the assessment of each specific IQ, for each indicator and context (i.e., descriptor, area).

Whilst the analysis method used in ES4 may be substantially different between candidate metric types, the comparison of metrics to enable selection requires that the quality assessment is standardized across these metrics within descriptors. We suggest that this quality scoring system provides this comparative basis. Various weighted and non-weighted scoring systems are possible in ES5. However, given that the key aims of this framework are the objective, transparent and repeatable evaluation and ranking of indicators according to quality criteria, we suggest that the use of a binary system (0,1) provides the most unambiguous statement of the assessment outcome: that the indicator does (1) or does not (0) meet the quality criterion tested. However, here, we compare this approach with that suggested by ICES (2013a,b, 2015), which includes an additional possible score (0.5) in IQs 2 and 4–8, expressing that a given quality criterion is partially fulfilled (three-way scoring system).

We suggest that once ES1–6 have been undertaken for a pair of candidate indicators (e.g., A and B) for a given descriptor, their total quality score (ES6) should provide a sufficient basis for a pair-wise comparison and selection, with preference given to the indicator with the highest score. This is a fundamental step toward an objective sorting and selection of candidate indicators, ensuring consistency, comparability, transparency and repeatability of the selection approach regardless of the indicator, descriptor, pressure, habitat, or biological component assessed. Overall, this general framework thus converts aspirational attributes (Table S1) associated with the definition of indicators into a series of defined, analytical steps to establish GEnS candidate indicator quality. IQ1 and IQ3 are seen as essential quality criteria in the assessment, such that failure to meet either of these criteria should render exclusion. In other words, IQ1 and IQ3 are "one-out-all-out" criteria. Overall score ties between candidate indicators (ES6) compared using this framework require expert judgment for selection (see also Table S1). Here too, the standardized format of the IQ-ES assessment could set a good basis to inform this decision because the quality assessment is broken down into its component criteria.

# The GEnS Indicator Quality Evaluation Steps

#### **IQ 1: Scientific basis (one-out-all-out criterion)**

**IQ1–ES1:** there is no scientific basis for the indicator.

**IQ1–ES2:** expert judgment/qualitative approach are adequate.

**IQ1–ES3:** publications evidencing the conceptual basis for using the indicator, stressing the existence of a general causal link between the indicator and a given pressure, highlighting an effect on the relevant descriptor. Peerreviewed publications are preferred but, in some instances, reports from governmental institutes or international institutions (e.g., ICES) may be more appropriate.

**IQ1–ES4:** the indicator must be reproducible, i.e., the conceptual basis and causality relationship have been published (preferentially in peer-reviewed literature) using multiple data sets, and this can be seen as a proxy for its wide acceptance within the relevant scientific community.

**IQ1–ES5:** the indicator scores 1 if the above can be verified. If the indicator scores 0 in IQ1, it is seen as failing in the quality assessment as this is a one-out-all-out quality criterion. Because of this, we consider that the three-way scoring system is not applicable to IQ1.

#### **IQ 2: Ecosystem relevance**

**IQ2–ES1:** there is no evidence linking the indicator to (a) ecosystem level processes and function (the non-anthropocentric perspective; e.g., indicators of processes undertaken by keystone species could be particularly relevant); and/or (b) ecosystem services (the anthropocentric perspective, i.e., societal relevance).

**IQ2–ES2:** expert judgment/qualitative approach are adequate.

**IQ2–ES3:** scientific, peer-reviewed evidence for the nonanthropocentric criterion and/or for the anthropocentric criterion.

**IQ2–ES4:** a literature review is a recommended approach to test IQ2. Evidence for the ecosystem relevance of the indicator should have been published in peer-reviewed literature. Within the anthropocentric perspective, the indicator must be explicitly listed within recognized ecosystem function/service typologies, or they have been linked directly to a monetary valuation. For instance, indicators listed under the Common International Classification of Ecosystem Services (Haines-Young and Potschin, 2013) or another equally widely applied typology are preferred.

**IQ2–ES5:** the indicator scores 1 if the above (IQ2–ES4) can be verified and 0 otherwise. The three-way scoring system could be applied to IQ2.

#### **IQ 3: Responsiveness to pressure (one-out-all-out criterion)**

**IQ3–ES1:** the indicator does not exhibit consistent and significant change as a result of a change in pressure, as listed within the recognized MSFD pressure list (EC, 2008), in the system of interest.

**IQ3–ES2:** a quantitative approach is adequate.

**IQ3–ES3:** the data used for testing should include some information about the natural baseline of the system, including information about its natural variability because this may confound the ability to detect a pressure driven effect. The drivers of the natural variability baseline of the indicator are known and understood. In case data for the area in question is not sufficiently comprehensive to allow proper pressure-response analyses, pressure-response analyses conducted for the same candidate indicator in comparable ecosystem(s) could be considered.

**IQ3–ES4:** the method of analysis must consider the impact/influence of natural variability (if any) on the response of the indicator (identify, estimate, and diagnose). The analysis must be appropriate for the complexity of the data to hand.

**IQ3–ES5:** the indicator scores 1 if a consistent and significant change is measured in response to the pressure (IQ3–ES4), and 0 if: (i) there is no change in response to pressure; or (ii) the change in the indicator in response to pressure is not consistent (across areas, scales); or (iii) the measured change in the indicator in response to the pressure is not statistically significant. If the indicator scores 0 in IQ3, it is seen as failing in the quality assessment as this is a one-out-all-out quality criterion. Because of this, the three-way scoring system is not applicable to IQ3.

#### **IQ 4: Possibility to set targets**

**IQ4–ES1:** a clear and unambiguous target cannot be defined for the indicator within a range with defined units of measurement.

**IQ4–ES2:** both expert judgment/qualitative approach and a quantitative approach can be adequate, depending on the indicator.

**IQ4–ES3:** information about the range of natural variability of the system is required, against which the target level is defined.

**IQ4–ES4:** the method of analysis must consider the impact/influence of natural variability (if any) on the response of the indicator (identify, estimate, and diagnose). The analysis must be appropriate for the type of data at hand (qualitative c.f. quantitative).

**IQ4–ES5:** the indicator scores 1 if a clear and unambiguous target can be defined with clear units of measurement, and 0 if: (i) a clear and unambiguous target cannot be defined; or (ii) there is not sufficient background information to define the range of the natural variability of the system (i.e., habitat and scale) within which the indicator is to be implemented. The three-way scoring system could be applied to IQ4.

#### **IQ 5: Precautionary capacity/early-warning/anticipatory**

**IQ5–ES1:** there is no immediate and measurable change in the indicator associated with a change in the pressure that anticipates ecosystem-level change in the system (see IQ2). **IQ5–ES2:** a quantitative approach is adequate.

**IQ5–ES3:** data that enables a quantification to be made about the time lag between pressure level and indicator response, and that between pressure change and ecosystem-level relevant change. Information must exist about a clear link between pressure level and ecosystem state. The indicator must be responsive to pressure (IQ3). These data are particularly important in instances where system collapse may occur. The rate of change in the indicator during impact and recovery phases may be distinct.

**IQ5–ES4:** any quantitative method of analysis that measures the lag time between pressure and indicator response, and the lag between pressure change and ecosystemlevel change. The indicator analysis method must be reproducible (IQ6).

**IQ5–ES5:** the indicator scores 1 if the lag time between pressure change and the detection of a measurable change in the indicator level is small and suitable to enable mitigation action to take place to prevent ecosystem-level change. The indicator scores 0 if the time lag between pressure change and indicator response is not sufficiently small to support action taking place within the system to prevent further ecosystem scale deterioration. The threeway scoring system could be applied to IQ5.

#### **IQ 6: Quality of sampling method: Concrete/measurable, accurate, precise and repeatable**

**IQ6–ES1:** the indicator is not concrete/measurable, accurate, precise or repeatable. Concreteness/measurability refers to whether the indicator can be quantitatively assessed. Accuracy refers to the closeness of an estimate of an indicator to the true value of the indicator. Precision refers to the degree of concordance among a number of estimates for the same population and repeatability to the degree of concordance among estimates obtained by different observers (Sokal and Rohlf, 1969).

**IQ6–ES2:** a quantitative approach is adequate.

**IQ6–ES3:** identification of whether an indicator is concrete/measurable requires availability of well-defined quantitative data. Testing for accuracy requires quantitative data to address the possibility of measurement bias. Testing for precision requires data covering spatial and temporal scales of variability and is necessary for quantifying how much sampling effort is required to identify an effect size of a defined level in the indicator in the context of the spatial- and temporal variability of the system being assessed. Testing for repeatability requires data allowing comparability of estimates obtained by two or more different observers.

**IQ6–ES4:** For the analysis of concreteness/measurability, any method that enables well-defined quantitative information on the indicator can be used. For testing accuracy and precision and repeatability, analyses of variability are suitable and these can be supplemented with power analysis and species area curves to evaluate the necessary sampling effort.

**IQ6–ES5:** the indicator scores 1 only in the case in which all analyses in IQ6–ES4 lead to the rejection of the null hypothesis set out by IQ6–ES1. The indicator scores 0 if the hypothesis cannot be rejected for one or more of the attributes (i.e., if the indicator cannot be positively identified as being simultaneously concrete, accurate, precise, and repeatable). In the case of score ties, indicators for which the most attributes in IQ6 could be validated are preferred. The three-way scoring system could be applied to IQ6.

#### **IQ 7: Cost-effective**

**IQ7–ES1:** the indicator is not cost effective.

**IQ7–ES2:** a quantitative approach is adequate.

**IQ7–ES3:** requires information about the levels of precision and accuracy required (IQ6), against which the costs of the necessary method of implementation of the indicator are calculated.

**IQ7–ES4:** any analysis that enables the establishment of the change in cost associated with an improvement in the criteria of accuracy and precision of the indicator.

**IQ7–ES5:** the indicator scores 1 if the cost associated with the desired level of precision and accuracy is manageable and 0 otherwise. The three-way scoring system could be applied to IQ7.

#### **IQ 8: Existing and ongoing monitoring data**

**IQ8–ES1:** the indicator is not currently used in ongoing monitoring program(s).

**IQ8–ES2:** a quantitative approach is adequate.

**IQ8–ES3:** requires information about the length of time during which the indicator has been in use within a monitoring program, and of the redundancy the indicator in relation others (if any) also in use within the scale of analysis of interest.

**IQ8–ES4:** any method that quantifies the above (IQ8–ES3). **IQ8–ES5:** the indicator scores 1 if is already in use in at least one monitoring program, and 0 otherwise. In a score tie, indicators with the longest use of application, exhibiting potential for application in the widest areas of interest, are preferred. The three-way scoring system could be applied to IQ8.

#### **ES6 sum of quality scores**

The scores given in ES5 in IQ1–8 are summed, ranging between 0 and 8.

#### Worked Examples

We exemplify the application of this framework in three case-studies, assessing potential candidate indicators of marine


ecosystem components ranging from nutrients and benthic vegetation to soft sediment faunal communities (**Table 2**). For practical reasons, we provide only one worked example in the main body of the text, analyzing two candidate indicators; two other worked examples are explored in the same level of detail in the Supplementary Materials Section.

#### RESULTS

# Worked Example I. Candidate Indicators for Descriptors 1 (Biodiversity) and 5 (Eutrophication): Presence of Keystone Kelp Species and Eelgrass Depth Limit

In this example, we comparatively evaluate the quality of two candidate indicators which could be used to monitor both descriptor 1 (Biodiversity) and descriptor 5 (Eutrophication), within Danish waters. Specifically, we compare the quality of: the presence of keystone kelp species (seaweeds) and the depth limit for eelgrass (a seagrass). This evaluation is summarized in **Table 3**.

#### IQ 1. Scientific Basis

Both candidate indicators and their general responses to human driven nutrient loading pressure (causing eutrophication) are conceptually well founded in the scientific literature. More specifically, Duarte (1991) and Duarte et al. (2007) demonstrated a global trend that deeper seagrass meadows occur in clearer waters. This relationship is supported by studies in Danish coastal waters, where the depth limit of eelgrass is largest in the clearest waters with lowest nutrient concentrations (Nielsen et al., 2002; Greve and Krause-Jensen, 2005; Krause-Jensen et al., 2011). Markedly deeper meadows than those found at present were found during past periods of lower nutrient inputs (Boström et al., 2014). Similarly, spatio-temporal data from Norway's coast indicate declines in kelp forests in response to nutrient loading causing eutrophication (Moy and Christie, 2012). Therefore, literature exists that has linked both of these candidate indicators to eutrophication, which is listed by the MSFD as reflecting poor GEnS (descriptor 5). In addition, kelp forests and seagrass meadows constitute habitat for a vast diversity of species (Gutiérrez et al., 2011; Boström et al., 2014). Therefore, both indicators are also linked to the descriptor 1 (Biodiversity). Both candidate indicators therefore scored 1 in IQ1 (**Table 3**).

#### IQ 2. Ecosystem Relevance

Kelp forests and seagrass meadows are so-called keystone species and ecosystem engineers, providing a whole range of additional ecosystem functions and services including coastal protection, seafloor stabilization, carbon and nutrient retention, and promotion of water clarity (Costanza et al., 1997; Gutiérrez et al., 2011; Duarte et al., 2013). Both candidate indicators therefore scored 1 for IQ2 (for both descriptors), fulfilling the criterion of ecosystem relevance, from both anthropocentric and non-anthropocentric perspectives.

#### IQ 3. Responsiveness to Pressure

The trend of deeper seagrass meadows in clearer and less nutrient-rich waters has been demonstrated in the case-study system (Danish waters, Nielsen et al., 2002; Greve and Krause-Jensen, 2005; Krause-Jensen et al., 2011; Riemann et al., 2016) and globally (Duarte et al., 2007). It is, however, important to note that while response to increased nutrient pressure may be quick, the recovery of this vegetation following reduced nutrient inputs may require long time frames (Krause-Jensen et al., 2012; Duarte et al., 2015; Riemann et al., 2016). Hence, eelgrass depth limits have been found to exhibit no signs of improvement after 15 years of nutrient input reductions in a shallow German bay (Munkes, 2005) while in Danish coastal waters, recovery has been observed more than 2 decades after nutrient input reductions (Hansen, 2013; Riemann et al., 2016). Several sources of variability have been tested for eelgrass depth limits (a requirement to meet this IQ in the present framework), the most important being spatial variability, which must be carefully addressed in the planning of monitoring programs (Balsby et al., 2013). Hence, with respect to the responsiveness criterion, eelgrass depth limits scored 1 in IQ3.

With respect to the presence of kelps, spatio-temporal data from Norway's coast indicate declines in kelp forests in response to nutrient loading (and warming) causing eutrophication (Moy and Christie, 2012). By contrast, a recent Danish study showed no response of the presence of kelps to varying nutrient concentrations (Krause-Jensen et al., 2015) indicating that this candidate indicator is not sufficiently sensitive near the geographical distribution limit, where low salinity and high summer temperatures constrain growth (Nielsen et al., 2014). Kelp presence scored 0 in the binary scoring system. As this is one of the most important quality criteria (i.e., one of the two "one-out-all-out" criteria), the presence of kelps as indicators for descriptor 1 (and 5) of GEnS would be rejected under the current assessment framework.

#### IQ 4. Possibility to Set Targets

Historical information on eelgrass depth limits from a period with limited nutrient input can form a suitable basis for establishing targets for eelgrass depth extension in Danish coastal waters, and pressure-response relationships can also be used for target-setting (e.g., Carstensen and Krause-Jensen, 2009) whilst considering the natural variability of this candidate indicator. Conversely, no clear pressure-response relationship between presence of even the most common kelps in the area [Saccharina latissima (Linnaeus) and Laminaria digitata (Hudson)] and nutrient pressure can be established at present to support target setting for this candidate indicator. Therefore, seagrass depth scored 1 in IQ4, whilst keystone kelp presence scored 0 in this particular example. Targets for both species should always be identified for the particular areas of interest.

#### IQ 5. Precautionary Capacity/Early-Warning/Anticipatory

The early warning capacity of both candidate indicators assessed is limited. Eelgrass depth limits scored 0 on this criterion


#### TABLE 3 | Summary of quality assessment of the candidate benthic vegetation indicators "Presence of keystone kelp species" and "Eelgrass depth limit," both relating to the MSFD indicator category "Distributional pattern (1.4.2), in association with the GEnS descriptors 1 and 5.

*(Continued)*

#### TABLE 3 | Continued


*ES1 is summarized in the text.*

because of the slow response to nutrient input reduction as that recorded in Danish coastal waters (Riemann et al., 2016). This likely reflects a slow recovery of light conditions and general environmental conditions including sediment quality, suggesting feed-back mechanisms of the degraded ecosystem in play that maintain a degraded state (e.g., van der Heide et al., 2011; Duarte et al., 2015; Riemann et al., 2016). Kelps are relatively long-lived and have complex life cycles. However, there are examples from Skagerrak of disappearance as well as of recovery of S. latissima stands within 1 year (Moy and Christie, 2012). Therefore, the presence of kelps are scored higher than eelgrass depth limit in IQ5: 1 in the binary system and for S. latissima in this particular example; or 0.5 if in the three level system (ICES, 2013a), because the re-colonization potential depends on distance from source populations. Eelgrass depth limit is scored 0.

#### IQ 6. Quality of Sampling Method: Concrete/Measurable, Accurate, Precise, and Repeatable

Both candidate indicators are concrete/measurable and repeatable. The actual measurement methods involved in the quantifications of the candidate indicators rely solely on adequately measuring depth of seagrass meadows in one case, and identifying kelp species in the other. Both approaches are common enough in the scientific community that IQ6 should be met. Precision in the identification of response to pressure (a requirement defined for this IQ in the present framework) requires addressing factors contributing to the variability in the estimates. Several sources of variability have been tested for eelgrass depth limits, the most important being spatial variability which must be carefully addressed in the planning of monitoring programs (Balsby et al., 2013). As mentioned above, for kelp forest, the factors associated with variability are particularly relevant at the edge of their geographical distributions and this should be considered in any assessment. Given this analysis, both indicators are scored 1 in IQ6.

#### IQ 7. Cost-Effective

Both candidate indicators can be monitored either by diving or by the use of under-water video surveys, the latter speeding up the assessments and, in themselves, serving as documentation for the assessment. The design of monitoring programs can be optimized by combining information on sources of variability and cost assessments, as has been exemplified for eelgrass depth limits (Balsby et al., 2013). Video surveys could be preferred to diver-based surveys, because of the lowering cost of good quality imaging technologies. However, specialized operators are still required to identify the presence of seagrass species, and the acquisition of general habitat information. The presence of kelp is assigned a 0 score in IQ7 because the required effort to acquire data is seen as being too high given the context dependence of pressure-response relationships (see IQ3). Eelgrass depth limits are assigned a score of 1 in the binary system, and 0.5 in the three-way scoring system (ICES, 2013a), because responsiveness to pressure is good but the cost associated with data acquisition is still relatively high.

#### IQ8. Existing and Ongoing Monitoring Data

Data on both candidate indicators have been collected continuously since 1989 as part of the Danish National Aquatic Monitoring and Assessment Programme (DNAMAP) and regional monitoring activities. Therefore, both candidate indicators scored 1 in IQ8.

#### ES6. Sum of Quality Scores

Overall, eelgrass depth limit scored 7, and presence of keystone kelp scored 5 in the binary system. The corresponding scores were 6.5 and 4.5 in the three-way scoring system (**Table 3**). This quality analysis indicates that eelgrass depth limits is the preferable of the two candidate indicators for descriptors 1 and 5 (**Table 3**) responding to nutrient pressure in this case-study area. This results from a better pressure-response relationship and possibilities for target setting for eelgrass depth limits, although the presence of keystone kelps may potentially have better capacity as indicator of system recovery under these descriptors.

#### DISCUSSION

The worked examples (Section Results and Supplementary Information) demonstrate the application of the proposed quality assessment framework for distinct types of candidate indicators and separate descriptors of GEnS. Despite these differences, the application of the framework was possible, and the worked examples are expected to provide guidance in future uses of this tool by highlighting the types of data sought, and how the evaluation steps should work. The structure of the quality assessment is particularly clear in tabular form (**Table 3**, and Tables S2, S3). The joint use of this format in support of the narrative form for reporting of the quality assessment is therefore recommended, because the former enables a quick and objective overview of the assessment process while detail is provided in the latter. This is seen as being particularly useful in the comparison of the quality of candidate indicators for the same descriptor within a region. In these cases, higher quality scoring is preferable because higher scoring within compared candidate indicators highlights which metric meets the MSFD assessment aims more closely.

However, implementation of the highest scoring candidate metric locally may not always be the preferred choice against, for instance, an overall aim to produce a standardized assessment across the MSFD participating parties. Specifically, it is likely that the quality score of individual metrics will vary between countries (and regions) given regional differences in data availability, skill set, costs, and resources available for data collection and analysis, among other constrains. Therefore, this testing framework would best support the decision making process, and indicator selection, if the approach was applied to candidate metrics at least at the country level, and ideally at sub-assessment region level. In this way, it could support a standardized indicator selection process through the determination of which specific candidate metrics score the highest across participating parties for each given descriptor. The clear representation of this quality assessment provides a consistent and objective structure to inform about what desired quality attributes each candidate indicator does or does not meet in each case, and the potential need for specific development in each case. A standardized format for the assessment table could be implemented to facilitate the application of the IQ-ES protocol within the MSFD assessment across the participating parties.

The structure imposed by the IQ-ES framework requires that the quality assessor maintains focus on what each IQ represents, and the provision of information about each assessment in a transparent manner, easily understandable by a third party. These characteristics are seen as being particularly useful in the implementation of the MSFD, in which at least some crossborder use of the same indicators will no doubt be necessary to ensure consistency within a standardized assessment. For instance, this quality assessment protocol (and particularly the tabular reporting of the IQ-ES assessment) is well placed to support the call of the Intersessional Correspondence Group on the Coordination of Biodiversity Assessment and Monitoring of the OSPAR Convention for the Protection of the Marine Environment of the North-East Atlantic, to ensure consistency in the testing of all common indicators. Indeed, the format for testing of candidate biodiversity indicators developed by that group fits well with the assessment structure presented here. In this study, as a starting point, we have applied this testing protocol successfully for three distinct descriptors (1, 5, and 6). Further testing could support its applicability to the other eight descriptors.

Scoring allows for similar indicators to be separated based on an objective analysis of their overall performance with regard to the aims of the MSFD assessment. This would allow MSFD parties considering candidate metrics available to them within their assessment region to determine their readiness to assess each descriptor of GEnS. To ensure continuity of the assessment between involved parties, the scoring system used for the quality assessment should exclude as much as possible user subjectivity, and the binary system used here could be seen as its simplest form. We compared this system with the three-way scoring system (ICES, 2013a,b) within the worked examples. For instance, the two benthic vegetation candidate indicators compared exhibited similar spread using both scoring systems (worked example I). It therefore seems that, despite the relatively higher complexity and subjectivity of the three-way scoring system compared to the binary system, the ability to discriminate quality between candidate metrics did not increase. Further testing could be used to determine the relative merit of the two systems within a wider basis of ecosystem components, descriptors and pressures considered by the MSFD, but our overall assessment is that the binary system would be preferred if the aim is to reduce user subjectivity in the quality evaluation.

Although a standardized approach is seen as being necessary to objectively assess the quality of GEnS indicators in support of the MSFD, additional weight associated with IQs 1 and 3 is acknowledged here ("scientific basis" and "responsiveness to pressure," the one-out-all-out criteria). I.e., failing these IQs is seen here to preclude a failure to meet essential quality standards required for MSFD implementation. We recommend that even when IQs 1 and 3 are fulfilled, an indicator meeting only half or less of the IQs should, however, probably not be considered for implementation, unless no better alternatives exist. Overall, one of the main benefits of using quality scoring is that a minimum score could potentially be defined as the minimum quality standard below which the evaluated metric is not a suitable route to support the MSFD assessment. We suggest that this threshold could be 4 because a candidate indicator with a lower score only meets less than half of the components of quality desired within the assessment community. However, we stress that the use of this framework is not intended to define what is or is not an adequate GEnS indicator or to determine the outcome of the selection procedure, which will be constrained by a number of additional parameters and aims. What the IQ-ES framework provides is a transparent, standardized structure to enable comparison of the quality of candidate indicators and in this way support the decision making process leading to indicator selection.

The objective quality testing protocol suggested here, and the standardized format for the reporting of this assessment we propose, could guide parties seeking better indicators for a given descriptor toward solutions in indicators scoring high in quality in other regions, and further support consistency of the assessment across parties. Through its structure, the use of the IQ-ES framework could help to inform about what types of additional information or method development are lacking within the assessment of individual parties, once local-specific constrains have been identified.

We identify IQs 3 and 4 ("responsiveness to pressure" and "possibility to set targets") as potential stumbling blocks in the quality assessment, and thus the comparison and selection of indicators. The outcomes of the evaluations of these two criteria may be more dependent upon the choice and adequacy of the analytical approaches employed, than on the indicator and data used in those assessments. Issues such as comparability of datasets between systems, the identification of effect sizes that account for natural variability, non-linear pressureresponse relationships, uncertainty and spatial and temporal autocorrelation may require the use of robust quantitative data analysis methods. Generalized additive modeling (Hastie and Tibshirani, 1990), generalized linear models (Dobson, 2001), mixed effects modeling (Pinheiro and Bates, 2000), Meta-analysis statistics (Borenstein et al., 2011), mechanistic modeling and data assimilation (Hyder et al., 2015) and many other methods are therefore likely to be needed in many instances. In addition, high frequency data (e.g., those based on remote sensing) may require the application of suitable techniques such as spectral methods, to identify harmonic structures (Bloomfield, 2004). Whether the analysis technique used is adequate to the complexity of data at hand, the IQ tested for, the scale covered by the analysis (e.g., local c.f. regional), and the resources and expertise available in

#### REFERENCES


Finally, despite its timeliness and contribution toward objectivity within the MSFD indicator selection process, this study is not sufficiently comprehensive to cover the diversity of data, indicator, pressure, and habitat types associated with the 11 GEnS descriptors. However, it highlights important aspects requiring consideration within the assessment, which will only be as good as the indicators chosen and the strategies employed to monitor GEnS. Overall, standardized approaches such as this will be required to ensure consistency, and facilitate cross-border development and the sharing of knowledge during the MSFD implementation.

#### AUTHOR CONTRIBUTIONS

AQ, DK, JS, KM, PS, JB, and JC conceptualized the manuscript. DK, AB, JC, RC, HN, SC, and AQ provided the worked examples. All authors contributed to the text.

#### ACKNOWLEDGMENTS

This study was conceptualized and undertaken through the DEVOTES (Development of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), www.devotes-project.eu. AQ and PS further acknowledge funding support from the Marine Ecosystems Research Programme (jointly funded by the UK Natural Environment Research Council and the UK Department for Environment, Food and Rural Affairs, contract agreement NE/L003279/1).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00073

Informe de la Fundación AZTI para la Dirección de Biodiversidad del Departamento de Ordenación del Territorio y Medio Ambiente, del Gobierno Vasco, p. 115


1: Developing Proposals for Potential Indicators. SAMS/MBA/QUB/UAber for JNCC Peterborough.


estuaries and coasts," in Treatise on Estuarine and Coastal Science, Vol. 7, eds E. Wolanski and D. S. McLusky (Waltham, MA: Academic Press), 53–81.


Teixeira, H., Berg, T., Fürhaupter, K., Uusitalo, L., Papadopoulou, N., Bizsel, K. C., et al. (2014). Existing Biodiversity, Non-Indigenous Species, Food-Web and Seafloor Integrity GEnS Indicators (DEVOTES Deliverable 3.1) DEVOTES FP7 Project. 198. Available online at: http://www. devotes-project.eu

United Nations (1992). Convention on Biological Diversity. Rio de Janeiro, 28.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declares that, despite being affiliated to the same institution as authors NP and MP, the review process was handled objectively.

Copyright © 2016 Queirós, Strong, Mazik, Carstensen, Bruun, Somerfield, Bruhn, Ciavatta, Flo, Bizsel, Özaydinli, Chuševe, Muxika, Nygård, Papadopoulou, Pantazi ˙ and Krause-Jensen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Approach for Supporting Food Web Assessments with Multi-Decadal Phytoplankton Community Analyses—Case Baltic Sea

Sirpa Lehtinen<sup>1</sup> \*, Sanna Suikkanen<sup>1</sup> , Heidi Hällfors <sup>1</sup> , Pirkko Kauppila<sup>1</sup> , Maiju Lehtiniemi <sup>1</sup> , Jarno Tuimala<sup>2</sup> , Laura Uusitalo<sup>1</sup> and Harri Kuosa<sup>1</sup>

*<sup>1</sup> Marine Research Centre, Finnish Environment Institute (SYKE), Helsinki, Finland, <sup>2</sup> Finnish Tax Administration, Helsinki, Finland*

#### Edited by:

*Jacob Carstensen, Aarhus University, Denmark*

#### Reviewed by: *Alberto Basset,*

*University of Salento, Italy Lumi Haraguchi, Aarhus University, Denmark*

\*Correspondence: *Sirpa Lehtinen sirpa.lehtinen@ymparisto.fi*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *14 June 2016* Accepted: *25 October 2016* Published: *10 November 2016*

#### Citation:

*Lehtinen S, Suikkanen S, Hällfors H, Kauppila P, Lehtiniemi M, Tuimala J, Uusitalo L and Kuosa H (2016) Approach for Supporting Food Web Assessments with Multi-Decadal Phytoplankton Community Analyses—Case Baltic Sea. Front. Mar. Sci. 3:220. doi: 10.3389/fmars.2016.00220* Combining the existing knowledge on links between functional characteristics of phytoplankton taxa and food web functioning with the methods from long-term data analysis, we present an approach for using phytoplankton monitoring data to draw conclusions on potential effects of phytoplankton taxonomic composition on the next trophic level. This information can be used as a part of marine food web assessments required by the Marine Strategy Framework Directive of the European Union. In this approach, both contemporary taxonomic composition and recent trends of changes are used to assess their potential consequences for food web functioning. The approach consists of four steps: (1) long-term trend analysis of class-level and total phytoplankton biomass using generalized additive models (GAMs) and calculating average biomass share of each phytoplankton class from the total phytoplankton biomass, (2) comparing the current phytoplankton community composition and its long-term changes with non-metric ordination analysis (NMDS) of genus-level biomass, (3) describing which taxa (the most accurate taxonomic level) are primarily responsible for forming the biomass and for causing the possible changes, and (4) interpretation of the phytoplankton results to assess the potential effects on the next trophic level. Within step 4, special attention is given to the following characteristic of taxa: potential suitability or quality as food for grazers, harmfulness, size, and trophy. These characteristics are selected based on existing scientific knowledge on their relevance to the higher trophic levels. In this article, we present the concept of the suggested approach and demonstrate the phytoplankton analyses with multi-decadal monitoring data from the northern Baltic Sea. We also discuss the future development of the approach toward a food web index by combining or replacing the taxonomic analyses with functional trait-based approaches.

Keywords: food web, phytoplankton community composition, marine strategy framework directive, long-term monitoring, environmental assessment

# INTRODUCTION

In marine pelagic ecosystems, phytoplankton is the key organism group responsible for practically all primary production. In the Marine Strategy Framework Directive of the European Union (MSFD; 2008/56/EC), and subsequent Commission Decision (2010/477/EU), the requirements for assessing the status of marine food webs were set. Looking back in time, the role of phytoplankton as the foundation of food webs was one of the main motivators for the first large-scale phytoplankton investigations undertaken in northern seas, among them the Baltic Sea, already in the early 1900s (cf. Kyle, 1910; Richardson, 2002).

On a general level, primary production, often calculated based on algorithms using surface concentration of chlorophylla derived from remote sensing images on ocean color as an important parameter, is considered to be a good predictor of the potential fisheries yield of the world's oceans (Chassot et al., 2007, 2010). Chassot et al. (2007) found also in the European seas, including the Baltic Sea, a strong linkage between primary productivity (estimated from chlorophyll-a derived from ocean color) and fisheries yield over long time scales from several years to decades. On the other hand, Friedland et al. (2012) found primary production alone to be a poor predictor of global fishery yields, but instead their results showed that chlorophyll-a concentration, particle-export ratio, and the ratio of secondary to primary production were positively associated with yields. However, chlorophyll-a concentration is a proxy for total phytoplankton biomass. It does not indicate taxonomic composition. Phytoplankton biomass may be formed by high or low-quality food or by toxic or nontoxic species, potentially differing greatly from each other as a food source for the higher trophic levels (Olli et al., 1996; Kozlowsky-Suzuki et al., 2003; Uronen et al., 2005; Sopanen et al., 2009). Thus, analyzing phytoplankton community composition reveals the ability of the primary producers to sustain effective trophic transfer, which is the basis for zooplankton and fish growth.

Prey size is one of the primary characteristics which determine the next trophic level (grazers such as mesozooplankton) (Sommer et al., 2000; Katechakis et al., 2002; Stibor et al., 2004). It is known that microzooplankton feed on phytoplankton with cell volumes <500 to 1000µm<sup>3</sup> (Sommer et al., 2005), while copepods are known to feed on both microzooplankton as well as medium to moderately large-sized phytoplankton (100–100,000µm<sup>3</sup> , Sommer and Sommer, 2006). In addition to creating optimal prey size spectrum for different grazers, cell size affects physiological, and ecological processes such as light absorption, nutrient uptake, and sinking (Kriest and Oschlies, 2007; Finkel et al., 2010; Acevedo-Trejos et al., 2015). The dominance of small phytoplankton is the basis for enhanced cycling through the microbial loop and less efficient transfer of production to higher trophic levels (Glibert, 2016).

In addition to cell size, the suitability of phytoplankton as food for the next trophic level is affected by its life form (colonies, filaments etc.), and cell morphology as well as its biochemical properties, e.g., the amino acid, vitamin, sugar, fatty acid, mineral, and toxin content (Koski et al., 1998). A complicating factor is that differences in the presence and concentration of these compounds are partly species-specific (or even strain-specific; Md Amin et al., 2011) and partly related to the physiological state of cells, thus varying with phytoplankton growth rate and cell age (Koski et al., 1998). Different grazer species also react differently to the same phytoplankton food (Engström et al., 2000; Md Amin et al., 2011).

Toxins produced by phytoplankton vary widely in their composition and effects (Granéli and Turner, 2008). In the Baltic Sea, although current knowledge suggests that the transfer rate of phytoplankton toxins through food web is low (Karjalainen et al., 2005, 2007; Setälä et al., 2011), toxic phytoplankton are considered a potential risk for co-occurring organisms, as well as for high-trophic-level consumers through toxin bioaccumulation in the food web (cf. Kuuppo et al., 2006; Sipiä et al., 2006; Setälä et al., 2009, 2014). In the Baltic Sea, phytoplankton toxins have been found in e.g., copepods (Lehtiniemi et al., 2002; Setälä et al., 2009; Sopanen et al., 2011), bivalves (Sipiä et al., 2001; Setälä et al., 2014), Baltic herring, flounder and roach, as well as eider (Sipiä et al., 2006; Karjalainen et al., 2008) with immediate effects of these compounds including reduced feeding and growth rates in fish larvae (Karjalainen et al., 2007), and even mortality in copepods (Sopanen et al., 2008) and fish (Lindholm and Virtanen, 1992). Allelopathy, i.e., the production of allelochemicals which negatively influence the growth and survival of other phytoplankton species, may have an effect on phytoplankton composition and thus affect grazers by modifying the availability of their preferred food (Reigosa et al., 2006).

Mixotrophy is a common feature in phytoplankton, and it is considered to be an important indicator of the efficiency of food webs (Mitra et al., 2014). Mixotrophic phytoplankton is capable of utilizing dissolved and/or particulate organic matter, including bacteria, for their nutrition in addition to phototrophy. Even though a mixotrophy-dominated food web may be more efficient than a traditional phototrophy-based food web in nutrient depleted situations (Mitra et al., 2014), the change from a phytoplankton-based food web toward a bacteria-based food web might yield considerably lower fish productivity (Berglund et al., 2007).

N2-fixation by the diazotrophic cyanobacteria may be an important function for the entire food web (Montoya et al., 2004; Karlson et al., 2015). In the Baltic Sea, it is has been shown that ca. 40–80% of the fixed nitrogen is released as dissolved bioavailable nitrogen for redistribution in the food web (Ohlendieck et al., 2007; Wannicke et al., 2009, 2013; Ploug et al., 2011). Larsson et al. (2001) have estimated that N2-fixation in the Baltic Sea Proper is 180– 430 kt N year−<sup>1</sup> , and this amount would be sufficient to sustain 30–90% of the pelagic net community production during summer. Still, based on results by Olli et al. (2015), the effects of the N2-fixing cyanobacteria on individual cooccurring phytoplankton taxa include both negative and positive effects, with no obvious phylogenetic or functional trait-based patterns.

In this article, we present an approach for using the phytoplankton taxonomic composition on evaluating its potential effects on the next trophic level. The approach consists of four steps: (1) long-term trend analysis of class-level and total phytoplankton biomass using generalized additive models (GAMs) and calculating average biomass share of each phytoplankton class from the total phytoplankton biomass, (2) comparing the current phytoplankton community composition and its long-term changes with non-metric ordination analysis (NMDS) of genus-level biomass, (3) describing which taxa (the most accurate taxonomic level) are primarily responsible for forming the biomass and for causing the possible changes, and (4) interpretation of the phytoplankton results to assess the potential effects on the next trophic level. Potential suitability as food for grazers, harmfulness, cell size, and trophy are the characteristics of the dominant or increased or decreased taxa which are specifically considered when interpreting the results (step 4), based on existing knowledge on their relevance to the next trophic level (e.g., Koski et al., 1998; Sommer et al., 2000; Berglund et al., 2007; Sopanen et al., 2008). Even though we demonstrate the approach with northern Baltic Sea phytoplankton data, the approach can be used for other sea areas as well since the methods are applicable with any long-term data and the functional characteristics which are specifically considered (quality as food, harmfulness, trophy, cell size) are common to all phytoplankton communities.

# MATERIALS AND METHODS

#### Concept

The aim of the present approach is to obtain an overview of the existing phytoplankton community composition and its possible ongoing changes and draw conclusions on their potential effects on the next trophic level in order to use this information as a part of marine food web assessments required by the MSFD. The approach consists of four steps (**Figure 1**). While interpreting the results (step 4), characteristics of taxa which are specifically regarded include potential suitability or quality as food for grazers, harmfulness, size, and trophy. A conceptual model presenting linkages between functional characteristics of phytoplankton taxa and high and low trophic transfer efficiency in pelagic food webs is presented in **Table 1**.

The approach requires quality-checked, comparable long-term quantitative phytoplankton biomass data. By phytoplankton, we mean microscopic planktonic auto- and mixotrophic algae and cyanobacteria which can be recognized using a light microscope (i.e., picoplankton is excluded, and trophy is assigned based on light microscopy). The data should be collected at least yearly from a geographical area (can include several stations) where the phytoplankton community composition and seasonal progression are similar. Only data collected during the same phase of the seasonal succession should be analyzed together to avoid adding seasonal variance in the results. Seasonal period when both phytoplankton and zooplankton are abundant (and trophic coupling between phytoplankton and zooplankton is potentially the highest) should be preferred. The number of samples per year should remain the same in the long-term analyses to ensure equal representation of the years.

#### Data

The proposed approach is demonstrated with Finnish national marine monitoring data collected as part of the HELCOM COMBINE monitoring program (HELCOM, 2015). Phytoplankton samples (n = 286) were collected once a year between July 15th and September 15th in 1979–2014 from 10 offshore monitoring stations situated in the Bothnian Bay, Bothnian Sea, Åland Sea, Gulf of Finland, and northern Baltic Proper (**Figure 2**). The sampling season was late summer, i.e., the period when zooplankton abundance and biomass are the highest in the area (Ojaveer et al., 1998), following the warming of the water and development of thermocline in the surface layer, but before the downwelling period which breaks up the thermocline. The data are stored in the Finnish national database OIVA (http://www.syke.fi/en-US/Open\_information; in Finnish), the ICES database (http://ecosystemdata.ices.dk/ inventory/index.aspx), and the COPEPOD database (http:// www.st.nmfs.noaa.gov/copepod/data/fimr/index.html).

The methodology followed the HELCOM COMBINE manual (HELCOM, 2015): integrated water samples were taken from the surface layer (0–10 m) by mixing equal amounts of water from the depths of 1, 2.5, 5, 7.5, and 10 m. Samples were preserved with acidic Lugol's solution (1 ml per 300 ml sample), and kept refrigerated (+4 to +10◦C) in the dark prior to microscopic analysis within a year of sampling. Microscopy was done with an inverted light microscope using the Utermöhl method (Utermöhl, 1958). A volume of 50 ml (or 25 ml, depending on the density of cells, HELCOM, 2015) of sample was settled in a settling chamber. A magnification of 125x was used to count the species larger than 30µm as well as taxa belonging to the order Nostocales; 250x magnification was used to count the 20–30µm sized species, colonies belonging to the order Chroococcales with a cell size larger than 2µm, as well as taxa belonging to the order Oscillatoriales; and 500x magnification was used to count species smaller than 20µm as well as Chroococcales colonies with cells smaller than 2µm. With each of the three magnifications, 60 ocular squares were analyzed, aiming to count at least 400 counting units with each magnification. Picoplankton (cells <2µm) counting is not possible with this technique.

During microscopic analysis and when converting the counting results into biomass (wet weight µg per liter), the taxon-specific counting units, size classes, and biovolume formulae of the HELCOM PEG (Phytoplankton Expert Group) taxon and biovolume list v. 2014 were used (Olenina et al., 2006; the annually updated biovolume list is available at http://helcom.fi/helcom-at-work/projects/phytoplankton).

Only taxa estimated to be auto- or mixotrophic (based on light microscopy and the HELCOM PEG taxon and biovolume list) were included in the analyses, while heterotrophic taxa, cysts, and benthic taxa (which sporadically occur in the plankton) were excluded. Unidentified <10µm autotrophic monads (unicellular) and flagellates were grouped into "Unidentified." The nomenclature of the HELCOM PEG biovolume list follows that of the World Register of Marine Species (WoRMS, http://www.marinespecies.org/about.php).

TABLE 1 | A conceptual model of the linkage of phytoplankton community properties (defined as functional characteristics) to high and low trophic transfer efficiency in pelagic food webs (DOM = dissolved organic matter).


# Step 1: Class-Level and Total Biomass Trend Analyses

The statistical analyses were performed using the R software (R Core Team, 2014). Time series for phytoplankton class biomasses in each area were analyzed using Generalized Additive Models (GAM, R package "mgcv," Wood, 2014). GAMs are well-suited to analyze long-term trends in phytoplankton biomasses (Hastie and Tibshirani, 1990). A GAM is a generalized linear model with a linear predictor involving a sum of smooth functions of covariates, and by specifying the model only in terms of smooth functions, rather than detailed parametric relationships, it allows for rather flexible specification of the dependence of the response on the covariates (Wood, 2006). Curves estimated with GAM are plotted on the data to visualize the direction of the statistically

significant long-term changes (i.e., decreasing, increasing, or non-linear trends).

We used class-level data for the GAMs since classes combine taxa with some similar characteristics into a convenient number (ca. 10) of groups. The autotrophic endosymbiont-bearing ciliate Mesodinium rubrum was only included in the phytoplankton counts since 1986, and therefore its trend was analyzed only since that year and its biomass was not included into the trend analyses of total phytoplankton biomass. Classes Chlorophyceae and Charophyceae were grouped into phylum Chlorophyta. In addition to the classes, biomass trends of unidentified taxa, and the total phytoplankton biomass were analyzed separately. Biomass data was modeled as annual averages of all stations within a sea area calculated from the late summer samples. The possible autocorrelation between years was modeled with AR1 (autocorrelation structure with lag 1). Curves estimated with GAM were plotted on the data for visually checking the direction of the significant long-term changes (plots not shown). The average total phytoplankton biomass and average biomass share (%) of each phytoplankton class from the total phytoplankton biomass was calculated based on the whole long-term data set (1979–2014), except for M. rubrum, whose average biomass share was calculated using the total phytoplankton biomass (including M. rubrum) during 1986–2014.

#### Step 2: Genus-Level Community Analysis

The Non-metric Multidimensional Scaling (NMDS, function metaMDS, R package "vegan," Oksanen et al., 2016) was used to make a visual ordination of samples based on the similarities and dissimilarities in the genus-level phytoplankton community composition. NMDS is commonly considered as the most robust unconstrained ordination method in community ecology (Legendre and Legendre, 1998; McCune and Grace, 2002). NMDS projects the observed community dissimilarities nonlinearly onto an n-dimensional (usually 2-dimensional) ordination space and it can handle nonlinear taxon responses. NMDS visualizes the phytoplankton community composition by positioning the samples in the ordination space based on their taxon-specific biomass composition. The names of the taxa characterizing the samples can likewise be plotted. The NMDS ordination graphs thus give an overview of the phytoplankton community composition and its spatio-temporal changes, to support the results of GAMs which reveal changes in the biomasses of different phytoplankton classes separately.

We used genus-level biomass data for the NMDS, since consistent species-level identification is not always possible and genus-level data may be more robust to differences in skill and effort among the individual phytoplankton analysts. Genus-level data is also recommended over class-level data since notable genus-level changes may occur even though class biomasses and their shares remain unchanged. Since some genera were not identified consistently by the different microscopists, they were grouped into order-level or into a taxa complex for the NMDS, even though they had been stored into the OIVA database separately: all cryptophyte genera were grouped into the order Cryptomonadales, all genera belonging to the order Chroococcales except for the genera Snowella and Woronichinia were grouped into the order Chroococcales, all genera belonging to the order Ochromonadales were grouped under the order name, and the genera Koliella, Monoraphidium, and Nephrodiella were collectively named the "Monoraphidium complex." Due to the properties of the community analyses (Legendre and Legendre, 1998) i.e., in order to improve the comparability of the data across the time series and to avoid that sporadically occurring genera confuse the results, genera which were present in less than 5% of the samples (with very low biomasses in all cases) were excluded from the NMDS analyses, resulting in a total of 53 taxa (genera, orders, and complexes) included in the analysis. Taxa which were excluded from the NMDS were acknowledged within the step 3 (most accurate taxonomic level examination). Biomass values were square-root transformed, and the Bray-Curtis dissimilarity was used as the distance metric.

# Step 3: The Examination of the Dominant Taxa on the Most Accurate Taxonomic Level

The most accurate taxonomic level data was analyzed by simple biomass ratio analyses showing which taxa dominate the biomass of each phytoplankton class (step 1, GAMs). The role of the dominant taxa in each class was confirmed by running a separate GAM for these taxa to see if the result agreed with that of the total class. Using the conventional methods of phytoplankton monitoring (i.e., light microscopic analysis of preserved samples), not all taxa can be determined to species level, and thus it was necessary for the analyses to consider some higher than species-level taxa in the same manner as the actual species. Within step 3, taxonomic level was anyways more detailed than in the community analysis (step 2, NMDS) to be able to acknowledge, e.g., only sporadically occurring taxa.

#### Step 4: Interpretation of Results

Within interpretation, all results from steps 1–3 are considered. The taxon-specific (mostly species-specific) characteristics specifically considered when interpreting the results are the potential quality as a food source for grazers, harmfulness, size, and trophy (**Table 1**). Since these characteristics may be affected by even the life stage of the cells or vary within strains, only the potential of taxa to possess the characteristics can be considered when interpreting the results. If the class-level GAM results were based primarily on taxa for which there exists knowledge on these functional properties, the statistically significant long-term trends (p < 0.05) may be used to indicate if the ongoing changes are positive or negative for grazers. For taxa which are considered low-quality food, as well as for taxa which are potentially harmful or toxic to other organisms of the food web, the preferred trend is "decreasing or no change," while for taxa which are considered high-quality food the preferred trend is "increasing or no change."

Careful interpretation of the results is important. Even though communities differ geographically and with seasons, the same types of characteristics (quality as food, harmfulness, trophy, size) are common to all phytoplankton communities. Factors possibly affecting the phytoplankton community or causing changes in it are not studied within the approach, but existing studies on physical, chemical, and other biological data can be discussed.

# RESULTS

# Step 1: Class-Level and Total Biomass Trends

In our demonstration data set from the northern Baltic Sea, the average total phytoplankton biomass during the study period (1979–2014) was the lowest in the Bothnian Bay (191 ± 267µg l −1 , mean ± S.D.), and the highest in the Gulf of Finland (average 520 ± 483µg l−<sup>1</sup> ). The average total phytoplankton biomass was 427 ± 355µg l−<sup>1</sup> in the northern Baltic Proper, 294 ± 212µg l−<sup>1</sup> in the Bothnian Sea, and 365 ± 159µg l −1 in the Åland Sea. The Bothnian Bay differed from the other areas also based on its phytoplankton composition. For example, the average share of cyanobacteria was there only ca. 2% of the total phytoplankton biomass, while the average share of cyanobacteria was ca. 27–37% in the other sea areas (**Table 2**).

The class-level data was analyzed for long-term trends in each of the five sea areas using GAMs and the results are summarized in **Table 2**. Statistically significant increasing trends were found for cyanobacteria (class Nostocophyceae) in the Bothnian Sea, Åland Sea and the Gulf of Finland, for prymnesiophytes (class Prymnesiophyceae) in all sea areas but the Bothnian Sea, euglenophytes (class Euglenophyceae) in the Åland Sea, and prasinophytes (class Prasinophyceae) in the northern Baltic Proper. The autotrophic ciliate M. rubrum increased in the Bothnian Sea and northern Baltic Proper. Cryptophytes (class Cryptophyceae) decreased in all sea areas except the Bothnian Sea, and diatoms (class Diatomophyceae) in the Bothnian Bay. The biomass of unidentified taxa decreased in all sea areas, and biomass of total phytoplankton in the Bothnian Bay. Statistically significant, but non-linear variability was shown by diatoms and prasinophytes in the Bothnian Sea.


TABLE 2 | Results of the generalized additive models (GAMs) for detection of long-term trends (p-values; bold = significant trend, p < 0.05; direction: blue, decreasing; red, increasing; purple, non-linear).

*The average biomass share (%) of each phytoplankton class from the total phytoplankton biomass is also given. Samples were collected from the Finnish HELCOM COMBINE offshore monitoring stations once a year between July 15th and September 15th in 1979–2014. The recording of Mesodinium rubrum started in 1986, and thus trends in its biomass were calculated for the period 1986–2014, and the species is not included in the biomass of the total phytoplankton community, except for calculation of its biomass share from the total phytoplankton biomass (including M. rubrum).*

*<sup>a</sup>Biomass trends for Mesodinium rubrum cover the period 1986–2014 only.*

*n* = *number of sampling years (*\* = *number of sampling years for Mesodinium rubrum).*

#### Step 2: Genus-Level Community Changes

Based on the NMDS analysis, community composition was clearly different only in the Bothnian Bay compared to the other sea areas (**Figure 3**). Chlorophyte (phylum Chlorophyta in the GAM) genera Desmodesmus, Elakatothrix, Dictyosphaerium, and Botryococcus, as well as the diatoms (class Diatomophyceae) Diatoma and Skeletonema characterized the Bothnian Bay samples. Nevertheless, the composition changed simultaneously in the same direction during the study period 1979–2014 in all sea areas (**Figure 3**). The genera Aphanizomenon, Nodularia, Chrysochromulina, and Cryptomonas were shown to be primarily responsible for the biomass formation and the statistically significant trends of cyanobacteria (class Nostocophyceae in GAM), prymnesiophytes (class Prymnesiophyceae), and cryptophytes (class Cryptophyceae), respectively (**Tables 1**, **2**).

#### Step 3: Most Accurate Taxonomic Level Examination

The taxa primarily responsible for the biomass formation and the statistically significant trends in each class are listed in **Table 3**. Most important characteristics of the taxa are also included in the **Table 3**.

#### Step 4: Interpretation

Total phytoplankton biomass decreased in the Bothnian Bay, but other significant trends in the total phytoplankton biomass were not observed (**Table 2**). In addition to the lowest total phytoplankton biomass, the Bothnian Bay differed from the other areas also based on its phytoplankton composition (**Table 2**). However, the community analysis demonstrated an ongoing change toward the same direction in all five sea areas, also in the Bothnian Bay (**Figure 3**). Suikkanen et al. (2013) found a significant increasing trend for chlorophyll-a concentration during the study period 1979–2011 in the same monitoring stations situated in the northern Baltic Proper, Gulf of Finland, and Åland Sea. In the Bothnian Sea, there was a significant increasing trend for chlorophyll-a (GAM, p < 0.001, n = 27) during 1979–2012 (unpublished data), In the Bothnian Bay, no trend in chlorophyll-a was observed (GAM, p = 0.101, n = 27) during 1979–2012 (unpublished data), Thus, our results showed no trends for total phytoplankton biomass (excluding picoplankton) in areas where chlorophyll-a increased, and a decreasing trend for total phytoplankton biomass for the Bothnian Bay where chlorophyll-a showed no trend.

Of the classes with statistically significant long-term changes, cyanobacteria, prymnesiophytes, and cryptophytes are the ones with potentially the most important food web effects in terms of harmfulness, food quality, and trophy in our study area. Both species of cyanobacteria, Aphanizomenon flosaquae and Nodularia spumigena, primarily responsible for the observed increasing trends of the class Nostocophyceae are N2-fixing, i.e., diazotrophic (**Table 3**). N. spumigena produces hepatotoxin, nodularin, which accumulates in the pelagic and benthic food web and are toxic for mammals (Sipiä et al., 2001; Karjalainen et al., 2007; Sopanen et al., 2009; Karlson and Mozuraitis, 2011), while the Baltic Sea isolates of Aphanizomenon have proven nontoxic, despite the toxicity of several freshwater strains (Lehtimaki et al., 1997).

The most important genus explaining the increasing trends in prymnesiophytes, Chrysochromulina spp. sensu lato, includes potentially harmful algal bloom species which can form fishkilling ichtyotoxins as well as allelopathic substances which

FIGURE 3 | A demonstration of the non-metric multidimensional scaling (NMDS) results based on northern Baltic Sea phytoplankton monitoring data. NMDS was used to cluster samples (A) based on genus-level biomass composition. The 53 genera (and orders and complexes) that the analysis is based on are plotted separately for clarity (B). The color scale represents sampling years from 1979 (red) to 2014 (blue). The HELCOM sea areas investigated were BOB, Bothnian Bay; BOS, Bothnian Sea; GOF, Gulf of Finland; NBP, Northern Baltic Proper; ÅS, Åland Sea. Taxa: ACTI, *Actinocyclus*; AKSH, *Akashiwo*; AMPH, *Amphidinium*; APHA, *Aphanizomenon*; BACI, Bacillariales; BOTR, *Botryococcus*; CHAE, *Chaetoceros*; CHROO, Chroococcales; CHRYROM, *Chrysochromulina*; CRYPTO, Cryptomonadales; CYCL, *Cyclotella*; CYLI, *Cylindrotheca*; DESM, *Desmodesmus*; DIAT, *Diatoma*; DICT, *Dictyosphaerium*; DINB, *Dinobryon*; DPHYS, *Dinophysis*; DOLI, *Dolichospermum*; ELAK, *Elakatothrix*; EUPO, Eupodiscales; EUTR, *Eutreptiella*; GLEN, *Glenodinium*; GONY, *Gonyaulax*; GLES, Gymnodiniales; GYMN, *Gymnodinium*; GYRO, *Gyrodinium*; HETE, *Heterocapsa*; MANT, *Mantoniella*; MICR, *Micromonas*; MCPLX, *Monoraphidium* complex; NEPH, *Nephroselmis*; NITZ, *Nitzschia*; NODU, *Nodularia*; OCHR, Ochromonadales; OLLI, *Ollicola*; OOCY, *Oocystis*; OSCI, Oscillatoriales; PERLES, Peridiniales; PLNE, *Planctonema*; PLNG, *Planktolyngbya*; PROC, *Prochlorothrix*; PROR, *Prorocentrum*; PROT, *Protoceratium*; PSAN, *Pseudanabaena*; PSELLA, *Pseudopedinella*; PSFI, *Pseudoscourfieldia*; PYRA, *Pyramimonas*; SKEL, *Skeletonema*; SNOW, *Snowella*; THAL, *Thalassiosira*; UNID, Unidentified monads and nanoflagellates; UROG, *Uroglena*; WORO, *Woronichinia*.

are harmful for other phytoplankton species (Reigosa et al., 2006; Granéli and Turner, 2008). In case of toxicity, we used the precautionary principle, i.e., expecting that taxa including potentially toxic strains may be toxic even though we cannot define from the monitoring data if the toxicity was actually present in the community. Another important group of phycotoxin producers is dinoflagellates, but their biomass did not show any statistically significant late-summer trends.

In addition to the increasing risk of potential harmful algal bloom effects in the ecosystem, the observed phytoplankton community changes can have direct food web effects through the changes in the food quality for micro- and mesozooplankton grazers. Cyanobacteria and prymnesiophytes have been shown to be low-quality food for herbivorous zooplankton (de Bernardi and Giussani, 1990; Sopanen et al., 2008), while cryptophytes, which decreased in most of the study area, are considered highquality food (Lehman and Sandgren, 1985). On the other hand, the cyanobacterium N. spumigena is known to be a good thiamine source for zooplankton (Sylvander et al., 2013), and thus optimal food may contain a small share of it.

Since Chrysochromulina spp. sensu lato includes mixotrophic species, its increase may indicate a shift from an autotrophic, phytoplankton-based food web toward a more mixotrophic, bacteria-based food web. The reason for increasing mixotrophy (importance of the microbial loop) may be either availability of extra energy to the food web due to additional dissolved matter from land, or less efficient food web functioning if the dissolved matter originates from the food web (e.g., direct DOM excretion, decomposition of cyanobacterial blooms, "sloppy feeding" of zooplankton). Based on a study by Berglund et al. (2007), a shift toward a more bacteria-based food web may reduce pelagic productivity at higher trophic levels in the Baltic Sea, since in the bacteria-based food web carbon passes additional trophic levels through flagellates and ciliates before reaching mesozooplankton, while in the phytoplankton-based food web there is a direct pathway from phytoplankton to mesozooplankton.

In the demonstration, all five sea areas were analyzed together in the NMDS to point out that the phytoplankton community composition is quite similar in all studied offshore areas except in the Bothnian Bay, but the ongoing community change was toward the same direction in all five sea areas. The comparison of results of the trend analyses and the community analysis showed that taxa with statistically significant GAM trends (**Table 3**) were located quite in the middle of the NMDS ordination plot (genus-level, **Figure 3**) suggesting that their importance in the study area as a whole has not changed markedly during the study period despite the distinct significant increase or decrease in their biomass in particular sea areas. Thus, there is obviously an ongoing phytoplankton community change in the northern Baltic Sea area which cannot be fully explained by changes in biomasses of single taxa in the different sea areas. Based on the recent study by Suikkanen et al. (2013), ongoing changes in the northern Baltic Proper, Gulf of Finland, and Åland Sea are most probably due to complex interactions


TABLE 3 | Taxa causing the statistically significant trends shown in Table 2 (Colors are explained in Table 2).

*For each class with significant trends according to GAM, the share (%) of main taxa of the total class biomass in each sea area is indicated, followed by the p-value of the GAM run for that individual taxon. Some characteristics of the main taxa are also listed.*

between warming, eutrophication and increased top-down pressure.

In conclusion, in the Baltic Sea phytoplankton, certain taxonomical groups have a direct link to functional characteristics. Cyanobacteria and prymnesiophytes are low-quality food and potentially harmful, and cryptophytes are considered high-quality food. The community analysis (**Figure 3**) and some trends (**Tables 2**, **3**) in our data show an ongoing change into an unsatisfactory direction. In the next EU MSFD assessment in 2018, phytoplankton class-level trends with statistically significant p-values in the offshore Gulf of Finland, the Åland Sea, and the northern Baltic Proper should be negative (instead of the current positive) for cyanobacteria and prymnesiophytes, and positive (instead of the current negative) for cryptophytes. In the Bothnian Sea, the trend for cyanobacteria should be negative (instead of the current positive) and new unwanted changes should not appear. In the Bothnian Bay, the trend for prymnesiophytes should be negative (instead of the current positive), and the trend for cryptophytes should be positive (instead of the current negative), and new unwanted changes should not appear. In addition, the results of the community analysis should also be supportive for the results of the trend analyses in 2018.

# DISCUSSION

#### Motivation for the Approach

In this article, we present a novel approach for using the phytoplankton taxonomic community composition to draw conclusions on its potential effects on the next trophic level, the goal being to facilitate the use of this information as a part of the assessment of the structure and functioning of the pelagic marine food web as required by the Marine Strategy Framework Directive (MSFD) of the European Union. Within this approach, a number of phytoplankton properties (potential suitability or quality as food for grazers, harmfulness, size, trophy) can be used to assess the potential efficiency of the pelagic food web, which cannot be deducted from other monitoring data. This supplements the currently insufficiently utilized bottomup approach, which can then be combined with the results of the present zooplankton indicators (Gorokhova et al., 2015) for a more holistic assessment (cf. Gowen et al., 2011; Pyhälä et al., 2014). The analyses of pressures and management options will follow the holistic analysis. Developing phytoplankton indicators has proven to be challenging (HELCOM, 2013), but it is definitely necessary at least for the food web assessments (Rogers et al., 2010). Currently, an indicator based on phytoplankton community composition does not exist in the Baltic Sea area, instead chlorophyll-a concentration is the only phytoplankton-based indicator used to assess the environmental status in the Baltic Sea (HELCOM core indicators, http://www.helcom.fi/baltic-sea-trends/indicators/).

# Evaluation of the Strengths and Weaknesses

The main strength of the present approach is the possibility of applying it to all kinds of quantitative phytoplankton biomass data (as long as data within one analysis follow harmonized methods and taxonomy), since the approach does not include ready-made presumptions of any certain indicator taxa or taxonomic groups forming life forms (Tett et al., 2008) or size categories (Lugoli et al., 2012; Roselli and Basset, 2015). Instead, we point out some functional characteristics which should be considered. Those functional characteristics (potential suitability as food for grazers, harmfulness, size, trophy) are common to all phytoplankton communities, and were selected based on existing knowledge on their relevance to the next trophic level (e.g., Koski et al., 1998; Sommer et al., 2000; Berglund et al., 2007; Sopanen et al., 2008). Using these functional characteristics within the interpretation of the taxonomic results is novel compared to some other recent approaches on analyzing longterm phytoplankton monitoring data (e.g., Suikkanen et al., 2013; Godhe et al., 2015; Haraguchi et al., 2015). Finally, the simple analyses can be done using the freely available R software. The only slight downside of the presented approach is that it will never be an "insert data, push the button, and get the results" type of an indicator: since the assumptions concerning the phytoplankton community composition are not fixed, interpretation of the results is an extremely important part of the approach and requires expert knowledge on local phytoplankton ecology.

Reporting consistent and detailed metadata and complementary information of the procedures enables selecting comparable data for the analyses (Zingone et al., 2015). Sampling, preservation, storage, analysis, taxonomical identification, nomenclature, and biomass calculation need to follow the same procedures throughout the data used in an analysis. In the Baltic Sea area, using phytoplankton monitoring data is feasible since harmonized methods for sampling, microscopy, and biomass calculations developed within the HELCOM PEG group are followed in most of the surrounding countries (HELCOM, 2015). Within the Baltic Sea area, microscopists partaking in HELCOM monitoring are trained annually in the HELCOM PEG workshops, and they participate regularly in species identification and counting proficiency tests (e.g., Vuorio et al., 2015). This is important since in a study including seven European sea areas, the main proportion of the recorded variation between cell densities was explained by the variation between the taxonomists counting the samples (Dromph et al., 2013). In Europe, also the Biological Effects Quality Assurance in Monitoring (BEQUALM) program, using the scheme developed by the UK National Marine Biological Analytical Quality Control (NMBAQC), develops quality standards for community structure analysis and organizes phytoplankton proficiency tests.

When performing the analyses for the first time for an area, a multi-decadal data should be used whenever possible, in order to facilitate distinguishing actual trends from interannual variation. Long-term analyses may also enable detecting a period or periods when community composition changed abruptly, indicating possible regime shifts (Möllmann et al., 2015). In addition, multi-decadal data series may in some cases help to estimate the community composition during the time when it was less affected by anthropogenic activities (i.e., being more close to reference conditions or pristine status). A suitable updating frequency of the analyses of presented approach is at least not shorter than 6 years, in accordance with the reporting period of the EU MSFD. When estimating how many years of monitoring data are required for the analyses, it should be considered that single samples are only random fractions representing the continuously fluctuating and dynamic phytoplankton community (Dromph et al., 2013). Thus, low sampling frequency may be a weakness when using phytoplankton monitoring data in assessments. In our demonstration data, the sampling frequency was only once per year but the study period was as long as 36 years. A higher sampling frequency would possibly allow detecting changes already within a shorter monitoring period. Sampling should cover the periods of tightest coupling between phytoplankton and grazers. In the northern Baltic Sea, for example, late summer is the period of the highest zooplankton productivity (Ojaveer et al., 1998) and therefore the season to be focused on.

Offshore and coastal areas should be analyzed separately, because phytoplankton composition in coastal waters usually differs from that in the open sea (e.g., Griffiths et al., 2016). In coastal areas, environmental conditions as well as phytoplankton communities may vary significantly within short distances (Griffiths et al., 2016), and thus it needs to be considered if coastal phytoplankton communities should be analyzed separately even for each station. Data from different offshore stations located within the same sea area may be analyzed together to describe community changes in the area. In that case, annual biomass averages for each season and sea area can be used in the trend analyses.

We recommend using phytoplankton biomass (wet weight per volume) as the input for the analyses because it is often more relevant from the food web perspective than abundance (counting units per volume). The size of different phytoplankton species, and consequently the biovolume of the food sources, varies considerably, which is not evident when using abundance data. Furthermore, biomass data are conveniently converted into carbon biomass data (Menden-Deuer and Lessard, 2000), which are usually utilized in food web models (e.g., Lignell et al., 2013).

The results of trend analyses (GAMs) and community analyses (NMDS) should be interpreted together since their results are complementary to each other and may reveal different aspects. Trend analyses study each taxon separately while community analyses aim at a more holistic view. The reason for using different taxonomic levels in the analyses is due to differing properties of the analyses.

If the possible bottom-up and top-down factors (e.g., Ware and Thomson, 2005; Casini et al., 2008; Prowe et al., 2012) affecting the phytoplankton community are to be discussed within the interpretation of the results (step 4), existing knowledge on those is needed. However, this is not a requirement for using the suggested approach since the analyses of pressures and management options should follow only after a holistic analysis including also other compartments (physical, chemical, and biological) in addition to phytoplankton community composition.

#### Northern Baltic Sea As an Example Area

Northern Baltic Sea was selected as an example area, since there is almost 40 years of phytoplankton monitoring data from that area and its ecology and phytoplankton and zooplankton dynamics are well studied (e.g., Wulff et al., 2001). Recent studies have reported long-term changes in the Baltic Sea phytoplankton and zooplankton communities (Suikkanen et al., 2013), Secchi depth (Dupont and Asknes, 2014), and several physical, chemical, and biological parameters (Lennartz et al., 2014). Changes have been linked to interactions between warming, eutrophication, and increased top-down pressure (e.g., Suikkanen et al., 2013; Elmgren et al., 2015). Despite the special characteristics (brackish water, clear seasonal succession) of the Baltic Sea, it is a suitable sea area for the demonstration since the functional characteristics which are specifically considered within the suggested approach (quality as food, harmfulness, trophy, cell size) are common to all phytoplankton communities, also for the northern Baltic Sea.

#### Future Development

The next step will be to compile information on food quality traits, test different trait-based methods (Litchman and Klausmeier, 2008; Litchman et al., 2012, 2015; Barton et al., 2013; Edwards et al., 2015), and finally develop a widely applicable phytoplankton community composition index based on the functional properties. To be able to proceed in this, quantitative

#### REFERENCES


information on the biochemical properties of phytoplankton taxa as well as on the specific nutritional needs of the higher trophic levels is required, including information on direct toxicity and harmfulness. Species-specific trait analysis should be supplemented with detailed cell size structure analysis, since pelagic predator-prey size ratios are variable (Hansen et al., 1994; Wirtz, 2012). Based on the results of our demonstration and earlier studies with different approaches and end results (e.g., Berglund et al., 2007; Mitra et al., 2014; Hoikkala et al., 2015), food web modeling would be extremely beneficial for understanding food web interactions connected to auto- and mixotrophy and optimal grazer feeding dynamics.

#### AUTHOR CONTRIBUTIONS

Each of the authors has contributed in writing this manuscript. Responsibility on planning was mainly on SL, SS, HH, PK, ML, LU, and HK. Data processing and statistical analyses were performed by SL, SS, and JT.

#### ACKNOWLEDGMENTS

We thank personnel of the Finnish Institute of Marine Research and Finnish Environment Institute for collection and analysis of the phytoplankton monitoring samples. This work was supported by the DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status, www.devotes-project.eu) project, funded by the European Union under the 7th Framework Programme, 'The Ocean of Tomorrow' Theme (grant agreement no. 308392), Academy of Finland research grant 259357, and BIO-C3 (Biodiversity changes -causes, consequences and management implications, www.bio-c3.eu) project, belonging to BONUS, the joint Baltic Sea research and development programme (Art 185), funded jointly by the Academy of Finland (grant number call 2012-107) and by the European Union's 7th Framework Programme for research, technological development and demonstration. We thank the two reviewers for their constructive comments, which improved the manuscript.


production in Baltic food webs. Ambio 44, 413–426. doi: 10.1007/s13280-015- 0660-x


assessment of ecological status in marine and transitional waters. Ecol. Indic. 23, 338–355. doi: 10.1016/j.ecolind.2012.03.030


Zingone, A., Harrison, P. J., Kraberg, A., Lehtinen, S., McQuatters-Gollop, A., O'Brien, T., et al. (2015). Increasing the quality, comparability and accessibility of phytoplankton species composition time-series data. Estuar. Coast. Shelf Sci. 162, 151–160. doi: 10.1016/j.ecss.2015.05.024

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer LH and handling Editor declared their shared affiliation, and the handling Editor states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Lehtinen, Suikkanen, Hällfors, Kauppila, Lehtiniemi, Tuimala, Uusitalo and Kuosa. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Light Thresholds to Prevent Dredging Impacts on the Great Barrier Reef Seagrass, Zostera muelleri ssp. capricorni

Kathryn M. Chartrand1, 2 \*, Catherine V. Bryant <sup>1</sup> , Alex B. Carter <sup>1</sup> , Peter J. Ralph<sup>2</sup> and Michael A. Rasheed<sup>1</sup>

*<sup>1</sup> Centre for Tropical Water and Aquatic Ecosystem Research, James Cook University, Cairns, QLD, Australia, <sup>2</sup> Plant Functional Biology and Climate Change Cluster, University of Technology Sydney, Sydney, NSW, Australia*

#### Edited by:

*Jacob Carstensen, Aarhus University, Denmark*

#### Reviewed by:

*Nomiki Simboura, Hellenic Centre for Marine Research, Greece Nuria Marba, Consejo Superior de Investigaciones Cientificas, Spain Peter Anton Staehr, Aarhus University, Denmark*

> \*Correspondence: *Kathryn M. Chartrand Katie.Chartrand@jcu.edu.au*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *14 April 2016* Accepted: *08 June 2016* Published: *08 July 2016*

#### Citation:

*Chartrand KM, Bryant CV, Carter AB, Ralph PJ and Rasheed MA (2016) Light Thresholds to Prevent Dredging Impacts on the Great Barrier Reef Seagrass, Zostera muelleri ssp. capricorni. Front. Mar. Sci. 3:106. doi: 10.3389/fmars.2016.00106* Coastal seagrass habitats are at risk from a range of anthropogenic activities that modify the natural light environment, including dredging activities associated with coastal and port developments. On Australia's east coast, the tropical seagrass *Zostera muelleri* ssp. *capricorni* dominates intertidal mudbanks in sheltered embayments which are also preferred locations for harbors and port facilities. Dredging to establish and maintain shipping channels in these areas can degrade water quality and diminish light conditions that are required for seagrass growth. Based on this potential conflict, we simulated *in-situ* light attenuation events to measure effects on *Z. muelleri* ssp. *capricorni* condition. Semi-annual *in situ* shading studies conducted over 3 years were used to quantify the impact of prolonged light reduction on seagrass morphometrics (biomass, percent cover, and shoot density). Experimental manipulations were complimented with an assessment of 46 months of light history and concurrent natural seagrass change at the study site in Gladstone Harbour. There was a clear light-dependent effect on seagrass morphometrics during seagrass growing seasons, but no effect during senescent periods. Significant seagrass declines occurred between 4 and 8 weeks after shading during the growing seasons with light maintained in the range of 4–5 mol photons m−<sup>2</sup> d −1 . Sensitivity to shading declined when applied in 2-week intervals (fortnightly) rather than continuous over the same period. Field observations were correlated to manipulative experiments to derive an applied threshold of 6 mol photons m−<sup>2</sup> d <sup>−</sup><sup>1</sup> which formed the basis of a reactive light-based management strategy which has been successfully implemented to ensure positive ecological outcomes for seagrass during a large-scale dredging program.

Keywords: seagrass, shading, light attenuation, thresholds, dredging management, Zostera muelleri, indicators

# INTRODUCTION

Seagrasses cover 38,079 km<sup>2</sup> of habitat on Australia's east coast within the boundary of the Great Barrier Reef World Heritage Area (GBRWHA; Coles et al., 2015). Coastal seagrasses are an integral part of the health and ecosystem function of the GBRWHA and provide key habitat linkages, feeding grounds for globally threatened turtles and dugong, habitat for commercially important fisheries, sediment trapping and stabilization, effective nutrient filtering from coastal inputs, and carbon sequestration (Hemminga and Duarte, 2000; Jackson et al., 2001; Orth et al., 2006; Romero et al., 2006; Heck et al., 2008; Duarte et al., 2010). Despite being highly valued globally for their contribution to ecosystem services, seagrass habitats are threatened by a range of anthropogenic activities including coastal development and declining water quality from poor catchment management activities (Waycott et al., 2009; Grech et al., 2012; Costanza et al., 2014). Anthropogenic pressures on seagrasses are often compounded by natural events such as severe storms and flooding that may cumulatively lead to widespread seagrass decline. This has occurred on the tropical and subtropical east coast of Australia where severe tropical storms have contributed to widespread seagrass declines in recent years (Devlin et al., 2012; Rasheed et al., 2014).

A major cause of seagrass losses globally relates to human induced changes to the inshore environment that reduce available light, the primary driver of seagrass growth and distribution (Dennison, 1987; Duarte, 1991; Ralph et al., 2007). The risk of these types of impacts along the Great Barrier Reef (GBR) coast tends to be highest in areas where urban development and port infrastructure have a strong foothold (Grech et al., 2011). In the GBRWHA, extensive seagrass meadows commonly occur in proximity to large port facilities (Grech and Coles, 2010). Recent, well-publicized port expansions (BREE, 2012; Grech et al., 2013) place adjacent seagrass meadows under increased pressure. The capital works required for port developments can include largescale dredging programs, which can have negative impacts on seagrass through direct burial and/or physical removal, and indirectly from turbidity plumes and the associated reduction in available light (Erftemeijer and Robin Lewis, 2006). In the GBRWHA, recent studies have shown that these plumes can have a substantial impact on seagrass (York et al., 2015). While physical damage to seagrass is relatively easy to quantify or directly avoid, it is the potential for large and persistent sediment plumes which are much harder to effectively forecast the scale of impact or to mitigate against seagrass loss.

The impact of dredge plumes are typically managed using measures not directly related to the ecological requirements of marine plants, such as reference to a background level of turbidity (Sofonia and Unsworth, 2010). Using the plant's light requirements to ensure minimal impacts is seldom attempted, largely due to a lack of understanding on what the in situ light requirements are for most seagrass species (Ralph et al., 2007). Turbidity can provide a measure of added pressure from dredging activity to the ecosystem, but does not necessarily have any direct biological relevance or account for the in-built resilience of an organism or whole system over short timescales (Sofonia and Unsworth, 2010). Adopting a direct measure of available light as a threshold for seagrass management is directly related to the plant's growth requirements making it far more preferable to turbidity.

Determining an appropriate light threshold for seagrasses involves several challenges: the light environment can be naturally highly variable over multiple timescales; plants can have dramatically different light requirements depending on time of year (Staehr and Borum, 2011); seagrasses can tolerate periods of time below their minimum light requirement without long-term impacts; and a range of other environmental parameters including water temperature and sediment chemistry can further influence in situ light requirements (Koch, 2001; Lee et al., 2007). The plant response to fluctuating light begins with explicit gene regulation driving changes in photosystems and pigment composition before growth rates and eventual plant morphology or meadow scale reductions become apparent (Abal et al., 1994; Collier C. J. et al., 2012). While laboratory experiments have helped to resolve the fundamental timeline of many of these responses (Abal et al., 1994; Collier C. J. et al., 2012; McMahon et al., 2013), the actual timeline of in situ seagrass growth dynamics is likely to be quite different due to additional extrinsic factors that cannot easily be replicated in laboratory or mesocosm trials such as nutrient availability, water temperature, hydrodynamics, epiphyte loads, water column oxygen fluxes and sediment chemistry (Carruthers et al., 2002; Waycott et al., 2005; Raun and Borum, 2013). In situ shading studies provide an empirical approach to measuring impacts of prolonged incident light attenuation and identify potential warning signs of decline in meadow-scale seagrass health as related to dredging or other anthropogenicinduced light reduction under realistic field conditions (Longstaff and Dennison, 1999; Collier C. et al., 2012).

Identifying the relevant timeframe to elicit a negative response by local seagrasses is a key component of developing a regionallyspecific light threshold. Most seagrasses can tolerate periods of time below their minimum light requirement without longterm impacts (Alcoverro et al., 1999; Collier C. J. et al., 2012). Short-term re-allocation of carbon from storage tissues and adjustments to photosynthetic machinery can help bide time until conditions improve (Alcoverro et al., 2001; Cayabyab and Enríquez, 2007). A light threshold must establish the juncture at which compensatory physiological mechanisms are superseded by plant-scale declines (Collier C. J. et al., 2012). An applied light management strategy must consider the light quantity, quality and duration of light that is required to sustain local seagrass populations.

Many coastal seagrass species are well-adapted to the variable conditions that occur in a near-shore environment, including naturally turbid waters related to runoff, large tidal fluxes, complex hydrodynamics and oscillating temperatures creating constantly shifting optical and metabolic challenges (de los Santos et al., 2010; Collier et al., 2011; Petrou et al., 2013). Strategies to tolerate temporary light reduction are broadly the same for all species: adjusting light harvesting capacity and the efficiency of light use (Abal et al., 1994; Enriquez, 2005); adjustments to rates of growth and plant turnover (Collier et al., 2009; Collier C. J. et al., 2012; and drawing upon carbohydrate reserves to maintain a positive carbon balance (Burke et al., 1996; Touchette and Burkholder, 2000). While seagrasses adapted to marginal environments may be tolerant of wide fluctuations in light, they can also be acutely sensitive to reductions in light beyond the natural range of conditions (Ralph et al., 2007). When light drops below a critical level, seagrass productivity is compromised and significant physiological, biochemical and structural changes begin to take place eventually manifesting into broader meadow-scale losses with consequences for ecosystem function (Lee and Dunton, 1997; Ralph et al., 2007; Hughes et al., 2008).

Zostera muelleri ssp. capricorni is a key coastal seagrass species found along the tropical east coast of Australia (Waycott et al., 2004) and occurs in the muddy, inshore estuarine environments few other seagrass species inhabit (Lee Long et al., 1993; Carruthers et al., 2002). In port areas of the GBRWHA it is often the dominant species present, including in the Gladstone region, where it is found in monospecific intertidal meadows covering up to 40 km<sup>2</sup> within the port limits (Thomas et al., 2010; Supplementary Figure 1). With no known functional replacement, a large-scale dieback due to a stress event such as dredging could have wider implications for the ecological success of the inshore marine community.

The goal of this study was to develop a species-specific, light threshold for the effective management of Zostera muelleri ssp. capricorni in Gladstone, Australia. Recent expansion of port infrastructure and shipping channels around Gladstone has involved large-scale dredging and the removal of ∼26 million m<sup>3</sup> of sediment over 3 years. In situ shading studies were used to elicit a response in a local seagrass population to determine a light threshold at which seagrasses will decline and over what time scale a decline is detectable in plant abundance. The approach used does not attempt to simulate a given dredging scenario but rather to apply information on how locally-adapted seagrasses withstand constant light attenuation or how regular short-term reprieves from light attenuation events affect the overall seagrass condition and its' recovery in order to better manage threats from dredging related turbidity plumes. This information was used to apply a management-based light threshold to protect seagrasses from light stress during dredging. Long-term monitoring of the seagrass meadow at an adjacent site also provided information on the status and trend of local seagrass in relation to seasonality, light history, and water temperature. The adjacent site also provides a testing ground to assess the suitability of our light threshold against seagrass condition over the long term.

Our study focused on the development of locally-relevant light thresholds that can be applied for effective management of coastal and port development activities in a way that maintains seagrass health. The term threshold, as used here, is defined as the point at which a change in external conditions causes a significant negative change in seagrass physical condition, i.e., above-ground biomass, cover, or shoot density. It is important to note that this is different to defining a minimum light requirement (MLR) for effective seagrass photosynthesis. Rather, the goal is focused around developing a biologically relevant management tool, which incorporates other local environmental drivers such as tidal cycles, seasonality and sediment chemistry dynamics that influence seagrass condition together with light in vivo.

#### MATERIALS AND METHODS

#### Shading Study Experimental Design

This study was conducted at Pelican Banks, Gladstone Harbour (151◦ 18′ 30′′E, 23◦ 45′ 58′′S), Australia (see Supplementary Figure 1) from 2010 to 2013. At Pelican Banks the tropical subspecies Z. muelleri ssp. capricorni forms a predominantly monospecific intertidal seagrass meadow on intertidal mud banks. Studies were carried out during two growing seasons for local seagrasses (ca. July to December) and two senescent seasons (ca. January to June) when seagrasses naturally decline with the onset of the tropical monsoon and subsequent cooler months in the austral winter (Mellors et al., 1993; McKenzie, 1994). Studies are described accordingly: growing seasons 1 and 2 (G1 and G2) and senescent seasons 1 and 2 (S1 and S2). The study location was chosen for its accessibility, semi-firm sediment composition for repeated measurements during emergence at low tide without compromising site integrity, and year-round seagrass cover to assess seasonal effects. A semi-diurnal tide cycle with a maximum range of 5 m meant seagrasses were exposed at least fortnightly, depending on the time of year.

The study site was ∼30 × 20 m with experimental plots randomly assigned to each of three shade treatments or as controls (n = 4). Vertical isolation borders (sever root connection between shaded and non-shaded areas) were inserted for the shade experiments by hammering 0.25 m<sup>2</sup> quadrats with a 0.25 m depth into the sediment until flush with the sediment surface to isolate plots where seagrass would be measured. This ensured seagrass outside of the experimental plot could not translocate nutrients/carbohydrates to seagrass within treatment plots. Plots were also "gardened" around the isolation border perimeter prior to each sampling event to prevent seagrass growing over the border and into experimental plots. Aluminium frames were secured into the sediment and covered with 1 m<sup>2</sup> neutral density polyethelene shade cloth of varying intensities fixed 0.15 m above the sediment surface. Shade treatments were used to assess three levels of reduced light on seagrass health; high, medium and low shade, equivalent to ∼15, 30, and 45% of incident benthic light, respectively. Control plots were established using quadrats with steel frames and isolation borders but without shade screens. No control was used for the effect of rhizome severing based on the work of Rasheed (1999) which found no border effect using an identical experimental design and field materials to measure shading effects on the same species. Controlling for the additional effect of shade screens on water movement was not possible without creating additional shading or fouling over control plots (see Fitzpatrick and Kirkman, 1995). Shade screens were changed and cleaned fortnightly to reduce the effects of fouling on shade treatments. Light intensities under shade treatments fluctuated with natural insolation but maintained consistent patterns among treatments and relative differences to naturally occurring benthic light, indicating that fouling of the shade screens was minimal. Shade screens were removed at the end of each experiment to track potential recovery from treatment conditions.

Experimental plots were randomly assigned to varying durations of continuous shading (between 1 and 3 months) during each seasonal study (**Table 1**). This variation in shading study duration and tracking of recovery was necessary to align the program with expected timeframes for managing impacts to seagrass health during dredging operations as required by managers and regulators. Therefore, comparison among seasonal



*Shade treatments included high shade (H), medium shade (M), low shade (L), and control (C). N is the number of replicates per shade treatment for each study.*

studies was limited to shading durations comparable between studies. In addition, fortnightly cyclic shading was carried out during G1 to assess the impact of periodic turbidity plumes (i.e., shorter periods of reduced light and subsequent respites) on seagrass condition.

#### Light Climate

Light (photosynthetically active radiation, PAR) was measured within the seagrass canopy and under shade treatments using 2π cosine-corrected irradiance loggers (Submersible Odyssey Photosynthetic Irradiance Recording System, Dataflow Systems Pty. Ltd., New Zealand) calibrated using a cosine corrected Li-Cor underwater quantum sensor (LI-190SA; Li-Cor Inc., Lincoln, Nebraska USA) and corrected for immersion using a factor of 1.33 (Kirk, 1994). Loggers were deployed on site for the duration of shading and maintained using automated wiper units. Readings were made at 15 min intervals and used to measure total daily light (mol photons m−<sup>2</sup> day−<sup>1</sup> ) reaching seagrasses under each shading treatment.

Substantial tidal flux in Gladstone Harbour leads to dramatic shifts in daily light intensities on the intertidal banks due to fortnightly intertidal exposure cycles and this has the potential to control light availability to the plant (Koch and Beer, 1996). To evaluate light over a practical timeframe for measuring impacts, light data was integrated as a rolling 14 day mean of the total daily benthic light under each shading treatment, controls, as well as the long-term monitoring site (detailed below). Current understanding of seagrass response indicates under low light stress conditions, physiological adjustments first occur over a matter of days, whereas plant-scale changes take place after a number of weeks and are a reflection of the integrated light history over that period rather than short term daily fluxes (McMahon et al., 2013). This 2 week rolling average incorporated spring and neap tide conditions, variation in tide height, and the associated degree of exposure that affects the light conditions reaching the seagrass. An assessment of integrated light over a 2-week period is therefore in line with both tidally-driven fluxes in light, as well as a period of time preceding apparent morphological changes to seagrass.

#### Seagrass Morphometrics

Seagrass above-ground biomass, percent cover and shoot density were measured at fortnightly or monthly intervals in each treatment plot during S1 and G1 studies, while only biomass and percent cover were recorded during S2 and G2 studies. Above-ground biomass was measured using a "visual estimates of biomass" technique (Kirkman, 1978; Mellors, 1991; Rasheed, 1999). Biomass was estimated for each plot by an experienced observer recording a rank of seagrass biomass from photographs of each plot taken during sampling. Biomass ranks were assigned in reference to a series of photographs of similar seagrass habitats for which above-ground biomass has previously been measured. The same observer was used for the duration of each study to remove any inter-observer variability. At the completion of recording ranks, the observer ranked a series of additional photographs that had been previously harvested, dried, and weighed and which represented the range of seagrass biomass in the survey. A regression of ranks and biomass from these calibration quadrats was generated for each observer (r <sup>2</sup> = 0.97; see Supplementary Figure 2) and applied to the measuring plot ranks to determine above-ground biomass estimates. Biomass ranks were then converted into above-ground biomass estimates in grams dry weight per square meter (g DW m−<sup>2</sup> ). Shoot density was estimated by counting all shoots within a mini-quadrat (0.01 m2 ) randomly placed three times in each measuring plot except where total-plot shoot density was less than 30 shoots and all shoots were counted within the 0.25 m<sup>2</sup> plot. Seagrass percent cover estimates were made for each plot by an observer using a standardized photo guide sheet.

## Light History, Environmental Conditions and Seagrass Trend in the Meadow

A monitoring site was established in the Z. muelleri ssp. capricorni meadow adjacent to the shading study site to assess incident light and temperature at the seagrass canopy and its potential influence on seagrass meadow condition over longer time scales under natural harbor conditions. Light was recorded continuously between November 2009 and September 2013. Light loggers were deployed and operated in the same manner as in the shading studies through June 2012. From July 2012, irradiance loggers were replaced with LiCor underwater sensors with inbuilt wiper units and customized telemeted systems (Vision Environment QLD., 2013) to ensure continuous data collection and immediate availability of data during dredging operations. Water temperature was measured in the seagrass canopy (Thermodata Pty Ltd, Melbourne, Australia), daily rainfall (Bureau of Meteorology Australia<sup>1</sup> ) and total hours of daytime tidal air exposure of the meadow (Maritime Safety Queensland, Department of Transport and Main Roads) were also collected.

Seagrass condition was assessed at three 50 m transects nested in two 50 x 50 m sites. Sites were selected within a relatively homogenous section of the Z. muelleri ssp. capricorni meadow. Seagrass above-ground biomass was estimated within a 0.25 m<sup>2</sup> sampling quadrat placed at 0 m and then every 5 m along each transect (eleven sampling points per transect) using the same technique described above (observer regression of ranks, r 2 = 0.95). Mean biomass was calculated for each sampling event (n =

<sup>1</sup>www.bom.gov.au

66 quadrats) with change in biomass calculated from consecutive sampling events.

#### Data Analysis

All values displayed are means ± standard error (SE). Differences in morphological responses of seagrass among shading treatments and over time were assessed using repeated measures analysis of variance (rmANOVA). Data were checked for homogeneity of variance by assessing residual plots. Significant deviations from normal variance were found in G1 biomass data which were log-transformed prior to analysis. If data still did not meet the criteria, the p-value was set to 0.01 to minimize the risk of a Type I error (Underwood, 1997). For repeated measures ANOVAs, matrices were tested for sphericity using Mauchly's test. If the assumption of sphericity was not met (p < 0.05) the Greenhouse-Geisser (G-G) epsilon adjustment was applied to the numerator and denominator degrees of freedom. Differences among treatment effects at a given sampling time were compared using Tukey's post-hoc analysis. For data collected during the "recovery phase," a one-way ANOVA was performed when a single recovery time point was measured with shading intensity as a fixed effect and tests for homogeneity of variance and transformation applied as previously described. Statistical analyses were performed using Statistica 7.0. When multiple recovery period measurements were taken, rmANOVA methods as described for the shading period were applied.

# RESULTS

# Seagrass Morphometrics

Shading treatments did not have a significant effect on Z. muelleri ssp. capricorni morphology during either senescent season study (S1 and S2). However, after 1 month of shading there was a significant increase in shoot density during S1 (p < 0.05), but no significant changes in biomass or percent cover (p > 0.05, **Table 2**; **Figures 1**–**3**). Above-ground biomass and percent cover declined significantly over the 12 weeks of shading among all treatments during S2 (both p < 0.001); significantly lower aboveground biomass and percent cover in treatments compared to

TABLE 2 | Repeated measures ANOVA of the effects of shading treatment (among groups effect) and time (within groups effect) for biomass, percent cover and shoot density during senescent seasons 1 and 2 (S1, S2) and growing seasons 1 and 2 (G1, G2).


*The ANOVAs were not significant (ns), or significant at* \**p* < *0.05,* \*\**p* < *0.01,* \*\*\**p* < *0.001. Probability values are Greenhouse-Geiser adjusted p values.*

<sup>∧</sup>*Log transformed; †Not recorded.*

control plots; this was apparent from the start of the study (both p < 0.05, **Table 2**; **Figures 1**–**2**).

Shading had a detrimental effect on Z. muelleri ssp. capricorni above-ground biomass during the growing seasons (G1 and G2, shade × time interaction p < 0.001, **Table 2; Figure 1**). During both growing season studies, biomass was significantly lower by the 8 week sampling under high shade treatments compared to controls and other treatments (**Figure 1**). This occurred between 4 and 8 weeks in G1 and 6 and 8 weeks in G2. There was significant loss of above-ground biomass under all treatments compared to control plots by 12 weeks during G1, including near total loss of above-ground biomass under high shade plots (**Figure 1B**). Within 4 weeks of shade removal, above-ground biomass under low shade treatments recovered to control levels, whereas biomass under medium and high shade treatments remained significantly lower than control plots (p < 0.001; **Figure 1B**). Control plots did decline somewhat from a peak at 4–16 week measurements, likely due to the onset of characteristic seasonal senescence which occurred toward the end of the study (Jan–Feb 2011). Similarly, above-ground biomass under high shade was significantly lower than under control, low and medium shade treatments by 8 weeks of shading during G2. Declines in above-ground biomass and percent cover from mid-November in G1 and G2 across controls and all treatment

plots are consistent with seasonal declines with the onset of the senescent season (**Figures 1B,D**, **2B,D**).

Negative effects of shading on percent cover during both growing seasons were similar to those recorded for aboveground biomass (both p-values for shade × time interaction <0.001, **Table 2**; **Figure 2**). Percent cover was significantly lower under high shade treatments compared with control, low and medium shade treatments for G1and G2 within 8 and 6 weeks, respectively, (**Figures 2B,D**). Within 12 weeks percent cover under all shade treatments was significantly lower than control plots during G1 (**Figure 2B**). Recovery of seagrass during G1to a percent cover similar to control plots occurred within 4 weeks of shades being removed for the low shade treatment, but there were no similar signs of recovery for treatments that had been under medium or high shade treatment (**Figure 2B**; **Table 3**). Percent cover of seagrass under high shade similarly demonstrated no sign of recovery 2 weeks following shade removal during G2 (**Figure 2D**; **Table 3**). High shade plots were nearly devoid of seagrass cover 4 weeks after shade removal for G1 and G2 (**Figures 2B,D**).

Shoot density was less sensitive to shading than percent cover and above-ground biomass. Seagrass shoot density decreased significantly by 12 weeks under the high shade treatment compared with control and low shade treatment plots during the

Results). Grayed area represents shading periods and white area represents monitored recovery periods where data was recorded. Data represent mean ± SEM (*n* = 4).

growing season (G1 study, shading x time interaction p < 0.05, **Table 2**; **Figure 3B**). There were no signs of recovery to control levels 4 weeks after shades were removed (**Figure 3B**). Shading had no significant effect on temporal fluctuations in shoot density during the senescent season (S1 study, p > 0.05, **Table 2; Figure 3A**).

Seagrass was less sensitive to fortnightly cyclic shading than to continuous shading when tested during G1. Above-ground biomass data is only presented, but shoot density and percent cover results were analogous. Above-ground biomass under all shade treatments was similar to control plots for the first 8 weeks of the study; however, by week 12 biomass under all shade treatments was equally and significantly lower than under control plots (two-way rmANOVA, shade x time interaction, p < 0.01, **Figure 4**). After 4 additional weeks without shading (weeks 12–16), no biomass recovery occurred under high shade treatments relative to controls (p < 0.05). While seagrass loss was delayed under cyclic shading, the magnitude of impact of these treatments was similar to those found under continuous shading after 12 weeks.

Above-ground biomass and percent cover in control plots throughout all studies was similar to that measured at the nearby long-term monitoring site (see **Figure 6**) indicating no effect of the physical presence of frames holding shade screens otherwise on the experiment.

# Light Climate in Relation to Morphometric Results

During both senescent season studies (S1 and S2), light levels were strongly attenuated under all shade treatments compared to controls, while no measured loss of seagrass biomass, percent cover or shoot density was recorded after 4 and 13 weeks, respectively, when shades were in place (**Figures 5A,C**). Light intensities measured under S1 and S2 shades were generally between 2 and 6 mol photons m−<sup>2</sup> d −1 , a similar range recorded during the G1 study under the same shading treatments.

During the first growing season (G1), light intensities under the high shade treatment measured consistently below 2 mol photons m−<sup>2</sup> d −1 leading to significant declines in aboveground biomass and percent cover recorded by 8 weeks (**Figure 5B**). Light remained at or below 2 mol photons m−<sup>2</sup> d −1 for the remaining 4 weeks of shading over which time seagrass was completely lost from high shaded plots. Light under medium shade treatments was higher and more variable over the course of G1, but generally stayed above 4 mol photons m−<sup>2</sup> d −1 for the initial 10 weeks of the study, while light under low shades remained above 6 mol photons m−<sup>2</sup> d <sup>−</sup><sup>1</sup> during the same period. Light declined between weeks 10 and 12 of the experiment across controls and all treatments during a period of high rainfall in November and December 2010 (Australian Bureau of Meteorology<sup>2</sup> ). Light levels were consistently below 4 mol photons m−<sup>2</sup> d <sup>−</sup><sup>1</sup> under all shade treatments in the fortnight leading up to the 12 week sampling event, when biomass and percent cover were significantly lower for all treatments compared with control plots (**Figure 5B**). Four subsequent weeks with shades removed (recovery; weeks 12–16) were insufficient reprieve for biomass, percent cover or shoot density to recover under medium and high shade treatments while low shade treatments recovered when returned to ambient light conditions (**Figures 1B**, **2B**, **3B**).

<sup>2</sup>www.bom.gov.au/climate/data/


TABLE 3 | Repeated measures and one-way ANOVA of recovery from shading treatments (among groups effect) and time (within groups effect) for biomass, percent cover and shoot density during senescent seasons 1 and 2 (S1, S2) and growing seasons 2 (G2).

*The ANOVAs were not significant (ns), or significant at* \* *p* < *0.05,* \*\* *p* < *0.01,* \*\*\* *p* < *0.001. Probability values are Greenhouse-Geiser adjusted p values.*

<sup>∧</sup>*Log transformed; † Not recorded;*

*# Not tested, one-way ANOVA applied.*

During the second growing season (G2), light under high shaded plots was less than 5 mol photons m−<sup>2</sup> d −1 in the fortnight leading up to detection of a significant decline in seagrass percent cover at 6 weeks (**Figure 5D**). Light declined further to <4 mol photons m−<sup>2</sup> d −1 for the fortnight leading up to sampling at 9 weeks, when significant declines in percent cover and above-ground biomass were detected. Light under low and medium shade treatments mostly stayed above 5 mol photons m−<sup>2</sup> d −1 for the duration of the G2 shading study; one exception was when light dropped below 5 mol photons m−<sup>2</sup> d <sup>−</sup><sup>1</sup> under medium shade for ∼1 week at week 9; although with no detectable change in seagrass biomass or percent cover recorded. In contrast, significant declines in seagrass biomass and/or percent cover were recorded following more prolonged periods of light <5 mol photons m−<sup>2</sup> d <sup>−</sup><sup>1</sup> under high shade treatments at weeks 6, 9, and 10.

# Climate History and Seagrass Trend

From September 2009 to September 2013, seagrass above-ground biomass at the monitoring site followed a typical oscillating seasonal pattern. Z. muelleri ssp. capricorni reached maximum biomass between October and December each year which coincided with higher water temperatures and ambient light (**Figure 6**). Light levels in the meadow were relatively high during the growing season which paralleled net positive growth. Light intensities remained above 8 mol photons m−<sup>2</sup> d −1 ; well above the levels at which significant impacts were measured under shade treatments. Annual seagrass senescence began at approximately the start of the year when temperatures consistently reached >30◦C in the meadow and the onset of rain and flooding events led to reductions in light (**Figure 6**). The relationship between seagrass above-ground biomass and mean maximum daily water temperature for the month prior to sampling in the growing period likewise indicated water temperature correlated with seagrass biomass (p < 0.01, r <sup>2</sup> = 0.55) until water temperature exceeded 30◦C and seagrass declined, despite high light intensities over the same period. Seagrass abundance typically reached a minimum by April/May after which a return to growth and increased seagrass biomass was observed around July each year.

#### DISCUSSION

Z. muelleri ssp. capricorni condition (biomass, shoot density and percent cover) was measurably driven by light reductions tested during the growing seasons but was unaffected by a reduction in light applied during either senescent season. Similar field shading experiments have demonstrated time-ofyear is a critical factor in defining the magnitude of the plant's response to reduced light conditions, linked to seasonal light and water temperatures (Lavery et al., 2009). We found that Z. muelleri ssp. capricorni declined in the growing season when light was ≤ 5 mol quanta m−<sup>2</sup> d −1 for periods of time exceeding 4 weeks. This was successfully used to develop a conservative management threshold to protect seagrasses during dredging operations by maintaining light levels above 6 mol quanta m−<sup>2</sup> d −1 .

The significant and consistent decline in Z. muelleri ssp. capricorni during the growing season shading studies highlights the sensitivity of this species during its period of peak productivity and expansion. Z. muelleri ssp. capricorni carbon fixation and above-ground biomass have been shown to significantly decline when grown under saturating or limiting light levels in conjunction with extreme temperatures (>33◦C; Collier et al., 2011) and for temperate Z. muelleri when grown under 30◦C conditions (York et al., 2013). Similar results have been found for the congeneric northern hemisphere species, Zostera marina, with summertime declines coinciding with low light and high temperatures (Zimmerman et al., 1989; Olesen and Sand-Jensen, 1993).

The high metabolic demand that comes with warmer conditions was typically supported by higher light (approximately July to December) at our study site (**Figure 6**). This likely allowed an increase in photosynthetic processes to keep up with rising seasonal temperatures up until a point, after which respiration would continue to increase without a concomitant increase in photosynthesis (Bulthuis, 1987; Lee et al., 2007). When such an imbalance occurs this can lead to die-off, whether seasonal or driven by episodic reductions in light. It was likely that Z. muelleri ssp. capricorni was not meeting its metabolic requirements during these warmer months when subjected to reduced light levels, leading to a dieback under our shading treatments. Similar trends were seen at our permanent monitoring location adjacent to the study site where seasonal cycles of seagrass growth and decline paralleled temperature and light regimes (**Figure 6**).

Seasonal seagrass growth rates are closely linked to light and temperature patterns (Lee et al., 2007). Intertidal Z. muelleri ssp. capricorni meadows along the Queensland coast follow typical seasonal fluctuations in condition linked to light, temperature and tidal exposure (Mellors et al., 1993; McKenzie, 1994; Carruthers et al., 2002; Petrou et al., 2013). From August to December, clearer waters and warmer temperatures spur rapid growth and expansion of seagrass meadows in the Gladstone region before typical dieback in late austral summer with the onset of high temperatures and wet season conditions.

The lack of a low light response in the senescent season could be due to a decrease in extrinsic energy requirements due to the lower seagrass standing crop and preferential use of carbohydrate reserves to support seagrass metabolic requirements (Burke et al., 1996; Touchette and Burkholder, 2000). Lavery et al. (2009) also found shading imposed over winter did not produce morphological changes; in contrast to their late summer results. They associated the effect of temperature on gross photosynthetic requirements of the plant to explain the disparity in seasonal effects. The saturating irradiance for photosynthesis (Ik) and respiration typically increase with temperature (Masini and Manning, 1997; Lee et al., 2007) equating to higher overall light requirements during summer growing periods compared to cooler months.

When light levels are sufficient, carbohydrate reserves are enhanced which help offset periods of high light attenuation by supporting short-term energy demands of the plant. In the first growing season study, medium shaded plots were not measurably affected until the 12 week sampling event and did not recover from losses within 4 weeks. While light under medium shaded plots during the first 10 weeks (4–5 mol photons m−<sup>2</sup> d −1 ) sustained Z. muelleri ssp. capricorni in vivo, it was likely near its' light requirement limit and may have exhausted energy reserves, making recovery unachievable in the short-term once shades were removed. Alternatively, light during G1 under low shaded plots, which received by and large > 6 mol photons m−<sup>2</sup> d −1 during the study, likely enabled excess energy to be stored in the plant and used to support recovery when shades were removed. These differences in treatment response illustrate that conditions leading up to an acute stress event are important in determining recovery success. Ensuring light is maintained at a level that not

FIGURE 5 | Fourteen day rolling mean benthic light recorded under shade treatments across four shading studies. (A) Senescent season 1 (S1); (B) growing season 1 (G1); (C) senescent season 2 (S2); (D) growing season 2 (G2). Grayed area represents when shades were over experimental plots and white area when shades were removed. White vertical lines indicate sampling days; asterisks overlaying shade treatment light data indicates a significant reduction in seagrass above-ground biomass and percent cover relative to control for that sampling event (percent cover only for week 6 in G2); dashed lines indicate a biologically significant light threshold based on shading study results; solid black lines denote the derived management light threshold.

only sustains seagrass cover, but also provides energy reserves to be maintained or increased when conditions are good is likely important to ensure short-term stress events do not push the plant past a point of no return.

The quality of the light environment reaching seagrasses may be as important as the quantity of light received. Dredging, for example, typically increases particulate matter in the water column which affects spectral quality (Kirk, 1994). The size and type of particles re-suspended by dredging activity alter PAR transmission in a non-linear manner, with some wavelengths being more attenuated than others, resulting in a reduced light environment with a shift toward yellow wavelengths (Kirk, 1994; Gallegos et al., 2009). Therefore, a light threshold value used for monitoring seagrass health during a dredging campaign, as determined according to the full PAR spectrum available, may overestimate the actual light available for photosynthesis as PAR measurements do not distinguish spectral shifts (Van Duin et al., 2001; Zimmerman, 2003). Light quality in Gladstone waters has explicit spatial variability, with broader spectral transmission in the outer harbor compared to the inner harbor, yet dredging had no effect on these spectral signatures when measured during the dredging campaign that occurred during this study (Chartrand et al., 2012). The region is naturally highly turbid and therefore already exhibits a yellow-enhanced light signature due to the particle load in the water column and was not further skewed with additional sediment re-suspension from the dredge operation. While a more accurate threshold applying photosynthetic usable radiation (PUR) in place of PAR could resolve any effects of wavelength-specific water column absorption we did not need to alter light threshold values to incorporate spectral shifts from dredging in this instance.

Short term repeated shading and respite (fortnightly) in the present study was carried out to mimic repeated acute attenuation events from turbidity plumes followed by subsequent "relief " intervals. In providing a 14 day period of respite after shading was applied, Z. muelleri ssp. capricorni appeared to cope for 12 weeks with even the highest shade treatment, which had significantly impacted treatment plots shaded continuously after only 6–8 weeks. A study by Biber et al. (2009) also explored extreme attenuation events interspersed with recovery periods of varying length. They found that recovery intervals at least equal to the period of light deprivation were essential for long term survival.

Other investigations into in situ light requirements on Zostera spp. agree with the measured light effects and management threshold derived in this study (Dennison and Alberte, 1985; Moore et al., 1997; Thom et al., 2008; Collier C. J. et al., 2012). Collier C. J. et al. (2012) tested reduced light conditions during laboratory shading experiments on Z. muelleri ssp. capricorni also collected from Gladstone Harbour and found shoot density declined after 8.7 weeks under 4.4 mol photons m−<sup>2</sup> d −1 and 10.6 weeks under 9.5 mol photons m−<sup>2</sup> d −1 . For the congeneric Z. marina, Dennison and Alberte (1985) found a significant reduction in Z. marina production rates with average daily scalar light levels of ∼3.7 mol photons m−<sup>2</sup> d <sup>−</sup><sup>1</sup> under shades compared to unshaded controls (8 mol photons m−<sup>2</sup> d −1 ) during critical summer growing conditions. Moore et al. (1997) found similar results where sites with high light attenuation (2.7 mol photons m−<sup>2</sup> d −1 ) over 30 days was lethal to Z. marina transplants compared to those with higher water clarity (13.4 mol photons m−<sup>2</sup> d −1 ). More recent work on Z. marina found light requirements for long-term survival is 3 mol photons m−<sup>2</sup> d −1 and at least 7 mol photons m−<sup>2</sup> d −1 for non-light-limiting growth conditions during critical growing months (Thom et al., 2008).

#### Deriving a Light Threshold for Management

Developing effective management tools and appropriate mitigation strategies to protect seagrasses from a large-scale dredging campaign requires information on the distribution, light requirements and tolerances of local seagrass communities. Shading studies and the 4-year seagrass and light monitoring program provided the means to develop an effective and ecologically-derived management threshold. A 14 day integrated daily light value was used to establish a light threshold, which if maintained, would allow sufficient light to maintain local Z. muelleri ssp. capricorni seagrass condition in Gladstone Harbour during dredging.

With no significant effects of shading on seagrass growth during either of the senescent seasons, a seagrass light management threshold was only defined for the growing season when Z. muelleri ssp. capricorni was sensitive to shading treatments. Both growing season studies clearly indicated light below 4 mol photons m−<sup>2</sup> d <sup>−</sup> is insufficient to maintain seagrass growth and or survival. In the second growing season study, light levels 2 weeks prior to a decline in seagrass measured between 4 and 5 mol photons m−<sup>2</sup> d −1 , indicating morphological changes in Z. muelleri ssp. capricorni can take place in Gladstone at light intensities of ≤ 5 mol photons m−<sup>2</sup> d −1 .

While the time to measurable loss in the first growing season was between 4 and 8 weeks, more frequent sampling during the second growing season documented appreciable declines in seagrass cover as early as 6 weeks under light limiting conditions. A study by Adams et al. (2015) found the timeframe over which light history and Z. muelleri above-ground biomass best correlated was from 8 to 35 weeks, however, they recognized management actions also should be triggered well before these measured reductions in biomass occur.

A range of bioindicators have been reviewed for use in seagrass monitoring programs to measure environmental pressures such as dredging (McMahon et al., 2013). While some metrics may be more sensitive on shorter time scales (e.g., rhizome sugars or ETRmax) to changes in the light climate (reviewed in McMahon et al., 2013), the ability to measure changes rapidly in relation to anthropogenic pressures (i.e., dredge operations) is important to apply an appropriate and timely management response. In the current study, above-ground abundance (either biomass or percent cover) reacted to light conditions within a timeframe that would allow a management response to be applied that could abate seagrass loss (i.e., move dredge to a new location), whereas shoot density was less sensitive to attenuated light. Other studies have also found shoot density to be a less sensitive metric; Z. muelleri ssp. capricorni alters leaf morphology before shoot loss under reduced light treatments, making above-ground biomass or cover a more sensitive indicator of change than shoot density as a consequence of environmental conditions (Rasheed, 1999; Collier C. J. et al., 2012).

As a conservative approach to protecting seagrass, a management light threshold needed to provide >5 mol photons m−<sup>2</sup> d <sup>−</sup><sup>1</sup> with some degree of buffer from potential impact to the plants and to ensure the plants not only maintained physical presence, but could generate energy stores. The threshold needed to ensure protection of seagrasses from deteriorating light conditions, while also having a credible fit with natural background light variability within the local meadow. If the threshold value was set too high and therefore routinely breached without measureable impacts to seagrass condition, it would be ineffective as a management tool. Conversely, a value too low that was never measured in situ in spite of concurrent declines in seagrass cover would likewise be inappropriate. A light threshold of 6 mol photons m−<sup>2</sup> d <sup>−</sup><sup>1</sup> was therefore used in a compliance framework by government regulators and management authorities to prevent measurable loss of seagrass from dredge related light attenuation in required management zones during dredging activity in Gladstone Harbour. This light threshold was considered in parallel with turbidity monitoring to ensure effects of turbidity related to the dredge vs. background conditions could be resolved (GPCL, 2012b). During the dredging campaign light was maintained above the management threshold for the growing season at all of the prescribed seagrass management zones (GPCL, 2012a). This coincided with the presence of the largest seagrass meadows in the greater region during and post-dredging (Carter et al., 2015) and provides confidence that the approach used could be applied elsewhere for managing seagrasses.

While much research is focused on quantifying seagrass light requirements (Dennison, 1987; Staehr and Borum, 2011; Collier et al., 2016), this work has focused on the application of seagrass light requirements for use in a management setting of a largescale dredging program. The absolute threshold value detailed here is not as critical as the approach used to derive a light-based model for seagrasses. The successful approach developed could readily be applied in other settings with sufficient knowledge of local seagrass dynamics and light conditions.

A range of additional measures would further improve the use of light thresholds to effectively manage seagrasses during dredging and other anthropogenic activities impacting on the light environment:


and how light reduction affects oxygen transport and below ground viability is vital to understand whether thresholds are in line with whole plant coping strategies.

4. Modification of light requirements under cumulative long-term impacts-Poor water quality prior to a major development may exacerbate efforts to manage additional impacts on already chronically stressed seagrass. Prolonged physiological strain from cumulative pressure over time may alter the plant's capacity to cope with further reduced light and may influence the light levels required for recovery.

# CONCLUSION

This study characterized the tolerance of Z. muelleri ssp. capricorni to light attenuation on an intra- and inter-annual cycle using in situ shading studies and light history monitored over a 4-year period. This information was used to develop a locallyrelevant management plan to protect seagrasses from dredgingrelated impacts to the light environment. A light threshold of 6 mol photons m−<sup>2</sup> d <sup>−</sup><sup>1</sup> was successfully trialed as part of a compliance program for mitigating dredging impacts. This minimized the risk that Z. muelleri ssp. capricorni, the dominant local species, was affected by dredge turbidity plumes within prescribed management zones. When implementing a light management strategy it is critical that local conditions, species and context are considered.

# AUTHOR CONTRIBUTIONS

KC, MR, CB, and PR together designed the research project. KC led the study and drafted the manuscript with the assistance of MR and AC. CB and AC provided major assistance in field execution and data analysis. All co-authors commented on and approved the final manuscript draft.

#### ACKNOWLEDGMENTS

We would like to thank Leonie Andersen and Vision Environment Pty Ltd for light data and assistance in maintaining field equipment. We would also like to acknowledge Brett Kettle from Babel-Sbf Pty Ltd and Queensland Gas Corporation who initially supported and commissioned this research. Further funding and support was provided by Gladstone Ports Corporation Pty Ltd and Australian Research Council Grant LP110200454. We thank James Cook University TropWATER staff for field support and data collection and K Petrou and I Jimenez for their invaluable input in the field and larger research program.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00106

# REFERENCES


services. Glob. Environ. Change 26, 152–158. doi: 10.1016/j.gloenvcha.2014. 04.002


Australia. J. Exp. Mar. Biol. Ecol. 235, 183–200. doi: 10.1016/S0022-0981(98) 00158-0


**Conflict of Interest Statement:** Funding for this research came from two industry bodies as detailed in the Funding Statement, creating a perceived conflict of interest. However, all data, results, analysis and conclusions were delivered through an independent government-mandated Dredge Technical Review Panel with a suite of scientific experts and engineers appointed to establish potential impacts of dredging on local seagrasses and to implement (as mandated under permit approvals) a light-based approach to dredge management.

Copyright © 2016 Chartrand, Bryant, Carter, Ralph and Rasheed. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Chemical Assessment of Ballast Water Exchange Compliance: Implementation in North America and New Zealand

#### Monaca Noble<sup>1</sup> \*, Gregory M. Ruiz <sup>1</sup> and Kathleen R. Murphy <sup>2</sup> \*

*<sup>1</sup> Marine Invasions Research Laboratory, Smithsonian Environmental Research Center, Smithsonian Institution, Edgewater, MD, USA, <sup>2</sup> Water Environment Technology, Department of Civil and Environmental Engineering, Chalmers University of Technology, Gothenburg, Sweden*

Fluorescence by naturally occurring dissolved organic matter (FDOM) is a sensitive

#### Edited by:

*Angel Borja, AZTI-Tecnalia, Spain*

# Reviewed by:

*Rafael Riosmena-Rodriguez, Universidad Autonoma de Baja California Sur, Mexico José Lino Vieira De Oliveira Costa, Centre of Oceanography of the Faculty of Sciencies of the Lisbon University, Portugal*

\*Correspondence:

*Monaca Noble noblem@si.edu; Kathleen R. Murphy murphyk@chalmers.se*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *19 February 2016* Accepted: *18 April 2016* Published: *09 May 2016*

#### Citation:

*Noble M, Ruiz GM and Murphy KR (2016) Chemical Assessment of Ballast Water Exchange Compliance: Implementation in North America and New Zealand. Front. Mar. Sci. 3:66. doi: 10.3389/fmars.2016.00066* indicator of ballast water source, with high FDOM in coastal ballast water decreasing typically dramatically when replaced by oceanic seawater during ballast water exchange (BWE). In this study, FDOM was measured in 92 ships arriving at Pacific ports on the US west coast and in New Zealand, and used to assess their compliance with ballast water regulations that required 95% replacement of port water to minimize invasive species risks. Fluorescence in many ships that reported BWE was significantly higher than is usual for oceanic seawater, and in several cases, significantly higher than in other ships with similar provenance and ballast water management. Pre-exchange source port conditions represented the largest source of uncertainty in the analysis, because residual coastal FDOM when highly fluorescent can significantly influence the fluorescence signature of exchanged ballast water. A meta-analysis comparing the intensities of FDOM in un-exchanged ballast tanks with calculated pre-exchange intensities assuming that ships all correctly implemented and reported BWE revealed notable discrepancies. Thus, the incidence of high-FDOM port waters was seven times lower in reality than would be expected on the basis of these calculations. The results suggest that a significant rate of reporting errors occur due to a combination of factors that may include inadequate BWE and unintentional or deliberate misreporting of ballast water management.

#### Keywords: Pacific Ocean, fluorescence spectroscopy, FDOM, invasion vectors, invasive species, AIS, NIS, CDOM

# INTRODUCTION

The transfer of ballast water between ports is an effective mechanism for moving a diverse assemblage of marine and estuarine organisms around the globe, posing considerable risk to the marine environment (Carlton and Geller, 1993; Ruiz et al., 1997; Roman and Darling, 2007). In the United States, controlling ballast water discharge is viewed as an important factor in the management of bays, estuaries, and the Great Lakes (Costello et al., 2007; Bailey et al., 2011). In New Zealand, economically and socially important fisheries are threatened by large volumes of ballast water discharged each year (Hewitt and Campbell, 2007). In both countries, ballast water is the suspected vector for several marine introductions. Damage caused to the Great Lakes by the Zebra Mussel, including extensive fouling and clogging of water intake pipes and impacts on native species, led in 1993 to the first ballast water exchange (BWE) requirements for ships entering the Great Lakes from outside the US exclusive economic zone. This authority was soon extended to other regions of the country by the National Invasive Species Act of 1996 (H. R. 4283, 104 Congress of the United States).

Ballast water is carried by vessels to provide stability and trim during sailing and during loading and unloading operations. It is usually loaded at the same time that cargo is unloaded and discharged in exchange for cargo, but may also be transferred between tanks within a vessel and carried for up to several months or even years. During BWE, port water within ballast tanks is replaced with oceanic water sourced outside of the coastal zone, preferably at least 200 nautical miles (nmi) from shore, although coastal BWE is often performed along routes that remain closer to shore (Miller et al., 2011). Depending on a range of factors including the tank design, type of exchange method used, and characteristics of individual species, BWE is capable of reducing concentrations of coastal organisms by 80– 95% (Gray et al., 2007; Minton et al., 2015). The effectiveness of current BWE policy at reducing invasion rates is difficult to evaluate (Costello et al., 2007) and policy efforts over more than a decade have been directed toward replacing BWE with better technological solutions (Briski et al., 2015) and concentrationbased performance standards (Albert et al., 2013). However, a range of setbacks have hampered the widespread adoption of new treatment technologies and performance standards with the result that BWE is still the only ballast water treatment method in widespread use (Minton et al., 2015).

Both the United States and New Zealand governments require commercial vessels arriving from overseas to treat or exchange their ballast water before discharge to reduce the risk of releasing invasive coastal species (MAF, 2007; Miller et al., 2011; United States Coast Guard (USCG), 2012a,b). Despite the legislative requirement for BWE in both countries, it is difficult to evaluate ships' claims regarding the origin and management of ballast water. In the United States, the process for determining whether a ship has conducted BWE are detailed in the US Coast Guard's Navigation and Inspection Circular 07–04, Ch-1. Ballast water management records may be examined, and salinity readings may be taken if non-compliance is suspected. In New Zealand, the Ministry of Primary Industries Biosecurity Division prohibits the discharge of ballast water into New Zealand waters without the permission of an inspector (MAF, 2005, 2007). To obtain permission, the vessel's Master must provide a signed declaration that the ballast water was subject to mid-ocean BWE. Inspectors approve ballast water discharge based on a combination of factors including agreement between ballast management records and salinity. In both countries, ballast water with salinity between 30 and 40 is considered consistent with BWE. However, this criterion fails to reliably detect ballast water originating in Pacific rim ports, since many ports in this region have high salinities either seasonally or year-round (Doblin et al., 2010).

Previous research indicates that fluorescence by naturally occurring dissolved organic matter (FDOM) is a robust coastal tracer, with sensitivity that exceeds many other chemical tracers including salinity and trace elements (Murphy et al., 2008a, 2013; Doblin et al., 2010). FDOM quantifies the organic matter fraction that absorbs light and reemits the radiation as fluorescence (Lakowicz, 2006). In estuaries, FDOM intensities vary with salinity gradients and biological activity as well as anthropogenic factors such as industrial effluent, and agricultural and urban runoff (Coble, 1996; Stedmon and Markager, 2005; Walker et al., 2009; Guo et al., 2011). Moving offshore away from terrestrial sources and as a result of exposure to sunlight, FDOM derived from terrestrial materials decreases (Duursma, 1974; Blough and Del Vecchio, 2002; Murphy et al., 2008b; Nelson et al., 2010). Because oceanic levels of FDOM are very low relative to concentrations at the coast, it can be deduced that samples with high FDOM are of coastal origin.

Previous studies have used fluorescence excitation-emission matrix spectroscopy to identify wavelengths most appropriate for measurement (Murphy et al., 2004, 2006). These found long-wavelength fluorescence associated with terrestrial organic matter to be an effective indicator of BWE. In shipboard experiments conducted in the North Pacific and Atlantic oceans, Murphy et al. (2006) determined that a threshold of 0.7 QSE (parts per billion quinine sulfate equivalents) measured at the C3<sup>∗</sup> wavelength pair (λex/λem = 370/494 nm) discriminated between exchanged and unexchanged ballast water in >95% of tests (N = 40 ballast tanks), some of which were in the range of oceanic salinities. An extensive survey (>2000 samples) of C3<sup>∗</sup> in ports and at varying distances from land confirmed that large differences in coastal vs. oceanic FDOM levels hold in the Pacific Ocean (Murphy et al., 2013). However, natural variability in coastal FDOM levels, which may legally represent as much as five percent of the water in an exchanged ballast tank, make it difficult to rely upon a simple C3<sup>∗</sup> threshold. For example, assuming oceanic C3<sup>∗</sup> levels of 0.5 QSE, any ship carrying ballast originally from a location where C3<sup>∗</sup> exceeds 4.5 QSE will exceed 0.7 QSE even after performing 95% BWE.

In practice, given incomplete knowledge of FDOM distributions in coastal environments on a global scale, reliable chemical assessments of BWE must rely upon a forensic approach, in which multiple lines of evidence feed into the judgment of a vessel's compliance. Assuming that FDOM levels that were present in the ballast water tanks prior to BWE are unknown, then port survey data and/or data from other vessels with ballast from the same location can help to constrain estimates of the likely contribution of port water to the measured FDOM signal upon arrival. To test this approach, FDOM was measured in a diverse cohort of vessels (N = 92 ships) boarded by inspectors at various ports along the US west coast and New Zealand. The results were used to assess BWE compliance of individual ships and to gauge the overall level of compliance among the vessel cohort.

# MATERIALS AND METHODS

### Experimental Design

Replicate ballast water samples were collected from 99 ballast tanks in 92 ships arriving to the United States or New Zealand. In the United States, ballast water samples were collected from 73 vessels that arrived at ports in California (47), Oregon (10), and Washington (16) in 2008 and 2009. Samples were collected by ballast water inspectors from three state agencies: the California State Lands Commission (CSLC), the Oregon Department of Environmental Quality (ODEQ), and the Washington Department of Fish and Wildlife (WDFW). In New Zealand, ballast water samples were collected from 19 vessels that arrived at the ports of Auckland (17), Tauranga (1), and Taharoa (1) in May, 2010. Sampling was performed by Ministry of Primary Industries (MPI, formerly Ministry of Agriculture and Forestry MAF) biosecurity inspectors, assisted by one researcher. Vessels of a range of types and trading histories were selected in an effort to maximize sample diversity. Ballast water source and management was self-reported by the vessel.

#### Sampling

Similar sampling methodologies were implemented in the United States and in New Zealand. Ballast water samples were collected through an open manhole from a single tank per vessel in the United States and one or two tanks per vessel in New Zealand. Three replicate samples were collected using large Clear-ViewTM PVC bailers (45.72×2.54 cm, 342 mL) from the vertical midpoint of the accessible sampling depth. The bailers have a stopper ball which allows them to collect samples from select depths. Water flows through the tube as the bailer is lowered into the tank, then when the bailer is retrieved the stopper-ball drops to the bottom of the tube sealing it. Once filled, the bailers were drained into a 60 mL syringe then filtered using Whatman 0.45µm PVDF syringe filters into pre-ashed 125 mL amber glass bottles. All equipment was subject to stringent cleaning prior to sampling, bailers and syringes, and filters were acid washed (10% HCl) and rinsed with 18 M deionized water and air dried in a laminar flow hood. Salinity was measured using a hand-held refractometer.

For all tanks scheduled for discharge, data regarding ballast water sources and management were obtained from ballast water reporting forms, which constitute legal declarations to the National Ballast Water Information Clearinghouse in the US and to MPI Biosecurity in New Zealand. For those tanks that were not to be discharged in the sampling port, source and management data were collected from the vessel's log books by the ballast water inspector. On the basis of these reports, each sampled tank was assigned to one of four management categories: exchanged in mid-ocean >200 nmi from shore (BWE, n = 57), exchanged <200 nmi from shore (BWEc, n = 19), filled from empty in the mid-ocean (FS, n = 11), or carrying unexchanged port water (none, n = 12).

#### Laboratory Analyses

FDOM fluorescence was measured using a benchtop Fluorologr-3 spectrofluorometer (Horiba Jobin Yvon, Edison, NJ). Undiluted filtered seawater samples were analyzed in ratio mode using a 0.5 s integration time and a 1-cm quartz cell held at 20◦C. Fluorometer bandpasses were set to 5 nm for both the excitation and emission monochromators. The Fluorolog-3 is configured with a single excitation monochromator (1200 grooves/mm) blazed at 330 nm and a dual emission monochromator (1200 grooves/mm) blazed at 500 nm, a watercooled, red sensitive photomultiplier tube and a 450-watt Xenon arc lamp.

Data were corrected for instrumental and lamp variability and normalized to quinine sulfate fluorescence intensity as previously described (Murphy et al., 2010). Fluorescence can be suppressed by absorbing species in the sample matrix, in a phenomenon known as the inner-filter effect (IFE). Suppression is below 5% at wavelengths where total absorbance (A) is below 0.042 in a 1 cm cell (Kothawala et al., 2013). Absorbance at 370 nm measured using a Cary 4E UV–Visible spectrophotometer was always below 0.015 m−<sup>1</sup> so no inner filter correction was necessary. Fluorescence intensities were calibrated against a quinine sulfate dilution series and are expressed in units of concentration (ppb quinine sulfate equivalents, QSE). An approximate conversion of these data to Raman Units (RU, normalized to the area of the Raman peak in a clean water blank excited at 350 nm) is obtained by dividing intensities in QSE by 100 (Murphy et al., 2010). Data are reported here for a single wavelength pair, C3<sup>∗</sup> (λex/λem = 370/494 nm) that has been extensively studied in the context of BWE, and for which BWE thresholds have already been developed and tested (Murphy et al., 2006, 2013; Doblin et al., 2010).

#### Chemical Assessments of Compliance

Since terrestrially derived FDOM in the open surface Pacific Ocean far from land is low and relatively stable compared to at the coasts (Nelson et al., 2010), then a lower bound for C3<sup>∗</sup> prior to BWE can be deduced from measured C3<sup>∗</sup> following BWE (Equation 1)

$$\text{C3}^\*\_{pre\ BWE} = \frac{\text{C3}^\*\_{post\ BWE} - \varepsilon \ast \text{C3}^\*\_{ambient}}{(1 - \varepsilon)}$$

In Equation (1), C3<sup>∗</sup> post BWE is the measured fluorescence intensity in a ballast tank was reported as having undergone BWE, C3<sup>∗</sup> pre BWE is the calculated fluorescence intensity prior to BWE, and ε is the BWE efficiency. C3<sup>∗</sup> ambient is the fluorescence intensity in the ambient ocean where BWE was performed.

In the calculations, BWE efficiency (ε) was assumed equal to the minimum level specified by law (95%), except in the case of ballast tanks filled from empty in the ocean (FS). For these a higher exchange efficiency (99%) was assumed based on earlier studies (Cohen, 1998; Drake et al., 2007). Filling at sea is relatively efficient because the only sources of port signals are residual volumes of unpumpable ballast water and sediments. C3<sup>∗</sup> ambient was assumed equal to 0.5 QSE in the open ocean, and =1 QSE in coastal exchange zones. These levels are consistent with surveys in the North Pacific (Murphy et al., 2013) and are probably conservative (i.e., represent upper limits) except when BWE was performed north of 45◦N where oceanic CDOM is relatively elevated (Nelson et al., 2010). If FDOM at the site of BWE was actually higher than the assumed level, this would result in C3<sup>∗</sup> pre BWE being slightly overestimated, of if lower then C3<sup>∗</sup> pre BWE would be slightly underestimated. However, a large over- or under-estimation is unlikely because even a 50% error in the assumed oceanic C3<sup>∗</sup> represents no more than a small absolute difference in post-exchange C3<sup>∗</sup> . Conversely, C3<sup>∗</sup> pre BWE is very sensitive to BWE efficiency since a decrease from 95% to 90% efficiency doubles the influence of the residual port signal.

Calculated C3<sup>∗</sup> pre BWE was used in two ways to assess compliance by individual vessels. First it was compared with measured C3<sup>∗</sup> at the port of origin, when port data were available from earlier surveys and published reports. Second, it was used in comparisons with measured C3<sup>∗</sup> in other ships that loaded ballast water in the same location at approximately the same time (within 2 weeks). To assess compliance by the cohort as a whole, the distribution of calculated C3<sup>∗</sup> pre BWE was compared with the measured distribution of C3<sup>∗</sup> in ballast tanks that were reported as having not undergone BWE (n = 48). The sample size for this comparison was increased by including data from any randomly-sampled tank containing unexchanged ballast water in our databases (n = 36). To avoid biasing the results, ships in our database that were deliberately targeted on the basis of source characteristics were excluded from this comparison.

#### RESULTS

**Table 1** summarizes C3<sup>∗</sup> fluorescence and salinity measurements for each sampled tank, classified by ballast water source and reported ballast water management (N = 99 tanks from 92 ships). The majority of tanks (88%) reportedly underwent some type of ballast water management. Most were exchanged in midocean more than 200 nmi from land (57%) or in coastal waters (20%), and 11% were filled from empty at sea. All ballast tanks reportedly sourced or exchanged at least 200 nmi from land (BWE and FS categories) had salinities between 31 and 41, i.e., within the range of salinities considered by regulatory agencies to be consistent with oceanic sources.

**Figure 1** shows the distribution of fluorescence intensities among tanks sampled in each management category. Intensities are shown as multiples of the BWE threshold, tc. As expected

a multiple of the BWE threshold (*t*c = 0.7 QSE) proposed by Murphy et al. (2006). Management categories are unexchanged (none), coastal exchange (BWEc), mid-ocean exchange (BWE), and filled at sea (FS), with number of tanks in each category listed in parentheses.

in ships that reported no BWE, C3<sup>∗</sup> always exceeded tc, while in half of the tanks, t<sup>c</sup> was exceeded by more than five times. Conversely, fluorescence intensities in exchanged ballast tanks were frequently much higher than expected. Among tanks that reportedly underwent mid ocean BWE or were filled at sea (BWE and FS, respectively), 54% of tanks had C3<sup>∗</sup> fluorescence exceeding t<sup>c</sup> and 25% of tanks had fluorescence exceeding 3tc. Among 19 tanks that reportedly underwent coastal exchange (BWEc), 36% exceeded 3tc, and 26% exceeded 4tc.

In **Figure 2**, fluorescence intensities measured in ships' ballast are mapped according to the reported geographical source of the ballast water. For unexchanged ballast water, the reported source was in a port, and for exchanged ballast water, the reported source was the offshore location where BWE took place. Blue symbols indicate low fluorescence consistent with oceanic sources, and orange and red symbols indicate high fluorescence consistent with coastal sources. C3<sup>∗</sup> fluorescence was typically highest in tanks ballasted near land and lowest in ships that reported oceanic BWE. However, a significant number of tanks that were reportedly exchanged in the open ocean far from land stand out as obvious exceptions to this rule.

**Table 1** contains the measured and reported data for each sampled ballast tank. Additionally, the final column contains calculated source intensities for reportedly exchanged ballast tanks, i.e. estimates of C3<sup>∗</sup> prior to BWE deduced using Equation (1), assuming BWE was performed properly. These data are used in **Figure 3** to compare the distribution of calculated source intensities with the measured distribution of source intensities in unexchanged ballast tanks. **Table 1** shows that many calculated source intensities (Cases 3, 19, 21-23, 27, 32, 38, 46, 56, 58, 60– 64, 83, 89, 97) represent extreme outliers. Most would remain outliers if the assumptions of the calculation were relaxed by assuming that C3<sup>∗</sup> at the exchange location had been 50% higher and BWE efficiency were below 85%. Overall, these data suggest that in many cases BWE was either misreported or undertaken with much less than the mandated 95% exchange efficiency.

A number of ships in this survey originated from ports that have previously been surveyed by our group. These port survey data can be used to explore whether high C3<sup>∗</sup> might reasonably be explained by residual (<5%) quantities of port water. Cases 3 and 4 represent two ballast tanks on the same ship ballasted in the port of Melbourne and later reportedly exchanged. Port surveys of FDOM in Melbourne do not support this reporting: C3<sup>∗</sup> in both tanks (1.4 and 3.2 QSE) was within the typical range measured at the port of Melbourne during winter and spring surveys in 2007 whereas calculated pre-BWE C3<sup>∗</sup> (9.6 and 54.4 QSE) greatly exceeded this range (Doblin et al., 2010). Similarly, Cases 57–67 represent ships that reportedly filled empty tanks in the Pacific Ocean at least 200 nmi from land, where C3<sup>∗</sup> should have been extremely low. However, measured C3<sup>∗</sup> intensities are consistent with predominantly open ocean sources in only two cases (57 and 65, with C3<sup>∗</sup> ≤0.55). In six other cases, C3<sup>∗</sup> intensities were in the range of 1.3–3.1 QSE, suggesting a moderate to large contribution by residual port water. Seasonal surveys at Los Angeles port and coastal waters in California have been conducted over several years by our group and indicate low background C3<sup>∗</sup> in the port (<2–3 QSE, Murphy et al., 2009)


TABLE 1 | Mean fluorescence intensities (C3\* = 370/494 nm) measured in randomly sampled ballast tanks in ships arriving to Pacific Ocean ports in this study.

*(Continued)*

#### TABLE 1 | Continued


*The number of days between loading and sampling of ballast water is indicated in the column "Age". Ballast water management is categorized as mid-ocean exchange (BWE), coastal exchange (BWEc), filled at sea (FS), or unexchanged (none). The final column contains calculated fluorescence prior to BWE (see main text). Missing data is shown as "nd."*

indicate intensities in multiples of the BWE threshold (0.7 QSE) developed by Murphy et al. (2006). Orange and red symbols indicate C3\* intensities that exceed the threshold by more than four and five times, respectively.

decreasing to below 0.8 QSE in the coastal ocean at distances exceeding 50 nmi from shore (Murphy et al., 2013). In Case 83, C3<sup>∗</sup> exceeded 4 QSE after reported 95% coastal BWEc, which would require that C3<sup>∗</sup> prior to BWE was around 30 times higher than the highest values measured during these earlier surveys.

The C3<sup>∗</sup> measurements in **Table 1** are organized geographically to facilitate comparisons between tanks having similar ballast water sources. When two ships ballast in the same port at around the same time and undertake similar ballast water management, C3<sup>∗</sup> intensities in both ships should be comparable. For example, cases 95 and 96 represent unexchanged ballast water obtained in Seattle by two different ships within a 3 week period and differ by <10%. Returning to Cases 3 and 4, these can be compared with Case 2, on another ship that ballasted in the port of Melbourne a few days earlier. For Case 2, C3<sup>∗</sup> after BWE was below t<sup>c</sup> as expected, and 3–6 times lower than in Cases 3 and 4. These results again suggest that BWE was undertaken in Case 2, but not in Cases 3 and 4. Similarly, Cases 74 and 75 from Kaohsiung are inconsistent because (1) despite tanks having been loaded and exchanged at nearby locations within a month of one another, C3<sup>∗</sup> was two-fold higher in Case 75, and (2) whereas for Case 74 the estimated pre-BWE C3<sup>∗</sup> is within the known range of Kaohsiung port (1–2 QSE, Murphy et al., 2009), for Case 75 it is a factor of two higher. Finally, Cases 95 and 96 with unexchanged Seattle water provide some support for the claim that BWE was attempted in Case 98, although it appears to have been much less than 95% efficient.

In most cases where fluorescence data were at odds with BWE reporting in this study, there was no evidence of irregularities in the ship's paperwork. However, the vessel involved in Cases 3 and 4 had serious enough paperwork irregularities that the port authority involved denied permission to discharge ballast water. Although our data were not the basis of this decision, the fluorescence measurements independently corroborated the inspector's suspicions regarding the integrity of the ship's records. Cases 16 and 83 also had inconsistent reporting and elevated fluorescence results.

An evaluation of reporting by the entire cohort is provided by **Figure 3**. Here, the distribution of calculated C3<sup>∗</sup> pre BWE (n = 72) can be compared directly with the measured distribution of C3<sup>∗</sup> in ships that did not report exchanging ballast water (n = 48). The calculated C3<sup>∗</sup> distribution has higher proportions of vessels in both the extremely low (<0.7 QSE) and extremely high (>20.7 QSE) fluorescence ranges. The low anomaly indicates that at least 10% of ships who reported BWE encountered C3<sup>∗</sup> levels in the ocean lower than those that were assumed in the calculations. The high anomaly indicates that the incidence of high-FDOM ports should be around an order of magnitude higher than it actually is, if ships were all correctly implementing and reporting BWE.

#### DISCUSSION

This study presents the first report of dissolved organic matter fluorescence intensities (C3<sup>∗</sup> = 370/494 nm) in ballast tanks of randomly-sampled ships arriving to Pacific ports. It was attempted to use these data to verify BWE when reportedly undertaken for those tanks, based upon reconciling fluorescence measurements with ships' reports without direct information regarding the chemical signatures of the ballast tanks prior to BWE. Previous research indicates that fluorescence is a stable and sensitive tracer of BWE in controlled experiments for which the source waters and treatments applied are able to be carefully monitored (Murphy et al., 2004, 2006). However, in a regulatory setting these data are usually unavailable or supplied by the ship and of unknown accuracy. Applying fluorescence as tool to verify BWE in a regulatory setting therefore introduces additional practical and technical challenges.

Applying a unilateral fluorescence threshold for determining BWE compliance, e.g., C3<sup>∗</sup> < 0.7 QSE, would be expected to fail in two main situations. First, if a ship ballasts in a clearwater port with little terrestrial input of organic materials, then fluorescence intensities may be low regardless of whether BWE takes place. According to **Figure 3**, ports with C3<sup>∗</sup> < 1 QSE account for <10% of cases in our dataset. Also, tanks sampled in this study were nearly all ballasted and exchanged in the Pacific Ocean which experiences low coastal influences compared to the Atlantic Ocean (Opsahl and Benner, 1997; Siegel et al., 2002). Low-CDOM ports are therefore likely to be less common in the Atlantic Ocean. Second, verification could fail if a ship ballasts in a humic-rich port and retains 5% of this water following BWE, since residual port water could significantly elevate the total ballast water signal. Assuming BWE were performed with 95% efficiency in the mid-ocean where C3<sup>∗</sup> is around 0.5 QSE, then ships that originally ballasted in ports where C3<sup>∗</sup> > 10 QSE would have C3<sup>∗</sup> above 1 QSE. Relatively high-CDOM ports with C3<sup>∗</sup>

> 10 QSE were uncommon in our dataset (<10% of measured tanks), although would presumably be more common had ships originated from Atlantic ports. To limit the loss of sensitivity that inevitably would result from a one-size-fits-all BWE threshold, a forensic approach considering multiple lines of evidence was employed in this study.

The chemical signature of exchanged ballast tanks was shown to be very sensitive to ballast exchange efficiency. Previous research indicates that BWE efficiencies vary by ship type and according to the method of exchange. Using the emptyrefill method, exchange efficiencies exceeding 98% are typical, however, flow-through exchange allows mixing between the incoming and outgoing water and often results in exchange efficiencies well below the mandated level. Increasing BWE efficiency from 95 to 98% decreases the port signal by more than half, whereas decreasing BWE efficiency from 95 to 90% doubles it. At the same time, biological risk is similarly sensitive to exchange efficiency. If the presence of 5% coastal organisms in ballast water represents the upper limit of acceptable risk, then accepting BWE with 90% efficiency results in twice the acceptable risk, and 85% BWE triples it.

The strength of the pre-BWE signal is also critical for determining the chemical profile of an exchanged ballast tank, even when oceanic water becomes 20 times more abundant than coastal water following BWE. Thus, for moderately fluorescent ports with C3<sup>∗</sup> = 5 QSE, a two-fold increase in pre-BWE C3<sup>∗</sup> has a similar effect on the post-exchange signal as a two-fold increase in open ocean C3<sup>∗</sup> . Accurately estimating the pre-BWE signal for individual ships is difficult, since the water quality conditions encountered by individual ships while ballasting in port are subject to a number of sources of uncertainty, including temporally and spatially variable processes affecting terrestrial inputs (Stedmon et al., 2006; Yamashita et al., 2008). The picture is further complicated in ships that top up or transfer ballast water between tanks, which produces a blended chemical profile of indeterminable origin. For these reasons, it is difficult to conclusively identify ships that misreport BWE except in relatively extreme cases or when directly comparable measurements happen to be available. Approximately 10% of ships fell into this category in this study, although due to the generally conservative assumptions used in calculations together with the high prevalence of relatively low FDOM ports along the Pacific Rim (Murphy et al., 2009; Doblin et al., 2010), this probably represents a lower limit of BWE reporting/implementation errors.

Whereas conclusively determining BWE compliance by specific ships is often difficult, a meta-analysis of the chemical data is consistent with the finding that 95% BWE is not being performed as frequently as ships report. If this were not the case, then the distribution of measured C3<sup>∗</sup> in unexchanged ballast tanks (**Figure 3**) should largely overlap with the pre-BWE C3<sup>∗</sup> distribution back-calculated from C3<sup>∗</sup> measured in exchanged ballast tanks. Instead, high-CDOM (C3<sup>∗</sup> > 15 QSE) source ports were at least seven times more common in the calculated vs. measured pre-BWE datasets. Overall, the results suggest that a significant rate of reporting errors occur due to a combination of factors, including inadequate BWE and unintentional or deliberate misreporting of ballast water management.

Experience from the Great Lakes of North America suggests that compliance by ships with BWE legislation is strongly linked to inspection effort (Bailey et al., 2011). Whereas, our earlier research established the scientific basis for using fluorescence spectroscopy to trace ballast water origin, this is the first study to move this technique to the level of implementation and demonstrate how the technology works when implemented by governmental inspectors. In-situ FDOM sensors have recently entered the market and offer the possibility of simple real-time measurements as long as instrument reliability, stability, and calibration issues are appropriately handled. Incorporating such measurements into inspection programs at Pacific rim ports could improve the detection of high-risk ballast water and the overall implementation of BWE in the region.

#### AUTHOR CONTRIBUTIONS

GR, MN, and KM conceived of the overall study and experimental design. MN performed the field trials and acquired the data in this study with assistance from others as described in the Acknowledgements. Statistical analyses were performed by MN and KM. KM and MN drafted the article and all authors revised it for intellectual content. All authors approve of the final version and are accountable for its accuracy.

#### REFERENCES


#### FUNDING

Funding for this project was provided by California State Lands Commission (CSLC), Washington Department of Fish and Wildlife (WDFW), Oregon Department of Environmental Quality (ODEQ), New Zealand Ministry of Primary Industries (MPI), US Coast Guard Research and Development Center (RDC), and National Sea Grant Ballast Water Demonstration Program, Department of Commerce Award # NA050AR4171066.

#### ACKNOWLEDGMENTS

The authors are indebted to many people and agencies that assisted with ballast water sampling during this project. We are grateful to Rian Hooff from ODEQ for assistance in Oregon. From CSLC we extend our thanks to Chris Beckwith, Tom Burke, Robert Chatman, Bob Chedsey, Nicole Dobrosk, Maurya Falkner, Ricky Galeon, Daphne Gehringer, Gary Gregory, Jackie Mackay, Chris Scianni, Bob Shilland, and David Stephens, amongst others. From WDFW we thank Gary Gertsen and Allen Pleus. From MPI we thank Clive Imrie, Stu Rawnsley, Greg Williams, Touzelle Batkin, Brendon Wakeman, Owen Aspen, Kevin Hawkes, Jeff O'Neil, Gary Higgins, Kristy Jacob, Tim Das, and others. Assistance with planning and organizing in New Zealand was provided by Chris Denny, Andrew Bell, Liz Jones, and Naomi Parker. Jennifer Boehme, Chris Brown, Darrick Sparks and Ashley Arnwine at SERC assisted with sampling and analyses.


invasions in the contiguous United States. Bioscience 61, 880–887. doi: 10.1525/bio.2011.61.11.7


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Noble, Ruiz and Murphy. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Long-term Patterns of Eelgrass (Zostera marina) Occurrence and Associated Herbivorous Waterbirds in a Danish Coastal Inlet

Thorsten J. S. Balsby 1, 2 \*, Preben Clausen<sup>2</sup> , Dorte Krause-Jensen<sup>1</sup> , Jacob Carstensen<sup>3</sup> and Jesper Madsen<sup>2</sup>

*<sup>1</sup> Department of Bioscience, Aarhus University, Silkeborg, Denmark, <sup>2</sup> Department of Bioscience, Aarhus University, Rønde, Denmark, <sup>3</sup> Department of Bioscience, Aarhus University, Roskilde, Denmark*

Seagrasses in coastal areas have substantial importance for the marine environment and also serve as food for herbivorous waterbirds. We investigated potential relationships between the autumn population of herbivorous waterbirds and eelgrass (*Zostera marina*) abundance in the EU protected area, Nibe-Gjøl Bredning, a broad of the Limfjorden estuarine complex in Denmark.This is an important site for migratory herbivorous waterbirds such as mute swan (*Cygnus olor*), coot (*Fulica atra*), brent goose (*Branta bernicla*), and wigeon (*Anas penelope*). We explored long-term (27 years) changes in eelgrass and bird-populations and relationships between eelgrass- and bird abundance. We applied trend- and correlative analyses of yearly monitoring data on eelgrass and waterbirds between 1989 and 2015 coupled with estimates of the potential grazing pressure exerted by the birds. Around 1990 eelgrass was abundant in this area covering more than 40 km<sup>2</sup> , but eelgrass coverage and biomass declined drastically around 1995 and remained low until 2011 when natural recolonization accelerated and by 2015 had restored the lost meadows. The number of herbivorous waterbirds also fluctuated substantially during the monitoring period with large abundance until the mid-end 1990s followed by reduced abundance in the 2000s and recovery after 2010. The number of bird-days showed a positive relationship with the same year's eelgrass abundance in the 1–2 m depth stratum. For the 0–1 m depth stratum, where the eelgrass meadows are most exposed to bird grazing but also to physical control from e.g., wind and ice, only a particularly detailed eelgrass data set available for a subset of the study period, showed a significant relationship with bird grazing. The potential waterbird consumption of eelgrass, estimated by multiplying average intake rate and number of bird-days for each species, ranged from less than 16% of the eelgrass biomass in most years to more than 40% of the eelgrass biomass in years with extremely sparse eelgrass populations. Hence, the study suggests that dense eelgrass populations stimulate herbivorous waterbirds whereas top-down control is only likely when abundant bird populations graze on sparse eelgrass populations.

Keywords: eelgras, herbivore waterbirds, plant herbivore interactions, eelgrass consumption, eelgrass biomass, eelgrass cover, waterbird consumption

#### Edited by:

*Maria C. Uyarra, AZTI Tecnalia, Spain*

#### Reviewed by:

*Anna R. Armitage, Texas A&M University at Galveston, USA Javier Franco, AZTI Tecnalia, Spain*

> \*Correspondence: *Thorsten J. S. Balsby thba@bios.au.dk*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *16 June 2016* Accepted: *19 December 2016* Published: *10 January 2017*

#### Citation:

*Balsby TJS, Clausen P, Krause-Jensen D, Carstensen J and Madsen J (2017) Long-term Patterns of Eelgrass (Zostera marina) Occurrence and Associated Herbivorous Waterbirds in a Danish Coastal Inlet. Front. Mar. Sci. 3:285. doi: 10.3389/fmars.2016.00285*

# INTRODUCTION

Seagrass meadows are important features of coastal ecosystems (Hemminga and Duarte, 2000) and are increasingly recognized for their vital role as ecosystem engineers, because their structure and biomass reduce hydrodynamic energy (e.g., Bouma et al., 2005), increase sedimentation (e.g., Gacia et al., 2003; Bos et al., 2007) and stabilize sediments (Fonseca, 1989), preventing coastal erosion (e.g., Adriano et al., 2005) and increasing water clarity (Maxwell et al., 2016). Moreover, seagrass meadows constitute significant carbon stocks (Duarte et al., 2013), serve as habitats and hatching/nursery areas for a wealth of organisms, and are an important source of food for herbivores such as nonbreeding herbivorous waterbirds (Baldwin and Lovvorn, 1994; Ganter, 2000; Heck and Valentine, 2006). However, seagrass meadows have declined in many parts of the world over the past decades (Waycott et al., 2009) in response to stressors such as eutrophication and sediment load from land (Orth et al., 2006), threatening ecosystem services provided by the meadows.

Eelgrass meadows in the temperate zone can be of considerable ecological importance for migratory waterbirds. They stage during autumn or spring migration or during winter for several months in areas with eelgrass meadows. This is indeed also the case in Denmark where thousands of swans, brent geese, dabbling ducks and coots congregate during autumn, winter and spring (Laursen et al., 1997). In most areas they feed on Zostera or other rooted macrophyte resources, especially Ruppia spp. Potamogeton pectinatus and Charophytes (Clausen and Percival, 1998; Madsen, 1998a; Holm, 2002; Meltofte and Clausen, 2011).

The availability of eelgrass as grazing resource is thus of significance for numerous migratory waterbirds, many of which are protected under international conventions and legislation. In Denmark a consequence of this is that a national comprehensive reserve network has been designed specifically to include important seagrass meadows as feeding habitats for these birds in Special Protection Areas under EU legislation designated for them (Madsen et al., 1998). If waterbirds follow an aggregative response (Hassell and May, 1973; Sutherland, 1983), where birds gather at sites with large food densities, then one would expect that herbivorous waterbirds would follow changes in eelgrass cover and distribution.

When the birds graze on seagrasses and other macrophytes it is evident that they can remove substantial amounts of biomass. Dos Santos et al. (2015 and references therein) reports values between 20 and 80% of the standing biomass, and comparable values have been found in Danish studies (Kiørboe, 1980; Madsen, 1988). Such grazing pressures could potentially, in turn, affect macrophyte cover leading to a top-down control of macrophytes by birds in addition to a potential bottomup control of birds by macrophyte abundance. However, lack of combined long-term data on waterbirds and macrophyte abundance has limited the number of studies investigating such relations with a few exceptions, which has found positive relations between the abundance of herbivorous waterbirds and macrophytes (e.g., Petersen et al., 2008; Meltofte and Clausen, 2011). Bird grazing could potentially also affect nutrient cycling in shallow bays by removing the nutrients contained in the grazed eelgrass biomass, while bird droppings are a source of nutrients.

We studied long-term changes in the abundance of eelgrass meadows and herbivorous waterbirds and potential relationships between them in Nibe-Gjøl Bredning (Limfjorden, Denmark) over a period of 27 years encompassing large fluctuations in both eelgrass and bird populations. We hypothesized that abundant bird populations related to rich eelgrass meadows, whereas birds could exert top-down control on sparse eelgrass meadows. We further hypothesized that potential effects of birds on nutrient fluxes are insignificant relative to nutrient loadings from land. We acknowledge that several other factors also affect eelgrass and herbivorous waterbirds. Hence, physical exposure to e.g., wave action affects eelgrass cover particularly in shallow areas and renders it highly variable, whereas light availability is a major regulating factor at greater depth (Krause-Jensen et al., 2000, 2003). Human disturbances and change in the flyway population may also affect the number of waterbirds. Such disturbance factors may disrupt a relation between eelgrass and waterbirds, but as the majority of these factors only affect either eelgrass or waterbirds, but not both, it is unlikely to drive the relationship between the two. Ice cover is an exception that may affect both eelgrass- and bird abundance, but it is only an issue from mid-December after most of the bird counts have been conducted. We tested the hypotheses based on trend-and correlative analyses of yearly monitoring data on eelgrass and waterbirds between 1989 and 2015 supplemented with estimates of the potential consumption of eelgrass by birds. First we describe the development in eelgrass cover and biomass, and the number of bird-days over the years. Then we compare the estimated eelgrass cover and biomass in Nibe-Gjøl-Bredning with the estimated consumption of eelgrass biomass by waterbirds. Finally we explore relationships between eelgrass and bird abundance in different depth strata. We analyzed the 0–1 m and the 1–2 m depth strata separately, because most bird species only graze eelgrass in the shallow stratum, which is also more exposed to wave action and ice scouring during winter, as opposed to the deeper meadows (Frederiksen et al., 2004). The deeper eelgrass meadows may further serve as a buffer zone stabilizing the shallow part of the eelgrass population (Olesen et al., 2016).

# METHODS

# Study Site

Nibe-Gjøl Bredning is situated in the eastern part of the large Danish Limfjorden estuarine complex (57◦ 02′N, 9◦ 37′E). It is the core area of the "Ulvedybet and Nibe Bredning" European Union Special Protection Area No. 1 and Ramsar site No. 7 in Denmark (Skov-og Naturstyrelsen, 1996). Most of this internationally protected area is situated in the Limfjorden, but adjacent saltmarshes, the brackish lake Ulvedybet with surrounding wetlands, and some agricultural areas in the upland are also protected. Most of the study area consists of shallow (<2 m deep) brackish estuarine waters (**Figure 1**), which around 1990 supported one of the largest known eelgrass Zostera marina beds in Europe. At this time, eelgrass covered almost the entire shallows, with dense populations at depths ranging from approx.

FIGURE 1 | Aerial photo of Nibe-Gjøl Bredning in 2014, with transects used for Zostera mapping in the NOVANA programme (blue) and by Madsen (1998b, red), with a few other mentioned site names. Inserted figure shows the cumulative area with substrate suitable for eelgrass for each 0.2 m interval. Stippled yellow lines encircles the main feeding areas for wigeon in the early 1990s, the herbivorous species in the study site that has the shortest neck and thus is most dependent on shallow water *Zostera* (redrawn from Madsen, 1998b).

25 cm and down to around 2 m, totaling ∼45 km<sup>2</sup> (Madsen, 1998a). Eelgrass populations experienced major decimation by the wasting disease in the 1930s and again due to eutrophication peaking in the 1980s (e.g., Krause-Jensen et al., 2012). Adjacent saltmarshes are typical Danish Juncus gerardi, Festuca rubra, and Puccinellia maritima dominated meadows (sensu Vestergaard, 1998), of which 52.9% were classified as well-managed by livestock grazing or hay-cutting in 2008 (Clausen et al., 2013a).

A comprehensive baseline study of impacts of humans on staging waterbirds in the area was conducted during 1985– 1988 (Madsen, 1998a). A following study examined the effects of banning or regulating the two primary sources of human disturbance (hunting and windsurfing) within the experimental reserves implemented during 1989–1992, and the establishment of a permanent reserve regulating both these activities from 1993 onwards (Madsen, 1998b; Clausen et al., 2014). The reserve establishment led to massive increases in numbers of staging waterbirds, especially quarry species, notably in the years 1990– 93 (Madsen, 1998b), but after the late 1990s, bird numbers fell dramatically (Clausen et al., 2014).

Three of the numerically most important bird species found in the area are herbivores, i.e., mute swan Cygnus olor, light-bellied brent goose Branta bernicla hrota and Eurasion wigeon Anas penelope, and although the similarly common Eurasian coot Fulica atra is an omnivore, we know from direct observations and distributional analysis from the area that all four species have a strong preference for feeding on Zostera in the study site (Madsen, 1998a; Clausen et al., 2013b), because it represents the most energetically favorable and easily accessible food source (e.g., Brunckhort, 1996; Clausen et al., 2013c).

#### Bird Counts and Bird-Days

Staging waterbirds have been counted in the area on an annual basis during 1985–2015. Intensive count coverage, involving one or more counts per month during August-November/December was established during the baseline-experimental reserve study years from 1985 to 1993 (Madsen, 1998a,b), and in conjunction with national reserve monitoring programmes in the years 1994–2003 and 2008–2010 (Clausen et al., 2014). During the years 2004–2007 and 2011–2015, the site was only counted in October as part of the annual national dabbling duck and brent goose count of the Danish National Monitoring and Assessment Program for the Aquatic and Terrestrial Environments (NOVANA)(Holm et al., 2015), when observers were also instructed to count swans and coots. The majority of counts were land-based total counts of waterbirds, where flocks of birds were identified, counted and drawn onto field-maps using 20–60× telescopes from a number of observation points in the upland, in some years supplemented with counts from observation towers (details in Madsen, 1998a; Clausen et al., 2014). Some counts were carried out by two observers from single-engine Piper or Cessna airplanes, using the "total count" method (Pihl and Frikke, 1992).

Annual autumn estimates of bird-days, i.e., the total number of birds observed multiplied by the numbers of days they were present, were calculated for the four herbivorous species. This was estimated by multiplying the mean count per month by the number of days in the month, and then summed for August-November for each year of the periods with intensive counts. For the remaining years, we used October counts as a predictor of total number of bird-days in the autumn, because October counts showed a significant positive linear regression with high explanatory power (R <sup>2</sup> > 0.83) with the total number of birddays per year in the years with intensive counts (**Table 1**). Based on this regression we estimated the total number of bird-days for the years where only October counts had been conducted, except for 2004–2007 where only wigeon and brent geese were counted, and estimates for mute swans and coot could not be made.

#### Eelgrass Data

The regional monitoring authorities conducted the eelgrass surveys in accordance with national guidelines for survey of eelgrass cover as part of the NOVANA programme. During all surveys, eelgrass cover at specific depths along transect lines was estimated visually in the field by divers or subsequently in the laboratory from underwater videos. In Nibe-Gjøl Bredning, a total of 9 different transects were surveyed over the period 1989– 2015. Between 1989 and 1997 only 2 transects were surveyed, whereas 4 transects were surveyed between 1998 and 2001, and 5 to 9 transects were surveyed per year from 2002 until 2015. Before 2001 eelgrass cover was estimated as an average for depth intervals along the transect lines, whereas estimates after 2001 were given as point observations recorded continuously along the transect lines with information on water depth for each point. For the analysis we divided the eelgrass in two strata 0–1 m and 1–2 m, relative to the normal water level (DVR90).

Additional detailed data on seagrass coverage were collected along 10 transects in Nibe Bredning and Gjøl Bredning (Madsen, 1998b and unpublished). These transects differed in geographical position and extent from the NOVANA transects mentioned above (**Figure 1**). Transects were surveyed in August in 1988– 1997 and 2001 (except 5 transects in Gjøl Bredning, where monitoring started in 1989). The cover and distribution of Zostera marina and Ruppia spp. (R. maritima and R cirrhosa, combined) were estimated at points with 100 m intervals along the transects with a radius of approx. 8 m from a boat using a large Aquascope Underwater Viewer. The position of stations was determined using a Decca (until 1992) or a GPS navigator (since 1993). We used these data to validate the less detailed eelgrass cover estimates from the NOVANA surveys prior to 1998. The eelgrass cover at 0–1 m depth estimated from the NOVANA survey showed a significantly positive relationship with the eelgrass cover estimated from the surveys by Madsen (1998b) (General linear model, R <sup>2</sup> = 0.427, F(1, 8) = 5.96, p = 0.041, slope = 0.679). On this basis we use the long-term NOVANA data throughout the study to describe eelgrass abundance in Nibe-Gjøl Bredning. However, as these transect were located within the known areas where herbivorous water birds foraged and since there was some variation between the two eelgrass surveys, we decided to also test these in relation to bird-days and consumption.

Based on the eelgrass cover data, we estimated the eelgrass biomass (g dry weight per square meter) using the empirical model by Carstensen et al. (2016). This model predicts eelgrass biomass from Secchi depth and depth-specific eelgrass cover. We used a Secchi depth of 2.9 m, which is the mean Secchi depth measured over several years in Nibe-Gjøl Bredning by the regional monitoring authorities. Model parameters were adopted from Carstensen et al. (2016). By combining the depth-specific



*The linear regressions were made without intercept.*

cover estimate with the area of each 0.5 m depth interval with suitable eelgrass substrate (i.e., soft and sandy substratum) we estimated the potential eelgrass biomass per year in the entire Nibe-Gjøl Bredning in October. This estimate assumes that all the substrate suitable for eelgrass was colonized by the percent eelgrass cover estimated for each year. However, observations confirm that the actual area colonized varied markedly over the study period. In 1993 about 45 km<sup>2</sup> was colonized by eelgrass whereas only 4.6 km<sup>2</sup> of the suitable area was colonized in 2001 (**Figure 2**). The estimates of eelgrass biomass are influenced by the variation in the area colonized by eelgrass. However, as we lacked annual data to quantify this variation we used assessments of the proportions of the suitable area being colonized by the eelgrass.

#### Water Level

For non-diving herbivorous waterbirds such as mute swan and brent goose, and poorly diving species such as coot, the water level determines, which areas are available for foraging (Clausen et al., 1996; Clausen, 2000). The same is true for the non-diving wigeon, which often feed on spilled plant-materials from foraging swans or coot (Holm and Clausen, 2009 and references therein). Fluctuations in water level are therefore important to quantify, in order to assess foraging abilities in Nibe-Gjøl Bredning. The water level at Nibe-Gjøl Bredning is mainly affected by wind. Winds from west push North Sea water into the Limfjorden through the opening at Thyborøn. With a tidal amplitude around 20 cm, lunar tides have marginal effect on the water level in the Limfjord, but wind surges often affects water levels and will occasionally invoke water levels below −50 cm (easterly winds) and above +150 cm (westerlies)(Clausen, 1998).

Water level has been measured at two nearby locations Øland/Attrup and Nibe. The Nibe station and the Attrup station are located on the south and north side of the Bredning, respectively. However, none of the stations cover the full duration of the study period (Øland/Attrup 522327 observations; from 19/4/1996 to 31/12/2013; Nibe: 364894 observations; from 5/2/1993 to 8/8/2007). Due to absence of data we could not assess water-levels effect on potential foraging days during 1989- 1992 and 2014-2015. Nibe is the station closest to many of the observed areas, and missing water level data for Nibe were estimated using a regression between Nibe and Attrup water level measures in August to December 1996 to 2007, when both stations were active. This relationship was linear for water levels between −50 cm and 100 cm relative to the normal water

level (DVR90, Danish vertical reference 1990. www.sdfe.dk). To estimate missing water level data for Nibe at extreme (and rare) water levels below −50 cm or above 100 cm at the Attrup station, we used the associated mean water level of the extreme water levels at Nibe in the years where in the years where it had been monitored.

# Eelgrass Consumption by Herbivorous Waterbirds

We estimated Zostera consumption by the four waterbirds using standard methods based on body mass, daily energetic needs and known digestion rates. We used body mass estimates from Cramp and Simmons (1977, 1980) for wigeon (0.70 kg) and coot (0.76 kg) and an average autumn mass of 1.6 kg for lightbellied brent goose (from Clausen et al., 2012), and subsequently estimated the birds daily energy expenditure, DEE in kJ/day by three different allometric relationships. The first is from Drent et al. (1978/79):

$$(1)\,\text{DEE} = 2.6^{\circ}\text{BMR},$$

where BMR is the birds basal metabolic rate. The BMR was estimated after the allometric regression for non-passerines in Lasiewski and Dawson (1967): BMR = 78.3∗Mb 0.723 in kcal/day <=> BMR = 327.6∗Mb 0.723 in kJ/day, where M<sup>b</sup> is the birds body mass in kg.

The second relationship follows Walsberg (1983):

$$\text{(2) DEE} = 12.84^\circ \text{M}\_\text{b}^{0.61} \text{ in kJ/day},$$

where M<sup>b</sup> is the birds body mass in g.

The third relationship follows Nagy (1987; allometric regression for non-passerines):

$$\begin{aligned} \text{(3)} \,\text{DEE} &= 10^{\log \text{FMR}} \text{ in kJ/day, where } \log \text{FMR} \\ &= 0.681 + 0.749^\* \log \text{(M}\_{\text{b}}\text{)}, \end{aligned}$$

where M<sup>b</sup> is the birds body mass in g, and the common logarithm log<sup>10</sup> is used for computations.

These estimates of DEE give slightly different values for the three species, and it is not obvious from the most recent paper or associated literature whether the one or the other is the more appropriate value. We therefore used the average value of the three DEE-computations for our estimates. The energy content of the food, E<sup>f</sup> , was set at 14.154 kJ/kg dry weight for Zostera marina leaves (Christensen et al., 1994).

The birds' daily Zostera food intake, DFI, in kg dry weight, to cover their daily energetic needs was then estimated as:

$$\text{DFI} = \text{DEE}^\* \text{PP}^\* 100 / \text{E}\_\text{f}^\* \text{D},$$

where D is the birds' digestion rate in % and PP is the proportion of the birds' food that we expect is derived from plant materials. Leaf-eating birds' digestion rates are generally relatively low (typically 25–40%), and we used a value of 36% for brent goose (average of 37% given by Drent et al., 1978/79, and 35% by Madsen, 1988), 46% for wigeon (Madsen, 1988), and 27% for coot (Hurter, 1979). The lower value for coot is probably explained by their mixed diet and shorter gut. For wigeon and brent goose we assumed a 100% seagrass diet, whereas for coot we used a value of 50%, as reported for coot feeding in a Danish estuary (Christensen et al., 1994). The resulting daily consumption estimates (dry weight/day) are 233 g for brent goose, 130 g for wigeon and 93 g for coot.

For mute swan we used a comprehensive study on the nutritional energetics of a seagrass-dependent and molting swan population in eastern Denmark in 1993–1995 providing a best estimate of daily seagrass consumption of 487.7 g dry weight/day (range 352.2–620.0 g dry weight/day, Clausen et al., 1996). Most of the variation of the estimate is caused by the fact that estimated daily energy expenditure for a 10.75 kg bird is quite different and lower if based on computations by Walsberg (1983), but higher if based on Drent et al. (1978/79) or Nagy (1987), whereas values for the other species are almost identical. The largest uncertainty in these calculations thus remains with the swan consumption.

Estimates of daily consumption rates and bird-days were multiplied to estimate consumption estimates for the autumn. In order to infer potential effects of bird-grazing on nitrogen (N) dynamics we used literature data on the N-content of eelgrass biomass.

#### Statistics and Data Analysis

We used mixed models to estimate the least square mean eelgrass cover per year in Nibe-Gjøl Bredning where year was a fixed factor and transect was a random factor. The mixed model allowed us to account for variation in the number of transects surveyed per year by including transect as a random factor in the model. We estimated the average eelgrass cover for the depth interval 0–1 m, with potential large grazing effects, and for the 1–2 m depth interval, which is less accessible for bird grazing except for mute swan. We choose to not use deeper strata as the eelgrass below 2 m would never be accessible to non-diving waterbirds.

The least square means for eelgrass estimates for all of Nibe-Gjøl Bredning for cover and for potential eelgrass biomass (estimated on basis of the cover) were related to the number of bird-days and total consumption per year. We used general linear models and mixed models to test the relations between eelgrass and herbivorous waterbirds. The model tested the relation between number of bird-days and eelgrass cover for both the 0–1 m and the 1–2 m depth intervals. We used both bird-days and estimated consumption to describe and quantify the potential relationships with eelgrass abundance, because species differ in consumption rates so some species would have larger potential impact than others. This analysis also included the number of bird-days and consumption from the gap-years, which were estimated using regression.

All tests have been conducted in SAS 9.3 (SAS Institute, Cary, NC) using proc glm and proc mixed.

# RESULTS

### Bathymetry and Water Level

Nibe-Gjøl Bredning consists of extensive shallow areas with substrate suitable for eelgrass. The majority of the suitable area is less than 1 m deep and covers 45 km<sup>2</sup> (65.5%) of our study area (**Figure 1**). Of the remaining area suitable for eelgrass, 10.6 km<sup>2</sup> (15.4%) occurs between 1 and 2 m depth, 9.1 km<sup>2</sup>

(13.2%) between 2 and 4 m, and the remaining 4.1 km<sup>2</sup> (5.9%) at greater depths.

A mute swan can feed down to 1.15 meters below the surface (Clausen et al., 1996) and a brent goose to 40 cm (Clausen, 2000). Analysis of the fluctuations in water levels in Nibe-Gjøl Bredning between August and December 1993 to 2013 showed that on average 50% of days had water levels below 0 (relative to normal), and another 30% in the range 0–0.20 m, whereas there were only 12.7 (range 3–24) days per autumn with water levels higher than 0.5 m. Assuming that eelgrass shoots typically are 0.5–1 m long (as found by Clausen et al., 1996; Clausen, 2000), this means that swans could reach the entire 0–1 m stratum most days, and even the deeper stratum out to at least 2 meters during some days. The other species feed at shallower depths, or by diving (coot) or by association with the swans (wigeon and brent geese). Fluctuations in water level did not cause major changes in accessibility of eelgrass in Nibe-Gjøl Bredning. So if all the suitable area were colonized by eelgrass, the majority of the eelgrass would be available to herbivorous waterfowl during staging between August and December.

## Fluctuations in Eelgrass Levels over the Years 1989–2015

In Nibe-Gjøl Bredning, eelgrass cover of the depth strata 0–1 m and 1–2 m varied hugely over the period 1989–2015 (**Figures 2**, **3**). The period 1989–2000 was characterized by wide fluctuations in both strata with peaks of about 45% cover in the shallow stratum around 1989 and 1998 and a peak of about 90% cover in the deeper stratum in 1993. From 2000, there followed a period with relatively stable low eelgrass cover of 10– 20% at 0–1 m and 30–40% at 1–2 m until around 2010. Since then eelgrass cover has increased steeply in both strata to the current levels of about 90% at 0–1 m and 75% at 1–2 m depth (**Figure 3**). Since 2000, eelgrass at 0–1 m depth has followed a similar trend to eelgrass at 1–2 m, whereas before 2000 the two strata showed opposite trends, i.e., when eelgrass cover increased in the 1–2 m stratum, it decreased in the 0–1 m stratum and vice versa (**Figure 3**). Although the eelgrass cover within the two strata was correlated (Pearson correlation r<sup>26</sup> = 0.40, p = 0.041) there were substantial differences between the two strata. Overall the eelgrass cover was significantly lower in the 0–1 m stratum compared to the 1–2 m stratum (paired t-test t<sup>25</sup> = 4.58, p < 0.001).

The potential eelgrass biomass per year in the 0–1 m stratum followed the pattern described for the eelgrass cover; whereas fluctuations in biomass were much smaller for the 1–2 m stratum, reflecting that the area of the suitable habitat is much smaller at 1–2 m compared to 0–1 m (**Figure 3**).

# Fluctuations in Number of Birds over the Years 1989–2015

The herbivorous waterfowl species coot, mute swan and wigeon all exhibited the highest number of bird-days before 1995, whereas brent goose peaked in 1998. The total number of autumn bird-days peaked in 1993 at 2.3 mio bird-days. The number of bird-days then declined steadily until 2002/2003, after which it remained low until at least 2011 (**Figure 4**). The peaks of wigeon,

cover used for these estimates was based on transects with a 0.5 m stratification. Note that these estimates of eelgrass biomass assume that all suitable areas have been colonized. However, for several years only parts of the suitable area was colonized.

coot, and mute swan bird-days in 1993 and 1995 were 1–2 orders of magnitude larger than for other herbivorous species, but after 2000 the total for all species declined to less than 86,000 bird-days.

The October counts followed the same pattern as described for the number of bird-days per autumn until 2011. Interestingly, a steep increase in the number of mute swans and especially wigeon was observed in Nibe-Gjøl Bredning in 2015 (**Figure 5**), whereas coot did not increase as steeply.

# Relationships between Eelgrass Cover and Waterbird Abundances and Estimated Biomass Consumption Based on Average Values for Nibe-Gjøl Bredning

The number of bird-days estimated from counts of herbivorous bird species and the associated estimated consumption of eelgrass (ton dry weight per year) related positively to the eelgrass cover at the 1–2 m stratum, with bird numbers and estimated bird consumption explaining 28 and 42% of the variation in eelgrass cover, respectively (**Figure 6**, **Table 2**). However, in the 0–1 m

stratum, which is primarily exposed to grazing, the bird-days and estimated consumption did not relate to the eelgrass cover (**Figure 6**, **Table 2**).

Bird-days, however, showed a significant positive relation with the more detailed eelgrass cover survey encompassing 5– 10 transects in 1988–1997 and 2001 [General linear model, Bird-days: R <sup>2</sup> = 0.62, F(1, 8) = 12.9, p = 0.007, slope = 36.7; Consumption: R <sup>2</sup> = 0.60, F(1, 8) = 11.9, p = 0.009, slope = 4.7] (**Figure 6**). These eelgrass transects were mainly located in the shallow areas with less than 1 m depth, and within the primary feeding distributions of the observed herbivorous birds (**Figure 1**, see **Figure 9** also).

Likewise, the number of bird-days and the related eelgrass consumption showed a significantly positive relation to the

estimated total eelgrass biomass at the 1–2 m stratum (**Table 2**), while there was no significant relationship for the 0–1 m stratum (**Table 2**). It should be noted that the estimated total eelgrass biomass assume that all suitable area would be colonized, which as previously mentioned was not the case in all years.

Repeating the above analyses by relating bird abundance and grazing to the previous year's eelgrass abundance, i.e., assuming a one-year lag in the birds' response, did improve the relationships slightly (**Table 2**).


TABLE 2 | Relations between bird days, consumption and eelgrass cover and estimated eelgrass biomass.

*The relations test the bird eelgrass relation within the same year, but also test if eelgrass affects the number of birds the following year. All tests were made using general linear models.*

# Consumption of Eelgrass by Waterfowl and Recycling of N

The estimated eelgrass consumption by herbivorous water birds decreased from around 200 tons dry weight/year in the beginning of the study period to <40 tons dry weight/year, between 2000 and 2014 (**Figure 7**).

The consumption of eelgrass can mobilize and recycle a substantial amount of N. An estimate of the N content is 2.37% of the dry weight in new leaves and 0.82% in old leaves (Pedersen and Borum, 1992). Using these estimates, the N removed by eelgrass consuming herbivorous waterbirds varied from as little as 0.03–0.08 tons N per year in the period 2000–2014 up to a maximum of 2.57–7.44 T N per year in 1993, the range reflecting the variability in eelgrass N-content with age (**Figure 7**).

#### Proportion of the Eelgrass Consumed

If we assume that all the suitable area is colonized by eelgrass and that at least the mute swan could reach the eelgrass down to 1.5 m, then the herbivorous waterfowl consumed between 0.2 and 9.4% of the available biomass in the area (**Figure 8**). The highest consumption occurred in 1993 and 1994, where we know at least in 1993 that most of the suitable habitat was indeed covered by eelgrass (**Figure 2**). However, during the years with declining and low eelgrass cover it is unlikely that all the suitable area was colonized by eelgrass. Furthermore, aerial photos suggest that the areas colonized by eelgrass went through substantial reductions in this period. Therefore, we estimated the available eelgrass biomass with scenarios where smaller proportions of the suitable area had been colonized (**Figure 8**). For the period with sparse eelgrass meadows exemplified by 1996 then the herbivorous waterbirds could consume up to 73% of the available biomass if only 20% of the suitable area had been colonized by eelgrass, and if they only foraged on eelgrass; and if 50% of the suitable area had been colonized, then the grazing pressure would reach up to 29% of the available biomass. However, for all years since 1998 the grazing pressure is less than 16% of the standing biomass if eelgrass colonized between 20 and 100% of the suitable area (**Figure 8**).

#### DISCUSSION

This study demonstrates major fluctuations in abundances of eelgrass and herbivorous waterbirds over a 27-year period in the protected area Nibe-Gjøl Bredning. The data suggests that dense

eelgrass meadows may stimulate the populations of herbivorous birds, even though our correlative approach does not document a causal relationship. On the other hand, the results indicate that during periods when large bird populations forage on sparse eelgrass meadows, the birds may exert top-down control with a grazing pressure of up to 73% of the standing biomass if eelgrass meadows colonized only 20% of the available area and a grazing pressure of up to 29% of the biomass if the meadows colonize 50% of the available area.

Our study demonstrates that the number herbivorous waters birds show a strong relation the availability of eelgrass in Nibe-Gjøl Bredning. Similar positive relationships between numbers of herbivorous waterbirds and benthic vegetation have also been demonstrated at other sites involving different species of birds and aquatic plants (Nienhuis, 1992; Petersen et al., 2008; Meltofte and Clausen, 2011).

## Relationships between Abundances of Eelgrass and Water Birds

Overall, the number of bird-days showed a positive relationship to the eelgrass cover at 1–2 m depth, but not to eelgrass cover at the most heavily grazed 0–1 m depth stratum for the NOVANA data. The number of bird-days and consumption showed a positive relation to the additional detailed eelgrass cover estimates made on the shallow foraging areas (**Figures 1**, **7**). This relation existed despite only 9 years of observations of eelgrass cover.

The lack of a relationship for the 0–1 m depth stratum for the NOVANA data may have several causes. One may be that the regression assumes that the relation between eelgrass cover and number of waterbirds is the same throughout the time series and this may not be the case, as birds could exert topdown control during periods with sparse eelgrass populations. Eelgrass meadows at 0–1 m are also more affected than those at 1–2 m by physical disturbances from ice scour in winter, wave action (Krause-Jensen et al., 2003) and possibly also from drifting macroalgae and burrowing fauna which may hamper the establishment of seedlings (Valdemarsen et al., 2010). As eelgrass meadows become sparse they also loose resilience, and feed-back mechanisms may act to maintain the state of reduced cover, e.g., through increased sediment resuspension (Maxwell et al., 2016). The temporal mismatch between the measure of eelgrass cover (generally monitored in August-September) and the grazing pressure, which peaks later in the season, may further reduce the chance of identifying potential relationships between the two factors. Large year-to-year variations in numbers of our four focal waterbird species also affect the relationship to eelgrass cover, with numbers of mute swans and coot in Denmark being heavily influenced by variations in mortality reflecting winter severity (Holm et al., 2015), numbers of wigeon being influenced by large annual variations in breeding success (Fox et al., 2016), and numbers of brent geese being affected by both these factors (Clausen et al., 1998). Hunting also add to the variability of bird populations because 1989, the first year in our time-series, was also the first year with a reserve, and there could be some initial lag-responses from the bird populations on the new reserve (Madsen, 1998b).

The location of the NOVANA transects relative to the feeding areas in the early study years could also be part of, and perhaps even the best explanation for the lack of a relation. This is so because the non-NOVANA transects surveyed in the shallow areas where birds were observed to forage did show a positive relation (**Figure 6**). This stress the importance of vicinity between eelgrass transects and the areas used by the herbivorous waterbirds.

The deeper eelgrass meadows, on the other hand, showed positive relationships to bird populations probably because these meadows were less at risk of top-down control, and also less affected by variability caused by physical exposure. The reduced physical stress and lower grazing pressure at 1–2 m depth where light is still available may also explain why eelgrass populations at this depth were more stable than the shallower populations, with average cover values never declining below 25%. The more stable populations at 1–2 m depth may serve as a buffer for the shallow populations by attenuating waves and by producing seeds that may facilitate recolonization (Olesen et al., 2016).

The steep increase in eelgrass cover since 2011 was not immediately reflected in increased numbers of herbivorous waterbirds, but the newest data from 2015 document the largest populations of herbivorous birds for 17 years. Hence, there seems to be a time lag in the birds' response to altered foraging possibilities, reflecting that the waterbird populations need time to discover and respond to the recovered eelgrass meadows. The eelgrass decline from the mid-end 1990s was followed by a decline in number of waterbirds a few years later, also suggesting a lagged response to the eelgrass decline. Introducing a general lag-phase of 1 year in the analysis of the birds' response to eelgrass abundance improved the relationships for the 1–2 m strata slightly, but had no effect on the 0–1 m strata (**Table 2**). This suggests that birds may learn and return to favorable foraging sites during migration in the following years. However, the birds' response time probably differs depending on species-specific habitat preferences and associated availability of alternative sources of food, or on Zostera beds outside our study area. Indeed, in 2008–2010, during the period with sparse eelgrass, brent geese and wigeon were more frequently observed foraging on neighboring salt marshes, or on Zostera beds near

FIGURE 9 | Two maps showing the relative distribution of wigeon, the most numerous herbivore, in our study site in 1991 (redrawn from Madsen, 1998b where maps for three other years are given), and 2008–2010 (average for 3 years, from Clausen et al., 2014). Note the different scales, and that the birds are plotted in a 500 m × 500 m grid for 1991, but in a 1 × 1 km grid for 2008–2010. Far more ducks were present in 1991. Reserve regulations are highlighted. In 1991 systematic mapped counts were only carried out in the area within the stippled rectangle, but gray-literature and citizen science portal data from Ulvedybet and the area west of Egholm suggest numbers in these areas were low during the 1990s, and most birds used the reserve in Gjøl Bredning (Clausen et al., 2014).

Egholm 5–10 km east of our study area (**Figure 9**, Clausen et al., 2014). The mute swans rarely and the coots only to a lesser extent use salt marshes as alternative feeding habitats, and have in the to a larger extent abandoned the overall site including the fjord habitats around Egholm (Clausen et al., 2014), just as it has been the case in Ringkøbing Fjord, where declines in numbers of these two species are more prominent than for brent geese and dabbling ducks during a period with reduced aquatic vegetation (Meltofte and Clausen, 2011).

# Grazing and Fluctuations in Eelgrass Populations

Already in 1992 and 1993 the local authorities reported brown or dead patches of eelgrass in shallow areas of Nibe-Gjøl Bredning as well as reduced depth limits in the broad and hypoxia in deeper Limfjord basins (Agger et al., 1994). Subsequently, in the midend 1990s the meadows of Nibe-Gjøl Bredning went through major declines (**Figures 2**, **3**), in line with the generally poor conditions of eelgrass populations in the eutrophic Limfjorden at the time (Krause-Jensen et al., 2012). Hence, it is evident that the eelgrass populations were unhealthy and that their marked decline was not caused by waterbird grazing. However, from 1995 to 1998 when the eelgrass meadows were under decline a substantial number of birds still foraged in Nibe-Gjøl Bredning. Although these may have foraged partly on tasselweed Ruppia sp., which temporarily showed increased cover in the area between 1994 and 1996 (Madsen, 1998b), it remains possible that the herbivorous waterbirds may have exerted a level of grazing pressure that could have accelerated the decline of the already weakened eelgrass population, since our analyses suggest a grazing pressure up to 73% of the standing biomass during this period.

For the 2000's when the eelgrass populations remained poor and bird populations were reduced, our analyses showed that the potential eelgrass consumption by waterbirds amounted to less than 16% of the standing eelgrass biomass, making it unlikely that grazing by waterbirds could hinder the recovery of the eelgrass population in general. Also the cumulative production of eelgrass biomass in Danish waters is 2.5 to 3.6 times the maximum biomass in a year (Sand-Jensen, 1975; Olesen and Sand-Jensen, 1994a), so the proportion of the production grazed was much less than the 16%.

The steep increase in eelgrass cover from 2011 to 2015 documents that substantial seed-based recolonization suddenly took place. Such fast recolonization can only have happened with the involvement of successful seed dispersal and establishment, as vegetative dispersal solely results in linear rates of expansion averaging 0.16 m per year from the edge of existing patches (Olesen and Sand-Jensen, 1994b). Evidence from aerial photos further documents the extremely fast spread of eelgrass in Nibe-Gjøl Bredning from 2011 to 2014 (http://arealinformation. miljoeportal.dk/distribution/).

Between 2012 and 2015, the eelgrass cover in the 0–1 m strata exceed the eelgrass cover observed in 1991 and earlier, whereas the eelgrass cover at the 1–2 m strata is within the range of earlier observations. The abundance of herbivorous waterbirds have not yet returned to the Nibe-Gjøl area in such numbers witnessed in the early 1990's, which may explain why the eelgrass at the 0–1 m stratum has reached such high coverage. Hence, it may be expected that if numbers of herbivorous waterbirds increase in Nibe-Gjøl Bredning, we may see a decline in eelgrass coverage at the 0–1 m strata. Such a decline may not necessarily imply that the eelgrass population is in a bad state, but may instead indicate that the ecosystem is approaching equilibrium between the standing crop and production of the eelgrass and numbers of grazing herbivorous waterbirds.

Several other studies have documented that grazing by herbivorous waterbirds affect the submerged vegetation in terms of leaf length, and below ground biomass (Bortolus et al., 1998), and above ground biomass (Rivers and Short, 2007; Dos Santos et al., 2015) but temporal removal of plant material does not necessarily have a permanent impact on the plants. Some aquatic plants such as Potamogeton pectinatus are in fact extremely "tolerant" to intense and annual grazing by swans, and may overcompensate and thus produce better in the presence of grazing (Nolet, 2004). In addition to removal of eelgrass, grazing also affects nutrient cycling. The birds remove nitrogen along with their removal of the biomass. While the waterbirds retain a proportion of the N they consume from the eelgrass, only a relatively modest proportion of the consumed N will be released immediately for recycling in the environment. Assuming digestability of soluble protein in eelgrass is comparable to meadow grasses, 61–80% will be digested by the birds (Buchsbaum et al., 1986) and used either as an energy source or to build tissue. Only the remaining 20– 39% will discarded as uric acid or undigested plant fragments in the feces. The amount of N consumed may amount up to 6 T per year, which is likely insignificant in comparison with the 1649–2438 tons of N supplied from the contributing catchment (Windolf et al., 2013).

In conclusion, the study area has undergone substantial fluctuations in abundance of both eelgrass and herbivorous waterbirds over the past 27 years. Our results suggest that dense eelgrass populations stimulate the number of herbivorous waterbirds, whereas the waterbirds may exert top-down control when eelgrass meadows are sparse and waterbird populations are large.

# AUTHOR CONTRIBUTIONS

TB, DK, PC, and JC analyzed the data and drafted the manuscript. JM started the bird counts and the eelgrass survey that was not part of the NOVANA survey. All authors contributed to writing the paper.

# ACKNOWLEDGMENTS

We would like to thank Jens Sund Laursen for finding the latest data on eelgrass in Nibe-Gjøl Bredning, Svend Aage Berntsen and Kirsten Elisabeth Broch, Naturstyrelsen for providing data on water level, the many observers counting waterbirds at Nibe-Gjøl Bredning, but especially Jens Peder Hounisen, Ebbe Bøgebjerg, Henrik Haaning Nielsen and Jørgen Peter Kjeldsen. We would like to thank Tony Fox for constructive comments on an earlier version of the manuscript and Jesper Bladt for assistance with ARCGIS. TB, DK, and JC received support from the DEVOTES project funded under the EC 7th framework program (grant

#### REFERENCES


agreement no. 308392). DK and JC also received support from the BONUS COCOA project funded jointly by the EC and the Danish Research Council, and DK also from the NOVAGRASS project funded by the Danish Council for Strategic Research. PC was part funded by the Danish Nature Agency.

and Conclusions. Commissioned report to Øresundskonsortiet. National Environmental Research Institute.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer JF and handling Editor declared their shared affiliation, and the handling Editor states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2017 Balsby, Clausen, Krause-Jensen, Carstensen and Madsen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Predicting the Composition of Polychaete Assemblages in the Aegean Coast of Turkey

Marika Galanidi\*, Gokhan Kaboglu and Kemal C. Bizsel

*Institute of Marine Sciences and Technology, Dokuz Eylül University, ˙ Izmir, Turkey*

Benthic infaunal species and communities have been extensively used to evaluate quality of the marine environment. Within the MSFD, community composition is addressed most commonly through Descriptor 6 (Seafloor integrity), criterion 6.2 (Condition of benthic communities). At the same time, the Directive has stipulations for addressing and assessing indicators linked with pressures in an explicitly spatial manner. At larger scales, achieving this through point sampling may be impractical or unfeasible; hence predictive methods are being increasingly employed to produce the large scale spatial data that are often required for marine spatial planning and management. The aim of the current work was to develop statistical and spatial modeling tools that can predict the distribution of soft-sediment benthic polychaetes in the Aegean coast of Turkey. To do that, we employed Species Archetype Models (SAMs), a novel analytical and modeling framework which uses mixture models to cluster species responses to the environment, producing a number of "archetypal" responses assumed to represent species with similar ecological/physiological tolerances. Polychaete presence/absence data were obtained from the literature and modeling was performed against environmental variables reflecting the main natural and anthropogenic gradients in the region. The resulting models are interpreted in light of the sensitivity/tolerance classification scheme for benthic invertebrates. Three Species Archetypes were identified through the analysis. In brief, Species Archetype 1 consists of the most prevalent species in the dataset and primarily follows the salinity and temperature gradients. Species Archetype 2, present in the central and southern Aegean, is dominated by sensitive and indifferent species and responds negatively to chlorophyll a, whereas Species Archetype 3 represents mostly tolerant and opportunistic polychaetes with increased probability of occurrence in eutrophic, shallow, inshore areas throughout the region. Predictive performance was constrained by the information contained in our data. These results from a limited data set show promise that SAMs as a modeling tool can offer valuable insights into patterns of benthic species distribution and coexistence and increase our capacity to provide predictive advice.

Keywords: polychaetes, species archetype model, composition, community-level model, Aegean, benthic invertebrates, ecological groups, soft-sediment

Edited by: *Angel Borja, AZTI, Spain*

#### Reviewed by:

*Silvana Noemi Raquel Birchenough, Centre for Environment, Fisheries and Aquaculture Science, UK Genoveva Gonzalez-Mirelis, Institute of Marine Research, Norway*

> \*Correspondence: *Marika Galanidi marika.galanidi@deu.edu.tr*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *15 June 2016* Accepted: *15 August 2016* Published: *31 August 2016*

#### Citation:

*Galanidi M, Kaboglu G and Bizsel KC (2016) Predicting the Composition of Polychaete Assemblages in the Aegean Coast of Turkey. Front. Mar. Sci. 3:154. doi: 10.3389/fmars.2016.00154*

# INTRODUCTION

The Marine Strategy Framework Directive is the current legal framework under which EU Member States are required to assess and protect the health of the marine environment with the ultimate target of achieving "Good Environmental Status" (EU, 2008). Understanding the links between the different components of marine ecosystems and the pressures resulting from human activities is a key element of the conceptual framework underpinning the MSFD and a prerequisite for the effective management of the seas (Smith et al., 2014; Berg et al., 2015). At the same time, the Directive has stipulations for addressing and assessing indicators linked with pressures in an explicitly spatial manner. This has a number of implications for the design of monitoring and assessment strategies, starting from defining meaningful and relevant ecological scales to demonstrate state-pressure links (Lynam et al., 2015) and establish the "naturalness" or natural background variability of the system in order to set reference conditions for assessment or environmental targets for management (Van Hoey et al., 2010). Strengthening this knowledge base can be greatly aided by modeling tools and predictive methods for the distribution of species and communities according to their responses to environmental parameters (Reiss et al., 2015).

Benthic infaunal species and communities, due to their space-use patterns and well-documented relationships with various forms of ecological stress (Gray and Elliott, 2009), are a biotic group that has been proven highly appropriate to demonstrate natural and anthropogenic impacts to the seabed and is extensively used to evaluate the quality of the marine environment (Muxika et al., 2005; Dimitriou et al., 2012; HELCOM, 2013). Community composition of the benthos is addressed within the MSFD most commonly through Descriptor 6 (Seafloor integrity), criterion 6.2 (Condition of benthic communities) and is currently better addressed by what we could call traditional ecological indices (Piroddi et al., 2013). These are based on the Pearson and Rosenberg paradigm (Pearson and Rosenberg, 1978) that describes the responses and succession of benthic invertebrates to the organic enrichment gradient and generally rely on the proportion of opportunistic to sensitive species in benthic samples (Carletti and Heiskanen, 2009). Modeling capabilities that can produce information and indicators to address community condition are currently lacking (Piroddi et al., 2015; Tedesco et al., 2016) and the development of habitat suitability models is encouraged in order to address this gap (Tedesco et al., 2016).

Current approaches for the assessment of soft-sediment benthic habitats within the framework of the MSFD include 3 steps, (i) defining habitat types, (ii) setting reference/target conditions taking into account the natural background variability of each habitat, and (iii) choosing suitable indicators for assessment (Van Hoey et al., 2013). With respect to the first step, predictive modeling of community composition has been extensively employed to integrate the biological components of the seabed with physical characteristics in order to define and map benthic habitats. This is usually accomplished with some application of the "assemble first–predict later" approach, sensu Ferrier and Guisan (2006), whereby communities or species groups are usually first delineated through algorithmic multivariate analyses, followed by modeling of these entities against environmental parameters (Degraer et al., 2008; Buhl-Mortensen et al., 2014, 2015; Gonzalez-Mirelis and Buhl-Mortensen, 2015; Gogina et al., 2016; Rubidge et al., 2016). While variations of this strategy have served the modeling community well, they have their shortcomings and limitations. For instance, distance-based methods have been criticized for confounding location and dispersion effects by misspecifying the mean-variance relationship in the data (Warton et al., 2012) with severe consequences in identifying "characteristic" taxa and detecting multivariate effects. As another example, sites that are not well classified in the "assemble" step and are commonly assumed to represent transitional communities result in entities that are either poorly predicted (Gogina et al., 2016) or are discarded from the "predict" step (Degraer et al., 2008), resulting in information loss that can be substantial for limited data sets.

A recent surge of interest in model-based methods for the analysis of multivariate data has resulted in an expanding number of statistical tools that offer an alternative to sitebased multivariate analyses by making species or species groups the response unit, thereby allowing formal description and inference about their relationship with environmental variables and a greater flexibility in modeling species co-existence (Warton et al., 2015b). These include methods for unconstrained ordination (Hui et al., 2014; Hui, 2015), correspondence analysis (Pledger and Arnold, 2014), exploring community-environment associations and fitting predictive models (Wang et al., 2012), model selection (Madon et al., 2013). A major challenge however when modeling multiple species responses to the environment is how to reduce the number of coefficients in the model based on ecologically meaningful criteria (Dunstan et al., 2013b; Warton et al., 2015a), both for computational but also for interpretability reasons. For soft sediment benthos this is even more pertinent since, with the exception of habitat forming and alien/invasive species, it is the response of the assemblage that is usually of greater interest.

A statistical approach that achieves model reduction by "seeking shared patterns in environmental filtering" (Ovaskainen et al., 2016) is Species Archetype Models (SAMs) (Dunstan et al., 2011). SAMs is a regression-based modeling method which uses mixture models to cluster species responses to the environment, producing a number of "archetypal" responses assumed to represent species that have similar ecological/physiological tolerances. As such they are likely to respond to pressures in a similar manner and consequently require similar management measures. SAMs have been used so far for purposes of bioregionalization (Woolley et al., 2013), exploring competing concepts of community assembly (Leaper et al., 2014) and investigating fish-assemblages' responses to bottom trawling pressure (Foster et al., 2015).

In the current study Species Archetype Models are employed to examine group responses of soft-sediment polychaetes to natural and anthropogenic gradients and produce modeling tools to predict their distribution in the Aegean coast of Turkey. Aiming to present modeling outputs that can be interpreted according to the same principles underlying the assessment of benthic community condition, the resulting models are presented in light of the sensitivity/tolerance classification scheme for benthic invertebrates (Grall and Glémarec, 1997). Thus, this approach brings together elements of habitat modeling with the concept of sensitivity to disturbance (used here as a collective term that encompasses a number of different impact sources eliciting similar responses–Rosenberg, 2001; Muxika et al., 2005; Josefson et al., 2009; De Backer et al., 2014). Combined with the ability of SAMs to model simultaneously species clustering with their response to the environment, it is believed that this modeling framework can offer valuable insights into the drivers of benthic assemblage composition and the scales at which significant patterns occur.

### METHODS

#### Data

#### Benthic Data

The biological data consist of a matrix of polychaete presence/absence (total of 327 taxa) by 52 sites. Samples were collected in August 2011 with a Van Veen grab sampling an area of 0.1 m<sup>2</sup> during a pollution monitoring project —see (Çinar and Dagli, 2013) for details–and the locations are presented in **Figure 1**. Species present in at least 5 stations were divided into Ecological Groups according to the sensitivity/tolerance classification scheme as summarized in Grall and Glémarec (1997).

Group I: Species very sensitive to organic enrichment and present in normal conditions.

Group II: Species indifferent to enrichment, always present in low densities with non-significant variations in time.

Group III: Species tolerant of excess organic matter enrichment. These species may occur in normal conditions but their populations are stimulated by organic enrichment.

Group IV: Second-order opportunistic species. These are the small species with a short life cycle, adapted to a life in reduced sediment where they can proliferate.

Group V: First-order opportunistic species. These are the deposit feeders that proliferate in sediments reduced up to the surface.

Species were assigned sensitivity scores (Ecological Group shown in Table S1) following a national database constructed specifically for the benthic invertebrates of Turkish waters (TUBITAK-MRC and MoEU-GDEM, 2014; Çinar et al., 2015). The choice of a local

classification database as opposed to a more widely used one (e.g., the AMBI classification database) was driven by findings that species can shift their tolerances between biogeographic regions with different dominant environmental gradients (Zettler et al., 2013) and the need to adjust species classification to ecological groups identified for the Aegean by Simboura and Reizopoulou (2007) and Çinar et al. (2012).

#### Environmental Data

A number of publically available gridded data were screened and predictors were chosen (**Table 1**) based on their relevance to benthic infauna distribution patterns. Sea Surface Temperature is used to define broad biogeographical regions (related to species distributions at geological timescales) and, together with salinity define different water masses, whereas bottom temperature and salinity are directly relevant to species' physiological tolerances. Salinity can be particularly important in areas of substantial freshwater input such as shallow and estuarine environments (Reiss et al., 2015). In this study, we used winter (February data) and summer (August data) bottom temperature and salinity in order to capture the seasonal gradients observed along the coast. Euphotic depth determines the depth at which photosynthesis can occur and macrophytes can exist. This can be important for the benthos, especially in the current study where a number of samples were collected from within or around Posidonia oceanica beds. Primary production is the most common descriptors of trophic condition for marine waters.

TABLE 1 | Environmental variables considered for modeling polychaete assembalges (abbreviations in brackets used throughout the document, variables highlighted with gray were removed from the analysis due to collinearity problems).


*a (Sbrocco and Barber, 2013) http://www.marspec.org/.*

*<sup>b</sup>http://www.emodnet-hydrography.eu/.*

*<sup>c</sup>http://emis.jrc.ec.europa.eu/.*

*d (Micheli et al., 2013) https://www.nceas.ucsb.edu/globalmarine/mediterranean. <sup>e</sup>http://msi.nga.mil/NGAPortal/MSI.portal?\_nfpb*=*true&\_pageLabel*=*msi\_portal\_page\_ 62&pubCode*=*0015.*

As far as pressure variables are concerned, chlorophyll a concentration was regarded as a measure of eutrophication. Demersal destructive fishing is one of the most important anthropogenic activities with well-known and documented impacts on the benthos (Kaiser et al., 2006; Hiddink et al., 2009). The fishing pressure index included in this study was constructed from spatial disaggregation of landings data and is used in full knowledge of its limitations and criticisms (Halpern et al., 2008; Heath, 2008) especially regarding its reliability to represent impacts on the seabed. Distance from major ports was considered as a pressure variable encompassing a number of processes, such as increased turbidity and disturbance from re-suspension, shipbased pollution and the presence of exotic/invasive polychaete species transferred by ballast water, an issue well-documented in Izmir Bay (Çinar et al., 2006, 2012).

All environmental layers we initially clipped to the extent of the eastern Aegean pilot area (**Figure 1**) defined for the DEVOTES project (http://www.devotes-project.eu/study-sites/), projected to geographic projection Lambert Azimuthal Equal Area centered on the Mediterranean sea and resampled by bilinear interpolation to a common resolution of 1 km<sup>2</sup> . Rasters were subsequently masked by the extent of the soft sediment areas at depths of 0–200 m, (layer extracted from https://www.nceas.ucsb.edu/globalmarine/mediterranean). Finally, model predictions were restricted to soft-sediment areas of the seabed extending up to 10 km from the shore, in order to reduce the extent of unsampled locations in covariate space. 10 km is an arbitrarily chosen distance which covers all sampled locations (largest distance from the coastline for any sampling site was approximately 9 km). Data coverage was not the same for all environmental layers; poor fit between the end of the predictor layer and the coastline were filled by interpolation. Point values of the predictor variables at the 52 stations were extracted from continuous layers with the extract function of the raster package (ref) in R. Spatial analysis was carried out in R (R Core Team, 2014) and SAGA GIS (Conrad et al., 2015).

Chlorophyll a and distance from ports were log(x+1) transformed prior to analysis. Collinearity of the environmental variables was investigated with the Variance Inflation Factor method with a cut-off threshold of 3 and with Spearman's correlation coefficient >0.7. Primary production, euphotic depth and chlorophyll a were highly collinear; retaining chlorophyll a in the predictor data set reduced the VIF to acceptable levels and was considered preferable as it indicates both trophic conditions and eutrophication pressure. Similarly, Sea Surface Temperature and Sea Surface Salinity were correlated with each other (at r > 0.7) and with bottom temperature and were discarded from further analysis, as was winter bottom salinity with summer bottom salinity. It was considered that high summer bottom salinity better reflects stress imposed on benthic organisms within the range of salinity values investigated and was thus included in the data set. For final mapped outputs of selected predictor variables (**Table 1**) see Figure S1.

#### Modeling

Multi-species responses to environmental variables were modeled by mixing generalized linear models (GLMs) employing finite mixture models in a method developed by Dunstan et al. (2011) and implemented in the R package "SpeciesMix" (Dunstan et al., 2013a). The finite mixture model has the capability to identify species with statistically indistinguishable responses to environmental gradients, parametrize the individual GLMs and cluster them into one or more common generalized linear models without any supervision. The resulting models are termed Species Archetypes and may represent one or many species that have similar ecological/physiological tolerances. The estimation of the group composition occurs simultaneously with the estimation of the shared response, such that the fitted models return both the probability of a species belonging to a particular species archetype and the model components (coefficients and standard errors) of the archetypal GLMs which describe the response of that entire group to the environment (Leaper et al., 2014).

Because species are classified into archetypes in a probabilistic manner (Hui et al., 2013), it is possible by a species to be represented by more than one archetype, and the probabilities of species membership to archetypes (tau) are an indication of how well a species response is aligned with each archetypal response. In order for all species to be assigned to an archetype, membership in the present study was defined based on a probability threshold of 0.5, however, the implications of different membership probabilities are discussed later in the manuscript, following Foster et al. (2015), who consider a probability>0.8 to mean that a species is strongly affiliated to the particular Archetype, whereas probabilities between 0.5 and 0.8 are interpreted as indicative of species membership to an Archetype.

In our analyses linear and quadratic terms were considered in order to increase the flexibility of the models to capture species responses (Leaper et al., 2014; Woolley et al., 2013) and covariates were standardized to zero mean and unit standard deviation to avoid dimensional issues. The biological response matrix included taxa present in at least 5 stations (109 taxa). Model selection was performed in two steps. (I) Determining the number of archetypes (G) was achieved by comparing models fitted with all covariates (including linear and quadratic terms) with different numbers of species archetypes; the model with the lowest Bayesian Information Criterion (BIC) is considered the most parsimonious and is used to select the number of archetypes. Multiple starts were performed in order to avoid convergence at local maxima (Dunstan et al., 2013b). (II) Further variable selection was based on minimisation of the BIC and examination of the Standard Errors (SEs) of the coefficients for the model terms (with optimum G determined in the previous step). Coefficients with high relative standard errors (RSE) are considered of lower importance. Moreover, coefficients which never exhibited low SEs and RSEs for any of the Species Archetypes were preferentially removed from the model. The model with the lowest BIC was retained for predictive mapping of the resulting Species Archetypes' distribution.

Covariate effects of the final model predictors were also visualized with partial effects plots were the effect of each covariate is modeled separately, while all other covariates are held at their respective means.

The performance of the models was evaluated for individual species. We did not fit individual species models. Rather, expected probabilities of occurrence were calculated for every species i per station j (εij) from the predicted probability of occurrence of each archetype per station and the species tau for each archetype (given in Table S1) according to the formula.

$$\mathbf{e}\_{i\circ} = \mathbf{p}\_{j\text{SA1}} \times \mathbf{t} \mathbf{u} \mathbf{u}\_{i\text{SA1}} + \mathbf{p}\_{j\text{SA2}} \times \mathbf{t} \mathbf{u} \mathbf{u}\_{i\text{SA2}} + \mathbf{p}\_{j\text{SA3}} \times \mathbf{t} \mathbf{u} \mathbf{u}\_{i\text{SA3}}$$

where p<sup>j</sup> is the probability of occurrence of each archetype for station j and tau<sup>i</sup> is the membership probability of species i for each archetype.

Predictions were obtained with the function predict. archetype in the SpeciesMix package from the fitted archetype models, hence no individual model selection was performed.

The terms were added, since the combined probability of presence of each species in all archetypes (sum of tausi) cannot be higher than 1 and the probability of presence of each archetype in every station is independent of the probability of presence of all other archetypes in the same location.

Once expected and observed frequencies were obtained and tabulated per species and station, model performance was assessed in two ways. Model accuracy was assessed with Spearman's correlation coefficient and discriminatory power (ability to correctly predict absences and presences) with the Area Under the receiver operator characteristic (ROC) Curve with the package modEvA (Barbosa et al., 2015).

#### RESULTS

Initial model selection was performed for values of G ranging from 1 to 8 and the number of Archetypes that minimized the Bayesian Information Criterion was G = 3 (BIC = 5727, Table S1). Subsequent variable selection for three Species Archetypes retained summer bottom salinity, winter bottom temperature, depth, chlorophyll a and distance from port as predictors. Quadratic terms for bottom temperature and salinity were also included in the most parsimonious model with BIC = 5714.

Archetype membership for most of the species was well estimated with probability values close to 1, particularly for Species Archetype 1 (Table S1). Out of the 109 modeled taxa only 8 were indicatively affiliated with an Archetype (0.5 < tau < 0.8) and most of them belong to Species Archetype 3. Species Archetype 2 (SA2) contains the highest number of species (S = 60), followed by Species Archetype 3 (SA3) with 28 species and Species Archetype 1 (SA1) which represents the 21 most frequently observed species.

In terms of composition, all three archetypes seem to be dominated by species belonging to ecological groups EGII (disturbance indifferent) and EGIII (tolerant) (**Figure 2**). What differentiates them is the higher relative representation of EGII in Species Archetype II, the relative proportions of the other ecological groups and the species identities of each archetype (see Table S1). A close inspection of histograms in **Figure 2** reveals two clear trends; an increase in disturbance sensitive and indifferent species as we move from SA3 to SA1 and finally SA2 and a respective decrease in the relative proportion

of tolerant and opportunistic species. Species identities are also important in understanding the composition of the archetypes; thus Species Archetype 1 contains no first-order opportunists but species indicative of transitional assemblages, such as Lumbrineris geldiay, Monticellina heterochaeta, Sigambra tentaculata, Aricidea claudiae. On the other hand, first-order opportunists (EGV), such as Prionospio fallax and Heteromastus filiformis, are found in Species Archetype 3, together with the second-order opportunists (EGIV) Mediomastus, Lanice conchilega, Pseudopolydora pulchra, and Podarkeopsis galangaui and many disturbance tolerant species. Two notable exceptions here are the species Schistomeringos rudolphi and Spio decoratus, which, although they are opportunists, were classified in Species Archetype 2 (SA2), where one can find all but one of the sensitive to disturbance species (EGI), a much lower proportion of EGIII species and only these two opportunists.

The regression coefficients and associated standard errors (**Table 2**) describe the relationship of the archetypes with the environmental variables, whereas the shape of the responses is illustrated in the partial effects plots of **Figure 3**. The importance of each covariate is considered to increase as its relative standard error decreases. Thus, Species Archetype 1 is predominantly determined by summer bottom salinity, displaying a strong quadratic response with highest probability of presence at intermediate salinity values. At the same time, it responds negatively to chlorophyll a and depth but these variables are of secondary importance. Species Archetype 1 is the most widespread of the three archetypes and follows mostly the large scale environmental gradients in the region (**Figures 4A,D**, Figure S1). Its strong affinity with bottom salinity results in the avoidance of the less saline Northern Aegean waters and a higher probability of occurrence in the central Aegean. It is also less likely to be encountered in the eutrophic waters of inner and middle Izmir Bay and in the deeper parts of the south-eastern Aegean.

The response of Species Archetype 2 to the environment seems to be better defined by the covariates, judging by the size of the relative SEs. Its probability of presence generally increases with winter bottom temperature to level off close to the high temperature values encountered in the southernmost region of the region (Figure S1). It is also more likely to be present in deeper waters, away from ports and responds negatively to chlorophyll a concentration (**Figure 4B**). Species Archetype 2 has a more patchy distribution, predicted with the lowest degree of uncertainty (prediction SE.SA2, **Figure 4E**). Even though it contains most of the modeled species (present in 5-21 stations), it is characterized by the lowest probabilities of occurrence. It shows a higher affinity for deeper, more exposed areas of the coastline, particularly of the central and southern Aegean and is predicted to be mostly absent from shallow, inshore areas.

Species Archetype 3 is characterized by weaker associations with the predictor variables, it does however respond positively to chlorophyll a and shows a similar response pattern to bottom salinity as SA1. Its probability of occurrence decreases linearly with depth but increases with increasing distance from ports. Species Archetype 3 has a very localized predicted distribution with moderately high probability of presence in a few inshore areas throughout the whole Aegean coast (**Figures 4C,F**). These are regions that coincide with areas of high chlorophyll a concentration (Figure S1). While there is a certain degree of overlap in the distribution of archetype 1 with the other two archetypes, model predictions for Species Archetypes 2 and 3 indicate that they almost never occur together. Uncertainty in the estimation of the model parameters was higher for archetype 3 and had a similar magnitude and spatial pattern as SA1.

The predictive accuracy of the Species Archetype models for individual species is generally moderate, with less than half the modeled species displaying significant Spearman correlation coefficients between observed and predicted values (**Table 3**, Table S1). Predictive performance increases not only with decreasing species prevalence, (with more species from Species Archetypes 2 and 3 being accurately predicted and very few from SA1), but also with the number of species represented by each archetype. Similar behavior is observed for the discriminatory power of the models, which is moderate for archetypes 1 and 3 but markedly better for archetype 2 with more than 60% of its species being very well predicted with AUC>0.7.

#### DISCUSSION

Species Archetype Modeling is a model-based approach that classifies species objectively, in an unsupervised way, into groups according to their responses to environmental (Woolley et al., 2013; Leaper et al., 2014) and/or pressure gradients (Foster et al., 2015). In the current study an attempt was made to integrate the influence of environmental and relevant pressure parameters on benthic species distribution patterns and assess the resulting predictive models. The Species Archetypes that emerged are interpreted on the basis of a long-standing classification scheme of their member species along the tolerance to disturbance gradient. Additional functional traits were not considered since the sensitivity scores essentially synthesize a number of morphological, life-history and life-style traits


TABLE 2 | Coefficients (coeff), Standard Errors (SE) and Relative Standard Errors (RSE) for the environmental predictors of the most parsimonious model for all Species Archetypes.

*Values in bold indicate variables with low relative standard error (RSE), considered as the most important in determining the Species Archetype. Abbreviations as in* Table 1*. <sup>2</sup>quadratic terms.*

(Paganelli et al., 2012) and can represent different combinations of their modalities.

The species archetypes arrived at by this modeling technique are not "traditional" communities as commonly defined per site based on species composition. They are groups of species with statistically similar responses to environmental parameters and their presence is not mutually exclusive. The final assemblage in a location will result from the co-occurrence of species from all the Archetypes that are likely to be found in that environmental setting. As such, it does not need to conform to pre-defined "community types" and this offers increased flexibility to predict co-occurrences in scenarios of environmental conditions that have not yet been encountered (Ferrier and Guisan, 2006). Polychaete species indifferent or tolerant to pollution are able to exist in a large range of environmental conditions and this is demonstrated by their common membership in all three Species Archetypes. The mixed composition of the archetype groups corroborates the long-standing knowledge that Mediterranean benthic fauna is generally evenly distributed with no one species naturally displaying strong dominance (Carletti and Heiskanen, 2009). In the case were presence/absence data is used instead of abundances, these differences may be even less pronounced (Muxika et al., 2012), particularly for EGII species which are "always present in low densities" and EGIII species, which "may occur in normal conditions but their populations are stimulated by organic enrichment." With the above in mind, the relative distribution and dominance of groups EGII and EGIII in the three archetypes is not surprising. However, it is the relative proportion of the other Ecological Groups that differentiates the three archetypes, as well as their responses to environmental variables, discussed in more detail below.

Species Archetype 1, with its widespread distribution, primarily reflects the main biogeographic gradients in the Aegean, determined by salinity, temperature and basin/sub-basin scale circulation patterns (Durrieu de Madron et al., 2011). The colder, more productive waters of the Northern Aegean above the Dardanelles, resulting from the inflow of riverine waters along the northern coast and brackish, rich Black Sea water through the Turkish Straits system, constitute a distinct sub-region of the study area (Velaoras and Lascaratos, 2010; Sayın et al., 2011), where SA1 is less likely to be found compared to the Central Aegean. Another faunistically distinct region is observed where the Aegean meets the Levantine Sea and extends to the Bodrum Peninsula. The south-eastern Aegean is the point of entry of warm, hypersaline Levantine waters, where the thermohyaline and atmospheric forcing result in a clear and oxygen rich water column and reduced organic carbon fluxes to the bottom (Lykousis et al., 2002), creating an "ocean margin" environment. SA1 therefore could be regarded as a persistent faunal group of the central Aegean coast of Turkey which avoids the most stressful environments. It may be argued that the delineation of Species Archetype1, consisting of the most common taxa, is largely determined by species prevalence. This modeling issue, resulting from mixing GLMs on all the parameters including the intercept, has been acknowledged in previous studies (Dunstan et al., 2013b; Hui et al., 2013) and may mask, to some degree, the strength of association of species based on shared responses. However, member species of archetype 1, such as Lumbrineris geldiay, Monticellina heterochaeta, Sigambra tentaculata, Aricidea claudiae have been reported to occur together and characterize transitional assemblages in previous studies of Izmir Bay (Dogan ˘ et al., 2005; Ergen et al., 2006; Çinar et al., 2012) and Edremit Bay (Albayrak et al., 2007) at locations that concur with the current distribution; it is hence believed that SA1 is a valid response group. Moreover, the absence of any second-order opportunists and the relative proportion of the other Ecological Groups would place it in the slightly to moderately disturbed conditions of a degradation model for benthic community health (see Borja et al., 2000; Simboura and Zenetos, 2002; Muxika et al., 2012).

Modeled species at the two ends of the disturbance tolerance spectrum are almost never grouped together under the same Archetype. Stress sensitive and indifferent species with high ecological requirements dominate Species Archetype 2 which is primarily characterized by its negative response to chlorophyll a, a reliable indicator of eutrophication (HELCOM, 2009) and its higher affinity for the south-eastern Aegean. Previous studies of the area have reported rich fauna with high diversity values (Ergen and Çinar, 1994; Pancucci-Papadopoulou et al., 1999;

Oku¸s et al., 2007) and attributed it to the hydrodynamic regime and the influence of Levantine waters. The composition of SA2 points to undisturbed/slightly disturbed conditions, or an assemblage of good to high status. Although the coexistence of opportunists (in low numbers) with sensitive species is predicted by degradation models for benthic faunal structure at slightly disturbed conditions (Grall and Glémarec, 1997; Borja et al., 2000), the inclusion of Schistomeringos rudolphi and Spio decoratus in SA2, both of which are documented opportunists and pollution indicators (Simboura and Zenetos, 2002; Çinar et al., 2015), is somewhat inconsistent with the general patterns of the current results and problematic for prediction purposes. A possible reason could be the use of presence/absence data in the current study that contains less information than abundance data, which, for opportunistic species in particular, is a determining property. Abundance data are anticipated to produce response curves and membership patterns that will more accurately represent species responses to pressures, particularly eutrophication. Another reason could be population fluctuations of these species (e.g., Dogan et al., ˘ 2005; Çinar et al., 2006, 2012, for comparison with previously reported data), and a lack of temporal replication to capture these fluctuations.

Species Archetype 3 is a species group devoid of sensitive taxa and with the highest overall representation in opportunistic species generally regarded as pollution indicators. In contrast

with the EGIV species of archetype 1, which are relatively long-lived, free living predators, the opportunists of SA3 are burrowing or tube-building deposit-feeders, characteristic of reduced sediments. The predicted extent of SA3 largely coincides with shallow, inshore areas in response to their high chlorophyll a concentration. This is particularly true for bays and gulfs with high urban pressure and around estuaries and lagoons, where sediment and nutrient loads are generally high, confirming expectations of where high numbers of tolerant and opportunistic species may be found. More specifically, Saros Bay receives nutrient rich discharges from the Evros (Meric) river and Black Sea water, Edremit Bay is impacted by increasing summer housing development, olive oil industry and bottom trawl fisheries (Kucuksezgin et al., 2013), Izmir Bay is an intensely urbanized and industrialized area with a major shipping port and the Büyük Menderes river delta receives large amounts of, mostly untreated, municipal and industrial waste through the adjoining river basin (Yesilirmak and Anac, 2008). The Dardanelles strait is a water body heavily impacted by sewage pollution and intense shipping traffic (Ate¸s et al., 2014) and, even though benthic data from the Dardanelles were not used for the parametrization of the models in the current work, existing studies have reported generally moderate to poor ecological status throughout the strait (Ate¸s and Katagan, 2011 ˘ ) and a comparable polychaete composition, with 16 out of the 28 SA3 species and 16 out of 21 SA1 species present in the northern part, close to the Marmara Sea (Çinar et al., 2011). Thus, SA3 is considered to represent moderately disturbed conditions based on summer samples (which generally display higher status, compared with autumn or winter samples - Çinar et al., 2012, 2015). Contrary to our expectations, this archetype's response to distance from port seems to be counter-intuitive in that it is predicted more likely to occur away from port areas. The environmental setting of the ports included in the analysis is far from homogeneous and, with the exception of the port of Izmir, which is a large international port situated in a shallow, sheltered bay, the rest are small to medium ports/harbors in relatively exposed areas with steeper seabed slopes. Thus, the initial assumptions about the processes this predictor would represent were either not met or were relevant at different scales in the majority of cases, in which case capturing the impacts of this pressure would require a different grain of sampling. Nevertheless, distance from port emerged as TABLE 3 | Summary of the predictive performance of the Species Archetype Models, evaluated by species and expressed as the number and the percentage of species with rs: Spearman's rank correlation coefficient, where \* denotes significant values at the p < 0.05 level and AUC: Area Under the ROC Curve (numbers in parentheses are total number of species for each archetype).


the most influential predictor of archetype 3 distribution and it is possible that its spatial pattern co-varies with the spatial structure of a different variable that truly affects the distribution of tolerant and opportunistic polychaetes in a negative way.

A number of environmental parameters with well established relationships with the benthos, namely hydrodynamic variables, such as current speed and wave orbital velocity (Jenness and Duineveld, 1985; Wright et al., 1987, 1997; Hall et al., 1994) and sediment characteristics, particularly granulometric composition and organic carbon content (Snelgrove and Butman, 1994; Degraer et al., 1999; Ellingsen, 2002; Van Hoey et al., 2004; Çinar et al., 2012, 2015) were not addressed in this study due to the lack of available data layers of sufficient resolution and information. Even though, at the scale studied here, the sedimentary environment may be of lesser importance for predicting distributions due to its small scale variability (Reiss et al., 2011), missing explanatory variables and the inherent uncertainty in the existing predictors have undoubtedly affected the uncertainty in predictions and subsequent model performance.

Uncertainty in SAMs is propagated through the analysis from the raw data to the final probability estimates, with standard errors of regression coefficients estimated from the variancecovariance matrix and standard errors of the fitted probabilities calculated from the model components and measures of uncertainty (Dunstan et al., 2011). It thus follows that better estimated archetypes with lower relative standard errors will display lower uncertainty values; such is the case for archetype 2, while uncertainty increases for archetypes 1 and 3. This can reflect how well the individual species' responses are aligned with their respective archetypical response (Woolley et al., 2013); indeed archetype 3 has the largest proportion of indicatively affiliated species (Table S1). Furthermore, uncertainty and predictive performance improve with the number of species represented by each archetype, in agreement with findings that the more observations a group response contains the better it is predicted (Elith and Leathwick, 2007; Gogina et al., 2016). For archetypes 1 and 3 it is possible that the smaller number of included species is not adequate for an accurate characterization of the group response or that these groups have a more heterogeneous species composition. More importantly, each archetype may be better defined by a slightly different set of covariates and variable selection simultaneously across all archetypes may fail to demonstrate that (Hui et al., 2015). As an alternative, separate models could have been developed with a different statistical method once the appropriate number of archetypes had been established. However, one of the attractive features in the SAMs approach is the unsupervised classification of species into group responses, which can change as variable selection progresses and predictor terms are dropped from the models (i.e., the posterior probability of group membership tau for each species may change and adjustments to the final species groupings will occur accordingly). Thus, it was considered that the benefits of proceeding with the SAMs analytical framework outweighed the potential drawbacks of simultaneous variable selection.

Species Archetype Modeling is a novel methodology that is continuously being developed and refined. There is certainly scope for improvement in many aspects, such as variable selection and model evaluation, use of species-specific intercepts in the models, adopting a different archetype model (e.g., Generalized Additive Model or Boosted Regression Tree in place of a GLM). While some of these issues are already being addressed (Dunstan et al., 2013b; Foster et al., 2015; Hui et al., 2015), they are still a work in progress and have not yet been implemented in an available software tool (Scott Foster, personal communication). Other modeling issues on the other hand, such as addressing species interactions through the inclusion of latent variables for instance, are more challenging and remain to be investigated (Dunstan et al., 2013b; Warton et al., 2015a). Nevertheless, SAMs have been shown to outperform single species GLMs in big data sets, especially when the ratio of the number of species to the number of stations is high, as they borrow strength from common species and even perform better than single GAMs for very rare species (Hui et al., 2013).

At appropriate scales, benthic species archetypes can complement existing efforts for the bioregionalization of the Aegean and the Mediterranean (Reygondeau et al., 2014), as Woolley et al. (2013) have demonstrated for south-western Australia and provide "an appropriate statistical method that can link the ecological information to the pressures" (Foster et al., 2015) as we have demonstrated here for benthic polychaetes and eutrophication. From a spatial management and planning perspective the use of archetype groups is attractive because it can summarize a large amount of complex information in a robust and parsimonious way that offers ease of interpretation. The archetypes of benthic polychaetes identified in this study may constitute a departure from our conventional scheme of delineating marine benthic communities, they do however offer an alternative and intuitive way to study and visualize the distribution of benthic species groups at scales that incorporate both biogeographical and more localized, often human-induced or enhanced, processes. In this regard, possibilities could be envisaged for the development of distributional (ICES, 2016)

or surveillance indicators (Shepard et al., 2015) from such a modeling approach.

Distributional indicators can help identify drivers and directions of change in a spatially explicit manner and serve as a first "alarm bell" that problems have occurred which require managers' attention and further study or mitigation measures (ICES, 2016). Moreover, they can provide valuable information on the natural background variability of different sections of an area under study. At the sub-regional scale the current SAMs confirm the ecological significance and the geographic extent of the three Aegean biogeographic areas (northern, central and southern) specifically for benthic polychaetes of shallow soft sediments, which are however a good proxy for the whole benthic community. At more local scales, they provide a good indication of areas where eutrophication impacts to the benthos are more strongly manifested. It is in these areas, where SA1 is complemented by or replaced with SA3 that distributional shifts or expansion of either of these two archetypes will be informative for management purposes. Similarly, areas of overlap, or substitution of SA1 with SA2 is where "mostly undisturbed" conditions may be sought for the central and southeastern Aegean, after water bodies and typologies have been established (for Turkish waters see TUBITAK-MRC and MoEU-GDEM, 2014). In contrast, our results indicate that "reference conditions" or "naturalness" for the northern Aegean will be somewhat different and mostly characterized by indifferent and tolerant species. It is recommended that monitoring schemes take into account the transition zones between the three Archetypes, where the direction of change in response to pressures may be better detected and visualized.

# CONCLUSIONS

Within the limitations of the information contained in our datasets, our application of SAMs captured both the natural and the adequately quantified pressure gradients, distinguished the responses of sensitive and opportunistic benthic polychaetes and

# REFERENCES


performed rather well in predicting their distribution. It thus serves as a first step to demonstrate the potential of this modeling framework to strengthen our knowledge base about the ways and the scales at which benthos respond to natural and anthropogenic gradients and offer insights into patterns of species co-existence, supporting the implementation of area oriented monitoring and assessment.

# AUTHOR CONTRIBUTIONS

Conceived and designed the study: MG, GK, and KB. Sourced and screened the data: KB, GK, and MG. Analyzed the data and wrote the paper: MG. All authors reviewed the manuscript.

# FUNDING

This manuscript is a result of the DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status - http://www.devotesproject.eu) project funded by the European Union (7th Framework Program "The Ocean of Tomorrow" Theme, grant agreement no. 308392).

# ACKNOWLEDGMENTS

The Turkish Ministry of Environment and Urbanization/General Directorate of Environmental Management is gratefully acknowledged for making available the report on the "Integrated pollution monitoring project in Aegean and Mediterranean Seas," coordinated by Derinsu Underwater Engineering.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00154


Maps of Potential Pressures in These Eco-regions. Perseus Deliverable 1.6. 1–45.


and adjoining estuaries and inner shelf. Estuar. Coast. Shelf Sci. 24, 765–784. doi: 10.1016/0272-7714(87)90151-X


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Galanidi, Kaboglu and Bizsel. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Mollusk Assemblages As Records of Past and Present Ecological Status

Gregory P. Dietl 1, 2 \* † , Stephen R. Durham2 †, Jansen A. Smith2 † and Annalee Tweitmann3 †

*<sup>1</sup> Paleontological Research Institution, Ithaca, NY, USA, <sup>2</sup> Department of Earth and Atmospheric Sciences, Cornell University, Ithaca, NY, USA, <sup>3</sup> Department of Natural Resources, Cornell University, Ithaca, NY, USA*

AMBI and Bentix are widely used benthic indices for guiding remediation decisions under two major pieces of environmental legislation in Europe—the Water Framework Directive (WFD) and the Marine Strategy Framework Directive (MSFD). These indices usually incorporate all living marine benthic invertebrates in a sample. Some recent studies, however, have applied these benthic indices to only mollusk species due to the ease of identifying a single taxonomic group to the species level and because death assemblages (accumulated dead mollusk shells in sediments) may be valuable sources of data for assessing baseline conditions. Although they found that ecological status differences can be detected by applying AMBI and Bentix to mollusks, these studies did not test whether mollusk-only index values, and the ecological statuses indicated by them, are equivalent to those calculated from the whole benthic community. To test this assumption, we performed a meta-analysis of data from 12 European benthic community studies comparing mollusk-only index values with whole-community values. Using five mollusk-only data sets, we also assessed whether application of AMBI and Bentix to molluscan death assemblages can be used to detect changes in ecological status over time. We show that the application of AMBI and Bentix to only the molluscan taxa in benthic communities is a viable method for determining the ecological status of water bodies. Our results also suggest that the application of benthic indices to molluscan death assemblages has great potential to (1) establish baseline conditions for assessing ecological status under the WFD and (2) estimate the natural range of variation of ecosystem attributes for defining sustainability thresholds under the MSFD. We outline three recommendations for the future use of mollusk-only AMBI and Bentix based on our results: (1) mollusk-only index values should be adjusted to facilitate comparisons with whole-community studies; (2) if possible, local ecological group assignments should be used; and (3) we encourage collaboration between paleoecologists and benthic ecologists to facilitate interpretations of index values from death assemblages. We conclude that mollusk-only benthic index assessments of molluscan death assemblages have the potential to be a powerful tool for guiding management decisions under the WFD and MSFD.

Keywords: AMBI, Bentix, bivalves, death assemblage, gastropods, geohistorical data, Marine Strategy Framework Directive, Water Framework Directive

Edited by:

*Maria C. Uyarra, AZTI Tecnalia, Spain*

#### Reviewed by:

*Nomiki Simboura, Hellenic Centre for Marine Research, Greece G. Lynn Wingard, United States Geological Survey, USA*

> \*Correspondence: *Gregory P. Dietl gpd3@cornell.edu*

*† These authors have contributed equally to this work.*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *16 June 2016* Accepted: *30 August 2016* Published: *30 September 2016*

#### Citation:

*Dietl GP, Durham SR, Smith JA and Tweitmann A (2016) Mollusk Assemblages As Records of Past and Present Ecological Status. Front. Mar. Sci. 3:169. doi: 10.3389/fmars.2016.00169*

# INTRODUCTION

The Water Framework Directive (WFD; European Commission, 2000), a major piece of environmental legislation implemented by the European Union in 2000, has led to the development of numerous benthic indices (e.g., Borja et al., 2000; Simboura and Zenetos, 2002; Rosenberg et al., 2004; Dauvin and Ruellet, 2007; Muxika et al., 2007) designed to determine the ecological status of European coastal and estuarine waters. Such indices are used to provide objective, data-based guidance for water body restoration decisions and extensive intercalibration exercises have been undertaken to ensure comparability of WFD ecological assessment results between countries using different indices. The Marine Strategy Framework Directive (MSFD; European Commission, 2008)—the oceanic counterpart to the WFD—requires the standardization of assessment criteria on a regional scale to avoid the need for expensive and challenging intercalibrations (Van Hoey et al., 2010). Some of the benthic indices developed under the WFD, however, have continued to be important tools for remediation assessments under the MSFD (Borja et al., 2011; Simboura et al., 2012; Spagnolo et al., 2014), particularly with regard to evaluating structural and functional aspects of sea-floor integrity—one of the 11 "quality descriptors" outlined in the MSFD for evaluating "good environmental status" (Van Hoey et al., 2010).

Benthic indices are typically based on the entire macroinvertebrate benthic fauna (e.g., annelids, crustaceans, echinoderms, mollusks), but some have been established based on subsets of these taxa. For example, the benthic opportunistic polychaetes amphipods index (BOPA), as the name implies, is calculated using only certain polychaete and amphipod taxa (Dauvin and Ruellet, 2007; see also the Foram Stress Index, Dimiza et al., 2016). Although BOPA is calculated based on the ratio of opportunistic polychaetes to sensitive amphipods, the index was calibrated to the five WFD ecological status categories using AMBI and Bentix, two popular whole-community benthic indices. Thus, BOPA's ecological status assignments were designed to approximate whole-community ecological status in accordance with the WFD (Dauvin and Ruellet, 2007). Indices that are calculated from a subset of taxa have advantages (e.g., reduced burden of taxonomic familiarity; Dauvin and Ruellet, 2007). However, as was done with BOPA, it is important that taxon-specific indices address the potential biases associated with assessments based on only subsets of taxa (e.g., the variability between taxa in responses to disturbances, variations in habitat, etc.; Van Hoey et al., 2010) so that ecological status assignments remain on the same scale as other whole-community indices.

Mollusks, which often comprise up to 20 and 25% of individuals in disturbed and undisturbed benthic communities, respectively (Stergiou et al., 1997), have a long history of use as ecological indicators (Zenetos, 1996; Mahmoud et al., 2010; La Valle et al., 2011; Coelho et al., 2014; Velez et al., 2016), making them good candidates for a taxon-specific approach. For instance, using AMBI and Bentix, Nerlovic et al. (2011) ´ found notable differences in the WFD ecological status categories of the bivalve community following anoxic events in the eastern portion of the northern Adriatic Sea. Similarly, again using AMBI and Bentix, Leshno et al. (2016) were able to detect the effects of pollution on the molluscan fauna off the Israeli coast. Although these studies demonstrated that differences in ecological status categories can be detected using only mollusks, Nerlovic et al. (2011) ´ and Leshno et al. (2016) did not confirm that benthic indices based solely on mollusks were correlated with those calculated from the whole benthic community. Thus, their conclusions may be biased toward either higher or lower ecological status by differences in the responses of mollusks to disturbance relative to the whole-community, and are not necessarily directly comparable with the results of other WFD studies.

Here, in the context of the WFD, we investigate whether ecological status assignments from mollusk-only assessments are equivalent to whole-community AMBI and Bentix analyses. We also assess whether AMBI and Bentix can be applied to molluscan death assemblages—the calcium carbonate shells of dead mollusks that accumulate in sediments over time—to detect temporal change in ecological status. Death assemblages commonly record average ecological conditions on timescales of decades to centuries (Kidwell, 2013). Application of benthic indices to molluscan death assemblages (or geohistorical records; NRC, 2005) may, therefore, have the potential to (1) establish baseline conditions for assessing ecological status under the WFD and (2) estimate the natural range of variation of ecosystem attributes that can be used to set sustainability thresholds during the implementation of the MSFD, among other needs that have been identified in the ecological assessment literature (Van Hoey et al., 2010; Borja et al., 2012).

# METHODS

#### Benthic Indices: AMBI and Bentix

AMBI and Bentix are calculated by assigning species to five and two groups, respectively, based on sensitivity to disturbance, such as eutrophication (Borja et al., 2000; Simboura and Zenetos, 2002; Munari and Mistri, 2010). Using data on species abundances and the represented ecological groups, AMBI applies the equation:

$$\text{AMBI} = [(0 \times \text{\%GI}) + (1.5 \times \text{\%GI}) + (3 \times \text{\%GIII})]$$

$$+ (4.5 \times \text{\%GIV}) + (6 \times \text{\%GV})]/100\tag{1}$$

GI through GV are ecological groups with increasing tolerance for disturbance. Resulting AMBI values range from zero to seven and correspond with the five WFD ecological status categories (High, 0 < AMBI < 1.2; Good, 1.2 < AMBI < 3.3; Moderate, 3.3 < AMBI < 4.3; Poor, 4.3 < AMBI < 5.5; Bad, 5.5 < AMBI < 7; Borja et al., 2004). According to the WFD, any water body ranked lower than "Good" requires remediation. Bentix assigns species to only two ecological groups, one for taxa sensitive to disturbance (GS) and the other for taxa that are tolerant (GT; Simboura and Zenetos, 2002). Bentix values can range from two to six (values of zero indicate azoic sediments) and are calculated using the equation:

$$\text{Benix} = (6 \times \text{\%GS} + 2 \times \text{\%GT})/100\tag{2}$$

Bentix values translate to the following WFD ecological status categories: Bad, Bentix = 0; Poor, 2.0 < Bentix < 2.5; Moderate, 2.5 < Bentix < 3.5; Good, 3.5 < Bentix < 4.5; High, 4.5 < Bentix < 6.0. Note that the AMBI scale is inversely correlated with ecological status, but the Bentix scale and ecological status are positively correlated.

The Bentix and AMBI indices are related (Bentix GS is equivalent to ecological groups I and II from AMBI and Bentix GT is equivalent to ecological groups III, IV, and V from AMBI) but draw from independent species lists for assigning taxa to ecological groups (Simboura and Zenetos, 2002; Munari and Mistri, 2010). Neither species list is exhaustive, but the AMBI list (n = ∼8000) is more inclusive than the Bentix list (n = 1250). Consequently, when calculating Bentix values, ecological groupings from AMBI were applied using the above conversion when a species was absent from the Bentix list. When assigning ecological groups to species using the AMBI list, if a species did not occur on the list, the ecological group was assigned using the following rules: (1) the ecological group listed for the genus was applied; (2) if the genus name alone was also not on the list, then the species was assigned the ecological group shared by the majority of congeneric species on the list; (3) if there were no congeners or there was no clear majority ecological group among the congeners on the list, then the species was not assigned an ecological group.

# Selection and Subdivision of Data Sets

We conducted internet searches of the published literature for papers reporting benthic community census data from European waters and contacted authors to obtain data sets from studies that did not report community abundance data. Our search yielded: (1) 12 live-only benthic community data sets from sites across Europe, including the English Channel, Baltic Sea, Bay of Biscay, Mediterranean Sea, Aegean Sea, and Adriatic Sea (Table S1.1 in Supplementary Material); and, (2) five live-dead mollusk-only data sets from the Mediterranean Sea (Table S1.2 in Supplementary Material). The live data sets were used to assess the correlation between whole-community and molluskonly index values (i.e., AMBI and Bentix) and subsequently the correlation between directly calculated whole-community and estimated whole-community index values. The live-dead data sets were used to examine the degree of variation in AMBI and Bentix values when comparing live assemblages (LA) with death assemblages (DA; Table S1.2 in Supplementary Material).

To evaluate the relationship between index values calculated from the whole community and those calculated from mollusks only, we compiled 91 stations with at least 10 mollusk species from each live-only data set (Table S1.1 in Supplementary Material) into one master data set and calculated AMBI and Bentix values for both the whole community and only mollusks for each station. AMBI values were calculated with the AMBI 5.0 software (species list v. Nov2014) and Bentix values were calculated using the Bentix Add-In v1.0 (© 2009 Hellenic Center for Marine Research, Institute of Oceanography) for Excel (Microsoft Corporation). The stations were then ordered by the whole-community AMBI values from largest to smallest and alternately assigned to group A or group B to ensure an even distribution of AMBI values. The same process was repeated using the whole-community Bentix values. We used data group A to examine the relationship between whole-community and mollusk-only index values, saving data group B as an independent data set to test the utility of the relationship for predicting whole-community values from mollusk-only values.

# Correlation between Whole-Community and Mollusk-Only Index Values

Using the 46 stations in data group A, index values calculated from the whole community were regressed on values calculated with just molluscan taxa. Note that because the AMBI software averages replicate samples and reports a single value for a given station, but the Bentix Excel script calculates a different value for each replicate, the total number of data points in the regressions differed between indices. For AMBI index values, a square-root transformation was applied as this adjustment has been shown to improve the results of AMBI (Tweedley et al., 2014). A regression equation with a slope of one indicates a perfect match between the whole-community and molluscanonly indices. If the slope is not one, then the molluscan community is either over-estimating or under-estimating the whole-community value.

# Correlation between Directly Calculated and Estimated Whole-Community Index Values

The 45 stations assigned to data group B were used to evaluate the concordance of mollusk-only index values with those of whole-community index values, using data independent of those used to produce the regression (i.e., group A). The regression equation produced from the stations in data group A was used to produce estimated whole-community values from the mollusk-only values for each station in data group B. We then regressed the resulting estimated whole-community values against the directly calculated wholecommunity values for each station in data group B to evaluate the relationship between the two, and in particular the concordance of the ecological status assignments. To account for our use of square-root transformed abundance data, we also adjusted the AMBI default ecological status category boundaries using the equation resulting from the regression of untransformed whole-community abundance data against square-root transformed abundance data (Tweedley et al., 2014). For the Bentix calculations, we assigned ecological status using the standard category boundaries because the abundance data were untransformed.

In order to evaluate the goodness of fit of the estimated whole-community index to the directly calculated wholecommunity index, three potential error types were quantified: (1) the proportion of stations where WFD ecological status categories were misclassified by the estimated whole-community indices; (2) the direction of misclassifications and their relative Dietl et al. Mollusk Assemblages As Records of Ecological Status

frequencies (i.e., the potential for bias toward over- or underestimates of directly calculated whole-community index values); and (3) the proportion of misclassified sites that were incorrectly classified into action (i.e., incorrectly classified below "Good") or no action (i.e., incorrectly classified above "Moderate") ecological status categories.

### Variation between Live and Death Assemblage Mollusk-Only Index Values

In order for comparisons of AMBI and Bentix values between a LA and DA to be meaningful from a management perspective, they must be capable of showing enough variation to indicate changes in ecological status (assuming changes have occurred). Therefore, using the five live-dead studies (Table S1.2 in Supplementary Material) found during our search of the literature, we calculated and plotted the resulting LA and DA index values by station to visualize the potential trajectory of ecological status for each station (either worsening or improving over time).

#### RESULTS

#### AMBI

Whole-community AMBI values were positively correlated (R <sup>2</sup> = 0.46) with mollusk-only AMBI values for the 46 stations included in data group A (**Figure 1**). The slope was less than one, however, suggesting that mollusk-only analyses tended to yield slightly higher AMBI values than analyses that included the whole benthic community (i.e., mollusk-only analyses tended to slightly underestimate the ecological statuses of the stations).

This pattern persisted when the estimated whole-community AMBI values calculated for the 45 stations in data group B were regressed against the directly calculated whole-community AMBI values (R <sup>2</sup> = 0.46; **Figure 2**). However, the ecological status assignments based on estimated and directly calculated wholecommunity AMBI values still agreed for the majority (78%; n = 35) of stations because ecological status ratings are based on ranges of AMBI values (**Figure 2**).

Estimated whole-community AMBI values misclassified the ecological group in 10 (22%) cases. The estimated whole-community AMBI values overestimated the ecological statuses of 13% (n = 6) of the stations and underestimated the ecological statuses of 9% (n = 4) of the stations (**Figure 3**). Because the "Moderate"-"Good" ecological status boundary is the cut-off for when remediation is required, eight of these 10 cases of differing ecological status ratings would have resulted in different decisions about the necessity of remediation (**Figure 3A**). Most of the differences in action would have been conservative. Eighty-three percent (n = 5) of overestimated ecological status ratings for estimated whole-community AMBI values were "Good" or better, when ecological status ratings based on the directly calculated wholecommunity AMBI values for the same stations were "Moderate" or worse (i.e., no action would be recommended although it would have been supported by the directly calculated whole-community calculation; **Figure 3B**). The ecological status ratings of three stations (75%) were underestimated by the estimated AMBI values as "Moderate", when the rating based on directly calculated AMBI values was "Good" (i.e., would have resulted in remediation, although it would not have been supported by the directly calculated whole-community calculation; **Figure 3C**).

#### Bentix

Whole-community Bentix values were correlated (R <sup>2</sup> = 0.3) with mollusk-only Bentix values for the 64 stations (or replicates) in data group A. Similar to the transformed AMBI regression, the slope was less than one, indicating that mollusk-only Bentix calculations will yield higher ecological status ratings than Bentix calculations that include the whole community (**Figure 4**). When estimated Bentix values were regressed against directly calculated whole-community Bentix values using data group B, the ecological status assignments

of estimated whole-community AMBI values that agreed with, overestimated, or underestimated the ecological status rating based on the directly calculated whole-community AMBI values, and the proportion of (B) overestimated and (C) underestimated AMBI values that would have resulted in the same or different conclusions about the need for remedial action (i.e., action when none is required or no action when remediation is necessary, for underestimated and overestimated ecological status ratings, respectively).

agreed for 73% (n = 47) of the 64 calculations (R <sup>2</sup> = 0.42; **Figure 5**).

Estimated whole-community Bentix values misclassified the ecological group in 17 (27%) cases. The ecological status ratings based on estimated whole-community Bentix values overestimated those based on directly calculated wholecommunity Bentix values in 11% (n = 7) and underestimated them in 16% (n = 10) of the calculations (**Figure 6**). Of the 17 instances where estimated and directly calculated whole-community Bentix values did not agree on ecological

FIGURE 5 | Regression of whole-community Bentix values calculated directly from all species against values estimated from the molluscan species at each station and the regression equation from Figure 4. Boxes indicate ranges of values falling into each of the Ecological Status classifications (Blue = High, Green = Good, Yellow = Moderate, Orange = Poor).

status ratings, 10 crossed the "Good"-"Moderate" boundary at which remediation is required. A need for remediation would have been missed (i.e., ecological status was overestimated using estimated whole-community Bentix) in four out of the seven (57%) overestimates, whereas erroneous ecological status ratings of "Moderate" or worse in six out of the 10 (60%) underestimates would have unjustifiably suggested that remediation was necessary (**Figure 6**). Only one of the 17 (6%) misclassifications was by more than a single ecological status rating using estimated as opposed to directly calculated wholecommunity Bentix (**Figure 5**).

# Variation between Live and Death Assemblage Mollusk-Only Index Values

When estimated whole-community AMBI and Bentix values were calculated using the data from five live-dead studies and the regression equations from **Figures 1**, **3**, respectively, two out of 18 (11%) Bentix values indicated changes in ecological status vs. seven of the 18 (39%) AMBI values (**Figure 7**). The most dramatic difference between Bentix values was from the data of Zenetos and Van Aartsen (1995), which suggested a decline in ecological status from Good to Moderate from the DA to the LA. For AMBI, the largest difference was from the data of Peharda et al. (2002), which suggested an increase in ecological status from Good to High from the DA to the LA at station 23. Overall, there was only rough concordance between the two indices. The AMBI and Bentix values agreed on the direction of change (positive or negative) in ecological status in seven out of 18 (39%) cases (**Figure 7**), and AMBI and Bentix values resulted in the same ecological status category for both LA and DA data in five out of the 18 (28%) cases (**Figure 7**). Either AMBI or Bentix indicated a change in ecological status had occurred between the DA and the LA in eight out of the 18 (44%) pairs of LA and DA calculations. However, there was only one (6%) station for which AMBI and Bentix both showed a change in ecological status.

# DISCUSSION

The estimated whole-community AMBI and Bentix values resulted in the same ecological status ratings as index values directly calculated from whole-community data in more than 70% of stations for each index. Further, although there were cases where estimated whole-community indices would have resulted in misleading ecological status ratings, all of the values that would have erroneously indicated a need for remediation were for stations that were already close to the Good-Moderate boundary based on the directly calculated whole-community calculations for both AMBI and Bentix (**Figures 2**, **5**). All of the cases where estimated whole-community index values substantially underestimated or overestimated ecological status either did not cross the Good-Moderate boundary (i.e., would not have resulted in different remediation recommendations), or crossed the boundary but overestimated the ecological status (i.e., no remediation recommended, although the directly calculated whole-community index values would have recommended it; **Figures 2**, **5**). Thus, it appears that estimated whole-community AMBI and Bentix values based on only the molluscan taxa in the community can be used to reproduce the ecological status ratings that would be indicated by directly calculated whole-community values, and when errors in ecological status assignments occur, they tend to be conservative with regard to remediation recommendations. The high performance of the estimated whole-community AMBI and Bentix indices tested here is encouraging but not perfect, reinforcing the recommendation that multiple types of metrics and indicators (e.g., physical, chemical, biological) should be used for environmental assessments to reduce uncertainty in results (e.g., Borja and Muxika, 2005; Dauvin, 2007; Teixeira et al., 2007; Kröncke and Reiss, 2010).

The results of applying estimated whole-community AMBI and Bentix to LA and DA abundance data also show promise for detecting trajectories of ecological status over time. There was variation in the index values that may be indicative of recent changes in ecological status for both AMBI and Bentix; however, because the majority of original studies were not focused on assessing ecological change, more information on the history of anthropogenic impacts at the study sites would be necessary to determine the cause of the variation. Further research will also be required to understand why the AMBI and Bentix values did not agree on the direction and magnitude of change in ecological status between some LAs and DAs.

# Precautions for Mollusk-Only AMBI and Bentix

Although our study results suggest that calculations of AMBI and Bentix using only the molluscan taxa in a benthic community will most often result in the same ecological status conclusions as whole-community calculations, there are a number of factors that must be considered to accurately interpret these values, particularly when involving DA data. For instance, our analysis clearly demonstrates that unadjusted mollusk-only AMBI and Bentix values are not directly comparable to those calculated from the whole-community. Mollusk-only values must be adjusted to estimated whole-community values for mollusk-only and whole-community ecological status ratings to be directly comparable. The lack of a one-to-one ratio of whole-community and mollusk-only AMBI and Bentix values is likely influenced by at least two variables: the percentage of individuals in the community that are mollusks, which varied from <1 to ∼59% in the data sets we compiled (Table S1.1 in Supplementary Material), and the ecological group of the most abundant mollusk species in the community (see Section 2 in Supplementary Material). First, mollusk-only values more accurately represent the whole community when more of the community's individuals are mollusks. For instance, the difference between the mollusk-only and whole-community index values decreases as the proportion of individuals in the whole community that are mollusks increases (Figure S2.1 in Supplementary Material). Second, mollusk-only and whole-community values are more consistent on average as the AMBI ecological group of the most abundant mollusk species in the community increases (Figure S2.1 in Supplementary Material). This pattern likely occurs because most mollusk species are categorized in low ecological groups (no mollusk species in the data sets we used were higher than AMBI ecological group 4). High variability in the "molluskonly − whole-community" difference can result if mollusks are rare in the community or the ecological groups represented are lower (i.e., more sensitive) than the ecological groups represented by other taxonomic groups in the community, which could include more disturbance-tolerant species, such as annelid worms (Figure S2.2 in Supplementary Material). This difference in ecological group distributions may help explain why mollusk-only analyses tend to overestimate the ecological

status of stations whose ecological status is already high and underestimate the ecological status of more highly disturbed stations (including natural disturbances; see Dauvin and Ruellet, 2007).

Additionally, given the importance of the distribution of individuals among ecological groups (especially when limited to a subset of taxa; i.e., mollusks), it is particularly necessary to correctly assign species to ecological groups (see Gillett et al., 2015 for an example). Concern over the potentially arbitrary nature of ecological group assignments was raised by Tweedley et al. (2014), who noted the strong family-level coherence of ecological group assignments between species, but found that family-level AMBI values could not accurately assess disturbance levels in estuaries outside of Europe.

Our results also suggest that comparisons between AMBI and Bentix values for LA and DA data are promising as indicators of ecological status changes through time, however, three important sources of bias in DAs—time-averaging, taphonomic inertia, and preservational bias (Kowalewski et al., 1998; Kidwell and Tomasovych, 2013)—must be considered when comparing LA and DA data. First, the degree of time-averaging, which is the accumulation and mixing of material of different ages into the same sedimentary layer (Kowalewski et al., 1998; Kidwell, 2013), at a given location can be highly variable depending on environmental factors such as water depth and sedimentation rate (Kidwell, 2013). Time-averaging can produce DAs that tend to be either young on average with less timeaveraging (i.e., decades to centuries), or older on average with greater time-averaging (i.e., centuries to millennia), for estuaries and continental shelf environments, respectively. Although these differences in temporal mixing can result in misleading abundance or species composition data (Kowalewski et al., 1998; Kidwell, 2007), when properly quantified, time-averaging can be advantageous by dampening out the short-term temporal variability that characterizes LAs. Thus, DAs can yield data on the mean conditions of the benthic fauna and environmental conditions for the time period over which the assemblage is timeaveraged and can indicate deviations from the mean conditions of the preceding decades or centuries relative to LAs (Kowalewski et al., 1998; Kidwell, 2007, 2013). The time-averaging process also tends to increase evenness in DAs relative to LAs because rare taxa will accumulate in a DA over time but occur too sparsely to be sampled in the LA (Kidwell, 2013). If unaddressed, this bias in the DA could decrease the accuracy of DA benthic index values, and cause misleading comparisons with LA data.

The second characteristic, taphonomic inertia (the lag in response of DA composition following changes in the LA; Kidwell, 2007) is sensitive to the degree of time-averaging. For instance, taphonomic inertia on the continental shelf is often greater than in estuaries, corresponding to the aforementioned difference in time-averaging. Additionally, taphonomic inertia is influenced by the balance between the gradual addition of dead remains to the seafloor and the constant reworking and removal of remains by biological, physical, and chemical processes such as bioturbation, wave action, and dissolution, respectively. Thus, the ecological signal of a DA (e.g., composition, abundance) lags behind the corresponding LA in time. For change in the DA composition to become evident, the signal from new material must overwhelm the existing time-averaged signal. Generally, it is assumed that similarity in metrics (e.g., species composition and rank-order abundance of species) between LAs and DAs indicates that there has been little disturbance in the ecosystem over long periods of time (Kidwell, 2007). Low taphonomic inertia can, however, lead to misleading conclusions in LA-DA comparisons. In such cases, the similarity between the LA and DA would not indicate a lack of disturbance in the LA, but simply that the DA reflects changes in the LA soon after they occur. Hence, it is important to consider the magnitude of taphonomic inertia to avoid misleading results from comparisons of LAs and DAs.

The third DA characteristic, preservational bias, is highly sensitive to the durability of molluscan remains, particularly when assemblages are time-averaged over long periods. For example, mollusk taxa that are small (<1.0 mm), fragile, or shellless rarely persist in DAs (Kidwell, 2001). Thus, DAs typically record only a fraction of the total living diversity, and how many taxa are preserved is both a function of the living diversity and characteristics of the preservational environment. Although such preservational bias restricts the diversity of higher taxa in DAs to varying degrees, the hard parts that remain intact to the point of final burial (the point at which they become buried deep enough that they are unlikely to be exhumed) can persist in the sedimentary record for millennia and provide ecologically meaningful data that are often the only source of local baseline information (Kidwell, 2013).

# Advantages of Geohistorical Data

AMBI and Bentix calculations on molluscan DAs have high potential value for benthic assessment and implementation of environmental legislation (e.g., WFD and MSFD). The difficulty of obtaining reference conditions (e.g., "near-pristine" areas, historical data) for most coastal and marine habitats is currently an obstacle to environmental assessment (Van Hoey et al., 2010). This issue, however, is a promising potential area of application for geohistorical data, such as those from molluscan DAs. Depending on the degree of time-averaging in a given DA, it can yield data to help address information needs for ecological assessment by: (1) increasing the availability of local baseline data against which ecological status in the WFD can be measured, especially where no largely undisturbed (i.e., "pristine") areas exist; (2) defining "naturalness" in an ecosystem by quantifying natural ranges of variability of ecosystem attributes in the past (including trajectories in those attributes over timescales beyond the reach of modern instrumental monitoring), which can be used to set sustainability thresholds during the implementation of the MSFD; (3) disentangling the relative importance of multiple ecological stressors responsible for benthic community changes, particularly for stressors acting over large temporal scales (e.g., climate change); and (4) identifying invasive species and estimating the duration of their presence in an ecosystem.

The most fundamental use for data from DAs is to improve local baseline data (Dietl and Flessa, 2011). Where the dead remains of benthic organisms with hard parts are easily buried and preserved, such as in coastal marine systems, locationspecific geohistorical data are often readily available. The timeaveraged and time-lagged nature of these data also means that they reflect environmental conditions from decades to millennia in the past. These attributes make geohistorical records, such as DAs, more useful sources of reference data for ecosystems than is generally realized in the restoration and conservation communities (Durham and Dietl, 2015; contra Borja et al., 2012). There is also abundant evidence that community attributes, such as species rank-abundances, which can be reconstructed from DAs, have high fidelity to corresponding undisturbed LAs (Kidwell, 2013).

By digging deeper into sediments, baseline information from multiple time intervals can be combined to document the natural range of variation of many ecosystem attributes and potentially also to document trajectories of change in the measured attributes during the recent past. This kind of information is becoming increasingly important for restoration and management planning activities (Wiens et al., 2012), particularly under the MSFD. Geohistorical data can provide information about this natural range of variation because they represent an average set of conditions from the preceding decades to millennia. They may also be more likely to reflect the ecological status of a specific habitat than reference conditions based on separate sites, avoiding the problem of comparing ecological "snapshots" from areas whose natural histories may differ. Thus, such data may help with defining regional and sub-regional sustainability thresholds for "good environmental status" because benthic indices are based only on the relative abundance of ecological groups in a sample (i.e., they are fundamentally ataxic in nature); that is, they are robust to changes in species composition of communities over time. Further, due to the decadal- to centennial-scale taphonomic inertia of most DAs, they can still be sampled to increase the temporal context and scope of baseline information from locations at which LA samples were already collected. These data could be used to refine sustainability thresholds that have already been defined, and may also be helpful for validating the results of intercalibration studies conducted under the WFD that may have lacked location-specific temporal context.

The data obtainable from DAs can also help to distinguish the relative importance of multiple stressors on an environment (Dietl et al., 2015), given that their onsets are unlikely to have been synchronous and DA data from multiple timescales may capture changes in ecosystem attributes related to the onset of each stressor. For instance, Casey et al. (2014) used fossil and archaeological data to show that major ecological changes in Long Island Sound, USA, such as the disappearance of oyster reefs, predated major eutrophication problems, but not overfishing, and showed that comparisons of LA and DA diversity did not follow the expected patterns based on a substantial east-west eutrophication gradient. These results strongly suggested that in the absence of efforts to address overfishing, pollution remediation may have only limited success in restoring the ecological condition of Long Island Sound (Casey et al., 2014). Further, multiple stressors may act on highly variable timescales that can easily exceed the amount of time typically accessible from instrumental and historical records (NRC, 2005). For instance, anchovy and sardine populations respond strongly to decadal-scale climatic cycles, but these natural population boom-bust patterns are difficult to distinguish from impacts related to overfishing without baseline data on the same timescales as the climatic cycle (Baumgartner et al., 1992; Valdés et al., 2008). Thus, in the absence of long-term baseline data, like those available from geohistorical records, it is very difficult (if not impossible) to disentangle the effects of multiple stressors on benthic communities.

The temporal context provided by data from DAs may also be very helpful for identifying invasive species and determining both the duration and effect of their presence in an ecosystem. For instance, the presence or absence of a presumed invasive or native species in geohistorical records of varying ages can help document the arrival times of the species (e.g., Chiba and Sato, 2014; Smith and Dietl, 2016). These records may also reveal simultaneous ecosystem changes with the arrival of the potential invasive species or other evidence to help evaluate whether an alien species qualifies as an invasive species under the MSFD, which requires that alien species cause harm in order to be termed "invasive" (Van Hoey et al., 2010). Distinguishing between species that are invasive vs. simply alien may be very difficult without the location-specific temporal context afforded by geohistorical records.

Finally, applying benthic indices to shallow DA samples requires relatively little additional cost or sampling effort, because DA material is often already collected in the process of sampling living benthic communities. For instance, many comparative studies of molluscan LAs and DAs bulk sample sediments using quadrat sampling, coring, or grab sampling methods, which sample both live and dead mollusks at a given station simultaneously. These bulk samples are then typically sieved through a screen and live and dead mollusks are retained for analysis, a very similar process to those already used to quantitatively sample living benthic communities. Due to this similarity in sample processing, the collection of DA data can also easily comply with existing LA sampling standards under the WFD and MSFD (e.g., for sample number, sieve sizes, gear types, etc.; Van Hoey et al., 2010).

# Recommendations for Use of Mollusk-Only AMBI and Bentix

We agree with Leshno et al. (2016) and Nerlovic et al. ´ (2011) that benthic indices applied to mollusks are useful for evaluating ecological status. In particular, our study supports the findings of Leshno et al. (2016) that applying benthic indices to DAs shows promise as a tool for helping to address some intractable problems in ecological assessments, such as a lack of local baseline information, clear stressorresponse relationships, and knowledge of the "naturalness" of an ecosystem (e.g., uncertainties regarding natural variability and thresholds of sustainability; Van Hoey et al., 2010). We have three recommendations for the future use of molluskonly AMBI and Bentix: (1) index values should be adjusted to estimated whole-community values to facilitate comparisons with other studies that analyzed the whole benthic community; (2) local ecological group assignments for species should be used whenever possible; and (3) given the complexities of DA formation and corresponding challenges of applying benthic indices to DA data, we encourage collaboration between paleoecologists and benthic ecologists.

First, we have demonstrated that the mollusk-only values must be adjusted before conclusions about ecological status from mollusk-only and whole-community analyses can be directly compared. For this purpose, we provide regression equations based on all of the data in our meta-analysis (groups A and B combined) for AMBI (RT = square-root transformed abundance data) and Bentix:

$$\text{AMBI}\_{\text{RT}} \colon \mathbf{y} = \mathbf{0}.6947\mathbf{x} + \mathbf{0}.9602\tag{3}$$

$$\text{Benrix: y = 0.44x + 1.8148} \tag{4}$$

To illustrate the need for adjusting mollusk-only index values, we converted the AMBI and Bentix values reported in Leshno et al. (2016) to estimated whole-community values. We applied **Equation 4** for Bentix and the regression equation based on untransformed abundance data (UT) for AMBI because Leshno et al. (2016) did not transform their abundance data. The UT regression equation is:

$$\text{AMBI}\_{\text{UT}} \colon \mathbf{y} = 0.7489 \mathbf{x} + 0.9096 \tag{5}$$

Leshno et al. (2016) did not report the Bentix values from their final analysis, so we used the Figure Calibration plugin (Hessman, 2009) for ImageJ 1.50e image processing software (Rasband, 1997) to estimate the Bentix values from their Figure 9. As expected based on our results, the estimated wholecommunity values were lower than the mollusk-only values for higher ecological status stations, and higher for lower ecological status stations (**Figure 8**). In general, this means that molluskonly index values in the "Good" and "Moderate" ecological status categories did not shift as much as stations in the "High," "Poor," and "Bad" categories, which can easily change ecological status categories when adjusted (**Figure 8**). Importantly, the apparent difference in ecological status between the impact and control stations was reduced following our adjustment (**Figure 8**), including narrower differences between DA and LA ecological status for all station and season combinations and a shift downward in the ecological status of values for the control stations (**Figure 8**). There was little agreement between the AMBIUT and Bentix values from Leshno et al. (2016) with regard to ecological status or the magnitude of change between DA and LA ecological status, although in all cases, AMBIUT yielded higher ecological status assignments than Bentix (this pattern also has been observed in the present study and in other studies comparing AMBI and Bentix; Simboura and Reizopoulou, 2007; Simboura and Argyrou, 2010; Leshno et al., 2016).

A full discussion of whether AMBI or Bentix is better suited to comparing DA and LA data is beyond the scope of this paper, but it is important to point out that decisions about the ecological group assignments of species can have a dramatic impact on the resulting ecological status assignments. For instance, when Leshno et al. (2016) altered the ecological group assignment of one dominant clam species, Corbula gibba, from "tolerant" to "sensitive," the ecological status ratings of the DAs at each station increased from about three (i.e., Moderate) to about five (i.e., High), and the ecological status ratings from the LAs increased substantially as well. Leshno et al. (2016) had an empirical reason for changing the ecological group of C. gibba—at the stations they sampled, the percent of individuals of C. gibba was positively correlated with ecological status. This example, among others (Tweedley et al., 2014; Gillett et al., 2015), suggests that the performance of AMBI and Bentix improves when ecological group assignments are based on local conditions and expertise. Hence, our second recommendation is that regional species lists for assigning ecological groups should be used whenever possible (e.g., Gillett et al., 2015).

Interpreting the results of DA analyses can be challenging due to the potentially biasing factors inherent to DAs (e.g., timeaveraging, preservational bias). Thus, our third recommendation is that the application of benthic indices to DAs may best be done collaboratively between benthic ecologists and paleoecologists who regularly consider these biasing factors. In fact, there are already paleoecologists who are interested in applying their skills to conservation and resource management (e.g., Dietl et al., 2015) and calls for such integration from ecologists (e.g., Price and Schmitz, 2016). Such collaborations would help address concerns about taphonomic bias in the DA data, allow for quantification of important factors such as DA age and degree of time-averaging through better access to geochronological dating methods, and bring expertise in paleoenvironmental interpretation to the environmental assessment.

# Future Work

There are several areas where further research is required. First, given calls for a better understanding of the cause of variable performance in benthic indices (Van Hoey et al., 2010), it may be helpful to further develop regional ecological group assignments for mollusks. Doing so will help determine whether differences in the performance of benthic indices between regions are due to variability in mollusks' tolerances to anthropogenic stressors or differing combinations of regionally acting stressors. Similarly, the sensitivity of mollusk-only benthic indices to different forms of environmental variability and anthropogenic disturbances, both between and within indices, demands further research. Van Hoey et al. (2010, p. 2191) pointed out the importance of using benthic indices with a "strong stressor-response relationship" to more confidently determine the ecological status of a location relative to reference conditions, and that indices will vary in their sensitivities to different kinds of stressors. Studies of the sensitivities of mollusk-only benthic indices could be accomplished by studying spatial variation in index values among stations with well-documented stress histories, or temporal variation using data that can be gathered from DAs on certain stressors. For instance, by analyzing trace elements in the mollusk shells themselves Gillikin et al. (2005) were able to track lead pollution over the past five decades in coastal waters near Cape Lookout, North Carolina, USA. Trace isotopic records of pollution or stress are independent and population-specific sources of data that could be used to help document the sensitivities of mollusk-only benthic indices to certain stressors, such as heavy metal pollution.

Further research into the application of benthic indices to molluscan DAs is also needed to help reconcile index results when different values are calculated for the same location, such as those of Leshno et al. (2016) for AMBI and Bentix. Integrating new research on index sensitivities, local information on anthropogenic impacts, and the application of multiple metrics to DAs, may be very helpful for interpreting apparently contradictory results of different mollusk-only benthic indices. Integrating multiple metrics and LA and DA data on multiple timescales may also be helpful for understanding the effects of multiple, potentially interacting, stressors on coastal and marine ecosystems and the corresponding benthic index results.

# CONCLUSIONS

The application of AMBI and Bentix to only the molluscan taxa in benthic communities is a viable method for determining the ecological status of water bodies under the WFD. In order to ensure fidelity to whole-community values, molluskonly results must be converted to estimated whole-community values. Also, although the application of benthic indices to geohistorical records, such as DAs, is in its infancy, the method has great potential to contribute local baseline information on multiple timescales. Such data can help address issues in ecological assessment, including improving our understanding of the natural variability of benthic ecosystems and environmental change through time. Further research is urgently needed to

guide decisions about selecting the most appropriate benthic index (or indices) and how to account for sources of bias in the outcomes of ecological assessments using only mollusks, both in living communities (e.g., taxonomic biases in ecological group distributions) and DAs (e.g., time-averaging, preservational bias, evenness bias). Addressing these issues will make molluskonly benthic index assessments of DAs a powerful tool for implementing environmental legislation.

# AUTHOR CONTRIBUTIONS

This paper is a product of a topics in paleoecology course (EAS 7650) at Cornell University. GD, JS conceived of the study. All authors contributed to the design of the study. SD, JS, AT collected the datasets. SD, AT analyzed the data. All authors contributed to data interpretation and writing of the manuscript.

#### ACKNOWLEDGMENTS

We thank those who generously provided the data sets that were utilized in this study. The Basque Water Agency (URA) provided data from the Basque Country (North of Spain), through Dr. Angel Borja (AZTI). The Northeast Atlantic Shelf Data were kindly provided by Dr. Richard Warwick (Plymouth Marine Laboratory). We also thank Nomiki

#### REFERENCES


Simboura and G. Lynn Wingard, whose comments improved the manuscript.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars.2016. 00169


Adriatic coastal transitional ecosystems. Mar. Pollut. Bull. 60, 1040–1050. doi: 10.1016/j.marpolbul.2010.01.022


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Dietl, Durham, Smith and Tweitmann. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# New Arrivals: An Indicator for Non-indigenous Species Introductions at Different Geographical Scales

Sergej Olenin<sup>1</sup> \*, Aleksas Naršcius ˇ 2 , Stephan Gollasch<sup>3</sup> , Maiju Lehtiniemi <sup>4</sup> , Agnese Marchini <sup>5</sup> , Dan Minchin1, 6 and Greta Srebalien ˙ e˙ 1

 *Marine Science and Technology Centre, Klaipeda University, Klaip ˙ eda, Lithuania, ˙ Open Access Centre for Marine Research, Klaipeda ˙ University, Klaipeda, Lithuania, ˙ GoConsult, Hamburg, Germany, <sup>4</sup> Marine Research Center, Finnish Environment Institute, Helsinki, Finland, <sup>5</sup> Department of Earth and Environmental Sciences, University of Pavia, Pavia, Italy, Marine Organism Investigations, Killaloe, Ireland*

#### Edited by:

*Michael Elliott, University of Hull, UK*

#### Reviewed by:

*Davide Francesco Tagliapietra, Istituto di Scienze Marine (ISMAR), Italy Angel Borja, AZTI, Spain*

> \*Correspondence: *Sergej Olenin sergej.olenin@jmtc.ku.lt*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *21 June 2016* Accepted: *05 October 2016* Published: *28 October 2016*

#### Citation:

*Olenin S, Naršcius A, Gollasch S, ˇ Lehtiniemi M, Marchini A, Minchin D and Srebalien ˙ e G (2016) New Arrivals: ˙ An Indicator for Non-indigenous Species Introductions at Different Geographical Scales. Front. Mar. Sci. 3:208. doi: 10.3389/fmars.2016.00208* Several legal and administrative instruments aimed to reduce the spread of non-indigenous species, that may pose harm to the environment, economy and/or human health, were developed in recent years at international and national levels, such as the International Convention for the Control and Management of Ship's Ballast Water and Sediments, the International Council for the Exploration of the Sea Code of Practice on the Introductions and Transfers of Marine Organisms, the EU Regulation on Invasive Alien Species and the Marine Strategy Framework Directive, the US Invasive Species Act, the Biosecurity Act of New Zealand, etc. The effectiveness of these instruments can only be measured by successes in the prevention of new introductions. We propose an indicator, the arrival of new non-indigenous species (*nNIS*), which helps to assess introduction rates, especially in relation to pathways and vectors of introduction, and is aimed to support management. The technical precondition for the calculation of *nNIS* is the availability of a global, continuously updated and verified source of information on aquatic non-indigenous species. Such a database is needed, because the indicator should be calculated at different geographical scales: (1) for a particular area, such as port or coast of a country within a Large Marine Ecosystem (LME); (2) for a whole LME; and (3) for a larger biogeographical region, including two or more neighboring LMEs. The geographical scale of *nNIS* helps to distinguish between a primary introduction and secondary spread, which may involve different pathways and vectors. This, in turn, determines the availability of management options, because it is more feasible to prevent a primary introduction than to stop subsequent secondary spread. The definition of environmental target, size of assessment unit and possible limitations of the indicator are also discussed.

Keywords: biological invasions, pathways and vectors, information system, large marine ecosystem

# INTRODUCTION

The Convention on Biological Diversity (CBD, 1992) set an ambitious goal "...significant reduction in the current rate of biodiversity loss...," calling to "...Prevent the introduction of, control or eradicate those alien species which threaten ecosystems, habitats or species...." Several legally binding and advisory instruments, aimed to reduce the spread of NIS species by particular vectors of introduction, were developed in recent years at international level. For example, the International Convention for the Control and Management of Ship's Ballast Water and Sediments (BWMC) (IMO, 2004), which shall come into force in September 2017 (IMO, 2016) defines procedures and sets technical requirements to reduce the threat of harmful aquatic organisms and pathogens transferred by ships ballast water. Another instrument is the Code of Practice on the Introductions and Transfers of Marine Organisms by the International Council for the Exploration of the Sea (ICES, 2005; Gollasch, 2007) recommends procedures and practices to diminish the risks of detrimental effects from the intentional introduction and transfer of marine and brackish water organisms.

There are numerous regional multi-lateral treaties, conventions, and agreements in place that address the issues of aquatic bioinvasions, such as the Barcelona Convention (Mediterranean Sea), the Helsinki Commission (Baltic Sea), the OSPAR Commission (North-East Atlantic including the North Sea), the UNEP regional Seas programs, the South Pacific Regional Environmental Program, and the Asia-Pacific Economic Cooperation (Hewitt et al., 2009 and references therein). Also, several nations have established regulatory frameworks for the prevention and management of intentional and accidental bioinvasions, for example, the US Invasive Species Act, the Biosecurity Act of New Zealand (Hewitt et al., 2009 and references therein).

At the European level, the EU Regulation on the Prevention and Management of the Introduction and Spread of Invasive Alien Species (2014) was adopted, indicating, inter alia, "...Prevention is generally more environmentally desirable and cost-effective than reaction after the fact, and should be prioritized...." Here, the clear distinction is made between the primary introduction of an alien species, which should be prevented, and its secondary spread within a region, which, in the aquatic world, practically seems to be unmanageable.

All the above legal and administrative, global and regional instruments require a robust, scientifically sound indicator(s) to measure their effectiveness in terms of reducing unwanted invasions. For example, the European Environment Agency (EEA) proposed an indicator "...Cumulative numbers of alien species in Europe since 1900..." to measure progress toward achieving the CBD goal (EEA, 2007). Counts from different countries were assigned to decades, data were provided by national authorities and only established species with selfsustaining populations were considered (EEA, 2012). In addition to the EEA proposal, the Marine Strategy Framework Directive (European Commission, 2008) includes within the 11 qualitative descriptors the non-indigenous Species (NIS) as one of the elements to be assessed to determine if an ecoregion is in good environmental status or not. To assess it, the European Commission (2010) proposed a series of indicators which include "...Trends in abundance, temporal occurrence and spatial distribution in the wild of non-indigenous species...," similar to the CBD indicator by EEA (2007). It was this indicator that was used by most Contracting Parties in their initial environment status assessments for the MSFD (ICES, 2016).

An elevated number of NIS generally indicates a greater level of exposure of a marine area to anthropogenic activity (Olenin et al., 2010). However, in contrast to most indicators of human impacts, the Cumulative number of NIS fails to show a direct correlation with environmental degradation gradients (MacDougall et al., 2006). Whether, or not, NIS become established is only partly related to the environmental status of an area; and it also depends on biological traits of the species (e.g., Cardeccia et al., in press), integrity of native ecosystems (Didham et al., 2005) and availability of resources (Davis, 2009).The "...Cumulative number of NIS..." as such, is of lesser importance for management than the "...Number of species transferred by a vector(s)...," which aids any practical prioritization of preventive measures. This is because for many early introductions taxonomic knowledge was incomplete and records were seldom kept (Carlton, 2009), also the presence of a NIS often remained unnoticed until they will have become obvious and created some nuisance impact (Olenin and Minchin, 2011). As it was shown in a recent regional overview (Ojaveer et al., accepted) even in a marine region with a long history of biodiversity research, such as the Baltic Sea, where due to natural circumstances, and recent geological history, species richness is low and any new arrival is likely to be more visible than elsewhere, there is a weak availability of introduction event records from before the 1950s. It is unfortunate that uncertainty is an inherent component in bioinvasion studies and as a result, the "Cumulative number of NIS" is compromised by gaps of knowledge, especially during the early periods when introductions were not effectively recorded. It is for this reason that the value of this indicator in measuring the response of marine systems to human pressures (sensu Borja et al., 2016) is limited.

We present a new indicator "Arrival of new NIS" (nNIS), aimed to establish "windows" (or hotspots) of primary introductions entering regional seas and to reveal the main pathways/vectors involved. The indicator is suitable for all easily recognizable taxa arriving at different geographical scales from a country coast within a Large Marine Ecosystem (LME, sensu Sherman, 1991) to an entire LME or a larger biogeographical level, that could involve two, or more, neighboring LMEs. We present the calculation method, show some applications of the indicator to a set of study-cases at different geographical levels and discuss its advantages and possible limitations.

# MATERIALS AND METHODS

# Information Support, Introduction Event, and Recipient Region

This study is based on data accumulated in the Information system on Aquatic Non-Indigenous and Cryptogenic Species– AquaNIS, where all geographic information is arranged in a hierarchical order ranging from oceans, ocean sub-regions, LMEs, sub-regions of LMEs to smaller entities, such as ports (Olenin et al., 2014; AquaNIS, 2016). All countries are linked to relevant LMEs or LME sub-regions. This provides database search combinations "country + LME" or "country + LME sub-region" for different coasts and for a country that borders different seas, e.g.,: "Germany within the LME 23 Baltic Sea," "Italy within the Adriatic Sea, a sub-region of LME 26 Mediterranean." Such data may also be aggregated at different geographical scales and in different combinations, e.g., "LME 22 North Sea + LME 23 Baltic Sea," or "Germany within both the North Sea and Baltic Sea coasts," which would be needed to define the level of primary introduction.

The basic data entry in AquaNIS is an introduction event record, documenting a species introduction into a recipient region, defined as a country or a country sub-area within an LME or LME sub-region. Registration of an introduction event includes the date of the first record when a species was noticed in a recipient region as well as pathways and vectors of introduction according to different levels of certainty. In addition, AquaNIS gathers and disseminates information on species biological traits, environmental tolerance limits, availability of molecular data for identification, habitats, etc. Moreover, the information system is equipped with a structured "search" function that allows for retrieving and organizing data by multiple and complex search criteria (for details see Olenin et al., 2014).

# nNIS, Assessment Unit, Initial, and Periodic Assessments

nNIS is the number of new NIS in an assessment unit, which were recorded and compared with the initial or periodic assessment. In this study, the assessment unit is equal to a recipient region as it is in AquaNIS. To illustrate such a calculation we selected a range of assessment units from different marine environments as examples, where all entered data has been checked for quality. The areas selected were the Baltic Sea and coastal waters of Italy and the records may be examined on-line (AquaNIS, 2016).

In the Baltic Sea, there are 10 recipient regions: eight bordering countries and the two separate regions of the Russian Federation, the Sankt-Petersburg area in the Gulf of Finland (RU\_S) and the Kaliningrad area in the south-eastern Baltic (RU\_K). In Italy, there are three recipient regions: the Adriatic Sea, Western Mediterranean (the western coast of the Italian mainland and north coast of Sicily) and Eastern Mediterranean (the south coast of the Italian mainland and south coast of Sicily).

The initial assessment is the first inventory of all NIS present in a recipient region. For example, most EU Member States will have performed an initial environmental status assessments under the MSFD and reported the cumulative number of NIS in the waters under their jurisdiction recorded by 2010. In the present study, all new NIS, arriving after this date, were counted. A periodic assessment is a record of new NIS arrived to a recipient region since the first inventory. The periodicity of the assessment may be defined by the management needs, for example, it will be 6 years for MSFD and in maximum 5 years for granting ballast water management exemptions under BWMC (Olenin et al., 2016).

# The Level of Primary Introduction and the Secondary Spread

A primary introduction is the first arrival of a NIS to a particular assessment unit, while the secondary spread is its further dispersal to other locations. The level of a primary introduction can be assessed at different geographical scales, from a recipient region to an LME or a larger biogeographical area. From the environmental policy point of view, more important are those primary introductions, which are new not only for a coast of a particular country (recipient region), but for an entire LME or, even for a larger biogeographical region, for example for two or more neighboring LMEs or LME sub-regions.

The levels of primary introduction should be defined for each case study separately, depending on the availability of data for larger geographical scales. In the Baltic case study, the lowest level of primary introduction (L1) is one of the 10 recipient regions, the next level (L2) is the entire LME (Baltic Sea), and the highest level (L3) is two neighboring LMEs (Baltic Sea and North Sea). Thus, nNISL1 shows how many new NIS were recorded in a particular country since the previous assessment, nNISL2 shows how many of them were new for the Baltic Sea, and nNISL3 indicates the number of NIS new for both the Baltic and the North seas. In the Italian case study, the lowest level (L1) is one of the three recipient regions, while the next level (L2) is all Italian marine areas together. The highest level here would be the entire Mediterranean Sea, the data for which currently is under development.

# Data Extraction Method

AquaNIS offers an opportunity to extract the value of nNISL1 directly, using the built-in "Search" function for the recipient region and year, from which the new arrivals should be calculated. The system can retrieve the number of species (i.e., nNISL1) and the number of introduction events. Data extraction for nNISL2 and L3 values involves several steps, using a combination of "Search" and "Comparison of search results" functions (**Table 1**).

The calculations of all nNIS values presented here are based on data that has accumulated in AquaNIS to July 28, 2016 (AquaNIS, 2016). All entries for cryptogenic species were not considered in our calculations.

# Level of Certainty

According to the AquaNIS (2016) definition, an introduction event should be ascribed to a pathway/vector with the defined level of certainty (**Table 2**).

# RESULTS

# The Baltic Sea Case Study

In total, 26 NIS involving 36 introduction events were recorded in 10 recipient regions within the Baltic Sea since 2010. Of these, 12 NIS are new to the Baltic (**Table 3**), while 14 were involved in secondary spread within the Baltic, i.e., previously were known from at least one of the 10 recipient regions. However, it is important to distinguish between the NIS, which were known in the Baltic Sea before and after the previous assessment. In this



#### TABLE 2 | Levels of certainty applied for pathways and vectors in AquaNIS\*.


\**Based on Minchin (2007) and Olenin and Minchin (2011).*

case, one species, the brackish water clam Rangia cuneata, first in the Baltic was recorded in 2010 in the Russian part of the southeastern Baltic (RU\_K) and during the assessment period have spread to three other recipient regions: Poland, Lithuania, and Estonia (**Table 3**). The sedentary polychaete Hypania invalida was found simultaneously in a water body shared between Germany and Poland (Szczecin Lagoon); therefore, the primary introduction is ascribed to 2 countries.

The difference between nNIS L1 and L2 for a recipient region indicates the number of species, which this particular region received during the assessment period due to secondary spread from other parts of the Sea. For example, since the initial assessment 12 NIS were recorded for Germany, i.e., nNIS L1\_Germany = 12 (**Table 3**). Of these, seven were primary introductions to the Baltic Sea (nNIS L2\_Germany = 7), and one of these six (the amphipod Echinogammarus trichiatus) is new at the level of the larger biogeographical region, comprising both the Baltic Sea and the North Sea LME (nNIS L3\_Germany = 1). In Poland, seven new species were recorded (nNIS L1\_Poland = 7). Two of them were new for the Baltic (nNIS L2\_Poland = 2): H. invalida and the tubificid oligochaete Limnodrilus cervix, while R. rangia which first appeared in 2011 was not counted at L2 as it was earlier recorded in the neighboring region of Russia. Finally, L. cervix was new at the scale L3, i.e., this species entered the Baltic Sea and North Sea via the Polish coast. In Sweden six new NIS appeared, however only one of them was new at the LME level, i.e., nNIS L2\_Sweden = 1.

The pathway analysis at the level of the Baltic Sea LME (L2) reveals that "Vessels" were responsible for 10 primary inoculations, involving ballast water, ballast tank sediments, hull fouling, etc. (AquaNIS, 2016; **Table 3**), with levels of certainty ranging from "Direct evidence" (the sea anemone Diadumene lineata found on ship hull) to "Very likely" for three and six "Possible" primary introductions. The pathway "Natural spread from neighboring regions," indicating secondary spread of NIS was ascribed for four primary introductions. The pathway "Culture activities" involving the vectors aquaculture equipment, stock movements and releases and escapees, was ascribed for two primary introductions.

At the scale of a larger biogeographical region, covering two neighboring LMEs (L3), four species were found to be new for both seas, while 18 were known earlier from the North Sea and may have spread from there to the Baltic Sea.

#### The Italian Seas Case Study

A total of 33 NIS were registered in the three Italian recipient regions since 2010, including 24 species previously not registered in Italy, and a further 9 NIS, introduced before 2010, which have spread further to adjacent coastal regions (AquaNIS, 2016). New


*introduction.*

arrivals have been recorded in similar numbers in all three coastal regions: 12, 14, and 15 NIS for the Adriatic Sea, Italian Eastern and Italian Western Mediterranean, respectively (**Table 4**). Some of these nNIS appeared in multiple regions along the Italian coast, having spread rapidly.

The number of likely pathways responsible for the new introductions is higher in Italy than in the Baltic Sea region: besides vessels and culture activities, Italian coasts have been receiving a worrying high number of species that have likely entered the Mediterranean through the Suez Canal, including species of high concern for human health (e.g., the stinging jellyfish Rhopilema nomadica and the toxic pufferfish Lagocephalus sceleratus), as well as species possibly associated with aquarium trade.

## DISCUSSION

## Defining the Environmental Target

In a recent review, Marchini et al. (2015), highlighted that inventories listing cumulative numbers of marine alien

TABLE 4 | New arrivals of non-indigenous species to Italian coastal seas (Adr., Adriatic Sea; E. Med., Eastern Mediterranean; W. Med., Western Mediterranean) since 2010.


*Year of the primary introduction indicates: underlined, new for entire Italy (L2); unformatted text, new for a recipient region (L1); (in brackets), recorded before the assessment period. Pathway (Aqua, Aquarium trade; Cult, Culture activities; Suez, Suez Canal; Vess, Vessels) and level of certainty (*\*\**Very likely,* \**Possible) are indicated only for the highest level of primary introduction (L2).*

species in the Mediterranean Sea present several types of uncertainty, unfortunately resulting in a confused picture of the phenomenon. Problematic species identifications, doubtful records and unknown native origin affect large portions of such inventories, thus preventing a comprehensive assessment of marine bioinvasions (e.g., Katsanevakis et al., 2016). However, modern taxonomy and molecular tools (Zaiko et al., 2015; Bucklin et al., 2016; Raupach et al., 2016; Viard et al., 2016), combined with a growing scientific concern for marine bioinvasions, are contributing to improve the quality of the current species records. In other words, while it is extremely challenging to reliably reconstruct the past history of redistribution of species due to human intervention (Carlton, 2009; Clavero, 2014), we now have more effective and accurate tools to measure the changes currently occurring. Modern records of new arrivals are often delivered with more in-depth analysis of possible vectors and more detailed knowledge on the NIS ecology (e.g., Reusch et al., 2010), and can therefore provide higher-quality knowledge to support scientifically based advice on management decisions. Therefore, an indicator based on new arrivals (nNIS), despite its inherent time-limitation, could offer a more reliable picture of the problem of bioinvasions and supply information essential for early warning initiatives, horizonscanning programs (sensu Roy et al., 2014) and identification of an environmental target for the MSFD Good Environmental Status (GES) descriptor 2 "Non-indigenous species" (European Commission, 2008). Further, such an indicator enables an assessment how effective implemented vector and pathway management measures are.

Defining the environmental target for the nNIS indicator, the following considerations should be taken into account:


Thus, in general, the environmental target for nNIS may be formulated for a country as "No new primary introductions of NIS by a particular pathway/vector per assessment period via the territory of that country." For example, it may sound like "No new primary introductions of NIS by ship's ballast water to the Baltic Sea via territory of Lithuania for the assessment period," i.e., the environmental target is: "nNISL2\_Lithuania (by ballast water) = 0." In this case, the environmental target will be achieved if during the assessment period no NIS, new for the entire Baltic Sea, entered the marine area under jurisdiction of Lithuania by ships ballast water. Thus, only primary introductions at the level of entire LME (L2) are counted, i.e., secondary spread is excluded.

The environmental target should be harmonized at the level of LME or a larger region, including several neighboring LMEs, where secondary dispersal of NIS may take place with currents and other natural means. For example, the Baltic Sea Action Plan (HELCOM, 2007) sets the environmental management objective "No introductions of alien species from ships." In fact, nNIS indicates the success or failure of the preventing measures and its highest target value could be set as "No new human-mediated primary introductions on the level of the European regional seas," although this seems to be impossible with the management options we have today.

# Defining the Size of Assessment Unit

The size of the assessment unit for the nNIS indicator may vary depending on the practical considerations, management needs, and, naturally, data availability. In this study, the assessment unit was equal to a country marine area within an LME or LME sub-region, i.e., to a recipient region as it is defined in AquaNIS (2016). Such subdivision is determined by practical needs, because the management decisions on preventive measures are taken at the level of national authorities.

The smallest possible level is a port and/or its vicinities, where it is practical to perform a NIS survey. Such biological surveys in ports are obligatory, for example, for taking decision on granting exemptions under BWMC (David et al., 2013; David and Gollasch, 2015; Olenin et al., 2016).

The largest level for the nNIS assessment is a marine region under a regional convention, e.g., North-East Atlantic (OSPAR) or Baltic Sea (HELCOM), where measures to prevent NIS introduction may be practically coordinated. The higher geographical scale to calculate nNIS including all regional seas of an entire continent so far is not achievable, because such datasets do not exist, or, at least are not publically available.

# The Technical Precondition and Possible Limitations of the Indicator

The technical precondition for the calculation of nNIS is the availability of verified and continuously updated source of information on introduction events, e.g., a NIS database. Depending of the size of the assessment unit, the geographical coverage of such a database may range from national to regional or interregional. Ideally, such information source should be global or, at least, continent-wide. For example, AquaNIS, used in this study, is being regularly updated by the ICES Working Group on Introductions and Transfers of Marine Organisms and contains information from other world regions as well.

Our results show that the average number of new arrivals having occurred in the past few years in different European sea regions is high, and therefore a reliable database can be achieved only by a continuous and scrupulous work of data checking and update. However, a scientific community needs to achieve the long-term maintenance and reliability of such databases, because they require frequent updating and corrections (Costello et al., 2014). Without continuous maintenance, update and data quality control, the usefulness of the database diminishes over time and its users may be hampered by outdated and therefore misleading information (Olenin et al., 2014) as it was shown in the Mediterranean Sea case (Marchini et al., 2015).

Ideally, all newly published records of NIS introduction events in journals, such as Biological Invasions, Aquatic Invasions, BioInvasion Records, Marine Biodiversity Records, Mediterranean Marine Science, shall be standardized and immediately entered in a global online NIS information system. That would speed up the transfer of knowledge on biological invasions and aid the analysis of new arrivals in relation to all other relevant data stored in the database.

Another technical precondition is that nNIS are scored for recipient regions or LMEs, where regular NIS surveys or, at least, a well-established long-term biological monitoring is in place (Olenin et al., 2011; Lehtiniemi et al., 2015). Taking into account that there are several international instruments (BWMC, ICES Code of Practice, MSFD, etc.) which include NIS monitoring for management purposes and for measuring progress toward achieving their goals, it would be feasible to coordinate NIS surveys regionally. For example, rapid assessment surveys focused on target species (e.g., Minchin et al., 2009, 2016) may be arranged simultaneously by several countries within an LME, in the same way as it is done for fishery surveys (ICES, 2014).

It should be taken into account, also that in some cases our ability to distinguish between the primary introduction and secondary spread may be limited. This is because, multiple introductions of a NIS from outside an LME area also possible as in the case of the American comb jelly Mnemiopsis leidyi "very-likely" spread via ballast waters from two distinct source populations from the western Atlantic to the Black Sea and the North Sea (Reusch et al., 2010). Multiple introductions make distinctions between primary introduction and secondary spread within an LME or larger biogeographical region more difficult. Development of eDNA techniques could assist in determining origin in the future (Rius et al., 2015).

# Risk Assessment, Management Implications, and Policy Relevance

It is difficult to predict those nNIS that may become invasive and cause problems to the environment, economy and/or human health in a recipient region. No control or eradication of invasive alien species without affecting other components of the ecosystem is feasible after an invasion process is underway and a species has become established within an ecosystem. Given the severity of problems that can be caused by biological invasions, it is mandatory for policy and management to focus on the pathways and vectors with the aim to prevent further introductions (e.g., European Commission, 2014).

The nNIS indicator evaluates the effectiveness of prevention measures where these can be employed, be it the ballast water management, precautionary approach in aquaculture or life food trade regulation. For example, strict ballast water management rules applied in recent decades in US and Canada resulted in no new fresh water introductions (i.e., nNIS L3 = 0) in the Laurentian Great Lakes Region since 2006 (Scriven et al., 2015). In contrast, the nNIS value for Italy, obtained in the present study, clearly shows a high exposure arising from a large geographical dispersion of introduction events and several pathways. For example, the relatively high number of newly arrived NIS that can be associated with aquarium releases (3 out of 24) indicates a requirement for a greater implementation of a code of practice in aquarium trade (Scalera et al., 2012), as well as a greater general public awareness. There is a need to review management for those NIS that might be prevented from becoming established where direct anthropogenic introductions can be regulated.

It is important to manage primary introductions at the scale L2 and L3, because the secondary spread, which can be inevitable, may seriously compromise the ability for any practical regulation. This is because a further spread may involve dispersal by the same pathway or by multiple pathways that might act in relay. What is not possible to manage are the natural processes involving tidal movements, alongshore drift, waterfowl, and other aquatic biota as a carrier of NIS either within an LME or from neighboring LME.

# CONCLUSION

The proposed nNIS indicator is clearly associated with anthropogenic pressure in terms of specific pathways/vectors involved and this may help to prioritize management actions. Regional Sea Conventions have been working on developments to harmonize the MSFD indicators, considering new arrivals as potentially useful parameter for environmental status assessments (e.g., HELCOM, 2012). This indicator provides a clear measure of effectiveness of legal and administrative instruments aimed at prevention of NIS species introductions, such as the International Convention for the Control and Management of Ship's Ballast Water and Sediments, the EU Regulation 1143/2014 on Invasive Alien Species, the MSFD and the ICES Code of Practice on the Introductions and Transfers of Marine Organisms.

# AUTHOR CONTRIBUTIONS

SO, ML, SG, and DM conceived the paper. SO and DM wrote the first draft, DM checked the language. The following authors provided the case studies: AM (Mediterranean sea), SO, AN, ML, and SG (Baltic sea). AN, AM, and GS contributed to the assessment analyses and presentation of the results. All authors contributed largely to the multiple drafts of the manuscript and approved its last version for publication.

# ACKNOWLEDGMENTS

The study was supported by (1) the DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status, http://www.devotesproject.eu) project funded by the European Union under the 7th Framework Programme, The "Ocean of Tomorrow" theme (Grant Agreement No. 308392 (for SO and AN); (2) BIO-C3 (Biodiversity changes—causes, consequences and management implications, http://www.bio-c3.eu) project funded

#### REFERENCES


by BONUS, the joint Baltic Sea research and development programme (Art 185), funded jointly by the Academy of Finland (Grant Agreement No. BONUS-1/2014) and by the EU 7th Framework Programme for research, technological development and demonstration (for ML); and (3) the Taiwan–Latvia– Lithuania Cooperation Project BALMAN "Development of the ships' ballast water management system to reduce biological invasions," project # TAPLLT-14-013 (for DM and GS). The authors are grateful to prof. Anna Occhipinti-Ambrogi for fruitful discussion on the bioinvasion indicators.

management of the introduction and spread of invasive alien species. Off. J. Eur. Union 57, 35.


Non-indigenous Species, Vol. 44. EUR 24342 EN. Luxembourg: Office for Official Publications of the European Communities.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Olenin, Naršˇcius, Gollasch, Lehtiniemi, Marchini, Minchin and Srebalien ˙ e. This is an open-access article distributed under the terms of the Creative ˙ Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# **Assessing the status in an integrative way**

# What Is Marine Biodiversity? Towards Common Concepts and Their Implications for Assessing Biodiversity Status

Sabine K. J. Cochrane1, 2 \*, Jesper H. Andersen<sup>3</sup> , Torsten Berg<sup>4</sup> , Hugues Blanchet 1, 5 , Angel Borja<sup>6</sup> , Jacob Carstensen<sup>7</sup> , Michael Elliott <sup>8</sup> , Herman Hummel <sup>9</sup> , Nathalie Niquil <sup>10</sup> and Paul E. Renaud<sup>2</sup>

*<sup>1</sup> SALT Lofoten AS, Svolvær, Norway, <sup>2</sup> Arctic R&D Department, Akvaplan-niva, Tromsø, Norway, <sup>3</sup> NIVA Denmark Water Research, Copenhagen, Denmark, <sup>4</sup> MariLim Aquatic Research GmbH, Schönkirchen, Germany, <sup>5</sup> University of Bordeaux, UMR EPOC, Pessac, France, <sup>6</sup> Marine Research Division, AZTI Tecnalia, Pasaia, Spain, <sup>7</sup> Department of Bioscience, Aarhus University, Roskilde, Denmark, <sup>8</sup> Institute of Estuarine and Coastal Studies, University of Hull, Hull, UK, <sup>9</sup> NIOZ Royal Netherlands Institute for Sea Research, Yerseke, Netherlands, <sup>10</sup> Centre National de la Recherche Scientifique/Université Caen Normandie, BOREA, Caen, France*

#### Edited by:

*Marianna Mea, Ecoreach srl, Italy; Jacobs University of Bremen, Germany*

#### Reviewed by:

*Ricardo Serrão Santos, University of the Azores, Portugal Christos Dimitrios Arvanitidis, Hellenic Centre for Marine Research, Greece*

\*Correspondence:

*Sabine K. J. Cochrane sabine@salt.nu; sc@akvaplan.niva.no*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *15 July 2016* Accepted: *14 November 2016* Published: *15 December 2016*

#### Citation:

*Cochrane SKJ, Andersen JH, Berg T, Blanchet H, Borja A, Carstensen J, Elliott M, Hummel H, Niquil N and Renaud PE (2016) What Is Marine Biodiversity? Towards Common Concepts and Their Implications for Assessing Biodiversity Status. Front. Mar. Sci. 3:248. doi: 10.3389/fmars.2016.00248* Biodiversity' is one of the most common keywords used in environmental sciences, spanning from research to management, nature conservation, and consultancy. Despite this, our understanding of the underlying concepts varies greatly, between and within disciplines as well as among the scientists themselves. Biodiversity can refer to descriptions or assessments of the status and condition of all or selected groups of organisms, from the genetic variability, to the species, populations, communities, and ecosystems. However, a concept of biodiversity also must encompass understanding the interactions and functions on all levels from individuals up to the whole ecosystem, including changes related to natural and anthropogenic environmental pressures. While biodiversity as such is an abstract and relative concept rooted in the spatial domain, it is central to most international, European, and national governance initiatives aimed at protecting the marine environment. These rely on status assessments of biodiversity which typically require numerical targets and specific reference values, to allow comparison in space and/or time, often in association with some external structuring factors such as physical and biogeochemical conditions. Given that our ability to apply and interpret such assessments requires a solid conceptual understanding of marine biodiversity, here we define this and show how the abstract concept can and needs to be interpreted and subsequently applied in biodiversity assessments.

Keywords: conceptual models, marine biodiversity, ecosystems, food-webs, components, assessment

# INTRODUCTION

The term "biodiversity", first used almost three decades ago as a derivative of "biological diversity" (Wilson, 1985, 1988) today is one of the most often cited terms in both ecological research and environmental management and conservation (i.e., 141,214 papers in ISI Web of Science, as consulted on 27th April 2016). However, its precise definition and our understanding of the concept varies widely both between and within disciplines. Biodiversity is recognized to encompass ".. the variability among living organisms from all sources including, inter alia, terrestrial, marine, and other aquatic ecosystems and the ecological complexes of which they are part; this includes diversity within species, between species and of ecosystems." (CBD, 1992). The elements of biodiversity are fundamental properties of an ecosystem, and, in the marine realm, these encompass all life forms, including the environments they inhabit, and at scales from genes and species to ecosystems (see Wilson, 1988; Boero, 2010). Biodiversity can be described as an abstract aggregated property of those ecosystem components (Bengtsson, 1998) and can relate to the structure or function of the community where structure relates to the system at one time whereas functioning relates to rate processes (Gray and Elliott, 2009). The structural aspect is represented by the various marine life-forms, ranging from the smallest prokaryote to the largest mammal, and inhabiting some of the most extreme environments. These species exhibit a diversity that probably exceeds that found in terrestrial environments (Heip, 1998, 2003). The functional aspect is represented by the relationships among and between these marine organisms and the environments they inhabit, and is defined in terms of rates of ecological processes (Strong et al., 2015); most notably they include physiological processes, predator-prey relationships, trophic webs, competition, and resource partitioning. These functions vary on both temporal and spatial scales (Solan et al., 2006), and include some of the most important ecosystem services, including oxygen provisioning, CO<sup>2</sup> sequestration, and re-mineralization of nutrients (Duarte and Cebrian, 1996; Costanza et al., 1997; van den Belt and Costanza, 2012). Both structural and functional elements contributing to biodiversity play a fundamental role in maintaining and defining healthy marine systems (Selig et al., 2013).

In essence, the marine ecosystem is comprised of three interlinked processes (Gray and Elliott, 2009). Firstly, the physico-chemical system creates a set of fundamental niches (most often the water column and substratum) which then are colonized by organisms according to their environmental tolerances—these may be termed environment-biology relationships. Secondly, the organisms interact with each other in, for example, predator-prey interactions, competition, recruitment, feeding, and mutualism—these are biologybiology relationships. Thirdly, the resulting ecology has the ability to complete the cycle with feedback loops and modify the physico-chemical system through bioturbation, space or material removal or change, bio-engineering, for example; these may be termed biology-environment relationships. Superimposed on these three systems are anthropogenic influences which then perturb the systems.

Human activities produce a range of pressures on marine systems, some of which may lead to irreversible changes (e.g., deyoung et al., 2008; Elliott et al., 2015). This may have immediate consequences for patterns of biodiversity and consequently for the critical ecosystem services they provide (Costanza et al., 1997, 2014; De Groot et al., 2002, 2010). Those ecosystem services can be grouped into provisioning, regulating, supporting and cultural ones which, after adding human complementary assets, in turn lead to societal benefits (Turner and Schaafsma, 2015).

In this context, the European Marine Strategy Framework Directive (MSFD) requires Member States to achieve Good Environmental Status (GES) (European Commission, 2008). The directive comprises 11 qualitative descriptors of GES, of which biological diversity is the first, but most if not all of the others can be considered to refer to some part of biodiversity in its broad sense, assuming we also consider habitats and their condition as being within the term; indeed it can be assumed that if the biodiversity descriptor has been satisfied then by definition all others are satisfactory and vice versa (Borja et al., 2013). In order to know whether the goal of GES has been achieved, an assessment needs to be performed that measures the current environmental status, hence this involves quantifying the abstract ecosystem feature biodiversity. For this, the European Commission has defined a number of GES criteria and indicators that represent and quantify various aspects of environmental status and biodiversity (European Commission, 2010). The available indicators in Europe, for the MSFD implementation, have been recently collated (Teixeira et al., 2016), and a method to select the most adequate has been proposed (Queiros et al., 2016). Then, some of them have been used in assessing the environmental status across regional seas (Uusitalo et al., 2016).

It is axiomatic that one cannot manage a system unless it can be measured and those measures require to be SMART (Specific, Measurable, Achievable, Realistic, and Time-bounded) otherwise it is not possible to determine whether management has achieved the desired result (Elliott, 2011). Hence the importance of quantitative indicators but these must be comparatively simple if they are to be operational (Rombouts et al., 2013; Borja et al., 2016), although many of these overlap, and such redundancies can compromise the efficiency and accuracy of assessments (Berg et al., 2015). The recent trend toward using long lists of indicators for an integrative assessment increases the risk of such overlaps (Teixeira et al., 2016). There are many potential combinations of study approaches and thus, before compiling the indicators, any large-scale or comparative assessment of biodiversity first requires a unified approach and a workable conceptual understanding of biodiversity.

Given the inherent complexity of biodiversity and the services which the ecosystems provide as a consequence of their biodiversity (see, for example, Heip, 2003; Bartkowski et al., 2015; Farnsworth et al., 2015), it is imperative to depict these into one or more simple conceptual models. There are many ways to view marine systems, depending on the questions asked, the management goals set and typically, as with any complex system, disaggregating the various levels of complexity allows us to better understand each of the components and their major interactions (Brooks et al., 2016). Consequently, an assessment of biodiversity used to answer a specific question will benefit from a set of conceptual models which together represent the various aspects of biodiversity. Together, these models provide a multi-faceted view of biodiversity and help users to identify the necessary elements to include in an environmental assessment by focusing on the aspects of biodiversity most relevant to the specific question and goal.

A common conceptual framework on marine biodiversity is presented here to facilitate integrative assessment of environmental status and implementation of the relevant legislation. We present a context-driven, multi-faceted view on biodiversity that will enable selection of the appropriate assessment elements and indicators. The framework is required to implement and further develop policies and practice to maintain biodiversity in the context of the sustainable management of human activities.

#### CONCEPTUAL VIEWS OF BIODIVERSITY

Marine biodiversity is an aggregation of highly inter-connected ecosystem components or features, encompassing all levels of biological organization from genes, species, populations to ecosystems, with the diversity of each level having structural and functional attributes (**Table 1**). Further, marine biodiversity, or any of its components, can be assessed at various temporal or spatial scales. A conceptual model of marine biodiversity and its interpretation therefore depends on the questions being asked, which of the different components are emphasized, and the information and understanding available, especially of the connectivity and feedbacks in the system. By definition, this involves the implicit understanding that the components are all part of a larger and inter-linked system, where changes in one element inevitably will produce knock-on effects elsewhere (Gamfeldt et al., 2015). These may be regarded as bottom-up processes, causing change from the cell to the ecosystem and from the physicochemical system to the landscape ("seascape") system. Similarly, they can be regarded as the responses in a top-down system focusing on the upper level (seascape and ecosystem) which is often the end-point of marine management and the focus of the current review. Accordingly, this review does not specifically address genetic, molecular, physiological, biochemical, population, and size-biomass-spectrum aspects of biodiversity (Zacharius and Roff, 2000; Kenchington, 2003; Palumbi, 2003; Gray and Elliott, 2009), as these are both intrinsic and implicit aspects within the concept of biodiversity, whichever viewpoint is emphasized. We thus specifically cover only the upper levels (**Table 1**, bold entries), but retain the understanding of the multi-level complexity within these.

Hence modeling such a complex system with a view to marine management requires (i) pragmatic simplifications through disaggregation of the elements into various conceptual viewpoints, followed by (ii) a context-driven re-aggregation of the necessary components. We here provide three illustrative examples of such conceptual upper-level views on marine biodiversity, where the information retrieved is restricted to that relevant to the main focus, or viewpoint (**Figure 1**). The first focuses on structural aspects using a classical taxonomic approach to biodiversity (structural taxonomic biodiversity). The second focuses on the functional aspects of biodiversity (functional ecosystem biodiversity), and the third illustrates food-webs as one of the most used types of a combined view on both structural and functional aspects of biodiversity (foodweb biodiversity). These examples only capture parts of the full complexity of biodiversity (**Table 1**) but are the most commonly found in specific user-driven contexts.

### Structural Taxonomic Biodiversity

Since the establishment of the hierarchical system of binomial nomenclature (Linné, 1735), a major focus of biological studies has been to categorize observed organisms into taxonomic units, and to describe new species as they are discovered. Quantitative taxonomic data sets are a useful tool in environmental assessments, with typical indicators being species (taxon) richness, and population abundance and biomass within a place, between areas or over time. This is especially important in nature conservation planning (Sarkar and Margules, 2002), notably because habitat destruction is a major driver of species extinctions, particularly those with narrow distribution ranges (Pimm et al., 2014), such that adequate knowledge of the structural taxonomic biodiversity of a particular area will help to preserve its endemic species. A taxonomic inventory and the associated habitats and their changes in space and time then becomes central to environmental impact assessments (Pearson and Rosenberg, 1978; Olsgard and Gray, 1995; Rosenberg et al., 2001; Borja et al., 2003), studies of marine protected areas (Klein et al., 2015) and the compliance with marine diversity and ecosystem health governance instruments such as the EC Habitats Directive (e.g., Boyes and Elliott, 2014).

The EU MSFD addresses biodiversity components within two main categories: (i) main species groups, and (ii) habitats and their associated communities (habitat diversity and mosaics) (see Cochrane et al., 2010; Hummel et al., 2015). The main species-level groups include mammals, birds, fish, cephalopods, and reptiles. Within the marine habitats, watercolumn communities comprise pelagic microbes, phyto- and zooplankton, whereas seafloor communities encompass benthic micro, macro- and mega- fauna as well as primary producers such as seagrasses and macroalgae. In addition, other species such as those included under the European Union legislation or international conventions, charismatic or non-indigenous species and genetically distinct forms (varieties or subspecies) of native species may be included, depending on the particular assessment area and questions being addresses. In the MSFD, the categories for birds, fish, and mammals are further sub-divided into main functional categories, mostly based on their feeding and/or depth preferences (**Table 2**). This, however, introduces a functional division into the otherwise purely structural view.

The predominant seabed and water column habitat types can effectively be characterized in terms of a pragmatic selection of the major categories under the European Nature Information System (EUNIS) scheme (Cochrane et al., 2010; Galparsoro et al., 2012, 2015) (**Table 3**). The biological communities associated with those habitats can then be addressed; thus extending the conceptual view from purely taxonomic entities to higher-level structural aggregations of taxa as part of their biotope (Olenin and Ducrotoy, 2006) (**Figure 2**). This structural view potentially omits the functional attributes or traits of the populations and communities associated with habitats although some of the structural attributes may be regarded as surrogates (proxies) for functional ones (Gray and Elliott, 2009). For example,


TABLE 1 | Structural and functional biodiversity examples across levels of biological organization (topics focused on in the current paper in bold) (extensively modified from Zacharius and Roff, 2000).

the benthic communities can be characterized in terms of proportional representations of different traits, feeding guilds, motility, burrowing activities etc. (Bremner et al., 2006a,b; Cochrane et al., 2012) but these have not previously been the main focus of structural biodiversity; most methods have centered on the plethora of quantitative means of defining benthic community structure (Gray and Elliott, 2009). However, recognizing and measuring functional diversity within the benthos also has become of increasing importance from a management perspective (Reiss et al., 2015).

A high biodiversity, including species richness, may enhance ecosystem processes and promote long-term stability by buffering, or insuring, against environmental fluctuations (Yachi and Loreau, 1999; Loreau, 2000). Conversely, a loss of biodiversity may impair ecosystem functioning, and thus also TABLE 2 | Predominant functional and/or feeding groups within the main biodiversity components for application in assessment of motile biodiversity components.


\**Annex III of the MSFD refers to "seabirds"; this term is commonly used to distinguish certain types of marine birds (petrels, gannets, cormorants, skuas, gulls, terns, and auks) from water birds (waders, herons, egrets, ducks, geese, swans, divers, and grebes). To avoid possible confusion with this narrower use, the term "birds" is used here. The ecotypes for seabirds (offshore and inshore) are as used by the ICES Working Group on Seabird Ecology for assessment of trends in seabird populations (ICES, 2009).* \*\**Species which depend upon ice and ice-driven biological processes for habitat, shelter, reproduction or feeding for at least some parts of the year, or for parts of their life-cycle.*

the services provided (Loreau and Hector, 2001). At least in the marine realm, habitat structure obviously influences the number of niches available for colonization and thus can indicate the number of types (species, traits, etc.) which can be supported

#### TABLE 3 | Predominant habitat types for application in assessment of Descriptor 1.


*<sup>1</sup>EUNIS 200611 version used.*

*Outline depth ranges are given for Atlantic waters for the shallow, shelf, bathyal, and abyssal zones. The precise depth ranges vary between subregions and also in the Baltic, Mediterranean and Black Sea Regions.*

within that habitat. Other community properties such as biomass and abundance are more dependent on ecological interactions such as predator-prey links and recruitment (Gray and Elliott, 2009). This biodiversity-stability relation is complex as it firstly requires a clear definition of what is meant by ecosystem temporal (dynamic) stability and/or the ability to withstand change through resistance and resilience (see McCann, 2000; Tett et al., 2013). Secondly, it requires understanding how biological diversity will enhance ecosystem stability (McCann, 2000; Hooper et al., 2005; Strong et al., 2015). There is a wealth of theoretical and empirical data to support the contention that biodiversity (numbers of distinct species, but also functional diversity) enhances both ecosystem productivity and its resistance to perturbation (e.g., Isbell et al., 2015a,b; Wang and Loreau, 2016). Habitats and species diversity are intrinsically intertwined, and baseline diversity is highly variable. For example, species diversity in seagrass meadows is greater than in adjacent non-vegetated areas (Hemminga and Duarte, 2000), but the lack of seagrass diversity makes these habitats more vulnerable to specific perturbations such as the Wasting disease and storms (Orth et al., 2006). However, this is not always the case as some lower diversity ecosystems, such as estuaries, have a high resilience conferred by the high tolerances and adaptability of the component species, a feature termed environmental homeostasis (Elliott and Quintino, 2007).

While structural taxonomic biodiversity may enhance ecosystem stability, it is not the structural biodiversity as such that causes stability, but the individual species and their role in the ecosystem. In order to understand which species or species groups are the major players within marine ecosystems and how they relate to the functioning of the ecosystem, the understanding of biodiversity would have less emphasis on recording all the taxa, but rather on including the main species within the different functional or feeding groups. This implies a redundancy in the ecosystem, the so-called "rivet hypothesis" (Gray and Elliott, 2009). This also emphasizes the need for a functional view of biodiversity.

#### Functional Ecosystem Biodiversity

By interpreting biodiversity from an ecosystem (top-down) entry point, the focus shifts from structural to functional aspects. In order to construct a simple-to-use view, it is necessary to distinguish between the terms functions and processes (**Figure 3**; rectangular and rounded boxes, respectively) of which there are three main categories of ecosystem functions: (i) Primary production; (ii) Secondary production (spanning from the herbivorous primary consumers to the top predators), and (iii) Nutrient cycling. Each of these major functions are carried out through many inter-linked processes, such as photosynthesis, particle flux (sedimentation, mixing, and resuspension) and consumption/respiration. Export of energy from the marine system to humans and birds through selective biomass extraction also is considered a process as is the re-introduction of nutrients through effluents/run-off and guano.

Documenting the biodiversity status of these three major ecosystem functions/processes, through which they are carried out, requires measurable parameters and indicators (diamond-shaped boxes in **Figure 3**). Most of the indicators currently, or potentially, used in environmental assessment are regarded as surrogates (proxies) of the three main ecosystem functions (see Uusitalo et al., 2016), but the extent to which these reflect the processes is variable, and often just reflect structural elements of the ecosystem. Measuring the abundance and/or biomass of microalgae, the content or concentration of chlorophyll or various proxies such as fluorescence is commonly used to represent the amount of primary producers in the system (Steele, 1962), even if these indicators do not always directly measure photosynthesis. Similarly, for nutrient cycling, appropriate indicators may include the abundance or biomass of microbes or the conservative or otherwise behavior of the different nutrient forms, but this may not give sufficient knowledge of microbial activity (Caruso et al., 2015, 2016). Secondary production, on the other hand, is more tangible, and there exist many indicators that are proxies for quantifying the distribution, population dynamics, abundance, and condition of the various categories of organisms, both in terms of functional traits and population and taxonomic composition (Diaz et al., 2004; Rice et al., 2012). Measuring the processes directly is somewhat more challenging because it often involves experimental approaches (for example respiration measurements), or long-term passive sampling (for example sediment traps) or repeated time-series of population dynamics, Allen-curves and biomass changes to allow production and productivity to be estimated (e.g., Crisp, 1984; Gray and Elliott, 2009), and these can be particularly time-consuming, expensive and not least of all, highly variable from daily, seasonal to annual scales (Bolam, 2014; Maire et al., 2015).

A unified approach to a biodiversity assessment with a functional ecosystem focus would therefore start by identifying indicators for the three main functions. Most assessment programmes will not include these functions, but their existence should at least be acknowledged. From there, the key processes and taxa within each of the major functions will be identified, first in general terms, and then in detail, specific to the assessment area in question. Furthermore, it is argued that there is an increasing emphasis in marine management, from the structural ecological approach in the EU Water Framework and Habitats Directives, to the more functional approach in the MSFD (Borja et al., 2010; Hering et al., 2010).

#### Food-Web Biodiversity

The food-web functional view (**Figure 4**) employs the three main ecosystem functions (primary production, secondary production and nutrient cycling) thus encompassing a range of processes (see Rombouts et al., 2013; Piroddi et al., 2015). The three ecosystem functions are carried out by various combinations of the structural components of biodiversity. Primary producers in the form of microorganisms, micro- and macroalgae as well as macrophytes (e.g., seagrasses), and including both photo- and chemosynthesis, exist in both the pelagic and benthic realms. Through the microbial loop and remineralization, microbes are responsible for the key function of nutrient cycling and make carbon available to the system (Azam et al., 1983; Fenchel, 2008). The primary herbivorous grazers such as copepods form the link between primary production and the rest of the food-web, although these also are transported out of the strictly marine

system through harvesting by seabirds and humans, as a source of omega-3 oil.

Thus, functional indicators of nutrient cycling can operate on microbes, primary production and secondary production to zooplankton, benthos and progressively higher-order predators. The processes typically are explored using more field-experimental, research-orientated indicators although the parameters or organisms to be measured within the three ecosystem functions depends on the biodiversity characteristics of the assessment area and the management questions being addressed.

In essence, a generalized food-web assessment requires indicators to cover all the major energy flow pathways throughout the system. Indicator selection would conceivably start at the producer level, such as abundance and biomass of phytoplankton and benthic algae, and also the basal zooplankton consumers. Indicators for motile components within the pelagic habitat would cover smaller components to top predators, assessed in categories appropriate to the survey area, but essentially covering, for example: (i) krill, gelatinous plankton, and juvenile fish, (ii) squid and small pelagic fish, (iii) large pelagic-feeding fish, reptiles, and mammals such as seals and finally (iv) large benthic feeding fish and mammals such as walrus and seals. The benthic secondary producing component can be seen in terms of functional groups, from herbivores (such as grazers), carnivores which actively seek prey and scavengers which consume both living and dead remains, to surface deposit feeders which consume material deposited from the planktonic realm, and filter-feeders that operate at the sediment-water interface, feeding on both settling particles as well as resuspended matter, the latter produced either through biological pumps or strong bottom currents.

# IMPLICATIONS FOR BIODIVERSITY ASSESSMENTS

Different management questions require different starting-points for selection of measurement parameters and indicators for biodiversity assessments (**Table 4**).

#### Structural Biodiversity Assessment

The structural view on biodiversity is typically used when nature conservation is the primary focus in preserving all (or at least those designated as being important) biotic components of a given ecosystem together with its characteristic abiotic features. For example, the EC Habitats Directive requires assessing the biodiversity status, especially for the conservation features for which an area was designated, by using the appropriate taxonomic and habitat quality indicators. This either ignores the functional relationships within the ecosystem or makes the assumption that the structural elements are proxies for

functioning. This can have implications for the management of such conservation areas since it may require manipulating the habitats and living conditions of certain species or communities when the assessment reveals a less favorable biodiversity status. In this case, ecoengineering may be required both to recreate and restore suitable eco-hydrological functioning (Type A ecoengineering) or to use the restocking or replanting to recreate populations (Type B ecoengineering) (Elliott et al., 2016). As an example, reef restoration is a measure to re-establish reef systems in places where these might have been damaged or lost. This requires the current habitat to be altered (e.g., from soft bottom to hard bottom) so it can support and promote the establishment of a new reef community. This structural change will be reflected in later biodiversity assessments and possibly document the increased biodiversity status. However, if the focus is on a structural view of biodiversity, it might not result in successful functioning and so this kind of biodiversity assessment will not be a holistic one. Hence, the context-driven approach maximizes taxonomical biodiversity but not necessarily ecosystem functioning. Although it can be assumed that biodiversity and ecosystem functioning relationships (BEF) will ensure that higher taxonomical biodiversity also produces higher ecosystem stability (in terms of resistance and resilience), there is insufficient evidence to support this assumption (Cardinale et al., 2012; Strong et al., 2015).

#### Ecosystem Assessments

Most management policies and assessments world-wide aim for some kind of ecosystem approach (Borja et al., 2008). The MSFD advocates an ecosystem-based approach, and many assessment and monitoring schemes exist aiming to integrate ecosystem functions and their values and services (see Atkins et al., 2011; Elliott, 2011, 2013, 2014; Laurila-Pant et al., 2015). However, as with the term biodiversity, the distinctions and uses of the terms Ecosystem Approach and Ecosystem-based management are far from consistent (see review in Borja et al., 2016). An Ecosystembased management strategy acknowledges the complexity of ecosystems and in particular: (i) the need to take into account both the structural aspects (e.g., life-forms present) and the interactions among organisms (especially inter-species relations) within ecological systems, (ii) the essence of connectivity between and within communities, ecosystems, habitats and biotopes, and (iii) that humans are a part of ecosystems thereby integrating human societies within biodiversity management (Elliott, 2011; Kelble et al., 2013; Long et al., 2015). This approach encompasses the structural and functional aspects of an ecosystem (its "emergent properties") as well as, at a smaller scale, the role of given subsystems or components from this ecosystem.

To that end, ecosystem assessments tend to employ at least two views on biodiversity: The structural taxonomic and the


TABLE 4 | Examples of common managerial questions and the appropriate conceptual viewpoints, as starting-points for indicator selection for biodiversity assessments.

functional ecosystem biodiversity. Both are used, or at least require to be used, in one single assessment, but require the need to keep overlaps minimal and to properly interpret the results when measures are to be taken on the basis of the assessment results. This, in turn, requires the need to interpret the resulting ecosystem status in both structural and functional ways so that managers can balance the different needs when planning management measures. As an example, Elliott (2011) proposed an ecosystem health assessment (or monitoring) programme consisting of four elements associated to the typical management cycle: (i) an analysis of main processes and structural characteristics of an ecosystem; (ii) an identification of known or potential stressors; (iii) the development of hypotheses about how those stressors may affect each part of the ecosystem, and (iv) the identification of measures of environmental quality and ecosystem health to test hypotheses. This encompasses and quantifies, from the socio-ecological system, the ecosystem services, and societal benefits approach (Atkins et al., 2011; Laurila-Pant et al., 2015). This approach has led to an extensive series of marine assessment systems which can include both the ecological health and societal well-being, for example the global Ocean Health Index (OHI) (Halpern et al., 2015; Borja et al., 2016).

In general, starting from the conceptual view of functional biodiversity, the clear distinction between ecosystem function and process (e.g., as proposed above) must be retained throughout the assessment and its interpretation when the terms are used to derive management actions from the indicators used to assess functions and processes. However, there is a notable lack of agreement throughout the literature regarding the terms "function" and "processes" when applied to ecosystems and their assessment; indeed the terms may be synonymous in that by definition a function is a rate process. In our functional ecosystem model, the three ecosystem functions (primary production, secondary production and nutrient cycling) together comprise holistic ecosystem functioning. These ecosystem functions are the sum of the physical, chemical and biological processes that transform and translocate energy and materials in ecosystems (Naeem, 1998; Paterson et al., 2012; Snelgrove et al., 2014; Borja et al., 2016).

Functions, and thus inherently also the processes by which they are carried out, are central to the "ecosystem services" which the marine environment provides for its own sustainability and human benefits. As indicated above (and also see Turner and Schaafsma, 2015), successful structure and functioning of the physico-chemical and ecological systems can produce intermediate and final ecosystem services: (i) provisioning, (ii) regulating, (iii) supporting (or habitat), and (iv) culture and heritage (Jax, 2005; De Groot et al., 2010). Complementary human assets are then required to extract societal benefits from such services (Atkins et al., 2014). Strong et al. (2015) listed five categories of "ecosystem functions," which also refer to processes: (i) production of biomass, (ii) (non-living) organic matter transformation, (iii) ecosystem metabolism, (iv) nutrient cycling, and (v) physical environment modification, for which they analyzed biodiversity.

Thus, there are many ways to refer to the functions and processes occurring within marine ecosystems, and in turn the services and societal benefits which they provide. Focusing our conceptual understanding of biodiversity from a functional ecosystem viewpoint on three main functions, driven by a range of processes, gives clarity about the logical basis for both selection of assessment parameters and interpretation of results. We recognize that the functions themselves are assessed by measuring some proxy of the processes, such as various qualities and attributes of the organisms which carry out those processes. With this understanding, we can select the indicators which represent the sections of the system which best address the questions asked, and at the same time retain an awareness of the information gaps which require us to extrapolate information from other measurements and to make appropriate inferences for ecosystem-scale assessments.

#### Food-Web Assessments

The conceptual view outlined in **Figure 4** provides the basis of a holistic food-web assessment. Typically, such assessments operate with a restricted set of parameters relating to predatorprey interactions, with a focus on abundance and population structure of commercially harvested species, and often also their main prey items. For example, the MSFD Descriptor 4 (trophic relations) adopted a pragmatic conceptual simplification in approach (Rogers et al., 2010; Rombouts et al., 2013). Two key attributes for food-webs were specified within the MSFD as: (i) energy flow in food-webs, i.e., from primary to secondary production, and (ii) structure of food-webs i.e., size and abundance of predators/prey (Rogers et al., 2010). Rombouts et al. (2013) argued that three main properties of foodwebs can be considered within the MSFD context: Structure, functioning and dynamics, with emphasis on the latter two and "the general principles that relate these three properties." The MSFD Descriptor 4 indicators for food-webs, such as the reproductive success of dominant piscivorous seabirds, are very much process-based and designed to capture responses to the multiple anthropogenic pressures that can affect food-webs, the main one being selective extraction of biomass (e.g., fishing).

The structuring influence of large predators on ecosystem stability, and the potential for human impacts thereon, can be illustrated, for example, by overfishing of the Atlantic cod, Gadus morhua which caused a notable increase in alpha and beta diversity of the remaining fish communities. These became more variable during periods where the cod no longer dominated the system (Ellingsen et al., 2015). This is an example of the difficulties a biodiversity concept will face when it becomes more complex. The overall assessment result will no longer be able to reflect both the structural and functional changes individually. The representability of an assessment of food-web status thus depends much on the indicators chosen and whether they are capable of capturing the "health" of the ecosystem, in terms of deviation from reference or target conditions (assuming these are in fact known and/or defined). Tett et al. (2013) emphasizes that the concept of ecosystem health is integral to management questions based on the overall assessment which thus encompasses an assessment of both biological diversity and the delivery of ecosystem services and societal benefits.

Where the aim of assessment is toward sustainable management, such as in the MSFD, or marine conservation, the selected food-web measurement parameters and indicators must focus on detecting the impacts of anthropogenic pressures (Coll et al., 2016). However, for a programme to understand the overall predator-prey structure in a system, all levels of interactions should be included into the underlying view on the biodiversity as the basis of the assessment. As with all aspects of biodiversity, changes in abiotic conditions such as climatic ones will also impact food-webs and create moving baselines against which changes in biodiversity are judged (Elliott et al., 2015). They are drivers for changes in species distributions, recruitment success and competition and so food-web indicators should operate at the species level (e.g., population indicators) but also at the ecosystem level when considering overall energy flow through the system.

The main practical challenge in finding fit-for-purpose foodweb indicators is the variability in pressure-impact relationships on their structure and functioning. An example on how to reach a more simplified generalization is the "fishing down the foodweb" rule (Pauly et al., 1998). It proposes that fishing a foodweb would first target larger and higher trophic level carnivorous fish and then progressively those at lower trophic levels, theoretically shortening food-webs. Thus, the mean trophic levels of consumers would be lower in an overfished food web, relative to an undisturbed one. An indicator reflecting the mean trophic level will adequately capture this aspect but other indicators will be needed when the aim of the assessment is not only to maintain sustainable fisheries, but also to preserve structural biodiversity. The corresponding conceptual view of biodiversity should be the basis of such preservation aims by including the relevant structural elements into the food-web but also assuming that such structural indicators are indeed proxies for successful functioning.

#### CONCLUSIONS

This review of the abstract concept of marine biodiversity is based on three conceptual views of the upper-level aspects of biodiversity (structural taxonomic, functional ecosystembased, and food-web biodiversity). They form the basis for constructing different biodiversity assessment types, depending on the context in which the assessment is used. The conceptual views serve as simplified common denominators from which can be developed a dialogue between both scientists and managers, balancing the needs for a sound scientific foundation and the pragmatic requirements for practical management of marine systems. The examples presented in this conceptual framework and the consequences for the assessment of biodiversity lead to three conclusions which improve the applicability and value of biodiversity status assessments and management.

Firstly, marine ecosystems are considered from different perspectives given the absence of a common and single understanding of what is marine biodiversity. The way in which we view this abstract biodiversity depends on various variables where this complexity can be simplified when focusing on the structural and functional elements of biodiversity that are important for the management question to be answered. This is best done using a carefully defined set of biodiversity elements to be assessed, knowing which elements to ignore and why and what consequences this has for the subsequent biodiversity assessment. This approach will allow for a context-driven assessment, where the meaning of the assessment result is pre-defined and derived from our applied understanding of biodiversity. The result does not need a special interpretation and is tied directly to the question we want to answer.

Secondly, we use the perspectives to construct a "management-friendly" assessment: A biodiversity status of "good" or "not good" needs a context for interpretation (see Mee et al., 2008). This context is given by the specific conceptual view. Together, this will provide information on what is the biodiversity status and how it can be improved by managing identified problems. Only an assessment that can explain the resulting biodiversity status and give insights into how the situation can be changed following management measures is useful for management. It is the conceptual view that leads to insights and measures to be applied by management thus emphasizing the need for knowledge on the biodiversity status and where and how it requires to be improved if it is considered to be degraded.

Thirdly, be aware of the limits and degree of quantification of the assessment: Since we know what has been omitted from our conceptual view, we also know what management cannot expect to achieve. Similarly, the success of management measures and their efficacy can only be determined by quantifying the conceptual approach. A primarily structural taxonomic view of biodiversity will not lead to an assessment that points to measures improving ecosystem functions. However, the conceptual view

#### REFERENCES

Atkins, J. P., Burdon, D., Elliott, M., and Gregory, A. J. (2011). Management of the marine environment: integrating ecosystem services and societal benefits with the DPSIR framework in a systems approach. Mar. Pollut. Bull. 62, 215–226. doi: 10.1016/j.marpolbul.2010.12.012

chosen allows us to determine the limits of our understanding of biodiversity and thus the possibilities of the management measures even before the assessment has been made. If the limits are clear and can be communicated, expectations are realistic whereas unrealistic expectations may arise from an incomplete conceptual approach or false assumptions of the links between structure and functioning.

A given conceptual view can always be expanded by including more elements and shifting the focus closer to the question asked. As one example, we can include activities which create the major pathways of human pressures, the state changes they involve in the marine system and the impacts this has on society, its welfare and well-being (Scharin et al., 2016; Smith et al., 2016). Such modifications will expand our understanding of biodiversity using the influential parameters relevant for the specific purpose of the individual biodiversity assessment.

# AUTHOR CONTRIBUTIONS

The basis for this manuscript was conceived during a pivotal discussion between SC, JA, TB, and P. Herman, Bilbao, November 2013, the first three of which produced the initial draft of the manuscript. The remaining authors each have contributed within various areas of expertise: HB, AB, JC, and HH (indicators and environmental assessments), ME (general concepts and management), NN (food webs) and PR (ecosystem functions and processes).

#### ACKNOWLEDGMENTS

This manuscript is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), www.devotes-project.eu. We acknowledge the life's work of the late Prof. Carlo Heip, for leading initiatives such as BIOMARE and the MarBEF network, which have sown the seeds for this present work. Further, we thank Peter Hermann, Anne Chenuil, and Chris Lynam. We also acknowledge the other members of the MSFD TG1 group, particularly David Connor and Per Nilsson, for together developing the criteria and indicators for biodiversity, adopted by the MSFD. The lead author sincerely thanks colleagues, collaborators and clients for all the countless discussions, understandings and misunderstandings, which have given rise to this manuscript, as well as Tom Pearson for past mentoring in benthic indicators and functional traits. Finally, thanks to two reviewers, particularly Christos Arvanitides, whose constructive criticism much improved this manuscript.

Atkins, J. P., Burdon, D., Elliott, M., Schaafsma, M., and Turner, R. K. (2014). Coastal and marine ecosystem services. Environ. Sci. 23, 26–30.

Azam, F., Fenchel, T., Field, J. G., Gray, J. S., Meyerreil, L. A., and Thingstad, F. (1983). The ecological role of water-column microbes in the sea. Mar. Ecol. Prog. Ser. 10, 257–263. doi: 10.3354/meps 010257


seas: present concerns and future needs. Mar. Pollut. Bull. 64, 2630–2638. doi: 10.1016/j.marpolbul.2012.10.010


assessing the quality of functional spaces. Global Ecol. Biogeogr. 24, 728–740. doi: 10.1111/geb.12299


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor MM declared a collaboration with the authors and states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Cochrane, Andersen, Berg, Blanchet, Borja, Carstensen, Elliott, Hummel, Niquil and Renaud. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

#### *Angel Borja1 \*, Theo C. Prins 2, Nomiki Simboura3, Jesper H. Andersen4, Torsten Berg5, Joao-Carlos Marques 6, Joao M. Neto6, Nadia Papadopoulou7, Johnny Reker 8, Heliana Teixeira9 and Laura Uusitalo10*

*<sup>1</sup> AZTI-Tecnalia, Marine Research Division, Pasaia, Spain*

*<sup>2</sup> Deltares, Delft, Netherlands*


*<sup>8</sup> European Environment Agency, Copenhagen, Denmark*

*<sup>9</sup> European Commission, Joint Research Centre, Institute for Environment and Sustainability, Ispra, Italy*

*<sup>10</sup> Finnish Environment Institute (SYKE), Helsinki, Finland*

#### *Edited by:*

*Stelios Katsanevakis, University of the Aegean, Greece*

#### *Reviewed by:*

*Angel Pérez-Ruzafa, Universidad de Murcia, Spain Henn Ojaveer, University of Tartu, Estonia Sabine Cochrane, Akvaplan-niva, Norway*

#### *\*Correspondence:*

*Angel Borja, AZTI-Tecnalia, Marine Research Division, Herrera Kaia, Portualdea S/N, 20110 Pasaia, Spain e-mail: aborja@azti.es*

Assessing the environmental status of marine ecosystems is useful when communicating key messages to policymakers or the society, reducing the complex information of the multiple ecosystem and biodiversity components and their important spatial and temporal variability into manageable units. Taking into account the ecosystem components to be addressed (e.g., biological, chemical, physical), the numerous biodiversity elements to be assessed (e.g., from microbes to sea mammals), the different indicators needed to be studied (e.g., in Europe, 56 indicators of status have been selected), and the different assessment scales to be undertaken (e.g., from local to regional sea scale), some criteria to define spatial scales and some guidance on aggregating and integrating information is needed. We have reviewed, from ecological and management perspectives, the approaches for aggregating and integrating currently available for marine status assessment in Europe and other regions of the world. Advantages and shortcomings of the different alternatives are highlighted. We provide some guidance on the steps toward defining rules for aggregation and integration of information at multiple levels of ecosystem organization, providing recommendations on when using specific rules in the assessment. A main conclusion is that any integration principle used should be ecologically-relevant, transparent and well documented, in order to make it comparable across different geographic regions.

#### **Keywords: ecosystems, marine, indicators, Marine Strategy Framework Directive, descriptors, criteria, assessment, integration**

#### **INTRODUCTION**

The requirement to assess the environmental status of marine waters is growing across continents (Borja et al., 2008). It is also one of the challenging tasks to be accomplished in Europe, within the Marine Strategy Framework Directive (MSFD) (European Commission, 2008). The different legislative mandates to asses status coming from the MSFD, Water Framework Directive (WFD) (2000/60/EC) and Habitats Directive (92/43/EEC) and other international initiatives have produced numerous methodologies that can be applied to different ecosystem components, such as various taxonomic or functional groups, habitats, traits, physical features, or to the whole ecosystem (Birk et al., 2012; Halpern et al., 2012). Despite this wealth of methods, determining environmental status and assessing marine ecosystems health in an integrative way is still one of the grand challenges in marine ecosystems ecology research and management (Borja, 2014).

Different attempts to understand, define and assess ecosystem health have been made in recent years (Costanza and Mageau, 1999; Ulanowicz, 2000; Mee et al., 2008; Ojaveer and Eero, 2011; Borja et al., 2013; Tett et al., 2013). The concept of "good environmental status" (GEnS) integrates physical, chemical and biological aspects, together with the services provided by ecosystems, including a sustainable use of the marine resources by society (Borja, 2014). However, synthesizing these aspects into a single value will never appropriately reflect all aspects considered to derive the value (Purvis and Hector, 2000; Derous et al., 2007). Still, this step is useful when communicating key messages to policymakers or the society, reducing the complex information of the multiple ecosystem components and their important spatial and temporal variability into manageable units, which can be used in ecosystem management. Following the recommendation from Mee et al. (2008), we use the GEnS acronym because the meaning of "environmental," within the MSFD, and "ecological" (good ecological status), within the WFD, is different (see Borja et al., 2010, for differences between both concepts), implying a different emphasis between these two major pieces of legislation.

In the case of the MSFD, an appropriate integration process might be even more complex, since the assessment of the status is based upon 11 qualitative descriptors (i.e., D1: biological diversity; D2: non-indigenous species; D3: exploited fish and shellfish; D4: food webs; D5: human-induced eutrophication; D6: seafloor integrity; D7: hydrographical condition; D8: contaminants; D9: contaminants in fish and seafood; D10: litter; and D11: energy and noise), which are further divided into 29 criteria and 56 indicators of health (European Commission, 2010). An overview of MSFD descriptors, criteria and indicators is shown in **Table 1**.

The aim of this work is to present an overview of the different methods currently available to synthesize the ecosystem complexity, by aggregating and integrating information when assessing the status, focusing mostly on the descriptors related to biodiversity, namely D1, D2, D4, D6 (Cardoso et al., 2010; Prins et al., 2014). This overview would assist managers, through the guidelines provided, in taking decisions for a better management of the marine ecosystems.

#### **ECOSYSTEM COMPONENTS COMBINATION REQUIREMENTS IN ASSESSING THE STATUS**

There are different methods that can be applied to combine indicators and criteria within descriptors and across descriptors to eventually result in an assessment of GEnS for a specific geographic area. This combination both involves aggregation and integration. The term aggregation is here used for the combination of comparable elements across temporal and spatial scales, indicators and criteria, within a descriptor. The term integration is used for the combination of different elements (e.g., across descriptors). Both combination methods (aggregation and integration) may involve numeric calculations.

In Europe, the MSFD defines environmental status as "the overall state of the environment in marine waters, taking into account the structure, function, and processes of the constituent marine ecosystems together with natural physiographic, geographic, biological, geological and climatic factors, as well as physical, acoustic and chemical conditions, including those resulting from human activities inside or outside the area concerned."

Taking this definition into account, Borja et al. (2013) have proposed an operational definition: "GEnS is achieved when physicochemical (including contaminants, litter and noise) and hydrographical conditions are maintained at a level where the structuring components of the ecosystem are present and functioning, enabling the system to be resistant (ability to withstand stress) and resilient (ability to recover after a stressor) to harmful effects of human pressures/activities/impacts, where they maintain and provide the ecosystem services that deliver societal benefits in a sustainable way (i.e., that pressures associated with uses cumulatively do not hinder the ecosystem components in order to retain their natural diversity, productivity and dynamic ecological processes, and where recovery is rapid and sustained if a use ceases)."

This latter definition includes all MSFD descriptors. Hence, to assess whether or not GEnS has been achieved, some aggregation within and integration across the 11 descriptors is required to move from the evaluation at the level of indicators (the 56 indicators and 29 criteria described in the Commission Decision (European Commission, 2010, see also **Table 1**) to a global assessment of status, as mentioned also in Cardoso et al. (2010). The problem is how to deal with the complex task of combining a high number of indicators and descriptors. To develop a common understanding on this, it is important that Member States are transparent on (i) the process of selecting the indicators to be monitored; (ii) the approaches and combination methods they have used; and (iii) the uncertainties in their indicators and methods.

#### **GENERAL PRINCIPLES FOR COMBINATION**

Based on a literature review, we identified a number of different approaches for combining a number of variables (which could be metrics, indicators, or criteria) into an overall assessment. Some of them have been used within the WFD, others within the RSCs and some others in the MSFD. An overview of the methods is given in **Table 2**.

When considering the aggregation of indicators, an important factor to be taken into account is the reliability of the individual indicators to be aggregated. With each indicator, it is always possible to make a type I error, i.e., to get a non-GEnS result when the system in fact is in GEnS. The probability of this false positive (FP) signal varies (i) between indicators (Murtaugh, 1996), depending on the natural variability; (ii) with the amount of data used to define the indicator value; and (iii) with the target level compared to the situation in the nature. The risk of getting a FP from each of the individual indicators should affect the aggregation rule as well: if the risk of a FP is a uniform 5% per indicator, on average 1 out of 20 indicators is expected to give a FP; a problem if all indicators should in fact show GEnS. In order to come up with an aggregated assessment in which the risk level is within reasonable bounds, this aspect cannot be overlooked.

#### **ONE-OUT, ALL-OUT (OOAO)**

The OOAO approach is used in the WFD to integrate within and across Biological Quality Elements (BQEs) (CIS, 2003), in order to reach the ecological status of a water body. This approach follows the general concept that the ecological status assigned to a water body depends on the BQE with the lowest status, and consequently, the OOAO approach results in a "worst case."

A prerequisite for the aggregation of various indicators is that they are sensitive to the same pressure (Caroni et al., 2013). In such a case, different aggregation methods can be used to combine parameters (medians, means, etc.). Caroni et al. (2013) recommend an OOAO approach when the combination involves parameters/indicators that are sensitive to different pressures. The application of averaging rules may lead to biased results in those cases. The WFD Classification Guidance (CIS, 2003) also advises


**Table 1 | Descriptors, criteria and indicators selected by the European Commission (2010), for ecosystem-based assessment and management of European seas, within the Marine Strategy Framework Directive.**

*(Continued)*

#### **Table 1 | Continued**


*(Continued)*

**Table 1 | Continued**



**Table 2 | Approaches for combining different metrics, indicators or criteria to assess the status, including the advantages and disadvantages of each approach, as considered by the authors.**

*GEnS, Good environmental status.*

to use OOAO when combining parameters/indicators that are sensitive to different pressures.

Borja et al. (2009a) discussed the challenge of assessing ecological integrity in marine waters, and suggest that simple approaches, such as the "OOAO" principle of the WFD, may be a useful starting point, but eventually should be avoided. The ecological integrity of an aquatic system should be evaluated using all information available, including as many biological ecosystem elements as is reasonable, and using an ecosystem-based assessment approach. The OOAO rule can be considered a rigorous approach to the precautionary rule, in an ideal world where the status based on each BQE can be measured without error. It results in very conservative assessments (Ojaveer and Eero, 2011). In practice, the inevitable uncertainty associated with monitoring and assessment for each metric and BQE leads to problems of probable underestimation of the true overall status. The OOAO principle has therefore been criticized as it increases the probability of committing a false positive error, leading to an erroneous downgrading of the status of a water body as it has been observed especially within the WFD (Borja and Rodríguez, 2010; Ojaveer and Eero, 2011; Borja et al., 2013; Caroni et al., 2013). In the case of the MSFD, with such large number of descriptors, criteria and indicators, the probability of not achieving good status becomes very high and, probably, unmanageable in practical terms (Borja et al., 2013).

Alternative methods for integrating multiple BQEs in the WFD are currently being considered (Caroni et al., 2013).

#### **AVERAGING APPROACH**

The averaging approach is the most commonly used method to aggregate indicators (Shin et al., 2012) and consists of simple calculations, using methods such as arithmetic average, hierarchical average, weighted average, median, sum, product or combinations of those rules, to come up with an overall assessment value.

Ojaveer and Eero (2011) showed that in cases where a large number of indicators is available, the choice of e.g., either medians or averages in aggregating indicators did not substantially influence the assessment results. However, this might not necessarily be the case when only a few indicators are available. In such a situation, the result will depend to a larger degree on the distribution of the values involved. A skewed distribution reflecting some major factors and a few ones with very different values will result in very different assessment results for the median compared to assessments based on means. Apart from the mathematical applicability of either method based on the underlying data (e.g., homoscedasticity), the choice of the actual averaging method may be driven by policy decisions focusing on either central trends without much attention to extreme values (median) or focusing on weighting the individual values by their magnitude (arithmetic mean).

The way the indicators are hierarchically arranged influences the assessment results as well, but Ojaveer and Eero (2011) found that these effects were considerably less important than the effects of applying different aggregation rules.

Differential weighting applied to the various indicators can be used when calculating means or medians. An adequate basis for assigning weights is not always available and in such cases an equal weight is recommended by Ojaveer and Eero (2011). Assigning weights often involves expert judgment, and Aubry and Elliott (2006) point out that in some cases, expert opinions on weights can show important divergence.

#### **CONDITIONAL RULES**

Conditional rules (a specific proportion of the variables have to achieve good status) are an approach where indicators can be combined in different ways for an overall assessment, depending on certain criteria. This provides an opportunity to use expert judgment when combining indicators, in a transparent way. An example of this approach is the application of a conditional rule of at least two out of three indicators (one biotic index and two structural or diversity indices) should pass the threshold in order to achieve GEnS for benthic community condition under D6 in Hellenic waters (Simboura et al., 2012). Tueros et al. (2009) present another example of the conditional rule in which when integrating water and sediment variables into an overall assessment of the chemical status and only one sediment or water variable does not meet the objective, while the rest of the variables meet, the final chemical status achieves the objective. This work was also mentioned under the "two out, all out" approach considering the case when two variables do not meet the objective and the final status fails.

Breen et al. (2012) used several risk criteria rules and worstcase or integrated approaches when combining evidence before a final assessment. Following Cardoso et al. (2010) the integrated approach was applied to Biodiversity, Non-indigenous species, Eutrophication and Seafloor Integrity descriptors, while all other descriptors used a worst case approach following the OOAO principle whereby if one set of evidence suggested that the risk was "high" then "high" was automatically assessed for the entire descriptor.

#### **SCORING OR RATING**

In this method different scores are assigned to a status level (for example, ranging from 1 to 5), for a number of different elements. The scores are summed up to derive a total score which is then rated according to the number of elements taken into account. Different weights can be assigned to the various elements. This method was proposed by Borja et al. (2004) to calculate an integrative index of quality and is the basis of many multimetric indices used within the WFD and the MSFD combining different parameters or metrics using the weighted scoring or rating rule into one integrative multimetric index (Birk et al., 2012). It must be recognized here that this approach implies the score values being on a cardinal scale and acting as weighting factors. Otherwise, using an ordinal scale for the scores, summing up the individual elements is mathematically not defined.

Another example is the method developed by Borja et al. (2010, 2011b) for a cross-descriptor integration, combining the 11 descriptors of MSFD based on the WFD, HELCOM (2009a,b, 2010) and OSPAR (2010, 2012) experiences. An Ecological Quality Ratio (EQR) was calculated for each indicator of the various MSFD descriptors, with the EQR for the whole descriptor being the average value of the EQR of the indicators. Then, by multiplying the EQR with the percent weight assigned to each descriptor (and summing up to 100), an overall environmental status value was derived.

#### **MULTIMETRIC INDICES TO COMBINE INDICATORS**

Within the WFD there are many examples of multimetric indices developed for different biological elements, driven by the need to fulfill the detailed requirements of the WFD (see Birk et al., 2012 for a complete synthesis).

In addition, within the MSFD, the European Commission established a number of Task Groups consisting of technical experts to help inform the discussions on how to reach a common understanding of the 11 descriptors. Hence, Task Group 6 report on seafloor integrity (Rice et al., 2010) recommends the use of multimetric indices or multivariate techniques for integrating indicators of species composition attributes of this descriptor, such as diversity, distinctness, complementarity/(dis)similarity, or species-area relationships.

There are various other examples of multi-metric indices used to assess the status of the macrobenthos (see Borja et al., 2011a for an overview). Multimetric methods to combine multiple parameters in one assessment may result in more robust indicators, compared to indicators based on single parameters. However, scaling of a multimetric index may be less straightforward, and ideally the various parameters should not be inter-correlated (e.g., the discussion on the TRIX index in Primpas and Karydis, 2011).

#### **MULTIDIMENSIONAL APPROACHES**

Multivariate methods, such as Discriminant Analysis or Factor Analysis combine parameters in a multi-dimensional space. For assessment purposes, areas need to be classified into groups of GEnS and non-GEnS.

Multivariate methods have the advantage of being more robust and less sensitive to correlation between indicators. However, interpretation is less intuitive than other methods, as information on individual indicators in each ecosystem is lost (Shin et al., 2012) and links to management options are less obvious.

#### **DECISION TREE**

Decision trees provide the opportunity to apply different, specific, rules to combine individual assessments into an overall assessment. A decision tree allows implementing individual rules at each of its nodes and thus incorporates arbitrary decisions at each step within the decision tree. The decision rules can be quantitative or qualitative as well as based on expert judgment. This gives room for a high degree of flexibility in reaching the final assessment and can thus be used where the other principles fail to represent the intricate interactions, feedback loops and dependencies involved in ecosystem functioning between the ecosystem components.

A simple version of a decision tree involves only having a few conditional rules where a specific proportion or certain individually specified indicators have to achieve good status in order to achieve GEnS. Borja et al. (2013) implicitly propose using this kind of decision tree when they take the view that for biodiversity (D1) to be in good status, all other descriptors must be in good status and if one of the pressure descriptors fails, then D1 also fails.

Borja et al. (2004, 2009b) describe a methodology that integrates several biological elements (phytoplankton, benthos, algae, phanerogams, and fishes), together with physicochemical elements (including pollutants) into a quality assessment. The proposed methodologies accommodate both WFD and the MSFD. They suggest that the decision tree should give more weight to individual elements taking into account the spatial and temporal variability and the availability of accurate methodologies for some of them (i.e., benthos) and to individual assessment methods which have been used broadly by authors other than the proposers of the method, tested for several different human pressures, and/or intercalibrated with other methods.

#### **PROBABILISTIC APPROACH**

Each of the indicator results are uncertain, due to several factors e.g., natural variation in the sampling sites, random variation in the samples, insufficient scientific understanding about what should be the reference value for good status, etc. Some indicators are bound to include more uncertainty that others, due to differences in the amount of data used, the extent of scientific understanding regarding the issue, and the amplitude of natural variation. If these uncertainties can be approximated, this gives rise to the possibility of taking this information into account when integrating the indicators. The more uncertain indicators will get less weight in the integrated assessment, while the more certain ones will be more reliable and hence get more weight. The calculus of the integrated assessment can be based on Bayesian statistics, giving transparent and coherent rules by which the final score is calculated.

This approach can be combined to one or several of the abovementioned approaches: for example, conditional rules can be set in addition to the probabilistic integration rule to include expert judgment; and the principles outlined in the decision tree approach can be applied as well.

Barton et al. (2012) demonstrate how to use the probabilistic approach in the DPSIR framework in the case of eutrophication management. There are several other examples in the recent literature about how to evaluate various management measures under uncertainty to optimize one target, such as eutrophication (Barton et al., 2008; Lehikoinen et al., 2014) and oil spill severity (Lehikoinen et al., 2013). This approach could be expanded to include several descriptors or indicators.

Probabilistic combination of uncertain indicators would naturally lead to a probability estimate of how likely it is that a marine area is in GEnS; we would, for example, end up with an estimate that the sea area is in GEnS with 70% probability. The managers would then have to decide how much uncertainty they are willing to tolerate; i.e., are they happy if the probability of GEnS is above 50%, or whether they want a higher certainty?

#### **HIGH-LEVEL INTEGRATION**

An example of a high-level integration, where assessments for several ecosystem components are merged into a final assessment, is the HELCOM-HOLAS project (HELCOM, 2010). The report presents an indicator-based assessment tool termed HOLAS ("Holistic Assessment of Ecosystem Health Status"). The indicators used in the thematic assessments for eutrophication (HEAT), hazardous substances (CHASE) and biodiversity (BEAT) were integrated into a Holistic Assessment of "ecosystem health." The HOLAS tool presented assessment results for three groups: biological indicators, hazardous substances indicators and supporting indicators, and then applied the OOAO principle on the assessment results of those three groups for the final assessment (**Figure 1**).

This approach, which includes the selection of an agreed reduced set of indicators and agreed weighting rules, could be considered a pragmatic compromise, reducing the risks associated with OOAO while still giving an overall assessment.

An example of such a high level aggregation is the integrative method of Borja et al. (2010, 2011b), which includes a weighted scoring or rating method proposed for the MSFD in the southern Bay of Biscay. After aggregating the indicators within each descriptor, each descriptor was weighted according to the human pressure supported by the area. Then the value of each descriptor (i.e., an EQR) was multiplied by the weighting and added to obtain a final value between 0 and 1, being 0 the worst environmental status and 1 the best. This high-level integration was done at spatial and temporal scale. Although these authors combine values across descriptors, leading to a single value of environmental status, it could also be reported as "x out of 11 descriptors" having reached GEnS. In both cases, this allows to take management measures on those human activities impacting more in some of the descriptors or indicators not achieving good status, as shown in Borja et al. (2011b).

Halpern et al. (2012) developed another method, based more upon human activities and pressures, which presents a high-level integration at country level, using internationally available

datasets (Ocean Health Index http://www*.*oceanhealthindex*.*org). Similarly, Micheli et al. (2013) looked at cumulative impacts to the marine ecosystems of the Mediterranean and the Black Sea as a whole, while producing impact scores and maps for seven ecoregions and the territorial waters of EU Member states.

A Baltic Sea Health Index (BSHI) will be developed based on: (i) the existing HELCOM toolbox (HEAT, BEAT, CHASE and HOLAS), the MSFD (European Commission, 2008, 2010), and (ii) the Ocean Health Index (Halpern et al., 2012).

Finally, there is a recent high-level integration example in Tett et al. (2013), for the North Sea, which includes five steps in the calculation: (i) identify (spatial extent) of ecosystem; (ii) identify spatial granularity and extent of repetitive temporal variability, and decide how to average or integrate over these; (iii) select state variables; (iv) plot trajectory in state space and calculate Euclidian (scalar) distance from (arbitrary) reference condition; and (v) calculate medium-term variability about trend in state space, and use this variability as proxy for (inverse) resilience.

#### **CONSIDERATIONS AND RECOMMENDATIONS WHEN USING SPECIFIC RULES**

As shown in the previous section, the considerations to be used in combining values and assessing the environmental status are not easily defined. From the lessons learned above, some guidance can be offered:

(1) OOAO is appropriate when:

	- In cases where indicators show a high level of uncertainty, when various indicators are sensitive to the same pressure, etc. In practice, the uncertainty associated with monitoring and assessment for each indicator/descriptor leads to problems of probable underestimation of the true overall class. Hence, if the error associated to the method used to assess the status of each indicator/descriptor is too high the OOAO approach is not advisable.
	- Note: Often, not all indicators are in the same state of development, or are scientifically sound and fully tested. In some cases P-S-I (Pressure-State-Impact) relations are uncertain. Also, sometimes multiple indicators are used to describe state. While not all of those indicators may be equally important or even comparable, this is done to include indicators that are used as supportive indicators,

where P-S-I relations are uncertain. In those cases an aggregation rule such as OOAO should not be applied.

	- The methods to assess the status of the different indicators/descriptors are in different levels of development. In this case, consider giving more weight to those indicator/assessment methods which have been: (i) used broadly by authors other than the proposers of the method; (ii) tested for several different human pressures; and/or (iii) intercalibrated with other methods.
	- It is important to be able to track the different steps involved in the assessment, making the path to the final assessment result transparent.
	- Note: Consider different weights for individual indicators/descriptors taking into account the relationship with the pressures within the assessment (sub)region. E.g., if the area is under high fishing pressure the most affected descriptors will be D1, D3, D4, D6 and D11; in turn, D2, D5, D7, D8, D9 and D10 will be less affected.
	- Consider carefully the uncertainties related to all of the various parts of the problem; be sure not to overestimate the well-known uncertainties (e.g., natural variance and sampling bias) and underestimate the poorly known uncertainties (e.g., insufficient knowledge or competing hypotheses about ecological interactions; combined effects of various pressures that may be strengthen or weaken each other, etc.).
	- Consider using expert knowledge in evaluating the various uncertainties.
	- If using expert judgment to weigh the different indicators in addition to the uncertainty estimate, make sure that the weighing is based on the relative importance of the indicators, not on the perceived uncertainty; otherwise you will end up double counting the effect of uncertainty in the final evaluation.
	- Integrating several indicators of species composition or several indicators of eutrophication or seafloor integrity (e.g., in D1, D5, D6).
	- It is advisable to verify that stakeholders and managers can understand the interpretation of the results, and results must be presented in a clear way.

#### **APPLICATION OF COMBINATION RULES IN ASSESSMENTS**

As shown above, the WFD focuses on the structure of the ecosystem using a limited number of biodiversity components (the BQEs), that are combined through the precautionary OOAO approach (Borja et al., 2010). In contrast, the MSFD can be considered to follow a "holistic functional approach," as it takes into account not only structure (biodiversity components, habitats), but also function (e.g., food webs, seafloor integrity) and processes (e.g., biogeochemical cycles) of the marine ecosystems. The MSFD also uses descriptors that not only relate to biological and physicochemical state indicators but also to pressure indicators (Borja et al., 2010, 2013). The MSFD requires the determination of GEnS on the basis of the qualitative descriptors in Annex I, but does not specifically require one single GEnS assessment, in contrast to the WFD.

There are many methodological challenges and uncertainties involved in establishing a holistic ecosystem assessment, when it is based on the large number of descriptors, associated criteria and indicators defined under the MSFD. The choice of indicator aggregation rules is essential, as the final outcome of the assessment may be very sensitive to those indicator aggregation rules (Ojaveer and Eero, 2011; Borja et al., 2013; Caroni et al., 2013). As shown in the previous section, different methodologies can be applied for aggregating indicators, which vary, amongst others, in the way the outliers influence the aggregate value.

When aggregating indicators most researchers agree that multiple accounting should be avoided. For example, phytoplankton indicators under D1 should be indicative of biodiversity state while under D5 it should be an estimator of the level of eutrophication. Similarly, macroinvertebrates under D1 should represent biodiversity state and under D6 also the state change from pressures on the seafloor. In these cases, although the datasets used could be the same, the main characteristics of the indicators to be used within each descriptor should be different, e.g., the value of macroinvertebrates indicators under D1 (rarity of species, endangered species, engineer species presence, etc.) and the condition of benthic community under D6 (ratio of opportunistic/sensitive, multimetric methods to assess the status, etc.). Of course, for aggregating indicators within the same criterion it is important that all indicators have the same level of maturity and that sufficient data are available.

There are at least four levels of combination required to move from evaluation of the individual metrics or indicators identified by the Task Groups to an assessment of GEnS (Cardoso et al., 2010). As an example, using D6 (Seafloor integrity), **Figure 2** shows: (i) aggregation of metrics/indices within indicators (see names of indicators in **Table 1**); (ii) aggregation of indicators within the criteria of a descriptor (for complex descriptors), e.g., criteria 6.1 (physical damage) and 6.2 (condition of benthic community); (iii) status across all the criteria of a descriptor; and (iv) integration of status across all descriptors.

As one moves up the scale from metric/indicator level to overall GEnS, the diversity of features that have to be combined increases rapidly (**Figure 2**). This poses several challenges arising from the diversity of metrics, scales, performance features (sensitivity, specificity, etc.) and inherent nature (state indicators, pressure indicators, impact indicators) of the metrics that must be integrated.

#### **AGGREGATION OF INDICATORS AND CRITERIA (COMBINATION WITHIN A DESCRIPTOR)**

Cardoso et al. (2010) summarize the methods for an integration within a MSFD descriptor, categorizing them into two wider

categories: (i) integrative assessments combining indicators and/or attributes appropriate to local conditions; and (ii) assessment by worst case (in this context, "worst case" means that GEnS will be set at the environmental status of the indicator and/or attribute assessed at the worst state for the area of concern).

**Table 3** summarizes the approaches to aggregate attributes within each descriptor. In some cases the MSFD Task Groups propose deconstructing the ecosystem into "descriptor indicators" and then recombining them again to give a pass/fail for the GEnS, using (in four cases) the OOAO principle (**Table 3**). Borja et al. (2013) emphasize that such a "deconstructive structural approach" makes large assumptions about the functioning of the system and does not consider the weighting of the different indicators and descriptors. It implies that recombining a set of structural attributes gives an accurate representation of the ecosystem functioning.

An example of this accurate representation is shown by Tett et al. (2013), who assess the ecosystem health of the North Sea, using different attributes and components of the ecosystem. These components include structure or organization, vigor, resilience, hierarchy and trajectory in state space. All the information from the different components are combined and synthesized for a holistic approach to assess the ecosystem health.

Other approaches have been used in aggregating indicators within each descriptor. For example, Borja et al. (2011b) use the biodiversity valuation approach, in assessing biodiversity within the MSFD, integrating several biodiversity components (zooplankton, macroalgae, macroinvertebrates, fishes, cetaceans and seabirds). Biodiversity valuation maps aim at the compilation of all available biological and ecological information for a selected study area and allocate an integrated intrinsic biological value to the subzones (Derous et al., 2007). Details on valuation methodology can be consulted in Pascual et al. (2011) (see Figure 4 in that paper). This methodology provides information for each of the components and their integrative valuation, together with the

**Table 3 | Summary of Task Group approaches to aggregate attributes within a Descriptor (Cardoso et al., 2010).**


reliability of the result, taking into account spatial and temporal data availability (Derous et al., 2007). The advantage of this method is that the current information used to valuate biodiversity can be adapted to the requirements of the MSFD indicators. Moreover, this method can avoid duplication of indicators in two descriptors (e.g., D1 and D6), since the metrics used could be different. This information can be converted into environmental status values, as shown in Borja et al. (2011b).

#### **INTEGRATION OF DESCRIPTORS (COMBINATION ACROSS DESCRIPTORS)**

Discussion on how to integrate the results of each descriptor into an overall assessment of GEnS for regions or subregions was not part of the Terms of Reference for the Task Groups. However, work within Task Group 6 (Sea floor integrity) identified a method for integration and assessment that might also be appropriate, if applied across all descriptors, at a regional scale (Cardoso et al., 2010). As these authors pointed out, crossdescriptor integration at the scale of (sub)regional seas runs the risk of blending and obscuring the information that is necessary to follow progress toward GEnS and to inform decisionmakers about the effects and the efficiency of policies and management. It may lead to masking of problems within specific descriptors.

Borja et al. (2013) describe at least 8 options to determine GEnS in a regional sea context (**Table 4**). These authors detail the concept behind these options, and propose the decision rule more adequate for the assessment method to be used, depending on the circumstances i.e., data availability, lack of monitoring, etc. In addition, these authors consider what type and amount of data are required, and then discuss the pros and cons of the different options. The implementation of a complex directive, such as the MSFD, requires a high amount of data to assess the environmental status in a robust way. Hence, the options from 1 to 8 proposed in **Table 4** are sequentially less demanding of new data, and the degree of detailed environmental assessment is also decreasing.

As such, Option 1, which is most similar to the WFD approach, deconstructs GEnS into the 11 descriptors and then into the component indicators, assessing each components for each area before attempting to produce an overall assessment (**Table 4**). However, having a complete dataset covering all descriptors and indicators for the assessment is difficult, if not impossible to achieve in practical terms. The use of pressure maps as an estimator of the environmental status and possible impacts to marine ecosystems could be considered instead (see **Table 4**). This would, however, build on the substantial assumption that the level of pressure is adequately representing the current state on all different levels of ecosystem components. Option 7, in contrast, only uses published data for the activities, and then infers a static relationship between activity, pressures, state changes and impacts both on the natural and the human system. Here, the number of underlying assumptions is even larger than using pressure maps, since the method relies on predefined and static DPSIR relations. Between these extremes, there are several intermediate options to integrate and present information, each with its own requirements, pros and cons (**Table 4**).

**Table 4 | Options for determining if an area/regional sea is in Good Environmental Status (GEnS) (modified from Borja et al., 2013).**


*OOAO, "one out, all out" principle.*

#### *One-out, all-out (OOAO)*

Although the MSFD describes the GEnS individually for each of the 11 descriptors, this does not necessarily imply the ability to have GEnS at the level of all the descriptors, nor does it mean that each descriptor should necessarily be graded individually in a binary way (i.e., good or not good environmental status) (Borja et al., 2013).

It could be argued that the 11 descriptors together summarize the way in which the ecosystem functions in terms of the MSFD view. As Member States have to consider each of the descriptors to determine good environmental status, this could be interpreted as a requirement to achieve GEnS for each of these descriptors. In that case, applying OOAO is the only integration method that can be applied to arrive at an overall assessment of GEnS, leading to a high probability of not achieving GEnS.

This assumes that the 11 descriptors, and the associated indicators, can be considered a coherent and consistent framework that adequately reflects the environmental status. In that situation, state descriptors not achieving GEnS would be accompanied by pressure descriptors not achieving GEnS, if the reaction of the ecosystem components is immediate, acting on the same time scale as the pressures. If this is not the case, for example if a pressure descriptor (e.g., D5 or D8) indicates that the level of the pressure is too high to achieve GEnS, while state descriptors (e.g., D1 or D4) do not reflect this, there is clearly an inconsistency in the assumed MSFD assessment framework, indicating that it does not capture delayed responses of state indicators to changing pressure indicators. That could be interpreted as a need for further research on the nature of P-S-I relations and the consistency in environmental targets for the descriptors involved, since our current state of knowledge on quantitative causal relations between pressures, state changes and impacts is limited. In addition, nearly all ecosystem components are subject to the true cumulative effects of many simultaneous pressures related to a range of human activities (Crain et al., 2008; Stelzenmüller et al., 2010; Knights et al., 2013). This means that, for some descriptors at least, there is a large scientific uncertainty associated with the definition of environmental targets and GEnS. Uncertainties in target setting, in the performance of an action (e.g., ecosystem state post-management) or in the contribution of individual driver(s) causing state change can undermine decision making when implementing environmental policy and can limit our ability to identify what should be managed, and what the impact of management might be (Knights et al., 2014). Consequently, developing a consistent assessment framework for all descriptors and indicators is an extremely challenging task, and using the OOAO approach is not appropriate.

#### *Alternative approaches*

The usefulness of integrating descriptors to one single value (overall GEnS assessment based on combination of the 11 descriptors) is under discussion by the Member States and the European Commission groups for the implementation of the MSFD. An argument against integration across descriptors is that it may not be informative any more since it results in loss of information at a crucial level where different elements are combined that cannot be integrated without major concessions.

The abovementioned groups have suggested that an integration across the biodiversity-related descriptors (D1, D2, D4, D6) might be an option, splitting those descriptors into various groups (e.g., functional or species groups). If a species or species group is assessed under more than one descriptor different aspects should be considered (e.g., chlorophyll a under D5 and phytoplankton species composition under D1).

However, if an integration across all descriptors is decided, Borja et al. (2010) suggest that the 11 descriptors are hierarchical and do not have an equal weighting when assessing the overall GEnS. Hence, Borja et al. (2013) suggest that for biodiversity (D1) to be fulfilled requires all others to be met and similarly if one of the stressor or pressure-related descriptors (e.g., D11, energy including noise) fails then by definition the biodiversity will be adversely affected at some point. This approach addresses the conceptual drawback of the OOAO principle and allows to have delayed responses to changing pressure regimes without drawing false conclusions and still being precautionary.

In addition to the problem of combining indicators (seen in the previous section) and descriptors the MSFD requires Member States to integrate and geographically scale-up the assessments at the level of a region or subregion (Borja et al., 2010). This differs strongly from the approach under the WFD, which is restricted to quality assessments at the scale of a water body (Hering et al., 2010). This means that the GEnS assessments of the different Member States within a regional sea need to be comparable and should avoid anomalies at the borders of Member States in order to enable synthesizing of the assessments into a region-wide assessment (Borja et al., 2013). This requires both comparable methods and associated combination rules to ensure minimum standards for GEnS reporting across Member States. As such, we advocate a set of common principles (expanded from Claussen et al., 2011, as shown in Borja et al., 2013):


#### *Visualizing and communicating the status*

The outlined alternative approach also shows that concerns on integration across descriptors do not necessarily have to be a problem. There are some methods which have demonstrated that integrating the information into single values (Borja et al., 2011b), maps (HELCOM, 2010) or radar schemes (Halpern et al., 2012) is still helpful and informative for ecosystem management, despite the involved loss of information that is inherent to a single number. Information can be retained when always presenting that single number together with the main underlying data, ideally visualizing the different levels of aggregation, allowing the lookup of the status at any level and relating the status with the actual pressures that lead to the synthesized value.

As an example, the Ocean Health Index (Halpern et al., 2012) provides weighted index scores for environmental health, both a global area-weighted average and scores by country (**Figure 3**). The outer ring of the radar scheme is the maximum possible score for each goal, and a goal's score and weight (relative contribution) are represented by the petal's length and width, respectively. This way of visualizing the integration could be adapted for the MSFD, integrating at the level of region or subregion, but also showing the values within each descriptor. This would still allow managers to extract relevant information and take actions at different levels: small (or local) scale, large (regional) scale, integrative (whole ecosystem status), or for each descriptor.

Another example, applied specifically for the MSFD, using all descriptors and most of the indicators, can be consulted in Borja et al. (2011b). These authors studied a system in which the main driver for the whole area is fishing, whilst at local level some pressures such as waste discharges are important. Although the overall environmental status of the area was considered good, after the integration of all indicators and descriptors, two of the descriptors (fishing and food webs) were not in good status (**Table 5**). Interestingly, biodiversity was close to the boundary to good status (**Table 5**), suggesting that the system could be unbalanced by fishing, but affecting various biological descriptors to different degrees. This means that the pressure must be managed to avoid problems in the future, especially because the descriptors already in less than good status showed a negative trend (**Table 5**).

Hence, from the examples above and the given reasoning, both main choices are still useful: either integrate or not integrate information across descriptors. Irrespectively of which combination proposal(s) is adopted and at which level, the precautionary principle should always be followed in absence of more robust knowledge (Borja et al., 2013). As a summary, the pros and cons of each decision are shown in **Table 6**.


**Table 5 | Example of an assessment of the environmental status, within the Marine Strategy Framework Directive, in the Basque Country offshore waters (Bay of Biscay) (modified from Borja et al., 2011b).**

*EQS, Environmental Quality Standards; EQR, Ecological Quality Ratio, both based upon the Water Framework Directive (WFD); NA, not available; Trends: red color, negative; green color, positive (in both cases can be increasing/decreasing, depending on the indicator).*

#### **PROPOSED STEPS FOR COMBINATION**

As a possible approach for the combination of assessments we propose the following steps (**Figure 4**):

– Assessments start at a low level, viz. the level of indicators and spatial scales that were defined for each specific indicator. This would result in assessment results for each indicator and each assessment area incorporating the levels of spatial assessment that was described as a nested approach (Step 1—spatial scales).

– Within one descriptor, this could result in a number of assessments for the different indicators, that all use the same scales for their assessment areas. This could be the case for descriptors like D5 and D8. In those cases, the assessments at indicator

#### **Table 6 | Pros and cons of the decision of integrating the information across descriptors.**


level can be aggregated to assessments at descriptor level for each assessment area, using suitable aggregation rules (Step 2—aggregation within a descriptor). These steps are already commonly used procedures in OSPAR (2009) and HELCOM assessments for eutrophication and contaminants.

– For other descriptors, the spatial scales for indicators may not be the same for all indicators. This could be the case for biodiversity, where a different spatial scale may be used depending on the species or habitat. Although integration of different biodiversity components and functional groups is required, methods need further development, and a number of EU projects are focussing on this issue.

Aggregation up to this level gives a detailed assessment result that suits the information needs for identifying environmental problems and needs for measures. The result of those steps at European level would be a very high number of assessment results, for each descriptor and assessment area (comparable to presenting the WFD assessments at water body level).

The following steps could provide information at a higher level of integration presenting the required overview of the current status of the overall environmental state and the progress toward GEnS:

	- Generally, use of OOAO (if one assessment area fails GEnS, the whole subregion fails) is not useful, as it gives a very conservative result and is not informative. Also, if the pressure is highly localized this approach is not adequate, since the whole subregion could fail GEnS due to a single location (which, of course, will need specific management measures).
	- In some cases, for example if a pressure is more or less homogeneous across a whole subregion (fishing, shipping), it could be useful to apply OOAO.

For some descriptors, surface area may be a good measure to express status at a subregional level: for example, D5, D8, and D10. For other descriptors, surface area is not suitable but other metrics should be considered, e.g., D1: numbers of species/habitats failing to achieve favorable conservation status; D3: number of stocks failing to meet "Maximum Sustainable Yield."

The end result of Step 3 could present the level at which GEnS is achieved at subregional scale as a pie chart. The aggregation results of Step 3 could be integrated across descriptors in a final presentation per subregion, using methods such as radar plots, or methods similar to the Ocean Health Index (Step 4—aggregation across descriptors). In this step, weighted approaches as suggested in previous sections would be considered.

#### **CONCLUDING REMARKS**

From the information provided in this overview, some conclusions can be highlighted:


principles/rules should be available as a sort of third assessment or backlog.


#### **ACKNOWLEDGMENTS**

The opinions expressed in this document are the sole responsibility of the authors and do not represent the official position of the European Commission. This work has been done under Framework contract No ENV.D2/FRA/2012/0019 (Coherent geographic scales and aggregation rules in assessment and monitoring of Good Environmental Status—analysis and conceptual phase), of the European Directorate General of Environment; and DEVOTES project (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) funded by the European Union under the 7th Framework Program "The Ocean of Tomorrow" Theme (grant agreement no. 308392) (www.devotes-project.eu). JHA was supported by the WATERS project (Waterbody Assessment Tools for Ecological Reference conditions and status in Sweden). María C. Uyarra (AZTI-Tecnalia) and Mike Elliott (University of Hull) provided constructive comments to the first version of the manuscript. This is contribution number 674 from the Marine Research Division (AZTI-Tecnalia).

#### **REFERENCES**


and biota, in determining quality standards? *Mar. Pollut. Bull.* 49, 8–11. doi: 10.1016/j.marpolbul.2004.04.008


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 17 May 2014; accepted: 20 November 2014; published online: 05 December 2014.*

*Citation: Borja A, Prins TC, Simboura N, Andersen JH, Berg T, Marques J-C, Neto JM, Papadopoulou N, Reker J, Teixeira H and Uusitalo L (2014) Tales from a thousand and one ways to integrate marine ecosystem components when assessing the environmental status. Front. Mar. Sci. 1:72. doi: 10.3389/fmars.2014.00072*

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science.*

*Copyright © 2014 Borja, Prins, Simboura, Andersen, Berg, Marques, Neto, Papadopoulou, Reker, Teixeira and Uusitalo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Overview of Integrative Assessment of Marine Systems: The Ecosystem Approach in Practice

Angel Borja<sup>1</sup> \*, Michael Elliott <sup>2</sup> , Jesper H. Andersen<sup>3</sup> , Torsten Berg<sup>4</sup> , Jacob Carstensen<sup>5</sup> , Benjamin S. Halpern6, 7, 8, Anna-Stiina Heiskanen<sup>9</sup> , Samuli Korpinen<sup>9</sup> , Julia S. Stewart Lowndes <sup>7</sup> , Georg Martin<sup>10</sup> and Naiara Rodriguez-Ezpeleta<sup>1</sup>

*<sup>1</sup> Marine Research Division, AZTI-Tecnalia, Pasaia, Spain, <sup>2</sup> Institute of Estuarine and Coastal Studies, University of Hull, Hull, UK, <sup>3</sup> NIVA Denmark Water Research, Copenhagen, Denmark, <sup>4</sup> MariLim Aquatic Research GmbH, Schönkirchen, Germany, <sup>5</sup> Bioscience, Aarhus University, Roskilde, Denmark, <sup>6</sup> Bren School of Environmental Science and Management, University of California at Santa Barbara, Santa Barbara, CA, USA, <sup>7</sup> National Center for Ecological Analysis and Synthesis, University of California at Santa Barbara, Santa Barbara, CA, USA, <sup>8</sup> Department of Life Sciences, Imperial College London, Silwood Park, London, UK, <sup>9</sup> Marine Research Centre, Finnish Environment Institute (SYKE), Helsinki, Finland, <sup>10</sup> Estonian Marine Institute, University of Tartu, Tallinn, Estonia*

#### Edited by:

*Christos Dimitrios Arvanitidis, Hellenic Centre for Marine Research, Greece*

#### Reviewed by:

*Ana M. Queiros, Plymouth Marine Laboratory, UK Antoine Jean Grémare, Université de Bordeaux, France Tatiana Margo Tsagaraki, University of Bergen, Norway Mathieu Cusson, Université du Québec à Chicoutimi, Canada*

> \*Correspondence: *Angel Borja aborja@azti.es*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *11 November 2015* Accepted: *15 February 2016* Published: *01 March 2016*

#### Citation:

*Borja A, Elliott M, Andersen JH, Berg T, Carstensen J, Halpern BS, Heiskanen A-S, Korpinen S, Lowndes JSS, Martin G and Rodriguez-Ezpeleta N (2016) Overview of Integrative Assessment of Marine Systems: The Ecosystem Approach in Practice. Front. Mar. Sci. 3:20. doi: 10.3389/fmars.2016.00020* Traditional and emerging human activities are increasingly putting pressures on marine ecosystems and impacting their ability to sustain ecological and human communities. To evaluate the health status of marine ecosystems we need a science-based, integrated Ecosystem Approach, that incorporates knowledge of ecosystem function and services provided that can be used to track how management decisions change the health of marine ecosystems. Although many methods have been developed to assess the status of single components of the ecosystem, few exist for assessing multiple ecosystem components in a holistic way. To undertake such an integrative assessment, it is necessary to understand the response of marine systems to human pressures. Hence, innovative monitoring is needed to obtain data to determine the health of large marine areas, and in an holistic way. Here we review five existing methods that address both of these needs (monitoring and assessment): the Ecosystem Health Assessment Tool; a method for the Marine Strategy Framework Directive in the Bay of Biscay; the Ocean Health Index (OHI); the Marine Biodiversity Assessment Tool, and the Nested Environmental status Assessment Tool. We have highlighted their main characteristics and analyzing their commonalities and differences, in terms of: use of the Ecosystem Approach; inclusion of multiple components in the assessment; use of reference conditions; use of integrative assessments; use of a range of values to capture the status; weighting ecosystem components when integrating; determine the uncertainty; ensure spatial and temporal comparability; use of robust monitoring approaches, and address pressures and impacts. Ultimately, for any ecosystem assessment to be effective it needs to be: transparent and repeatable and, in order to inform marine management, the results should be easy to communicate to wide audiences, including scientists, managers, and policymakers.

Keywords: assessment, integration, status, health, indicators, ecosystem approach, science-based communication

# INTRODUCTION: WHY IS IT NECESSARY TO ASSESS THE STATUS OF MARINE ECOSYSTEMS?

Traditional and emerging human activities in coastal and coastal/open marine waters, including shipping, fishing, wastewater discharges, recreation, and renewable energy production, have increased greatly in recent years (OSPAR, 2009), in part due to increasing coastal populations worldwide (Halpern et al., 2015a) and the need for new resources to support that accelerated growth. Despite the benefits these activities deliver to humans, the resulting pressures, including noise, overfishing, habitat destruction, and pollution, alter marine ecosystems in a combination of synergistic and/or antagonistic ways (Crain et al., 2008; Ban et al., 2010; Piggott et al., 2015). In addition, the rapid increase in anthropogenic pressures has modified the types, frequency, extent, and duration of disturbances or impacts on aquatic species, communities, and ecosystems (Nõges et al., 2016).

Legislation at national or regional levels aims to control the potential adverse impacts of marine activities (Borja et al., 2008; Boyes and Elliott, 2014), thereby changing the paradigms of marine management from studying and managing individual pressures separately toward managing the cumulative and in-combination activities and their pressures in a holistic, ecosystem-based management approach (Agardy et al., 2011; **Box 1**). This represents one of the grand challenges in marine ecosystems ecology (Borja, 2014).

Healthy oceans provide multiple valuable ecosystem services, which in turn produce societal benefits through food provision, raw materials, energy and recreation (Costanza et al., 1997; Barbier et al., 2012; Turner et al., 2014; Turner and Schaafsma, 2015). Nevertheless, human activities can compromise the delivery of ecosystem services in the short or long term, prompting society (marine users, conservationists, policy makers, managers, and scientists) to respond. Thus ensuring that the benefits enjoyed by these stakeholders continues to rely on a scientific understanding of how various parts of the marine ecosystem are interlinked, affecting ecosystem services provision and hence human societies. Managing human activities impacting the marine environment will only be successful by undertaking a science-based integrated ecosystem approach (Agardy et al., 2011).

The Ecosystem Approach emanates from the original 12 principles defined in the Convention for Biological Diversity (CBD, 2000), which indicates that it is "a strategy for the integrated management of land," water and living resources that promotes conservation and sustainable use in an equitable way. The application of the Ecosystem Approach will help to reach a balance of the three objectives of the Convention: conservation, sustainable use and the fair and equitable sharing of the benefits arising out of the utilization of genetic resources' (CBD, 2000). In essence, this is taken to mean that the natural system structure and functioning are maintained and enhanced while at the same time the ecosystem will support human uses and deliver the ecosystem services and societal benefits required by society (Elliott, 2011). It has often been used to refer to a particular sector such as an "Ecosystem-based approach to fisheries" (Garcia et al., 2003) although the view here is that the true Ecosystem Approach cannot be sectoral but must cover all sectors. This true "Ecosystem Approach" to management requires several elements: (i) defining the source of the pressures emanating from activities; (ii) a risk assessment and risk management framework for each hazard; (iii) a vertical integration of governance structures from the local to the global; (iv) a framework of stakeholder involvement, and (v) the delivery of ecosystem services and societal benefits (Elliott, 2014). All of this may be regarded as a means of achieving both a healthy natural system and a healthy social system which is fit-for-purpose (Tett et al., 2013).

An important component of an integrated ecosystem approach to marine management is an adequate assessment of the actual environmental status, describing the health of marine ecosystems in an integrative way (Borja et al., 2013; Tett et al., 2013). Considering the spatial extent and complexity of marine ecosystems, a considerable amount of data is needed to assess the status of coastal and open seas systems with sufficient precision. For that reason cost-effective monitoring methods are needed, delivering harmonized data with an adequate spatial and temporal coverage (Borja and Elliott, 2013). To inform management planning adequately, it is especially important that assessment methods and management tools can incorporate new knowledge, new monitoring methods (to tackle the problem of covering large areas) and indicators into assessments, but still maintain comparability with previous assessments so that any change in the status can be measured and quantified.

In essence, the successful application of the Ecosystem Approach is centered around the concept of "health"—by achieving both the health of the natural, environmental system and the health of the human system (Tett et al., 2013). Health can be regarded as indicating the "fitness for survival of natural components" and maintenance of individual, population and societal well-being and so a healthy and sustainable ecosystem can also be described as one that is able to attain its full expected functioning (Costanza and Mageau, 1999). With regard to marine ecological functioning, marine monitoring should explicitly or implicitly encompass health at all levels of biological organization

#### BOX 1 | ECOSYSTEM APPROACH DEFINITION

The Ecosystem Approach [defined in CBD (2000)] is a management and resource planning procedure that integrates the management of human activities and their institutions with the knowledge of the functioning of ecosystems. In the management of marine ecosystems and resources, it requires to "identify and take action on influences that are critical to the health of marine ecosystems, thereby achieving sustainable use of ecosystem goods and services and maintenance of ecosystem integrity" (cf., Farmer et al., 2012, for a review of the concept of ecosystem approach in marine management). The Ecosystem Approach can be defined as the ability to fulfil the major aim of protecting and maintaining the natural structure and functioning while at the same time ensuring the creation of ecosystem services from which societal benefits can be obtained (Elliott, 2011).

(Elliott, 2011), from the health of the cell, to the tissue level, individuals of a population, populations, and communities, which is currently the most used form of ecological monitoring (Gray and Elliott, 2009; Borja et al., 2013).

In addition, as emphasized throughout all major pieces of marine governance, there is a duty to assess and ensure the health of the whole ecosystem—as ensuring protection against adverse symptoms of ecosystem pathology (Elliott, 2011; Tett et al., 2013). This allows the detection of anomalous or malfunctioning attributes as well as the ability of the ecosystem to withstand change (its resistance) and/or its ability to recover after being subjected to a marine stressor (its resilience; Borja et al., 2010b; Duarte et al., 2015).

Hence, if the marine system can produce the provisioning, regulating, cultural and supporting ecosystem services then such well-being will be guaranteed. The role of marine management then requires an ecosystem health assessment (or monitoring) programme which analyses the main processes and structural characteristics of the coupled socio-ecological ecosystem and identifies the known or potential stressors. This then requires the development of hypotheses about how those stressors may affect the ecosystem and identifies measures of environmental quality and ecosystem health to test hypotheses. Because of this we need indicators to describe the condition of ecosystem components, the extent of pressures exerted on these components and the responses to either the condition or changes to it.

Given these challenges of applying the science-based ecosystem approach which by definition integrates the natural and societal features of the system, the objective of this position paper is to review and summarize the current knowledge on the assessment of marine health status, focussing on the Ecosystem Approach. Although very many methods have been developed to assess the status of single components of the ecosystem (see a review in Birk et al., 2012), there are very few assessing multiple components to give a holistic view of the ecosystem (e.g., Borja et al., 2014).

# MEASURING THE RESPONSE OF MARINE SYSTEMS TO HUMAN PRESSURES

Understanding the response of marine systems to human activities and resultant pressures requires a good conceptual basis that links the causes and consequences of change. This has been encapsulated in the DAPSI(W)R(M) approach (**Figure 1**, defined below), an improved version of the much used DPSIR approach (Wolanski and Elliott, 2015; Burdon et al., in press). This framework takes into account the different spatio-temporal scales at which **D**rivers, **A**ctivities, **P**ressures on the system, **S**tate changes, **I**mpacts (on human **W**elfare), and management **R**esponses (as **M**easures) operate. The Drivers relate to basic human needs including physiological desires, the requirement for safety and protection, employment, cultural satisfaction, or demand for goods and energy. The Impacts on human Welfare encompasses the loss of ecosystem services and employment and the psychological effects of risks and hazards. The complexity of the estuarine and coastal environment results in multiple interactions between various DAPSI(W)R(M) elements, especially in multi-use/multi-user cases. Furthermore, the nested-DAPSI(W)R(M) framework specifically recognizes the impact of Exogenic Unmanaged Pressures (ExUP)—such as climate change—and Endogenic Managed Pressures (EnMP) on

the system—such as new port developments or fisheries (Elliott, 2011). This management framework quantifies and assesses the Pressures, State changes and Impacts on human Welfare but it manages (using Responses as Measures) the Drivers and Activities.

Determining the adverse effects of human activities and their resultant pressures on ecosystems is essentially a risk assessment and risk management framework (Cormier et al., 2013) that has been included in the framework of Environmental Impact Assessments (EIA) for many decades. Scientific studies of effects of single pressures on the marine environment are already wellembedded in assessments but Halpern et al. (2008) was the first to assess cumulative human activities and their potential impact at high spatial resolution. This triggered a series of national and regional studies on the effect of multiple stressors on ecosystem components (Crain et al., 2008; Ban et al., 2010; Coll et al., 2012; Korpinen et al., 2012; Micheli et al., 2013; Marcotte et al., 2015; Piggott et al., 2015; Nõges et al., 2016), with each one also aiming to improve the method and bridge caveats of the method (Halpern and Fujita, 2013).

The "cumulative impact method" itself (Halpern et al., 2008, 2015a) is a straightforward additive model linking pressures and ecosystem components over a grid of assessment cells and using expert-based weights to estimate the impacts of each pressure on specific ecosystem components (i.e., species, habitats, ecosystems). The formula is:

$$I = \sum\_{i=1}^{n} \sum\_{j=1}^{m} P\_i \times E\_j \times \mu\_{i,j} \tag{1}$$

where P<sup>i</sup> is the log-transformed and normalized value of an anthropogenic pressure in an assessment unit i, E<sup>j</sup> is the presence or absence of an ecosystem component j (i.e., populations, species, habitats, or broad-scale habitats), and µi,<sup>j</sup> is the weight score for P<sup>i</sup> in E<sup>j</sup> . As the source data are high-resolution spatial layers for pressures and habitats, the scientific interest has often focused on the production of the weighing scores. As weighting scores are determined for stressor-habitat combinations, for global analyses they can miss nuanced interactions that better maps can provide, which has been done in smaller-scale assessments.

At smaller scales, weighing scores can be developed using local knowledge of system interactions, which, combined with local spatial data, has been shown to have a more significant role in the assessment results than the weighted scores in the Baltic (Korpinen et al., 2012) and the Mediterranean and Black Sea (Micheli et al., 2013). In the North Sea, Andersen et al. (2013) introduced the probability of species occurrence to the index, which is particularly suitable for highly mobile species such as seabirds, marine mammals, and big fish. With regards to pressure data, fuzzy logic was used in the U.K. sea area (Stelzenmüller et al., 2010) and in Hong Kong (Marcotte et al., 2015) to estimate the occurrence of pressures and spatial extent of adverse effects in the grid cells. In the Dutch sea area, the effects on species populations have been linked to the population demography, which allowed ecologically more realistic impact assessments (de Vries et al., 2011). When applying the index to smaller geographic scales, the need to account for the environmental variability increases. In the Finnish Archipelago Sea, a pilot study evaluated the effects of water depth and wave exposure (i.e., benthic energy) on the cumulative impacts in the index method (Sahla, 2015). The role of the two factors had significant effects on the index results in the small-scale study area.

Cumulative impacts have become a widely used element of marine assessments. For example, in Europe, the Marine Strategy Framework Directive (MSFD) particularly requires "the main cumulative and synergetic effects" to be included in Member States' assessments of Good Environmental Status (GES; European Commission, 2008). This GES should be achieved within all European seas by 2020, i.e., an area is deemed by the use of operational indicators to be one side or the other of the boundary between meeting or not-meeting GES (European Commission, 2008), using a set of 11 descriptors (biodiversity, alien species, fisheries, foodwebs, eutrophication, seafloor integrity, hydrography, pollutants in seafood and environment, litter, and noise), which encapsulate the whole ecosystem function. The European Commission (2010) proposed a set of 56 indicators to assess environmental status.

# NEED OF INNOVATIVE AND COST-EFFECTIVE MONITORING

In determining the effects of pressures over large geographical scales, and taking into account the holistic view of the new integrative assessment methods, there is a clear need for developing new monitoring approaches and especially those which encompass and combine all the relevant features of ecosystems; despite this, deciding on what, where, how, when, and how often monitor is not always as obvious (Borja and Elliott, 2013). Similarly, the role of monitoring in marine management and the pros and cons of the possible monitoring framework have to be determined, including the ability of the monitoring to detect a signal of change against a background of inherent variability (the "noise" in the system; Nevin, 1969). Elliott (2011) considered 10 types of monitoring, focusing on (i) the ability to determine the overall status of an area and over a time period this includes surveillance monitoring and condition monitoring, i.e., to monitor the features of an area and its status and then a posteriori to detect a trend; (ii) the ability to determine whether an area or a time period meets a pre-determined and pre-agreed status such as a baseline, threshold, or trigger value, which may be defined in law or in licence conditions and hence a priori has the status defined—this includes compliance monitoring and operational monitoring, and (iii) once a difference has been detected between what is expected and what is found, i.e., change has occurred, then that sequence or trajectory of change, and its causes and consequences have to be determined—this requires investigative or diagnostic monitoring and possibly feedback monitoring and toxicity analyses in which the assessment has a direct and real-time link to management.

Taking this into account, here we summarize and focus on four main promising approaches, which can assist monitoring, with importance in marine systems: genomic tools, remote sensing, acoustic devices, and modeling, which can be combined in a novel way to cover the needs of monitoring large geographical areas.

Genomic tools are seen as a promising and emerging avenue to improve ecosystem monitoring, as these approaches have the potential to provide new, more accurate, and cost-effective measures. Several techniques have been identified as potential substitutes of traditional approaches for various applications (Bourlat et al., 2013), and some can even provide measurements that were not possible before the genomic era (**Figure 2**).

Meta-omic (metabarcoding, metagenomics, and metatranscriptomics) techniques are particularly appealing as they allow the analysis of environmental samples without the need to isolate organisms. Probably, the most promising, developed, and straight-forward genomic tool for environmental monitoring is metabarcoding (Cristescu, 2014; Chariton et al., 2015). This technique consists of taxonomically identifying the organisms present in a given sample based on a small DNA fragment (called a "barcode") that is unique to each species. Potential applications of metabarcoding in marine monitoring include calculating biotic indices based on taxonomic composition, detection of invasive species or understanding trophic interactions by analysing fecal samples or stomach contents (Aylagas et al., 2014; Chariton et al., 2015; Dafforn et al., 2015). However, the routine application of this technique still requires that standardized practices at each step of the procedure are developed. For example, sampling strategies, nature of the barcode selected, conditions of barcode amplification or available reference barcode library may affect the taxonomic composition inferred from genomic data (Aylagas et al., 2014). Several campaigns of sampling standardization have already been initiated, such as the Ocean Sampling Day (Kopf et al., 2015) for marine microbe sampling, and the use of Autonomous Reef Monitoring Structures (ARMS; http://www.pifsc.noaa.gov/ cred/survey\_methods/arms/overview.php) for sampling both prokaryotic and eukaryotic organisms. There is therefore an urgent need to compare both traditional and molecular based taxonomic composition inferences so that metabarcoding can be introduced as a regular tool in monitoring programs.

Satellite remote sensing is another promising monitoring approach. Although this has long been used to monitor chlorophyll a (Coppini et al., 2012), it has only recently been applied to determine phytoplankton size structure (Barnes et al., 2011; Brewin et al., 2011), composition and functionality (Moisan et al., 2013; Palacz et al., 2013; Rousseaux et al., 2013) and monitoring of harmful algal blooms (Frolov et al., 2013). However, there are still few studies which assess the ecological status of coastal an d open marine waters based on the phytoplankton component (Gohin et al., 2008; Novoa et al., 2012), thus requiring the development in support of assessments in large marine areas.

Acoustic devices are a monitoring approach built on the traditional use of benthic habitat mapping (see Brown et al., 2011), that can be used to determine the composition and abundance of different biodiversity components, especially fish and cetaceans (André et al., 2011; Denes et al., 2014; Fujioka et al., 2014; Parks et al., 2014). Again, there are few studies regarding the use of underwater acoustics to assess the status of diverse ecosystem components and indicators (Trenkel et al., 2011).

Furthermore, certain types of modeling provide a valuable accompanying approach to monitoring, for example to increase spatial coverage of environmental variables and predict spatial distribution patterns of different ecosystem components, i.e., through species distribution modeling (Reiss et al., 2015). Deterministic models can be used to predict physico-chemical characteristics such as water quality parameters or fish stock size,

FIGURE 2 | Genomic approaches (left) and their potential marine potential application (right). Metabarcoding, metagenomics, and metatranscriptomics consist respectively on sequencing a region of the genome, the genome or the transcriptome of a whole community; qPCR (quantitative PCR) and microarrays consist on measuring the quantity of DNA or RNA in a given sample at low and high throughput respectively; SNP genotyping consists on determining the genotype of selected Single Nucleotide Polymorphisms of individuals from the same species in order to estimate differences in allele frequencies among populations. Applications that cannot be performed using traditional techniques are underlined.

whereas empirical models are valuable to link species presence to habitat characteristics and thus extrapolate from a monitored area to the wider spatial coverage (Groeneveld et al., in press; Peck et al., in press). Ecological modeling is being used to describe or understand ecosystem processes, and is currently a valuable approach used to predict and understand the consequences of anthropogenic and climate-driven changes in the natural environment (Piroddi et al., 2015). Piroddi et al. (2015) have reviewed the most commonly used capabilities of the modeling community to provide information about indicators used to assess the status in marine waters, particularly on biodiversity, food webs, non-indigenous species and seafloor integrity. Ecosystem modeling has the potential to show the complex, integrative ecosystem dimensions while addressing ecosystem fundamental properties, such as interactions between structural components and ecosystem services provided (Groeneveld et al., in press). As such, some modeling tools (i.e., species distribution modeling) can be used in support of monitoring to predict the distribution of species in areas not monitored or to derive indicators in support of the assessment process.

Traditional monitoring tools (i.e., direct sampling, visual identification, etc.) and these new monitoring approaches are producing information to generate the indicators needed to assess the status of marine systems, as presented below.

## EXAMPLES OF HEALTH AND STATUS ASSESSMENT IN MARINE SYSTEMS

The following sub-sections give examples (in chronological order of publication) of integrative assessment methods. All can be applied to large marine areas in open and coastal waters. Most of the methods are motivated by international legislation or conventions and use various indicators to derive the status assessment. The most important differences are their choice of indicators and the way these are synthesized into the overall ecosystem health. **Table 1** summarizes the main characteristics of the methods described here.

### Ecosystem Health Assessment Tool

With the adoption of the HELCOM (Baltic Marine Environment Protection Commission - Helsinki Commission) Baltic Sea Action Plan, the Contracting Parties to the Helsinki Convention launched an ambitious Action Plan to restore ecosystem health of the Baltic Sea (HELCOM, 2007). As the Action Plan is based on the Ecosystem Approach, tracking, and documenting progress in meeting the vision and objectives was required. Hence, a plan for establishing a region-wide baseline was developed and implemented through the production and publication of an indicator-based assessment of ecosystem health in the Baltic Sea region (HELCOM, 2010a).

The ecosystem health is based on a Baltic-wide application of a multi-metric indicator-based assessment tool, the HELCOM Ecosystem Health Assessment Tool (HOLAS; HELCOM, 2010a). This is based on existing HELCOM tools for assessing "eutrophication status" (HEAT; HELCOM, 2009a and Andersen et al., 2010, 2011), "biodiversity status," (BEAT; HELCOM, 2009b and Andersen et al., 2014) and "chemical status" (CHASE; HELCOM, 2010b; Andersen et al., 2016). Currently, the HOLAS tool is under revision to ensure applicability for the MSFD assessments in the future. This will include revision of the aggregation rules for the indicators that have been developed and agreed in the HELCOM CORESET project (HELCOM, 2013) where the jointly agreed set of indicators is to finalized currently.

Three dilemmas were faced. First, using few groups of indicators (one or two) and averaging across many indicators may potentially lead to "thinning" and potentially to "upward"

#### TABLE 1 | Summary of the main characteristics of the methods described here.


*For the complete names of the methods, see text. MSFD, Marine Strategy Framework Directive, HELCOM, Helsinki Convention; OOAO, One out, all out. <sup>a</sup>For contaminants, target values are used instead of background values/reference conditions.*

misclassification (i.e., arriving at a better status classification compared to the use of more groups; lessons learned from the development of the CHASE prototype tool). Second, many groups of indicators and stringent use of the "one out, all out" principle, in which overall status of a region defaults to the status of the worst biological component (Hering et al., 2010), may potentially lead to "downward" misclassifications (i.e., arriving at a poor status classification compared to the use of fewer groups; lessons learned from HEAT and Borja and Rodríguez, 2010). The one-out-all-out principle has been adopted in the European Water Framework Directive (WFD; European Commission, 2000). Third, in some cases, good indicators and target values do not yet exist.

The HOLAS tool has four steps (**Figure 3**). In step 1, indicators are nested in three categories (CI: biology; CII: chemistry; CIII: supporting). In step 2, either an Ecological Quality Ratio (EQR) or a Chemical Score (CSchem) is calculated. For categories I and III, a weighted average Ecological Quality Ratio (EQRbio and EQRsupp; see Equation 2) is calculated (ranging from 0, bad status, to 1, high status, sensu the WFD, European Commission, 2000) and for category II, a Chemical Score (CSchem; see Equations 3 and 4) is calculated as the ratio of the status against a threshold value. In step 3, categories I, II, and III are classified in five classes (High, 0.0–0.5; Good, 0.5–1.0; Moderate, 1.0–5.0; Poor, 5.0–10.0; and Bad > 10.0). Finally, in step 4, category classifications are combined (using the lowest ranking classification cf. the "one out, all out" principle (see Borja and Rodríguez, 2010), into a final classification of "ecosystem health" (in 5 classes).

The applied assessment principles differ for category I and II indicators. For category II indicators, as well as category III, the assessment principles on the indicator level is straight-forward, the only difference relate to whether the response is numerically positive or negative to an increase in pressure:

$$\begin{aligned} \text{EQR} &= \text{RefCon} / \text{Obs} & \quad \text{(positive response)}\\ &= \text{Obs} / \text{RefCon} & \quad \text{(negative response)} \end{aligned}$$

where RefCon is the reference condition and Obs is the observed value. Detailed descriptions of the above principles as well as integration principles within groups of indicators can be found in HELCOM (2010a) and Andersen et al. (2010, 2011, 2014).

For category II indicators each indicator is simply assessed against a threshold level by calculating the ratio and the results of the indicators are then combined to obtain the status for each element. For each of the indicators (n) in an assessment unit (i.e., a spatial quadratic unit), the Contamination Ratio (CR) of the measured concentration (Cm) to a relevant assessment criterion for GES (CThreshold) is calculated using:

$$CR = \frac{C\_m}{C\_{Threshold}}\tag{3}$$

Integration of the CRs of the indicators is calculated as a Contamination Score (CS; Equation 4):

$$\text{CS} = \frac{1}{\sqrt{n}} \sum\_{i=1}^{n} \text{CR}\_i \tag{4}$$

high (blue), good (green), moderate (yellow), poor (orange), and bad (red) (see text and Andersen et al., 2014 for further details). EQR, Ecological Quality Ratio.

A detailed description of these assessment principles and calculations as well as their practical use can be found in Andersen et al. (2016). As such, the HOLAS tools has been tested and applied in the Baltic Sea (HELCOM, 2010a) for the classification of ecosystem health status in selected open and coastal waters (**Figure 4**).

# A Method for the Marine Strategy Framework Directive, Within the Bay of Biscay

The first attempt for assessing status according to the MSFD, using the 56 indicators proposed by the European Commission (2010), was undertaken in the southern Bay of Biscay (Borja et al., 2011). The approach was based on combining indicators, by grouping the marine ecosystem components into four distinct and interlinked systems: (i) water and sediment physico-chemical quality (including general conditions and contaminants); (ii) planktonic (phyto- and zooplankton); (iii) mobile species (fishes, sea mammals, seabirds, etc.), and (iv) benthic species and habitats. These ecosystem components, affected by different human pressures, are linked to the 11 MSFD descriptors and, as such, indicating the quality of the different indicators (see Borja et al., 2010a, 2011).

Borja et al. (2011) assessed each indicator and descriptor by deriving an EQR (as in the WFD and the HOLAS method, see Section Ecosystem Health Assessment Tool) in which monitoring data are compared with reference conditions of each indicator, a fundamental step in any quality status assessment (Borja et al., 2012).

After calculating a status value for each of the indicators, the method integrates the values at the level of single descriptors and then combines all 11 descriptors into a final assessment (**Table 2**). Weighting each descriptor has been proposed, and could depend on its relationships with dominant pressures in the study area. Weighting would thus emphasize certain descriptors, e.g., fishing in **Table 2** (see also recommendations by Borja et al., 2010a).

An environmental status value was derived by multiplying the weight by the EQR of each descriptor and dividing by 100, and an overall environmental status value was obtained by adding all the values for each descriptor. The indicators and descriptors that have values below GES (see Section Measuring the Response of Marine Systems to Human Pressures) require management action and can be easily identified (**Table 2**). Criteria for achieving GES can be found in Rice et al. (2012), Borja et al. (2013), and ICES (2013). The method also assesses the reliability of the result in a qualitative way, taking into account data availability and confidence in the methods used in assessing the status, and following the same approach as for the assessment.

# Ocean Health Index

The Ocean Health Index (OHI; Halpern et al., 2012) was a logical progression following the development of the cumulative impacts framework (Halpern et al., 2008), as the OHI includes not only the negative impacts exerted on the oceans but also captures the tangible and less-tangible benefits derived from the oceans. The OHI framework scores a suite of benefits ("goals") that are delivered to people by assessing the current status and likely future state (including pressures and resilience measures) of each goal for each region that together comprise the whole assessment area (**Figure 5**). A single OHI Index score is calculated by

FIGURE 4 | Classification of "ecosystem health status" in the Baltic Sea. In panel (A), classifications are spatially interpolated in order to illustrate that the impairment is a large scale problem. Panel (B) shown classification per sub-region [expressed as good (green), poor (yellow), poor (orange), or bad (red)], while panel (C) shows the confidence assessment of the classifications per sub-region [expressed as a high confidence (blue), a moderate but acceptable confidence (green), and a low confidence (red)]. See HELCOM (2010a) for details.


TABLE 2 | Example of an assessment of the environmental status, within the Marine Strategy Framework Directive, in the Bay of Biscay (modified from Borja et al., 2011).

*EQR, Ecological Quality Ratio; WFD, Water Framework Directive; MSFD, Marine Strategy Framework Directive; OSPAR, Oslo-Paris Convention; NEA, North-East Atlantic; M-AMBI, multivariate-AMBI; Green, good status; Red, less than good status. Yellow color show the values for indicators included within several descriptors (in blue).*

combining all goal scores with the following equation:

$$I = \sum\_{i=1}^{N} \alpha\_i I\_i,$$

where I1...<sup>N</sup> are the n goal scores and α<sup>i</sup> are the goal weightings (equal by default although can reflect relative importance of goals within the assessment area). Individual goal (and sub-goal) scores I<sup>i</sup> are based on the current status relative to its reference state along with the recent trend in status and the interaction of pressures and resilience measures. Assessments to date have generally evaluated 10 goals, some of which have sub-goals.

The framework can be used to assess areas with different spatial scales, characteristics and priorities as it is tailored to the specific context, such that only relevant goals are assessed. Furthermore, scores are calculated relative to reference points based on what is important within the assessment area. OHI assessments use existing information so that assessments reflect the best available knowledge of the system at the time of the assessment; this can require indirect measures to be included in assessments where the direct measures that ideally would be included are unavailable. Therefore, assessments not only produce scores that can be used to inform policy decisions, but they also identify knowledge gaps that can also be highly valuable to prioritizing further management action.

To date, 11 assessments have been completed for seven different locations: globally for all coastal nations and territories for each year 2012–2015 (Halpern et al., 2012, 2015b), Brazilian coastal states (Elfes et al., 2014), the U.S. West Coast states and sub-states (Halpern et al., 2014), Fiji (Selig et al., 2015), Israeli Mediterranean districts (Tsemel et al., 2014), Canada (in prep), Ecuador Gulf of Guayaquil (in prep), and Chinese coastal provinces (in prep). Because the global assessment has been repeated annually for 4 years (Halpern et al., 2012, 2015b; www.ohi-science.org), emerging trends and patterns in calculated scores are becoming apparent. For example, continued improvement in the global economy since the economic collapse of 2008 is reflected in improving coastal livelihoods and economy scores, and the steady increase in creating marine protected areas

Frontiers in Marine Science | www.frontiersin.org March 2016 | Volume 3 | Article 20 |

worldwide has increased part of the sense of place goal. Repeated assessments also incorporate newly available data (e.g., when new satellites are launched creating a new data source), and can be used to evaluate if or how well particular policy actions are performing in changing ocean health. But to be relevant for policy, assessments should be conducted at governance scales appropriate for management action. At a minimum, this usually requires assessments at the regional sea or national scale.

The OHI framework was first applied in two countries of highly different sizes, both relatively information-limited: Brazil and Fiji. In each case, it was found that individual goal models could be redeveloped or improved with at least some local information, while relying on inputs and models from global assessments for goals where such information was unavailable (Elfes et al., 2014; Selig et al., 2015). The framework was also applied to a data-rich setting, the U.S. West Coast assessment. In this case high resolution and quality data were available for nearly all goals and data components of the Index. Regionally-appropriate reference points for some goals were also developed, allowing the assessment to better reflect regionspecific preferences within the assessment area (Halpern et al., 2014).

Completion of the 11 assessments noted above as well as involvement in additional ongoing assessments has allowed refining and improving conceptual and technical aspects of the tools and resources available to conduct an OHI assessment (Lowndes et al., 2015). Computational and visual tools as well as instructions for their use have been developed, and these tools are shared and support is given with independent assessment efforts. As with the cumulative human impacts framework (Halpern et al., 2008), the OHI framework has also triggered independent groups to assess areas of interest using local input information representing local characteristics and priorities. Of the 11 completed OHI assessments, four have been independently-led. The first was led by the Israeli National Nature Assessment Program HaMaarag, assessing the Israeli Mediterranean coast and incorporating local measures, including tourism patterns and desalinated water and setting reference points based on local priorities (Tsemel et al., 2014). At the same time, a group funded by the Canada Healthy Oceans Network (CHONe) completed a feasibility study where they added attributes important to Canada and recalculated scores with methods from the global assessment. They also led a survey on how goals should be weighted and will be able to build from this initial work and calculate scores separately for each Canadian ocean. The most recently completed assessments were led by the governments of Ecuador and China. These assessments were able to use government statistics as input information and management targets as reference points

for many goals. Additional independent assessments are also currently underway, in Spain, the Baltic Sea, Chile, Colombia, the Arctic, Hawaii, Peru and British Columbia.

Each OHI assessment can build from past assessments, conceptually and technically, since all data, methods and code are freely available online (www.ohi-science.org). Such transparency allows interrogating methods and results, but perhaps most importantly facilitates repeated assessments within a given area, allowing managers, scientists, and stakeholders to track and compare scores through time. A single assessment provides an important baseline of overall ocean health and guidance on strategic actions to improve ocean health; repeated assessments allow determining the efficacy of management measures taken.

### MARMONI Tool

The MSFD Marine Biodiversity Assessment Tool (referred to as the MARMONI Tool) is a publicly available webbased application developed in the framework of the LIFE+ MARMONI project with the aim to perform MSFD compatible, indicator-based, integrated marine biodiversity assessment (www.sea.ee/marmoni/). It uses various indicators for the assessment area with several options for GES determination (see also Section A Method for the Marine Strategy Framework Directive, within the Bay of Biscay). The boundary value, determining when GES is attained, can be defined as a fixed value or an interval of values or through an acceptable deviation (value or percent) from reference condition, GES can also be defined as a direction of trend or by expert judgment (Auni ´ nš and Martin, 2015).

The MARMONI tool follows a hierarchical approach (**Figure 6**). The first level is the assessment of the operational indicators according to their specific methodology (indicator specific assessment methods including: either GES is defined through reference conditions and acceptable deviation or GES is defined by a range of values or GES is defined by trend direction), resulting in attributing either GES or non-GES status. The tool uses a binary approach where an indicator reaching GES is scored 100, while an indicator which does not reach GES is scored 0. The second level constitutes the aggregation of assessment results to each Commission Decision (CommDec) indicator (e.g., distributional range, distributional pattern, habitat area; European Commission, 2010). This is carried out by calculating the mean of individual indicator scores within each aggregation unit. The next aggregation is at CommDec criteria level (e.g., species distribution, population size, and habitat extent) followed by a final aggregation at descriptor level (biodiversity in this case; European Commission, 2010). The method includes the possibility of weighting different indicators in three classes and the ability to test different scenarios by excluding different indicators entered in the database for scenario testing.

A separate procedure is performed to estimate the uncertainty of the assessment across four different elements: (i) spatial

uncertainty; (ii) temporal uncertainty; (iii) uncertainty associated with the measurement of operational indicator, and (iv) uncertainty associated with defining its GES level or targets. The spatial representation aims to describe how well the data used for the indicator calculation cover the area of interest, whether the sampling is complete in terms of spatial coverage and whether all relevant habitats are well covered. Uncertainty connected to temporal aspects can come from different sources of temporal variability (i.e., within year or assessment season, seasonal variability and between year variability) as relevant. To assess the confidence level at each level of temporal resolution, a measure of variance needs to be calculated. The quality of assessment data depends on whether the indicator values are entirely based on objective measurements, subjective estimations or modeled indicator values. Uncertainty is low when the GES boundary or target is based on robust historical data. Each of these uncertainty elements is attributed to one of three uncertainty classes. At each level of aggregation the median of the uncertainty elements is calculated and presented on each level in the same way as the assessment score.

The tool displays information about assessment scores at Descriptor and CommDec criteria levels, the number of operational indicators for different CommDec criteria and indicators, the biological features that are covered by indicators and the source of the greatest gaps, and the overall uncertainty class at each assessment level (**Figure 7**). Although the resulting assessment is intended as a basis for drawing conclusions on whether the assessed area has achieved GES or not, there are no strict MSFD guidelines on this kind of decision (e.g., how many or what proportion of the indicators not being in GES are allowed, for the area to still be considered being in GES). The tool is designed to illustrate on how far is the study area away from achieving GES for all indicators/criteria and where are the gaps in monitoring rather than to provide an unambiguous answer to whether an area is in GES or not. This is further complicated by the fact that Member States have not yet decided on the aggregation rules for combining the assessments based on individual descriptors (Borja et al., 2014).

The MARMONI tool has been tested on data from four areas within the Baltic Sea (Martin et al., 2015) and shows that it is an easy-to-use and straightforward method to perform assessment of the status of MSFD Descriptor 1 (biodiversity). The main limitations for the practical application can be the lack of operational indicators and data covering different biodiversity components of the assessment area. Using more operational indicators as well the even distribution of them between different biodiversity components and assessment criteria will increase the confidence of the assessment result.

# NEAT (Nested Environmental Status Assessment Tool)

This is a tool developed by the DEVOTES project (http:// www.devotes-project.eu, based on Andersen et al., 2014) for assessing the environmental status of marine waters, within the European MSFD (European Commission, 2008). It focuses on

biodiversity status rather than the pressures leading to state changes. The indicators are thematically grouped, assigning them to the corresponding habitats, biodiversity components, spatially defined marine areas and pressures for which they are used (available as the DEVOTool software; Teixeira et al., 2014). This can be used to check for a suitable set of indicators in terms of coverage of all important biodiversity components and habitats within assessment. As NEAT is designed around the Ecosystem Approach (Tett et al., 2013), encompassing all ecosystem features relevant to the assessment (Gray and Elliott, 2009) can thus be safeguarded.

NEAT guides a user through the assessment process once the user defines the spatial scope of the assessment. This can be a regional sea or any other number of geographical entities and is based on Spatial Assessment Units (SAU). Since biodiversity is rooted in the spatial domain (without space there is no biodiversity; Sarkar and Margules, 2002), the indicators are assigned to a SAU and a habitat. To do this, multiple hierarchically nested SAUs can be used in one assessment and different indicators can be used for each of them. The tool includes a nested hierarchy of habitats from which to choose and each of the SAUs used in an assessment will thus be assigned to corresponding habitats. The combination of SAU and habitat then determines which indicators can be used in the chosen setting (**Figure 8**). Every indicator used in the tool also carries information on the numerical scale of its status classification (number of status classes, class boundary values).

The next step is to enter the observed indicator values for different combinations of SAUs and habitats. Indicator values are entered alongside with their classification scale. Before employing these values in the assessment calculation, they are mathematically transformed to a common normalized numerical scale (from 0 to 1). Furthermore, together with the indicator values, a value or judgment on their standard error must also to be entered to allow an integrated uncertainty assessment.

NEAT uses weighting factors in the assessment calculation but, in contrast to other tools, it does not weight the indicators. Instead, the weighting is done on the entities of interest, namely the important features of the ecosystem such as the SAUs, habitats or biodiversity components. By default, all SAUs are weighted equally but SAUs within the assessment may be weighted differently in order to emphasize the importance of specific parts of the whole assessment area. For this, the SAUs can be weighted using their area and/or by their quality giving, for example, the relative value of SAUs, a feature of the assessment as quality is an assessment criterion. Further, habitats can also be weighted by either their area or their quality.

Essentially, the final assessment value is calculated as a weighted average, where the final weights are combined with the observed indicator values. In this simple example of synthesis, no special rules are applied but the tool design allows assigning different aggregation rules at the various steps in the calculation of the overall assessment value. As an example, instead of using the default algorithm, specific needs may require to employ the one-out-all-out principle between partial results of the weighted indicator values.

In order to assess the uncertainty in the final assessment value and thus the uncertainty of the biodiversity state classification, the standard error of every observed indicator value is used. The observed value is assumed to represent the mean value of a normal distribution with the standard error being its standard deviation. The resulting probability distribution is used to run a simulated assessment using the Monte-Carlo technique with 10,000 iterations. During each iteration the indicator values are picked randomly from the given probability distributions and the final assessment value is calculated. The 10,000 realizations integrate the uncertainty of the overall status assessment and can be displayed as a histogram of simulation results falling into the various status classes.

FIGURE 8 | Conceptual model of the design of the Nested Environmental Assessment Tool (NEAT). Every Spatial Assessment Unit (SAU) may be assigned to several habitats, every SAU/habitat combination to several indicators. SAUs and habitats are characterized by their area and a weight/quality while indicators are assigned to biodiversity components or other ecosystem features. The subsequent algorithms combine the indicator values using the weighting of their corresponding SAUs and habitats and result in the overall biodiversity status.

# LESSONS LEARNED FROM COMPARING THE TOOLS

This review summarizes key attributes of some of the main tools and approaches currently available as an illustration of the means of assessing marine waters under an Ecosystem Approach. Such assessment relies on our ability to determine the source and effects of human activities which lead to pressures, by monitoring and assessing the status. While not detailing all methods, the aim of this overview has been to show tools which: (i) are fit for purpose; (ii) can cover the relevant temporal and spatial scales; (iii) have encompassed the range of marine responses to human activities and pressures, and (iv) have been tested with available data. In particular they have given assessments which are an integral part of making decisions and taking the necessary actions to ensure and/or improve that health. The assessment methods reviewed in this study share some common attributes, discussed below (see also **Table 1**), that provide lessons about key attributes needed for assessment of environmental status of open and coastal systems.

# Assessments Should Use the Ecosystem Approach

All methods presented here are designed around the Ecosystem Approach. In the case of European methods, the MSFD requires that the member states that share the same marine region (i.e., Baltic, Atlantic, Mediterranean, and Black Sea) should collaborate to develop marine strategies in order to ensure coherence in the assessment, setting environmental targets and monitoring programmes. The regional platforms for developing coherent marine strategies are the Regional Sea Conventions (RSCs), which are the required regional coordination structures. Similarly, the MSFD states that "Marine strategies shall apply an ecosystem-based approach to the management of human activities," but no clear definition of the Ecosystem Approach is provided in the MSFD, although it is described elsewhere (e.g., CBD, 2000). The KnowSeas project definition (Farmer et al., 2012) provides a simple definition as: "a resource planning and management approach that recognizes the connections between land, air and water and all living things, including people, their activities and institutions." However, this definition does not specify how and by which means the Ecosystem Approach will be applied and what targets will be used. Those targets are dependent on each specific case that may vary among sea areas.

Therefore, using the Ecosystem Approach requires a common and explicit vision of the desired status of the environment, and multiple stakeholders need to be involved in the definition of that status. Within Europe, all RSC have stated their visions of the marine environment (**Table 3**) which emphasize the protection of ecosystem health and biodiversity as well as the sustainable use of marine ecosystem resources, which are implicit in the definition of GES of the MSFD. The next step is to decide upon strategic goals for fulfilling different aspects of the vision (e.g., health, diversity, and sustainability aspects; **Table 3**), and operational objectives for the different goals (Backer and Leppänen, 2008). Those objectives can be both science-based, evolving from the ecosystem state evaluations, or society-based describing potential threats impacting ecosystems (Laffoley et al., 2004).

# Assessments Should Include Multiple Components of the Ecosystem

When applying an Ecosystem Approach in assessing environmental status, it is especially important to include both biotic and abiotic components of the natural system and a range of social components from the human system. The biotic components should be included in the assessment at different organizational levels (e.g., species, communities, biotopes) even though the assessments of the different levels may serve different purposes. For example, while information at the population level is required for stock evaluation, information at the community level is required for a broader biodiversity assessment. Similarly, as shown here, assessing community and ecosystem structure is central to surveillance monitoring, techniques for determining the cellular and individual health may be of more benefit in investigative or diagnostic monitoring (Elliott, 2011). The latter may also give early warning of change whereby deterioration in the health of a cell or individual, unless checked, will ultimately affect the population, community and ecosystem health (Tett et al., 2013). In turn, cellular (genomic) assessments as shown here may be of value in both explaining a likely response but also in predicting future changes to organisms and hence to populations and communities. Hence, the ecosystem level is represented by the combination of all species, habitats, communities, and their interactions, and the methods in this overview aim to include all these components.

In addition to the natural system, social components being monitored should include the many different ways that people interact with and benefit from natural systems. Of course, there are many potential indicators that can be used in the assessment of the components. In the case of the European MSFD, some of the 56 candidate indicators could potentially fulfill some of the desired criteria to be used and, at the same time, consider the characteristics, pressures, and impacts that are described in this directive (Teixeira et al., 2014).

# Assessments Should Use Reference Conditions or Baselines and Be Repeated to Track Changes

The importance of setting targets and reference conditions in assessing marine ecosystem quality has been highlighted several times (i.e., Mangialajo et al., 2007; Gray and Elliott, 2009; Borja et al., 2012; Andersen et al., 2014). It is especially important to track the changes in marine status due to management measures being taken to reduce human pressures. Hence, it is necessary to repeat assessments both to inform new management objectives and to detect whether existing policies are effective, by measuring the discrepancy between the values of the monitored indicators and the reference conditions or target values set; this has been defined as true monitoring as opposed to surveillance (Gray and Elliott, 2009). It is axiomatic that all environmental legislation aimed at preventing adverse effects due to human actions requires the current system to be assessed against what TABLE 3 | Comparison of the visions of the Good Environmental Status (GES) characterized by the regional sea conventions, OSPAR (The Convention for the Protection of the Marine Environment in the North-East Atlantic), HELCOM (The Convention on the Protection of the Marine Environment in the Baltic Sea Area—the Helsinki Convention), UNEP/MAP (The Convention for the Protection of Marine Environment and the Coastal Region of the Mediterranean—the Barcelona Convention, implemented in the framework of UNEP/MAP), BSC (The Convention for the Protection of the Black Sea—the Bucharest Convention, implemented by the Black Sea Commission), and the Marine Strategy Framework Directive (MSFD).


is expected in an area if the actions were not present. For example, EIA, the WFD and MSFD, in Europe, and the Clean Water and Oceans Acts, in the US, all rely on detecting change from a known baseline, target, threshold, or reference value or determining a trend against the preferred situation (Borja et al., 2008). All of the methods reviewed here rely on the use of reference conditions to assess and track changes in the status; in turn this requires methods and calculations that can be repeated to enable future assessments with new information to be comparable. Repeatability is thus one fundamental characteristic of an ideal assessment.

# Use an Integrative Assessment of All Components

We emphasize that by definition an integrative assessment must include multiple ecosystem components (e.g., biological, chemical, physical, social, economic), numerous biodiversity elements (e.g., from microbes to cetaceans), different assessment scales (e.g., from local, to regional and global sea scale), some criteria to define spatial scales and some guidance on integrating information (see a review in Borja et al., 2014).

Once the indicators, each with their specific targets or reference conditions, have been set, tested, and validated and the monitoring programmes implemented to provide data for those indicators, the assessment cycle can be completed (e.g., for MSFD; **Figure 9**). Thematic, holistic assessments need to integrate indicators addressing different aspects of the ecosystem, as shown by all the methods described here, to indicate the overall ecosystem level health of the marine region as well as the spatially expressed pressure and impact indices (Korpinen et al., 2012).

Some authors (Borja et al., 2014) have concluded that any integration and aggregation principle used should be ecologically relevant, transparent and well documented, to make it comparable across different geographic regions, as exemplified by the methods reviewed here although they do differ in the way in which this is achieved. Some of the methods rely on an overall thematic integration, for example, the HELCOM HOLAS tool uses the themes biology, chemistry, and supporting indicators.

The method from the Bay of Biscay groups indicators into four interlinked systems of ecosystem components. Another way of integration is to follow some external scheme such as the MSFD descriptors and subsequent criteria, as implemented in the MARMONI tool. Used in an unreflective manner, this can, however, involve some difficulties such as double counting the same ecosystem feature under different criteria (Berg et al., 2015).

Furthermore, there is the continuing discussion regarding whether an assessment of status should be a single value into which is embedded many descriptors or indicators or whether each element should be presented with its own quantified status. For example, in Europe, there is a continuing debate regarding whether the environmental status is presented as one single outcome (pass or fail), for a sea area by merging the assessments of all Descriptors, or whether each descriptor should be assessed independently and so a sea area would have 11 (one per Descriptor) indications of pass or fail at environmental status. The former approach has the benefit of simplicity in communicating the results (i.e., a sea area can be regarded as having passed or failed a definition of environmental status) whereas presenting 11 separate indications of the status allows a cause of failure to be readily identified (if, for example an area failed the Descriptor for seafood contamination but passed the other descriptors then management actions are more identifiable).

# Use a Range of Values for Capturing Status

A value for the "deviance from target" is needed for planning the programme of measures and management actions to reduce or remove human pressures by controlling societal activities and drivers. This means that the assessment methods should show the variation in the status value. Usually this can be done through continuous ranges between 0 (bad status) and 1 (high status), as in the case of the methods for the WFD (see Birk et al., 2013). It has been adopted also for several of the methods reviewed in this study, for example the OHI (Halpern et al., 2012, 2015b; in this case uses a range from 0 to 100). The only method which has no continuous range is MARMONI, employing the binary scheme of only 0 and 100 as distinct values.

The MSFD similarly and implicitly uses a binary scale as it classifies an area as either in or not in GES. Using a common scale has the advantage of making assessment methods comparable, through intercalibration exercises, as those organized in Europe for the WFD implementation (Birk et al., 2013). Surprisingly, and in contrast to the WFD, the MSFD does not explicitly require intercalibration but the inescapable conclusion from the analysis here is that any member States, region or sea area using different methods will require intercalibration to demonstrate the coherence in application and implementation.

#### Weighting Components When Integrating

Sometimes, weighting indicators when combining them allows comprehensive assessments to recognize and capture that some information is more relevant or directly related than other information. All tools reviewed here have a weighting option, allowing managers to give more weight to indicators or features taking into account: (i) the spatial and temporal variability of the indicator; (ii) the availability of reliable data; (iii) the accuracy of assessing methodologies for each indicator, and (iv) the differential response of each indicator to the main pressures in the area, among others. NEAT is the only method in this review not applying the weighting to the indicators but rather use ecosystem features for weighting. Thus, the weight (influence) of an indicator on the assessment result is determined by, for example, the size and/or quality of an area to which the indicator is applied (Probst and Lynam, 2016). This allows giving due weight to the major ecosystem components, which are much easier to characterize than the major indicators, although this depends on how the weight system is defined (Probst and Lynam, 2016).

As highlighted by Borja et al. (2014), an adequate basis for assigning weights is not always available and in such cases equal weighting is recommended by Ojaveer and Eero (2011). However, assigning weights often involves expert judgment and some degree of subjectivity, and Aubry and Elliott (2006) point out that in some cases, expert opinions on weights can show important divergence even though best expert judgment may be the most defendable and acceptable method.

# Calculate the Uncertainty Associated with the Assessment

Management of human activities to ensure GES naturally requires a solid foundation and a defensible approach, before decisions are made that may potentially have large economic consequences. Hence, it is important to ensure high confidence in the marine status assessment. Confidence quantification of the integrated status assessments has so far generally been neglected due to the complexity of such calculations. Only NEAT investigates the propagation of uncertainties from inputs of indicator values to the overall assessment in a quantitative way (using the Monte-Carlo method as described above), whilst the other methods assess uncertainty in a qualitative way. However, it is essential to associate indicator values with an uncertainty estimate which can be quantitative (as in case of natural variability) or qualitative (as in case of conceptual uncertainties behind the indicator). Unfortunately, most studies developing marine indicators do not consider indicator uncertainty or do not indicate how to calculate the uncertainty. The indicator uncertainty can be calculated based on estimates of various uncertainty components affecting observations used for the indicator, and the number of observations required to achieve a given accuracy and precision can be calculated (Carstensen, 2007). It is paramount that more focus is devoted toward quantifying the uncertainty of indicator values and how these affect the overall integrated assessment. Without knowing the confidence in marine environmental status assessments, or if the uncertainty is too large, decision-makers may decide not to adopt any measures to regulate human activities, due to the lack of precise information, especially if such measures have a high cost and uncertain outcome.

# Ensure Comparability across Regions and Time

All of the methods reviewed here allow spatial and temporal comparisons within and between regional seas but each have strengths and weaknesses which need to be considered to improve the assessments and their confidence in managing marine ecosystems. Give that the type of assessments described here are enshrined in marine governance (Boyes and Elliott, 2014), such as the European MSFD and the US Oceans Act, and in licensing or marine activities (such as national pollution control legislation) then again it is emphasized that it is increasingly possible that there will be legal challenges to the science being used. Hence, the methods have to be robust and legally defendable both inside and between countries and at one time and across various reporting periods (e.g., Hering et al., 2010).

# Use of Robust Monitoring Approaches and Data

As shown in the section Need of Innovative and Cost-Effective Monitoring, the monitoring methods are evolving and improving and thus the assessment methods or frameworks need to be sufficiently flexible to incorporate data acquired using new studies, instruments and methods, and which are used to derive new indicators with their own targets. The methods presented here can receive data from multiple sources and monitoring networks, making them sufficiently flexible to incorporate new indicators, for an Ecosystem Approach assessment.

# Approaches Should Address Pressures and Impacts

Elliott (2014) showed the need for a holistic marine management, which is focussed around a risk assessment and risk management approach, which accounts for vertical governance systems and horizontal integration across stakeholders. Successful and sustainable marine management relies on the detection of changes in pressures, state, and impacts on human welfare but then, following the implementation of responses and measures, it addresses the drivers and activities in the marine arena and the catchments affecting it. Reducing human impacts on marine ecosystems, produced by pressures, requires a scientific basis for any management measures and ultimately the need for spatial predictions of environmental status (Andersen et al., 2015). An independent verification of the cause of the problem requires pressure indicators especially as the presence of an activity cannot be assumed to cause a pressure. For example, seabed extraction of sand does not have to cause smothering if mitigation measures are employed. However, those pressure indicators have to accommodate the fact that the pressure impacts have different spatial and temporal scales depending on the activity footprints, the pressure types and trajectories and the species they affect and therefore the pressure-state link may not always be within detectable timeframes. Including the timescales to the assessment tools is, nonetheless, within our reach.

# CONCLUSIONS

Assessing the status of marine ecosystems under an Ecosystem Approach is fundamental to informing management decisions, and assessment frameworks have been developed to fit this need. As these frameworks are applied through time and to different regions, improvements with new information and

# REFERENCES


increased understanding will be incorporated. Characteristics that are paramount to marine assessment frameworks include (i) transparency in describing which decisions were made and why; (ii) being scientifically defendable by being based on a sound conceptual understanding; (iii) repeatability, so change can be tracked through time, through understanding and quantification of uncertainty via access to detailed methods and computational code, and (vi) communicability of methods and scores through distillation and visualization to wide audiences (modified and expanded from Lowndes et al., 2015). Conducting assessments with these characteristics will not only make future assessments comparable between marine regions and through time for management interpretation but will also reduce the time and resources required for subsequent assessments and at the same time make the assessments legally defendable.

# AUTHOR CONTRIBUTIONS

AB and ME conceived the paper. All authors have contributed in an equal manner to the preparation of the paper, the introduction and lessons learned. Then each author has had a part of the paper to write: ME, SK, and JA that of pressures, NR and AB that of monitoring, HOLAS method by JA, SK, and AH, Bay of Biscay by AB, OHI by BH and JL, MARMONI by GM, NEAT by JC, JA and CM.

# ACKNOWLEDGMENTS

Several authors of this manuscript are supported by the DEVOTES (Development of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), www.devotes-project.eu. This position paper has resulted from a Summer School organized in San Sebastián (Spain), from 9th to 11th June 2015, with the support of EuroMarine (Ref.: EM/PFB/2014.0015) and DEVOTES. The constructive comments from four reviewers have improved notably the first version of this manuscript. This paper is contribution number 755 from AZTI-Tecnalia (Marine Research Division).


Report from DCE – Danish Centre for Environment and Energy No. 18. Available online at: http://www.dmu.dk/Pub/TR18.pdf: 139 p


European Marine Strategy Framework and the Water Framework Directives. Mar. Pollut. Bull. 60, 2175–2186. doi: 10.1016/j.marpolbul.2010.09.026


**428**

faunal data for assessing ecosystem condition. Adv. Ecol. Res. 51, 1–40. doi: 10.1016/B978-0-08-099970-8.00003-8


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Borja, Elliott, Andersen, Berg, Carstensen, Halpern, Heiskanen, Korpinen, Lowndes, Martin and Rodriguez-Ezpeleta. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Integrated assessment of marine biodiversity status using a prototype indicator-based assessment tool

#### *Jesper H. Andersen1,2\*, Karsten Dahl 3, Cordula Göke3, Martin Hartvig4,5, Ciarán Murray3, Anna Rindorf 4, Henrik Skov6, Morten Vinther <sup>4</sup> and Samuli Korpinen2*

*<sup>1</sup> NIVA Denmark Water Research, Copenhagen, Denmark*

*<sup>5</sup> Centre for Macroecology, Evolution and Climate, University of Copenhagen, Copenhagen, Denmark*

#### *Edited by:*

*Christos Dimitrios Arvanitidis, Hellenic Centre for Marine Research, Greece*

#### *Reviewed by:*

*Christos Dimitrios Arvanitidis, Hellenic Centre for Marine Research, Greece Marco Sigovini, National Research Council of Italy, Italy Céline Labrune, Centre National de la Recherche Scientifique, France*

#### *\*Correspondence:*

*Jesper H. Andersen, NIVA Denmark Water Research, Ørestads Boulevard 73, 2300 Copenhagen S, Denmark e-mail: jha@niva-danmark.dk*

Integrated assessment of the status of marine biodiversity is and has been problematic compared to, for example, assessments of eutrophication and contamination status, mostly as a consequence of the fact that monitoring of marine habitats, communities and species is expensive, often collected at an incorrect spatial scale and/or poorly integrated with existing marine environmental monitoring efforts. The objective of this Method Paper is to introduce and describe a simple tool for integrated assessment of biodiversity status based on the HELCOM Biodiversity Assessment Tool (BEAT), where interim biodiversity indicators are grouped by themes: broad-scale habitats, communities, and species as well as supporting non-biodiversity indicators. Further, we report the application of an initial indicator-based assessment of biodiversity status of Danish marine waters where we have tentatively classified the biodiversity status of Danish marine waters. The biodiversity status was in no areas classified as "unaffected by human activities." In all the 22 assessment areas, the status was classified as either "moderately affected by human activities" or "significantly affected by human activities." Spatial variations in the biodiversity status were in general related to the eutrophication status as well as fishing pressure.

**Keywords: biodiversity, marine, integrated assessment, habitats, communities, species, Marine Strategy Framework Directive**

#### **INTRODUCTION**

Assessments of biological diversity have the ambitious objective of describing the state of an entire ecosystem, often by using only a few selected indicators. The challenge of this objective is to select a representative set of indicators, which fulfill the needs of science and marine policy. The EU Marine Strategy Framework Directive (MSFD) sets 11 qualitative descriptors for "good environmental status" (Anon, 2008), laying a common framework for all European marine biodiversity assessments. In this new assessment regime, biodiversity is considered to include not only the species diversity and the state of populations and habitats, but also seafloor integrity and food webs. Despite the detailed guidance on the selection of indicators (Anon, 2010), the MSFD does not provide a methodology to assess the overall state of marine ecosystems with the proposed criteria and indicators. Instead the EC tasked ICES with the production of detailed reports on the next steps of the implementation of the MSFD descriptors (see Cardoso et al., 2010 and relevant background reports).

Biodiversity assessments generally need to take into account the fact that marine biodiversity is sensitive to and also structured by natural factors such as salinity, currents, temperature, etc. More specifically, marine biodiversity assessments have been limited by the lack of integrated monitoring networks, highquality biodiversity indicators, and indicator-based assessment tools (Borja, 2014), partly a consequence of the vast nature of biodiversity. We hypothesize that all three deficiencies are related to two shortcomings in monitoring. Firstly, monitoring of marine biodiversity is often expensive compared to the monitoring of eutrophication and contamination and good proxies for biodiversity changes have not been developed. Secondly, for certain features of marine biodiversity, e.g., seabirds, monitoring is inadequately integrated with the existing marine environmental monitoring and, hence, resources are wasted in uncoordinated efforts.

Consequently, assessments of marine biodiversity are not as well-developed as other types of assessments, where multi-metric indicator-based assessment tools are commonly used (HELCOM, 2010; Andersen et al., 2011). The regional sea conventions in the Baltic Sea (HELCOM; www*.*helcom*.*fi) and North-East Atlantic (OSPAR; www*.*ospar*.*org) as well as EU Directives (Habitats Directive and MSFD) call for assessments of biodiversity, but only HELCOM has thus far made an attempt to develop an prototype indicator-based tool for an assessment of biodiversity (HELCOM, 2009b, 2010).

*<sup>2</sup> Marine Research Center, Finnish Environment Institute (SYKE), Helsinki, Finland*

*<sup>3</sup> Department of Bioscience, Aarhus University, Roskilde, Denmark*

*<sup>4</sup> DTU Aqua, Section for Marine Ecosystem-Based Management, Technical University of Denmark, Charlottenlund, Denmark*

*<sup>6</sup> DHI, Hørsholm, Denmark*

A few recent studies of marine biodiversity in Northern Europe are based on data addressing a wide range of biodiversity features (such as phytoplankton, benthic communities, fish, seabirds, marine mammals) and robust and transparent scientific methods, e.g., Certain et al. (2011), Ojaveer et al. (2010), and Ojaveer and Eero (2011). These studies do not, however, take into account numerical biodiversity targets, and this is a shortcoming in regard to assessment of biodiversity status in the context of the MSFD (Anon, 2008).

In this study, we introduce and describe a simple indicatorbased methodology (i.e., tool) for assessing the status of marine biodiversity. The tool is tested in Danish marine waters using provisional indicators with associated numerical target values and the results presented and discussed should accordingly be regarded as tentative. The assessment of biodiversity is made despite the lack of a commonly accepted definition of "marine biodiversity." Both the tool and the assessment are anchored in a Baltic Sea-wide conceptual understanding of "good biodiversity status" (HELCOM, 2010), where the overall vision is a healthy Baltic Sea with a favorable biodiversity status, including (1) natural marine and coastal landscapes, (2) thriving and balanced communities of plants and animals, and (3) viable populations of species. Hence, our understanding of "marine biodiversity" is broad and includes other elements than just a count of the number of species.

#### **METHODS**

We have developed a methodology for classification of "biodiversity status," employing a tool named Biodiversity Assessment Tool (BEAT) 2.0, which is an improved version of the HELCOM Biodiversity Status Assessment Tool (BEAT 1.0). This multimetric indicator-based tool was initially developed for integrated assessment of the status of biodiversity in the Baltic Sea (HELCOM, 2009a, 2010), but its updated version differs from its predecessor by having an improved fit with the EU MSFD descriptors, three status classes, a balanced approach to confidence rating as well as a more user-friendly appearance, where information about the Biodiversity Quality Objective (BQO) as well as interim (per category) and integrated classification results are presented.

BEAT 2.0 is an indicator-based assessment tool. For an individual indicator, synoptic information is required regarding reference conditions (RefCon), acceptable deviation from reference conditions (AcDev), and observations of the present state of biological diversity (Obs). AcDev is defined as a fraction or percentage of the RefCon, and is set site-specifically per indicator.

In calculating the status, we considered two types of indicators: (1) indicators that show a positive (+ve) response to human pressure factors, i.e., whose value increases with greater degradation in biodiversity (e.g., primary production, which is positively correlated to nutrient enrichment), and (2) indicators with a negative (−ve) response, i.e., whose value decreases with greater degradation (e.g., depth distribution of submerged aquatic vegetation, which is negatively correlated to nutrient enrichment or population size of a fish species, which is negatively correlated to fishing pressure).

As a first step, a BQO, which defines the border between "biodiversity status unaffected by human activities" (UN) and "biodiversity status moderately affected by human activities" (MO), is calculated per indicator:

$$\text{BQO} = \text{RefCon} \times (1 + \text{AcDev}) \qquad \text{( + ve response)}$$

$$= \text{RefCon} \times (1 - \text{AcDev}) \qquad \text{( - ve response)} \quad \text{(1)}$$

Step 2 is calculating the state value for each indicator through comparison with the BQO to determine indicator status. For example, for an indicator with +ve response, if the observed state (Obs) does not exceed the BQO, then the status "unaffected by human activities" is achieved. If the BQO is exceeded, the status is "moderately" (MO) or "significantly affected by human activities" (SI).


Step 3 is to calculate a Biodiversity Quality Ratio (BQR), which in principle is comparable with the Ecological Quality Ratio principle *sensu* the WFD (Anon, 2000; Andersen et al., 2011). The BQR approach used in this assessment marks the ratio (0–1) between Obs and RefCon. For indicators with a positive response the BQR is given by RefCon/Obs. For those having a negative response the BQR is the inverse, i.e., Obs/RefCon.

$$\text{BQR} = \text{RefCon} / \text{Obs} \qquad (+ \text{ve response})$$

$$= \text{Obs} / \text{RefCon} \qquad (- \text{ve response}) \qquad (3)$$

This step represents a transformation of indicator-specific information regarding the state of biodiversity to a numerical value, where the BQR values for different indicators can be compared and combined.

As a step 4, indicators are combined within four categories: (I) broad-scale habitats, (II) communities, (III) species, and (IV) supporting indicators. The classifications are based on a weighted average of the BQO and BQR values within each category. Weights are established by expert judgment and used to balance indicators among different biodiversity components or correlated indicators (e.g., several fish indicators are down-weighted against single indicators for seabirds or mammals). If not specified otherwise, the weighting is kept neutral by giving each of the indicators equal weights. On the basis of the BQR and AcDev values, each category is given a quantitative assessment according to the principles described above for a single indicator. Individual indicators have only two "classes," i.e., "unaffected" and "impaired/affected." There are three category classes from "unaffected," to "moderately affected" and "significantly affected" by human activities. Whilst the boundary between "unaffected by human activities" (UN) and "moderately affected by human activities" (MO) is a simple weighted average derived from the indicator-specific BQOs, the boundary between "moderately" and "significantly affected by human activities" (SI) is a value of two times the criteria-specific BQO.

At step 5, the results of the four categories are combined by applying the so-called "One out—All out" principle *sensu* the Precautionary Principle (MSFD Preamble, section 27; Anon, 2008) to the Categories I–IV. This implies that the category most sensitive to human activities, i.e., scoring lowest, defines the overall status of biodiversity within an assessment sector.

In addition to the above-described classification of biodiversity status, we estimate the confidence of the data and of the resulting classification by applying a simple scoring system (see Andersen et al., 2010). This system was initially developed for estimation of the confidence in eutrophication classifications but can be directly transferred and applied, when assessing biodiversity status. The approach, which scores the data on RefCon, AcDev and Obs gives equal weight to each of these three factors. In order to balance BQOs and Obs, we have modified the weighting of the factors with 25% to RefCon and AcDev and 50% to Status. The final confidence of the assessment can range between 100 and 0% and is according to Andersen et al. (2010) grouped in three classes: High (100–75%), Acceptable (75–50%), and Low (*<*50%). A description of the confidence rating method is available online as Supporting Material (Annex S3).

All calculations and subsequent classifications are made within a spreadsheet (see the Supplementary Material). Tracking calculations per indicator and also the integrations made per category and integration made in order to arrive at a final classification of biodiversity status is transparent and straightforward.

The BEAT 2.0 tool was tested and demonstrated using data from Danish marine waters, which are located in two distinct marine regions, the saline North Sea and the brackish Baltic Sea (**Figure 1**). Comprehensive descriptions of the study area and environmental status can be found in HELCOM (2010) and OSPAR (2010). The test was made on the basis of 22 assessment sectors in the Danish marine waters (**Figure 1**). The assessment sectors were larger in the offshore waters where spatial variation of the biodiversity indicators was considered smaller than in the coastal waters.

The data used for testing of BEAT 2.0 were compiled from various sources. Data on submerged aquatic vegetation as well as plankton (chlorophyll-a), benthic invertebrate communities, and nutrient concentrations originate from the Danish National Aquatic Monitoring and Assessment Programme (DNAMAP; see Conley et al., 2000; Carstensen et al., 2006; Dahl and Carstensen, 2008; Hansen, 2013). Data originates from three sources which are specific to the following areas: (1) offshore parts North Sea, Skagerrak and Kattegat (assessment sectors 1, 2, 4, 5), (2) offshore part of the Arkona Basin and Bornholm Basin, which are parts of the Baltic Sea (sectors 21 and 22), and (3) Danish coastal waters (sectors 3 and 6–20).

The indicators in regard to offshore fish, seabirds and marine mammals, which should be regarded as provisional, were developed specifically for this study and were also used for an interim assessment of biodiversity status in the North Sea (HARMONY project; unpublished data). Indicators used in previous assessments of the state of the North Sea (OSPAR, 2010) and Baltic Sea (HELCOM, 2010) were used for benthic and pelagic habitats and communities as well as supporting indicators. Detailed

**FIGURE 1 | Map of Danish marine waters.** The borders indicated in the map represent the current MSFD boundary between the North Sea region and the Baltic Sea region, relevant OSPAR boundaries, relevant HELCOM boundaries as well as relevant Habitats Directive boundaries for

biogeographical regions (BOR, Boreal; ATL, Atlantic; CON, Continental). Numbers indicates assessment sectors (see **Table 1** for names). Large circles indicate offshore assessment sectors, small circles coastal assessment sectors.

**Table 1 | Assessment and classification of biodiversity status in Danish marine waters.**


*For each assessment sector, the weighted Biodiversity Quality Ratio (BQR) is presented. These values represent the perturbation in regard to the reference conditions. C I, marine landscapes (broad-scale marine habitats); C II, communities; C III, species; C IV, supporting indicators; MO, moderately affected by human activities; and SI, significantly affected by human activities. The category being decisive for the outcome of the integrated assessment and classification is marked with an asterisk. See Online Supporting material for details.*

information about (1) the interim biodiversity indicators, (2) the sources for the monitoring data used as well as (3) the periods covered is available online as Supporting Material.

#### **RESULTS**

The average number of indicators per assessment sector was 10.2 (*n* = 22) ranging from 1 (no. 15 and 16) to 25 (no. 5). The average number of indicators in the four categories I–IV was 1.0, 4.0, 3.1, and 2.3, respectively. For the 6 offshore assessment sectors, the average number of indicators was 19.3 ranging from 8 (no. 22) to 25 (no. 2 and 5) and the average number of indicators in the four categories were 1.5, 5.8, 10.3, and 1.8 respectively. For the remaining 16 coastal assessment sectors, the average number of indicators was 6.8 ranging from 1 (no. 15 and 16) to 15 (no. 6) and the average number in the four categories were 0.9, 3.3, 0.3, and 2.4, respectively.

In the Danish marine waters, the average Biological Quality Ratio was 0.556, 0.595, 0.531, and 0.563 per category (**Table 1**). In category I, the BQR ranged from 0.230 to 0.862, in category II from 0.239 to 0.939, in category III from 0.370 to 0.656, and in category IV from 0.320 to 0.850.

For each assessment sector, a status classification was made per category and combined to a final integrated assessment of status per assessment sector (**Table 1**). The average of the lowest classified category was 0.433, ranging from 0.230 (sector 17: Southern Little Belt) to 0.639 (sector no. 1: North Sea, East+South). Areas with a BQR *<* 0.400 included Odense Fjord (sector 14), Little Belt (sector 17), and Bornholm Basin (sector 22), which all are significantly affected by eutrophication (HELCOM, 2010; Andersen et al., 2011). Areas with a BQR value above 0.600 were few and only found in the North Sea (sectors no. 1 and 2) and Isefjorden/Roskilde Fjord (sector 9). None of the assessment sectors were classified as unaffected by human activities. Three out of 22 assessment sectors were classified as moderately affected by human activities. The areas were Arkona Basin (no. 21), The Sound (no. 10) and Aarhus Bight (no. 12). The remaining 19 sectors were classified as significantly affected by human activities, and in 17 of these, the final classification was caused by categories I (broad-scale habitats), II (communities) or III (species). In two sectors, Hjelm Bight (no. 20) and Fakse Bight/Stevns (no. 11), the final classifications were a result of supporting indicators.

The confidence of the assessments was generally estimated to be above 50% and therefore considered acceptable (**Figure 2A**). However, two assessment sectors had a low confidence (no. 15 and 16: respectively, Sejerø Bay and Kalundborg Fjord) due to low number of indicators in the assessment in combination with challenges in regard to the setting of AcDev. Analysing the data per indicator revealed that monitoring data (State) and RefCon values on average had a higher confidence than the information on AcDev, which seemed to be slightly below the border between acceptable and low confidence (**Figure 2**). Scrutiny of the confidence per category revealed that all four categories on average had an acceptable confidence. All final classifications of the biodiversity status in the North Sea/Skagerrak area and the Kattegat had an acceptable confidence, while in the sub-division covering the Danish parts of the Baltic Sea, 2 out of 12 had an unacceptable confidence.

#### **DISCUSSION**

In this study we have presented a spreadsheet-based assessment tool for assessment of biodiversity, based on indicators, quantitative thresholds for good environmental status, and confidence rating. The assessment tool, tested by using both (i) existing and provisional indicators and (ii) recent data, showed that the marine biodiversity of Danish marine waters cannot be considered to be in good environmental status. The perturbations from reference conditions are indicative of human pressures in the assessment area (OSPAR, 2010; Korpinen et al., 2012).

Given the data and indicators available, we estimated the perturbations—understood as the deviation from reference conditions—represented by the lowest BQR values within an assessment sector. Parts of the North Sea and Skagerrak were less disturbed compared to the Kattegat and the Danish parts of the Baltic Sea (**Figure 3A**). The areas deviating most from reference conditions are all characterized by high nutrient inputs, high fishing pressure, and physical modification, sometimes caused by destructive fishing practices (HELCOM, 2010; Korpinen et al., 2012). Any measures to improve biodiversity status should as a priority address these key pressures.

An overview of the biodiversity status in the Danish marine waters revealed that a group of sectors being classified as moderately affected are interconnected (**Figure 3B**). The Sound is located downstream of Arkona Basin with a surface current from Arkona Basin to the west through Femernbelt between Denmark and Germany and to the north through the Sound. Hjelm Bight (sector no. 20) is located to the west and downstream of Arkona Basin. Fakse Bight/Stevns (sector no. 11) is located in between Arkona Basin and the Sound. The biodiversity status of the Arkona Basin and the Sound being classified as moderately affected by human activities is in line with the general understanding of the ecological status of these areas (HELCOM, 2010). Another sector having a slightly better status is Aarhus Bight (no. 12), where biodiversity status was classified as moderately affected by human activities in all the four categories. This, together with an estimated high confidence, does in our opinion confirm the classification. The reason for this slightly better status compared to adjacent sectors is most likely due to significant reductions in nutrient loads to Arhus Bight over past two decades (HELCOM, 2012).

Making an assessment without estimating the confidence of the result is a tendency, which in principle is unacceptable (**Figure 3C**). Estimating confidence is a statistical challenge, but the simple scoring system developed as a part of BEAT 2.0 overcomes this challenge in a non-statistical way and is able to cover confidence of threshold values, data and also the low number of indicators. This approach can be seen as temporary, leading to more sophisticated and data driven systems for assessment of confidence.

Many of the indicators in this assessment test have long traditions in previous assessments. Benthic communities and submerged aquatic vegetation have a long history in regard to assessments of eutrophication in the North Sea and Baltic Sea regions. Also indicators of fish communities have been used in previous assessments (Daan et al., 2005; Greenstreet et al., 2011), but reference levels had not yet been proposed for our study area, and for this analysis we used reference levels and acceptable deviations of 1 standard deviation based on the historic time series available.

Basin-wide biodiversity assessments have not hitherto included indicators for seabirds or marine mammals. The assessment in this respect can therefore be seen as a first attempt to use the trends in the population size of key species of seabirds or marine mammals as indicators of the status of the pelagic ecosystem in terms of habitat quality, food supply, and human-induced displacement. As the seabird data available for the assessment did not include data from the most recent period, the assessment used AcDev values of 50% and, hence, may give false positive impression of their status. Therefore, the reported changes in the abundance of fish-eating seabirds in the eastern parts of the North Sea, Skagerrak, and Kattegat should be regarded as strong indications of negative changes in the ecological status of these regions. Recent studies indicate that the regional reduction of fish-eating seabirds in the North Sea is mainly governed by changes in the large-scale abundance of herring (Fauchald et al., 2011). Reflecting the spatial caveats in the marine mammal data, the assessment used AcDev values of 50%. It is not known to what degree the impaired status of marine mammals in the eastern parts of the North Sea is a result of similar changes in the supply of pelagic fish which affected the abundance of seabirds in these regions. We did not include indicators for non-native species in this study. However, there is a growing understanding that, contrary to the normally negative perception of the ecological impact of non-native species, some species may provide significant ecosystem services in specific cases (Norkko et al., 2012).

In the current implementation process of the EU MSFD, there is a growing need to coordinate indicator development and agree on common sets of indicators, which allow coherent, trans-boundary assessments of the state of marine environment. By using existing indicators from the region, we noticed that several of the indicators were inherently correlated in nature (e.g., LFI and the slope of the size spectra, or chlorophyll a and Secchi depth) and using both as independent indicators in the present study may not be appropriate from a statistical point of view. In this study this correlation was accounted for by giving small weights to such indicators, but more stringent statistical

**FIGURE 2 | (A)** Confidence ratings made for (i) integrated assessments; (ii) information in regard to RefCon, AcDev, and AcStat of indicators, and (iii) categories I–IV. Values *>* 50% indicate an acceptable confidence (Andersen et al., 2010). **(B–D)**

Sub-region-specific confidence assessments for the North Sea and Skagerrak, the Kattegat including the northern parts of the Sound and the Belt Sea and the western Baltic Sea. Please confer with Supplementary Material for details.

consideration should be given to the issue before the next regional MSFD assessments.

We used supporting indicators to reflect changes in water quality in the Danish waters, which are affected by eutrophication (Ærtebjerg et al., 2003; Andersen et al., 2011). The eutrophication indicators indirectly reflect the condition of pelagic and benthic habitats and can, thus, indicate an overall status for a range of species and communities. Significant relations have been identified between nutrient loads and concentrations, chlorophyll-a concentrations, Secchi depth, depth limit of eelgrass (*Zostera marina*), total cover of macroalgae, and oxygen concentration in bottom waters (Conley et al., 2000; Nielsen et al., 2002a,b; Carstensen et al., 2004; Dahl and Carstensen, 2008). Thus, the water quality indicators can in a sense be called "true" indicators, as they can predict biological changes with simple methodology and relatively low costs. Nonetheless, in this study we considered them as "indirect" and prefer more direct measurements of biological parameters.

#### **CONCLUDING REMARKS**

Biological diversity in the Danish marine waters is significantly affected by human activities in most areas, but in a few sectors only moderately. None of the assessed sectors were classified as having a biodiversity status unaffected by human activities. The confidence of the assessments was estimated indirectly and generally regarded as acceptable, in a few cases even high. In two out of 22 sectors, the confidence was low indicating that monitoring of biodiversity in these sectors should be improved. The majority of the indicators were considered scientifically robust, but some indicators could, however, be further strengthened through production of peer reviewed scientific publications. Caution is also recommended in regard to the use of supporting indicators, especially in those few cases where they overrule biological indicators and thus determine the outcome of the integrated and final classification of biodiversity status. The BEAT 2.0 tool can support the EU Member States in the implementation of the MSFD, which specifically requires an overall assessment of the state of the marine environment as well as a specific assessment of biodiversity (Anon, 2008). The tool requires reliable indicators and quantitative thresholds for GES, but can function even with heterogeneous data availability. Assessments based on single indicators, though being simpler to link to human pressures, cannot reflect the variability and complexity of biodiversity responses required by the new assessments and therefore an integration of several indicators by an assessment tool is a prerequisite for the successful interface of science and environmental policy.

Finally, we would prudently like to remind the reader that there is no such thing as a perfect assessment tool. We do not promote the BEAT tool as such. We rather see this tool as a step for further development leading to better ecosystem-based tools for assessment, classification and adaptive management of marine biodiversity and human activities affecting marine life. The key challenges in regard to future integrated assessments of biodiversity status in marine waters are: (1) development of a wider range of biodiversity indicators representing different ecosystem components/food web categories, as well as (2) development of data driven methods for indicator integration and estimation of uncertainties.

#### **ACKNOWLEDGMENTS**

This article has been funded by the HARMONY project. The article has also been supported by the DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), www*.*devotes-project*.*eu. The authors would like to thank Johnny Reker and Joachim Raben as well as Stefan Heinänen, Alf B. Josefson, Alf Norkko, and Anna Villnäs. Martin Hartvig acknowledges the Danish National Research Foundation for support to the Center for Macroecology, Evolution and Climate. A prototype of the BEAT assessment tool was originally developed for HELCOM's integrated thematic assessment of biodiversity in the Baltic Sea and we would like to thank Hermanni Backer, Maria Laamanen, and Ulla Li Zweifel for constructive discussions of this prototype.

#### **SUPPLEMENTARY MATERIAL**

The Supplementary Material for this article can be found online at: http://www.frontiersin.org/journal/10.3389/fmars. 2014.00055/abstract

Additional supplementary material, i.e., Annex S1 containing detailed information on data sources and provisional indicators used for the testing of the tool presented in this study, Annex S2 a summary of the confidence rating methodology, Annex S3 containing 22 individual BEAT classifications, and Annex S4 being a step-wise BEAT 2.0 tutorial, is available to the online version of this article.

#### **REFERENCES**


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

#### *Received: 04 April 2014; accepted: 30 September 2014; published online: 29 October 2014.*

*Citation: Andersen JH, Dahl K, Göke C, Hartvig M, Murray C, Rindorf A, Skov H, Vinther M and Korpinen S (2014) Integrated assessment of marine biodiversity status using a prototype indicator-based assessment tool. Front. Mar. Sci. 1:55. doi: 10.3389/ fmars.2014.00055*

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science.*

*Copyright © 2014 Andersen, Dahl, Göke, Hartvig, Murray, Rindorf, Skov, Vinther and Korpinen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# Indicator-Based Assessment of Marine Biological Diversity–Lessons from 10 Case Studies across the European Seas

Laura Uusitalo<sup>1</sup> \*, Hugues Blanchet 2, 3, Jesper H. Andersen<sup>4</sup> , Olivier Beauchard<sup>5</sup> , Torsten Berg<sup>6</sup> , Silvia Bianchelli <sup>7</sup> , Annalucia Cantafaro<sup>7</sup> , Jacob Carstensen<sup>8</sup> , Laura Carugati <sup>7</sup> , Sabine Cochrane2, 9, Roberto Danovaro7, 10, Anna-Stiina Heiskanen<sup>1</sup> , Ville Karvinen<sup>1</sup> , Snejana Moncheva<sup>11</sup>, Ciaran Murray 4, 8, João M. Neto<sup>12</sup>, Henrik Nygård<sup>1</sup> , Maria Pantazi <sup>13</sup>, Nadia Papadopoulou<sup>14</sup>, Nomiki Simboura<sup>15</sup>, Greta Srebalien ˙ e˙ 16 , Maria C. Uyarra<sup>17</sup> and Angel Borja<sup>17</sup>

*<sup>1</sup> Marine Research Centre, Finnish Environment Institute (SYKE), Helsinki, Finland, <sup>2</sup> SALT Lofoten AS, Svolvær, Norway, <sup>3</sup> UMR 5805 EPOC, University of Bordeaux, Talence, Bordeaux, France, <sup>4</sup> NIVA Denmark Water Research, Copenhagen, Denmark, <sup>5</sup> Estuarine and Delta Systems Department, Royal Netherlands Institute for Sea Research, Yerseke, Netherlands, <sup>6</sup> MariLim Aquatic Research GmbH, Schönkirchen, Germany, <sup>7</sup> Department of Life and Environmental Sciences, Marche Polytechnic University, Ancona, Italy, <sup>8</sup> Department of Bioscience, Aarhus University, Roskilde, Denmark, <sup>9</sup> Akvaplan-niva, FRAM – High North Centre for Climate and the Environment, Tromsø, Norway, <sup>10</sup> Stazione Zoologica Anton Dohrn Napoli, Napoli, Italy, <sup>11</sup> Marine Biology and Ecology, Institute Of Oceanology - Bulgarian Academy of Sciences, Varna, Bulgaria, <sup>12</sup> Department of Life Sciences, IMAR - Institute of Marine Research, University of Coimbra, Coimbra, Portugal, <sup>13</sup> Hellenic Centre for Marine Research, Institute of Marine Biological Resources and Inland Waters, Athens, Greece, <sup>14</sup> Hellenic Centre for Marine Research, Institute of Marine Biological Resources and Inland Waters, Heraklion, Greece, <sup>15</sup> Institute of Oceanography, Hellenic Centre for Marine Research, Attica, Greece, <sup>16</sup> Marine Science and Technology Centre, Klaipeda ˙ University, Klaipeda, Lithuania, ˙ <sup>17</sup> Marine Research Division, AZTI Tecnalia, Gipuzkoa, Spain*

The Marine Strategy Framework Directive requires the environmental status of European marine waters to be assessed using biodiversity as 1 out of 11 descriptors, but the complexity of marine biodiversity and its large span across latitudinal and salinity gradients have been a challenge to the scientific community aiming to produce approaches for integrating information from a broad range of indicators. The Nested Environmental status Assessment Tool (NEAT), developed for the integrated assessment of the status of marine waters, was applied to 10 marine ecosystems to test its applicability and compare biodiversity assessments across the four European regional seas. We evaluate the assessment results as well as the assessment designs of the 10 cases, and how the assessment design, particularly the choices made regarding the area and indicator selection, affected the results. The results show that only 2 out of the 10 case study areas show more than 50% probability of being in good status in respect of biodiversity. No strong pattern among the ecosystem components across the case study areas could be detected, but marine mammals, birds, and benthic vegetation indicators tended to indicate poor status while zooplankton indicators indicated good status when included into the assessment. The analysis shows that the assessment design, including the selection of indicators, their target values, geographical resolution

Edited by: *Michael Elliott, University of Hull, UK*

#### Reviewed by:

*Jan Marcin Weslawski, Institute of Oceanology of the Polish Academy of Sciences, Poland Bernardo Antonio Perez Da Gama, Federal Fluminense University, Brazil*

#### \*Correspondence:

*Laura Uusitalo laura.uusitalo@ymparisto.fi; laura.uusitalo@iki.fi*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *08 June 2016* Accepted: *19 August 2016* Published: *06 September 2016*

#### Citation:

*Uusitalo L, Blanchet H, Andersen JH, Beauchard O, Berg T, Bianchelli S, Cantafaro A, Carstensen J, Carugati L, Cochrane S, Danovaro R, Heiskanen A-S, Karvinen V, Moncheva S, Murray C, Neto JM, Nygård H, Pantazi M, Papadopoulou N, Simboura N, Srebalien ˙ e G, Uyarra MC and Borja A ˙ (2016) Indicator-Based Assessment of Marine Biological Diversity–Lessons from 10 Case Studies across the European Seas. Front. Mar. Sci. 3:159. doi: 10.3389/fmars.2016.00159* and habitats to be assessed, has potentially a high impact on the result, and the assessment structure needs to be understood in order to make an informed assessment. Moreover, recommendations are provided for the best practice of using NEAT for marine status assessments.

Keywords: biodiversity, assessment tool, MSFD, environmental status, spatial aggregation, integration, indicator sensitivity

#### INTRODUCTION

Biological diversity is widely recognized as one of the cornerstones of healthy ecosystems (e.g., Worm et al., 2006). Diversity may safeguard ecosystems against undesired regime shifts (Folke et al., 2004) and guarantee the continued delivery of ecosystem goods and services (Duarte, 2000; Beaumont et al., 2007). The need to maintain biodiversity is also recognized by international legislation (e.g., Convention of Biological Diversity; UNEP, 1992); to European Union (EU) level, the Marine Strategy Framework Directive (MSFD; European Union, 2008) requires its member states to assess the status of marine biodiversity and take action to guarantee that it remains at, or is restored to, Good Environmental Status (GES). A definition of what can be interpreted as good status can be consulted in Borja et al. (2013).

In order to conduct an assessment of status, and to determine the effectiveness of any implemented remedial measures, we need a clear definition of biodiversity and a unified approach for its assessment. In the marine assessments like MSFD, biodiversity is defined on the level of species, communities, habitats, and ecosystems, as well as in the genetic level (Cochrane et al., 2010). Indicators that show the ecosystem response to human pressures form the basis of the tool kit with which we can describe environmental status (Borja et al., 2016). Based on qualitative environmental objectives, targets are set for each indicator which allow policy makers to implement management measures should these not be reached (Borja et al., 2012).

One of the challenges faced during the first round of MSFD initial assessments is the diverging data availability for biodiversity across highly variable systems, but yet an overarching need to conduct compatible assessments across European regional seas (Hummel et al., 2015). European marine ecosystems comprise a complexity and variability both in space and time, ranging from fully saline systems such as in Mediterranean and Atlantic waters to the brackish Baltic Sea, and exposed open water systems such as in the northern Norwegian and Barents seas to fully enclosed systems such as the Black Sea. The levels of available knowledge and data within these systems vary, as well as the biological parameters and indicators used for assessments (Hummel et al., 2015).

The conclusions of the European Commission, in their evaluation of the EU member states' reports on the initial assessment carried out in 2010–2012 was that there is an apparent lack of coherence and comparability in the indicators used and in the final evaluation of the overall status, between the countries and within all regional seas (Palialexis et al., 2014). Therefore, there is an urgent need for coherent frameworks and methodologies to allow consistent approach in biodiversity status assessment across the European Regional Seas. This would also be needed in order to allow coherence in the biodiversity assessments for the EU Birds and Habitats directives and the EU Biodiversity Strategy 2020.

While we could argue that we cannot compare studies if we do not have directly comparable datasets, in practice this is rarely possible, and certainly not at large spatial scales, or involving multiple research institutes and member states. Since there is no single way of describing biodiversity that fits all purposes, and since regional seas have intrinsic differences, we need a pragmatic selection of indicators which are appropriate to the specific questions asked, as well as a flexible and transparent indicator-based tool for assessment of biodiversity status. There is a large number of operational indicators, which have been used to describe the status in different types of aquatic systems (Birk et al., 2012; Borja et al., 2016). As biological diversity is multifaceted, including different taxonomic and functional groups, it cannot be expressed with a single indicator. Consequently, sets of different indicators are needed to cover the broad aspects of biological diversity and it is their combination into a single assessment that becomes a challenge (Borja et al., 2014; Probst and Lynam, 2016). In order to obtain a single overall assessment value, or conclusion, the results of the multiple indicators used in the assessment need to be aggregated, depending on the purpose of the assessment; e.g., if the aim is to inform different stakeholders and to set overall targets for the improvement of the marine environment, or depending on the assessment scale (Borja et al., 2014). Clear and transparent aggregation and integration rules are needed to interpret indicator information onto an environmental status assessment (see Borja et al., 2014 for a review on integration methods).

A variety of assessment tools enabling the integration of indicators already exists (see e.g., HELCOM, 2009a; Andersen et al., 2014; Borja et al., 2016). However, only few of them have treated biological diversity in a comprehensive way, have been tested broadly (i.e., outside the region in which they have been developed), or consider the complexity at an adequate level of detail for the spatial scale for which they are applied. To overcome these issues, in the context of the EU funded project DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing GES), the Nested Environmental status Assessment Tool (NEAT; Berg et al., 2016; Borja et al., 2016) has been developed to assess biodiversity status of marine waters under the MSFD. NEAT uses a combination of high-level integration of habitats and spatial units, and averaging approach (Borja et al., 2014), allowing for specification on structural and spatial levels, applicable to any geographical scale.

In this contribution NEAT is applied to the assessment of marine biological diversity in 10 different case studies distributed across the European regional seas (**Figure 1**). The assessment results are discussed, but the main focus of the paper is on: (i) analyzing the outcome of these assessments in light of the practical choices that have to be made to apply this tool, and (ii) proposing best practices for marine biological diversity assessment using this tool.

### MATERIALS AND METHODS

#### Case Study Areas

The case study areas were selected to represent a wide range of marine systems (**Figure 1**), with different climatic and hydrographic characteristics as well as exposure to different human activities and management challenges (**Table 1**). These areas represent a wide range of marine biogeographical areas from subtropical waters to temperate and Arctic, covering the four European regional seas (i.e., Mediterranean, Atlantic, Black, and Baltic Seas). The surface areas of these case studies varied from <3000 km<sup>2</sup> in Saronikos Gulf (Greece) to >820,000 km<sup>2</sup> in the Barents Sea (Norway; **Table 1**). Detailed descriptions of the case study areas, with relevant references, can be found in Supplementary Material (S1–S10).

## NEAT

NEAT is a structured, hierarchical tool for making marine status assessments (Berg et al., 2016; Borja et al., 2016), and freely available at www.devotes-project.eu/neat. In NEAT, the study area can be divided into hierarchical spatial assessment units


*(Continued)*

**442**

Frontiers in Marine Science | www.frontiersin.org September 2016 | Volume 3 | Article 159 |

TABLE 1 |

Characteristics

 of the case study areas.


TABLE

1


Continued


TABLE

1


Continued

(SAU) and habitat types (HBT); e.g., SAU "archipelago zone" could include "inner archipelago" and "outer archipelago" as lower-level SAUs, and they, in turn, could include, e.g., water bodies as yet lower-level SAUs. Similarly, the HBT "seafloor" could include HBTs "soft bottom" and "hard bottom," which again could be further sub-divided (**Figure 2**). NEAT classifies the status of each SAU based on indicators that have been defined for that SAU; if one SAU has indicators describing different HBTs, the status of each HBT within a SAU is assessed first, and each HBT is then given equal weight in assessing the status of the SAU. The overall assessment is an average of the SAUs, weighted by their surface areas (km<sup>2</sup> ). Other weighting schemes can be applied, if desired.

Each indicator must be explicitly linked to a SAU and a HBT the same indicator, e.g., "the maximum depth of seaweed," can be included multiple times for multiple SAUs and HBTs if it has been assessed for multiple areas. These instances of indicators are called "indicator values" in this paper, while the indicators describing a certain ecological concept, e.g., the growth depth of a macrophyte species, or the reproduction rate of a bird species, are called "unique indicators."

In order to aggregate indicators by weighted average, it is necessary to transform all indicators to a common scale. In NEAT, indicators are transformed into values that range from 0 to 1 using a continuous piecewise linear function. On this scale, the value of 0.6 corresponds to the boundary between good (>0.6) and not good (<0.6) status. Transformation to this scale is defined by specifying the values of the indicator in the original measurement scale, which corresponds to the transformed values of 0, 0.2, 0.4, 0.6, 0.8, and 1.0. Though the transformation function is piecewise linear, the definition of 5 segments allows a reasonable approximation to non-linear functions. These five segments are also used here for illustrative purposes, and they are called bad/poor/moderate/good/high classes, although it is recognized that the boundary between GES and non-GES lies between the "moderate" and "good" classes.

#### Indicator Selection and Specification

The indicators used for this assessment represent the best available data and expertise for the six biological descriptors of the MSFD [i.e., D1 (biodiversity), D2 (non-indigenous species), D3 (commercially important species), D4 (food webs), D5 (eutrophication), and D6 (sea floor integrity)] in each case study area. These indicators include the national and regional indicators used for the MSFD assessment, and indicators derived from scientific literature and expertise. They have been selected to be representative of various biodiversity components, habitats, and geographical areas relevant for each case study area; however it is possible that no indicators exist to be used for some relevant components. The list of indicators included in each case study is available in Supplementary Material S11.

Each indicator is associated to an ecosystem component class that describes the ecosystem component that the indicator describes. In this study, 12 ecosystem components were defined in order to accommodate all indicators used in all of the case studies. These components were phytoplankton, zooplankton, fish, reptiles, marine mammals, birds, benthic fauna, benthic vegetation, pelagic fauna (composite indicators consisting of data from multiple pelagic fauna groups), all taxa (composite indicators consisting of data from multiple taxa), benthic habitat, and water column habitat. The latter two components gathered indicators related to physico-chemical conditions of the habitat, necessary to maintain life (e.g., oxygen or nutrients), whilst the "all taxa," benthic fauna, and pelagic fauna groups included composite indicators encompassing many species groups; the other nine ecosystem components were taxonomic groups.

### Biodiversity Status

The status of the biological diversity was assessed for each case study area using NEAT. The analysis provides an overall assessment for each case study area and a separate assessment for each of the ecosystem components included in the assessment. The final value has an associated uncertainty value, which is the probability of being in a determinate class status (GES/non-GES). This uncertainty was determined by the standard error linked to the indicator values (Carstensen and Lindegarth, 2016).

# Evaluation of Assessment Design and Its Effects on the Status Assessment

The application of NEAT to a broad range of marine regions provides an opportunity to test and compare the NEAT assessment approaches and evaluate the consequences of design choice for the general environmental status assessment. How the available data are combined within the tool might have consequences on the results of the status assessment of biodiversity (Borja et al., 2014; Probst and Lynam, 2016). Therefore, one of our aims is to evaluate the consequences of the way the assessment was designed on the general assessment result.

NEAT gives a framework to organize the assessment, but it does not prescribe the number of assessment components, i.e., indicators, SAUs, HBTs, or ecosystem components to be used in an assessment. The user has the option to organize the different components of NEAT depending on the case, e.g., the morphological characteristics of the area, availability and resolution of data, and how the selected local indicators are defined.

In order to describe the assessment design, the following key components were summarized for each case study: (i) the total number of SAUs and how many hierarchical SAU levels there are, (ii) the total number of HBTs and their hierarchical levels, (iii) the number of ecosystem components covered by the indicators, (iv) the number of unique indicators (i.e., not repetition of the same indicator on a different spatial unit), as well as (v) the quantity of data, defined as the number of different indicator values (e.g., if the same indicator is defined separately for five different SAUs, they would comprise five indicator values).

NEAT assigns weights to the indicators based on the SAU and HBT that they represent (see Section Evaluation of the Assessment Results). The SAUs are weighted according to their surface area and the HBTs are weighed equally within a SAU. Therefore, the indicator values contribute to the assessment with different weights, the highest weight being assigned to an indicator representing a large SAU with a small number of indicators, and within it a HBT with a small number of indicators. The relative weights of the indicator values were used to identify the indicators that contribute 90% of the weight of the final assessment. In addition, the relative weight of each ecosystem component in each case study assessment was calculated. These summary statistics highlighted differences in aggregating information among case studies.

To test the sensitivity of the case study assessments to the selection and number of indicator values, a sensitivity analysis was performed by running the assessment using randomly selected indicator values. The number of indicator values included into the assessment varied from 1 to the maximum number of indicators in the case study minus one. This process was repeated 100 times for each number of indicator values. For example, take a case study with 120 indicator values. First, one random indicator value is selected and the assessment is done using only that indicator. This procedure is repeated 100 times. Then, two indicator values are picked at random, and the assessment is run using them; this again is repeated 100 times. This procedure is repeated for all numbers of indicator values up to 119. This results in a large number of values whose divergence can be analyzed to see if any patterns can be identified.

# RESULTS

#### Assessment Design

The number of SAUs as well as how many hierarchical levels were used in these varied widely between the case studies. The number of SAUs included in the Gulf of Finland and Portugal continental sub-division cases were much higher (>60) than in all other case studies which included, on average, 9 different SAUs. Excluding these two case studies, larger areas were usually assessed using more SAUs. The number of hierarchical SAU levels varied between 1 and 5, but in 7 out of 10 cases, there were 3 or 4 levels (**Table 2**, **Figure 2**). The total number of HBTs included in the assessment varied between 3 and 9, and 9 out of 10 case studies had 2 or 3 hierarchical HBT levels (**Table 2**).

Not all SAUs necessarily included all habitat types, and indicators or data may not exist for all defined HBT types for each SAU. The number of SAU-HBT combinations that were assessed by at least one indicator value, varied between 6 and 132 (**Table 2**).

The number of ecosystem components included in the analyses varied between 5 and 9, with an average of 7.3 (**Table 2**). It has to be noted that all ecosystem components identified in this study were not applicable to all areas; an example being reptiles that do not occur in most of the study sites.

The number of unique indicators applied in each case study area varied between 11 and 116 (**Table 2**, Supplementary Material S11). The number of indicator values varied greatly with 466


TABLE 2 | Synthesis of the structure used by the different case studies for the nested assessment.

*SAU, spatial assessment unit; HBT, habitats. The case studies are ordered according to their latitude.*

values at the higher end in Portugal continental sub-division and between 20 and 200 values in all other case studies (**Table 2**).

#### Biological Diversity Status

The summary of the test NEAT assessments of the 10 case study areas is presented in **Figure 3**. The assessment resulted in GES for the Basque EEZ and the Barents Sea-Lofoten, with 100 and 66% confidence, respectively, the remaining eight case studies presented non-GES (i.e., bad, poor, or moderate; **Figure 3**). Lithuanian coast has the potential for being in GES, but with a low confidence of 20% (**Figure 3**). For the other case studies, this probability of achieving GES was <1% (**Figure 3**).

The different ecosystem components showed different status in the case study areas (**Figure 4**). No strong pattern among the ecosystem components could be detected, but some commonalities were found: Indicators based on marine mammals generally indicated degraded situation in 6 cases out of 7 (**Figure 4**). When included, birds and benthic vegetation indicators as well as water column indicators of physicochemical status also indicated degraded situation in 5 cases out of 7. Indicators encompassing several ecosystem components ("AT," on **Figure 4**) always indicated degraded situations. On the other hand, indicators of benthic habitats' physico-chemical status and of zooplankton community status indicated GES when they were included in the assessment (**Table 1**, **Figure 4**).

# Relative Contribution of Indicator Values and Biodiversity Components

The indicator values contributed differently to the final assessment result (**Figure 5**); indicator values defined for larger SAUs tend to have more weight, particularly if there are only few indicators defined for these SAUs. In 7 out of the 10 case studies, <10 indicator values already contributed to more than 50% of the final assessment result. For 9 case studies, <50 indicator values contributed to >90% of the final assessment. This 90% of the final assessment was reached with <20 indicator values in five case studies (**Figure 5**). The five indicator values that made the highest contribution to the final assessments of each case study are listed in **Table 3**. These indicator values were dominated by mammal, bird, fish, and benthic fauna indicators.

The 12 different ecosystem components' contribution to the final assessment result did not correspond to the number of indicator values defined for each component (**Table 4**). For example, most case studies had a large proportion of benthic fauna indicator values (average: 22.4% of indicators values), which ultimately did not reflect proportionally in the final assessment (average contribution: 11.7%). In contrast, the proportion of fish and marine mammals indicator values were lower, but these components contributed to a higher proportion of the final assessment. In five case studies (i.e., Barents-Lofoten, Gulf of Finland, Dutch North Sea, Saronikos Gulf, and Adriatic Sea), "Benthic fauna" was the component with the highest proportion of indicator values (**Table 4**); the other five case studies each had a different component with the highest number of indicator values. However, in none of the case studies, benthic fauna was the component with highest contribution to the final assessment (**Table 4**); in five (i.e., Gulf of Finland, Dutch North Sea, Basque coast, Portuguese continental subdivision, and Black Sea coast) and two case studies (i.e., Barents Sea–Lofoten and Adriatic Sea) respectively, fish and mammals were the components carrying the highest weight to the final assessment (**Table 4**). However, other ecosystem components, that overall did not contribute to many case study assessments, were very relevant for specific case studies (e.g., the composite

group "all taxa" in the Saronikos Gulf and benthic habitat in the Lithuanian coast).

#### Sensitivity Analysis

The sensitivity analysis shows that there are major differences in how much the result varies if only a subset of the indicator values is included in the assessment (**Figure 6**). For example, if only a small number (close to 0) indicators were included, the assessment results in all studies could be anywhere between high and bad status, except in Barents Sea and Portuguese continental subdivision, where they could range from poor to high status. As more indicator values are added, the range of outcomes narrows down. However, how steeply that happens when indicator values are added varies between the case study areas (**Figure 6**).

# DISCUSSION

The current NEAT-based assessment demonstrates a largescale marine biodiversity assessment, providing a feasible solution to the apparent problem pointed out by the European Commission, in their evaluation of the EU member states' reports on the MSFD initial assessments carried out in 2010– 2012 (Palialexis et al., 2014). This problem was the apparent lack of coherence and comparability in indicators used and in the final evaluation of the overall status between the countries and within all regional seas (Palialexis et al., 2014). Despite the available guidance and Commission Decision (European Union, 2010) on GES descriptors, criteria and indicators, the overall picture in assessments was patchy and non-coherent (European Commission, 2014). The use of NEAT, and its validation in different regional seas and case study areas, is a crucial contribution from the DEVOTES project to provide a harmonized approach and methodology for a coherent and comparable environmental status assessment across the European regional seas. It also shows that although the regional seas have different characteristics and human pressures impacting those (Claudet and Fraschetti, 2010; Micheli et al., 2013a; Andersen et al., 2015), a coherent assessment framework can be employed to evaluate differences in the environmental status and the ecological components that are impacted by different pressures.

The study and the comparison of the case studies brought into light several issues that need attention in order to improve the coherent and comparable "biodiversity status" assessments of the European regional seas. These issues are related to the data and indicator availability, how the assessments are structured, how the integrative assessment should be structured, and how this structure should be taken into account when defining the spatial resolution and indicator selection of the assessments. The current study revealed that while these assessments could be carried out, there are two major problems in achieving the objectives of the MSFD assessments: (i) there are still multiple gaps in the availability and coverage of indicators in the various areas, and (ii) comparability of the status assessments across different regions would benefit from a more unified assessment framework, even if indicators suitable for each area remained different. NEAT provides a general framework that could be accompanied with guidelines for the selection of SAUs, HBTs, and indicators.

Each of the case studies was initially designed with the best available selection of spatial units, habitats, and indicators, adhering to the NEAT methodology but without specific guidelines for the indicator selection, target level setting, etc. This situation resembles the situation where the new users would start using NEAT on their area. For the purposes of this study, the assessments were evaluated and harmonized to some degree, e.g., if the same indicator appeared in multiple case studies, it was ensured that it was associated to the same biodiversity component (e.g., chlorophyll a levels would be assigned to phytoplankton). Despite this harmonizing, there were major differences in how the case studies were constructed in terms of spatial resolution, habitats, and indicator definition. The current assessment is based on best available data and evaluation of the experts participating within this exercise, and the biodiversity status results of this study should be considered as indicative, not definitive.

The indicators selected for the assessments are designed or adapted for each area separately, including the geographical and habitat specification and the target level, i.e., which values are considered good and which less than good in any given area and habitat. This means that the "good" status is scaled according to the area: In areas with a naturally low biodiversity, lower

biodiversity is also considered "good" than in areas with naturally high diversity. This makes the assessment relevant for each area, and the result must be interpreted to be in relation to undisturbed condition of that area rather than in absolute terms of diversity.

According to a categorization of rules or methods for combining or aggregating indicators or criteria within a given descriptor (Prins et al., 2013; Borja et al., 2014), NEAT is classified as a high-level integration method which reduces the risks associated to the "one out, all out" principle of the Water Framework Directive approach (Borja and Rodríguez, 2010) while giving an overall and specific (to descriptors and components) assessment.

According to the relevant guidance document for the MSFD (Prins et al., 2013), the spatial scales are not the same for all indicators within the biodiversity descriptor, where depending on the species or habitat a different spatial scale may be used. It is also recommended to address uncertainties and assess confidence of the classification result (as a secondary assessment). In our study, the NEAT software treats equally all assessment elements assigning equal weights, but gives more weight in cases of larger spatial coverage, with higher data representativeness, in that way

incorporating the spatial scales issue and the confidence level into the assessment. This could be the reason for which some ecosystem components (e.g., seabirds, mammals, and fishes) have more weight in the final assessment, since they are normally assessed at large scale spatial areas, which have more weight when aggregating (e.g., Saronikos gulf). However, NEAT also includes the possibility to weight indicators differently.

#### Implications of the Assessment Design

Most of the case study areas lacked indicators regarding one or several biodiversity components and habitats (**Table 1**, **Figure 4**), even those that were deemed important in the area. The lack of indicators stemmed either from lack of monitoring data regarding the area or biological diversity component (e.g., birds, reptiles, pelagic fauna), or from obstacles in the indicator development, including the lack of expert time to develop indicators, or insufficient knowledge about the target levels due to lack of long-term or reference condition data (Hummel et al., 2015). In some cases, more basic ecological research is needed in order to understand the ecological processes well enough to develop indicators. In fact, most of the assessments undertaken until now by member states is more qualitative than quantitative (Hummel et al., 2015), representing a challenge for the assessment.

The habitats and biodiversity components for which no indicators are available potentially affect the final assessment result. It is entirely possible that adding even one indicator that would represent a poorly-represented, large area or habitat, would change the overall assessment for better or for worse. Therefore, in order to make a reliable assessment of the status of the biological diversity, the critical gaps in each assessment case need to be evaluated for their potential to affect the overall result. If such highleverage gaps exist, the assessment result must be taken with caution.

Different indicator values and spatial assessment units had varying weights in the final assessment result in all of the cases (**Table 3**, **Figure 2**). The differences in the indicator value weights stem from the fact that the default NEAT assessment first assesses the result for each SAU, giving equal weight to each HBT with similar hierarchy, and combines these SAUs hierarchically so that each SAU is given weight according to its area. Therefore, if a SAU has a large surface area and only a small number of indicators per one or several of its habitat types, these indicator values end up contributing strongly to the final assessment.

This emphasizes the importance of the balanced nature of the indicator set, and particularly the reliable assessment of indicators that are used to assess the status of large areas, and particularly their habitats with only few indicators (Feary et al., 2014). Therefore, particular attention should be paid to both the observed value, the boundary values between the classes, and the uncertainty estimation of these most influential indicator values.

The fact that the SAUs are weighted according to their surface area in the default mode of NEAT also emphasizes the need for careful consideration of the definition of the SAUs. Ideally, the SAUs should be defined in the manner that an indicator value defined for a SAU can be expected to reasonably represent all of the SAU. On the other hand, if the assessment area is split into several sub-SAUs and only a fraction of them actually has indicator data, their value will be generalized to represent the whole super-area in the hierarchical assessment anyway.

#### TABLE 3 | List of the top-five indicator values contributing the most to the overall assessment for each case study.


*(Continued)*

#### TABLE 3 | Continued


*In case of equal contribution of several indicator values, all the indicator values are given. The contribution to the overall assessment (in %) of each indicator value is given. Numerical values are rounded. ES100, expected number of species in 100 individuals; AMBI, AZTI's Marine Biotic Index; BQI, Benthic Quality Index; DKI, Danish Index; CIMPAL, Cumulative IMPacts of invasive ALien species; M-AMBI, multivariate AMBI; GES, good environmental status.*

In NEAT, it is possible to weight the SAUs according to their perceived ecological relevance instead of their surface area; for example, biodiversity hotspots, important reproduction areas, marine protected areas, etc., could be given a higher weight than their area alone would imply. In this study, this option was not used in any of the case studies.

Uncertainty of the results is assessed based on Monte Carlo simulations, using the observed value as mean and the standard error value as the standard deviations, assuming a Gaussian distribution (Carstensen and Lindegarth, 2016). Based on these simulations, NEAT determines how often the sampled value falls into each of the five classes, and this distribution is reported. Therefore, the standard error values assigned to the indicators play a major role in the uncertainty associated with the final assessment result. This emphasizes the importance of careful evaluation of the standard deviation, particularly with indicators that have a high weight in the assessment.

#### Evaluation of the Assessment Results

There are other tools to assess the status of marine systems, e.g., the Ocean Health Index (OHI; Halpern et al., 2012). This index has different concept and a much broader spatial scale, and a comparison between NEAT and OHI results (BD values presented in Table S6 in Selig et al., 2013) shows that the results are quite different (**Table 5**).


The OHI tends to give a more reduced range of status values (74–97) than those provided by NEAT (0.37–0.69) for these areas. The OHI does not provide a GES/non-GES status, but in general provides higher values than those by NEAT. The OHI study (Selig et al., 2013) has been applied globally, and includes a large variety of worldwide cases with great differences in setting and problems. In that context, e.g., the Mediterranean and the Baltic Sea seem to be in a (seemingly more homogenous) better state than e.g., waters around Africa or Indonesia and Philippines.

An interesting observation is that there is a negative rather than a positive correlation between these results, and those areas ranked low in NEAT (such as the Gulf of Finland and Kattegat) get high scores in OHI, while the best-scoring area in NEAT (Basque EEZ) gets lowest score in OHI (**Table 5**). This discrepancy is partly due to the fact that the OHI scores are given by country, thus covering larger areas than the case studies assessed here with NEAT. Therefore, the local status of a case study area may be masked by the results from the rest of the country in OHI. The NEAT results are reported here for the entirety of each of the case study areas, but where the case study area includes smaller SAUs, the results can be viewed for each of them separately as well, yielding even a more detailed geographical resolution.

Another factor possibly contributing to this discrepancy is the use of different indicators; the OHI assessment used publically available data with little local/regional detail, which can vary the final assessment when applying to regional scales (Halpern et al., 2014), while the current NEAT assessment used indicators specifically designed for marine status assessment. The species scores of OHI focused on the extinction risk of marine species (Selig et al., 2013), while the indicators in the NEAT assessments included a wider spectrum of indicators of species status. The OHI habitat scores were based on condition estimates of mangroves, coral reefs, seagrass beds, salt marshes, sea ice, and subtidal soft-bottom (Selig et al., 2013) while the NEAT assessments were tailored for each area.

The NEAT assessment results were in most cases in line with previous regional/local assessments, understanding, or known pressure gradients (**Table 1**, **Figure 4**). For example, The Baltic Sea biodiversity has been assessed by HELCOM (2009a, 2010) to be in poor to bad status in all of the three Baltic case study areas included in this analysis (Gulf of Finland, Lithuanian marine waters, Kattegat), being similar to the NEAT results but not to the OHI assessments. The difference between the NEAT and OHI results in these cases is probably largely due to eutrophication, which is documented to be major pressure threatening the ecosystem functioning of the Baltic Sea (HELCOM, 2009b, 2010). While it is reflected in the status of phytoplankton and water column habitats, and also affects the higher trophic levels of the food web (Österblom et al., 2007) and the seafloor (Karlson et al., 2002), it is not likely to be strongly reflected in the extinction threat of marine species (used in OHI), although it does affect the habitat scores, particularly seagrasses (Table S1 in Selig et al., 2013).Another factor affecting the discrepancy in the case of Finland is that the Gulf of Finland area has poorer biodiversity status than the Finnish marine waters on average (HELCOM, 2010).

continental subdivision study included 466 indicator values in total.

In the North Sea, fishing is considered the main pressure, and the results show fish to be the ecosystem component in poorest status; the other assessed ecosystem components (birds, mammals, benthic fauna, and phytoplankton) were assessed to be in GES, with the exception of zooplankton that showed sub-GES (moderate) status (**Figure 4**). The Black Sea Coast case results obtained in this study also corresponded very well to known pressure gradients, such as nutrient enrichment affecting the status of the plankton community (**Figure 4**). Phytoplankton and benthic vegetation assessments correspond to category "poor" in the Varna Bay itself (Dencheva and Doncheva, 2014; Moncheva et al., 2015) as the most affected by anthropogenic pressure among the BSC sub-SAUs (Shtereva et al., 2012). The lowest benthic fauna score is also found there, which is fully in compliance with recently published results (National Report on the State and Protection of the Environment in Bulgaria, 2014). Similarly, the Basque area, which was previously assessed as being in good status, using a different methodology (Borja et al., 2011) also results in good status after applying NEAT; only mammals were assessed to be in sub-GES status (**Figure 4**).

In Saronikos Gulf the assessment results correspond to the ecological status categorization according to the WFD which is

TABLE 5 | Comparison of the biodiversity assessments obtained using the Ocean Health Index (OHI; data from Selig et al., 2013) and the Nested Environmental status Assessment Tool (NEAT) (this study) in the countries for which NEAT has case studies.


poor in the sewage outfall area and moderate in the inner central gulf (Simboura et al., 2014, 2015, 2016). Aliens, fish including threatened sharks, and mammals contributed to the moderate status seen for the outer Saronikos and overall Saronikos. In general, the respective assessment results, although not definitive, are in line with pertinent studies (Frantzis, 2009; Katsanevakis et al., 2013; Papaconstantinou, 2014; Vasilakopoulos et al., 2015; Zeneto<sup>6</sup> et al., 2015; Simboura et al., 2016) regarding the Greek marine waters. The Saronikos Gulf result obtained in this analysis was lower than the OHI assessment of the Greek waters, which was to be expected, as the Gulf is intensely exploited.

Results from the Norwegian part of the Barents sea indicated a general good status, which is in accordance with indicators of fish status on exploited large marine ecosystems (Kleisner et al., 2014; Coll et al., 2015), the report on the Barents Sea management plan (Sunnana et al., 2010) and the work from Certain et al. (2011). Nevertheless, several indicators indicated potentially degraded situations both in the coastal area and in the area of seasonal ice presence: (1) Along northern Norway coast, the current extent of kelp forest, an important component of fjords ecosystem and coastal landscape, cannot be considered as good in northern Norway. Kelp forests along the Norwegian and Russian coast were indeed dramatically grazed during the early 1970s and replaced by barren grounds dominated by sea urchins (Norderhaug and Christie, 2009). Though a progressive northward recovery of kelp forests extent is observed, its recovery status is still partial in northern Norway (Sivertsen, 2006; Rinde et al., 2014). (2) In northernmost part of the Barents sea, sea-ice extent is undergoing a particularly dramatic decrease (Parkinson et al., 1999) with a significant decrease rate of −3.5% per decade of winter ice extent (Sorteberg and Kvingedal, 2006) as a response to climate warming (Boitsov et al., 2014). This dramatic loss of habitat has consequences on the associated communities (Kovacs et al., 2011) as well as in the functioning of the Barents sea ecosystem as a whole (Wassmann et al., 2006). The growing evidence of impacts of climate change on this area rises the issue of exogenic unmanaged pressures on this system and the issue of shifting baselines for the definition of target values. In addition, there are still no indicators of the impact of trawling activities included in this assessment (see however Jørgensen et al., 2016).

For the Portuguese coast, the initial assessment officially provided in the scope of the MSFD (MAMAOT, 2012), presented a general environmental quality status higher than the NEAT results calculated in this study. This may be partly due to the fact that the present assessment did not include some special areas with a higher degree of protection (such as Berlengas' Marine Reserve and Professor Luiz Saldanha's Marine Park or Goringe Seafloor). These areas, which have restricted access by the public, are important for marine high trophic level species (e.g., marine birds, mammals), some of which were not included in the present assessment. Due to inconsistencies in the data (now being improved by projects such as MARPRO— Conservation of Marine Protected Species in mainland Portugal, http://marprolife.org), marine mammals, reptiles and benthic vegetation were not included in the current NEAT assessment, which may also contribute to the lower environmental quality results achieved by NEAT. The higher result reported by the OHI may be related to the methodology used for the scores' calculation, and may reflect more specifically the trend than the present environmental status.

An exception to the good correspondence between the current and previous assessments is the Adriatic Sea, where the assessment provided by NEAT appears too low considering the current trends, also reported in the scientific literature, and available information from expert opinions (Coll et al., 2010; Bastari et al., 2016). Despite the historical impacts on this shallow water basin, the Adriatic Sea is still characterized by a wide diversity of habitats, including rocky and soft bottoms, large estuaries and lagoons, seagrass meadows and in, its southern part, also deep-water environments. The habitat richness is reflected by a high biodiversity (Coll et al., 2012; Micheli et al., 2013b), with approximately 49% of the species described for the Mediterranean Sea (Boudouresque et al., 2009; UNEP, 2015) and a variety of endemic species (e.g., 18% of the endemic fish species of the Mediterranean; UNEP/MAP-RAC/SPA, 2015). Human activities and multiple stressors, and in particular bottom trawling, hydraulic dredging and habitat loss, are certainly still impacting the Adriatic Sea (Micheli et al., 2013a; Pusceddu et al., 2014). However, the overall environmental condition is not worsening with respect to the past decade. Eutrophication and dystrophic crises, related to the high nutrient discharge from the Po River combined with an alteration in water circulation, have caused hypoxia, anoxia and massive mucilage events, with consequent mortality of the benthic organisms, but the frequency of these events decreased significantly (or even disappeared) in the last decade (Degobbis et al., 2000; Danovaro et al., 2009). Thus, we hypothesize that the assessment of the environmental status obtained by using NEAT can be affected by the number and typology of data included in the specific exercise. An improvement of the number and type of the biological indicators (e.g., species or ecosystem functioning) could be crucial to obtain a more realistic classification of the marine environmental health of the Adriatic Sea.

Birds and mammals were found to be in poor status in many of the case study areas. This reflects the fact that seabirds are indeed considered as more threatened than any other comparable groups of bird species in general and display a faster trend of decline than other bird species during the last decades (Croxall et al., 2012). In addition, using IUCN Red list categories, it has been evidenced that, among seabirds, pelagic species of seabirds are disproportionately more threatened than coastal resident or coastal non-breeding visitor species (Croxall et al., 2012). Pelagic seabirds are particularly sensitive to disturbance as most species lay only a single egg, adults do not reproduce every year and usually reproduce several years after reaching sexual maturity (Furness and Camphuysen, 1997). Most seabird species display very large home range and thus integrate the state of the environment and impacts of pressures over larger scale.

The conservation status of marine mammals is of particular concern with an estimated proportion of threatened species ranging worldwide between 23 and 61% of species (Schipper et al., 2008). The North Atlantic region, which includes several of the cases studied here, is one of the areas where the proportion of threatened marine mammals is the highest, as shown by the low quality values in Barents Sea, Kattegat, and Basque case studies (**Figure 4**). The main reported threats explaining the bad status of marine mammals are a long history of harvesting, accidental mortalities through bycatch and collisions with vessel as well as a very large panel of pollutions (from sound pollution to contaminants and marine debris) and climate change (Schipper et al., 2008). The sensitivity of these species to changes in their environment might be related to their very slow population dynamics, low densities in correlation with their large bodysize (Cardillo et al., 2008). Those life traits are also related with relatively large home range. As a consequence, indicators of marine mammals are usually measured over large scale, and they are difficult to monitor with precision, leading to higher uncertainty on many indicators (Taylor et al., 2007).

In two of the areas (Lithuania and Basque coasts), the indicator contributing the most to the final assessment was "the extent of the seabed significantly affected by human activities," which is a direct indicator of pressure. This is interesting since some authors (e.g., Borja et al., 2013) have supported the use of pressures instead of assessing the environmental status, if there are not enough indicators. This should be done under the premise that if an area has no obvious pressures then any changes in the area must be due to natural changes which are outside the control of management and vice versa.

#### Sensitivity Analysis

The sensitivity analysis results show differences among the case studies in terms of how many indicator values are needed before the assessment results will show approximately the same results regardless of which indicator values are selected into the assessment (**Figure 6**). This implies that there is no universally sufficient number of indicator values needed to make a reliable assessment, but that the number varies among case studies. No clear patterns could be found among the 10 cases evaluated in this study that would indicate a number of indicator values of biodiversity components that can be considered sufficient regardless of the case study and its structure.

The variation in the assessment result depends on the set of indicator values that is available for the assessment. If the indicator values are close to each other, i.e., all indicating similar status, the variation in the results is naturally smaller. In contrast, if the different indicator values indicate very different status, e.g., some areas or biodiversity components are in good status while others are in bad, this naturally incurs a larger variation when a subset of these variables are selected, as e.g., in the Gulf of Finland.

These observations lead to the conclusions that if there is variation among the status of the geographical or biodiversity components in the study area, all of them should be covered by indicators if possible. Particularly the inclusion of high-leverage indicator values, i.e., those that have high weight and whose value differs from the overall mean, can change the assessment result. Therefore, the careful evaluation of the value and class limits of these indicators should be a priority.

# CONCLUSIONS

The structured assessment forces us to critically evaluate the available indicator set in terms of ecological and spatial representativeness of each indicator. This framework highlights the gaps in the assessment as well as those parts that are wellrepresented by current monitoring and available indicators. This, in turn, helps in determining the best way to improve the quality of the assessment: (i) via developing additional indicators to fill in the gaps within the ecosystem approach (i.e., if not all the important trophic levels of key species/ groups are covered in the existing indicator set), (ii) working to determine the optimal SAU for different categories of indicators that are targeted to assess various trophic levels and functions in the food web, as well as the HBT classification for each area, and (iii) working toward improving specificity, robustness, and pressure relevance of the indicators and enabling estimation of their standard errors.

The development of NEAT and this extensive testing with 10 case studies in very different European marine areas offers insight both to the status of the marine waters and to the stateof-the art of the available indicator assemblages as well as the development needs of the marine biological diversity assessment. The application of the tool will make the improvement and harmonization needs of the assessments visible and pave the way toward a harmonized assessment across large geographical scales.

In conclusion, we propose the following recommendations for the best practice in performing the environmental status assessment using NEAT:


of size and hierarchy of the spatial assessment units as well as the definition of habitats.


# AUTHOR CONTRIBUTIONS

LU, HB, and AB conceived the paper. The following partners provided the case studies: HB, SC (Barents Sea–Lofoten), LU, VK, HN (Gulf on Finland), GS (Lithuanian marine waters), CM, JA, JC (Kattegat), OB (Dutch North Sea), AB, MU (Basque Coast), JN (Portuguese continental subdivision), SM (Black Sea Coast), NP, MP, NS (Saronikos Gulf), RD, LC, AC, SB (Adriatic Sea). LU wrote the first draft. TB and CM contributed to the

# REFERENCES


calculations and sensitivity analyses. LU, HB, AB, SC, AH, and MU contributed largely to the introduction and discussion. All authors contributed to the last draft and to the discussions.

# FUNDING

This manuscript is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), www.devotes-project.eu. MU is partially funded through the Spanish programme for Talent and Employability in R+D+I "Torres Quevedo." Moreover, the monitoring of Saronikos Gulf was financed by the Athens Water Supply and Sewerage Company (EYDAP SA).

# ACKNOWLEDGMENTS

The authors would like to thank DEVOTES Advisory Board members Paul Snelgrove and Simon Greenstreet for their constructive comments at the early phase of the work.

### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: http://journal.frontiersin.org/article/10.3389/fmars. 2016.00159

overview of biological methods to implement the water framework directive. Ecol. Indic. 18, 31–41. doi: 10.1016/j.ecolind.2011.10.009


management," in Proceedings of the Twelfth International Conference on the Mediterranean Coastal Environment MEDCOAST'15, ed E. Ozhan (Varna; Dalyan: Mediterranean Coastal Foundation), 351–362.


Naturstyrelsen (2012). Danmarks Havstrategi. Basisanalyse, 100 (In Danish).


stress factors in Eastern Mediterranean coastal waters. Ecol. Indic. 70, 89–105. doi: 10.1016/j.ecolind.2016.05.018


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Uusitalo, Blanchet, Andersen, Beauchard, Berg, Bianchelli, Cantafaro, Carstensen, Carugati, Cochrane, Danovaro, Heiskanen, Karvinen, Moncheva, Murray, Neto, Nygård, Pantazi, Papadopoulou, Simboura, Srebalien ˙ e, ˙ Uyarra and Borja. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# **Socio-economic aspects and management**

# Assessing Costs and Benefits of Measures to Achieve Good Environmental Status in European Regional Seas: Challenges, Opportunities, and Lessons Learnt

Tobias Börger 1, 2 \*, Stefanie Broszeit <sup>2</sup> , Heini Ahtiainen<sup>3</sup> , Jonathan P. Atkins <sup>4</sup> , Daryl Burdon<sup>5</sup> , Tiziana Luisetti <sup>6</sup> , Arantza Murillas <sup>7</sup> , Soile Oinonen<sup>8</sup> , Lucille Paltriguera<sup>6</sup> , Louise Roberts <sup>5</sup> , Maria C. Uyarra<sup>7</sup> and Melanie C. Austen<sup>2</sup>

*<sup>1</sup> Department of Geography and Sustainable Development, University of St Andrews, St Andrews, UK, <sup>2</sup> Plymouth Marine Laboratory, Plymouth, UK, <sup>3</sup> Natural Resources Institute Finland (Luke), Helsinki, Finland, <sup>4</sup> Hull University Business School, Hull, UK, <sup>5</sup> Institute of Estuarine & Coastal Studies, University of Hull, Hull, UK, <sup>6</sup> Centre for Environment, Fisheries, and Aquaculture Science, Lowestoft, UK, <sup>7</sup> AZTI Tecnalia, Pasaia, Spain, <sup>8</sup> Finnish Environment Institute (SYKE), Helsinki, Finland*

#### Edited by:

*Jacob Carstensen, Aarhus University, Denmark*

#### Reviewed by:

*Joana Patrício, Executive Agency for SMEs (EASME), Belgium Berit Hasler, Aarhus University, Denmark*

> \*Correspondence: *Tobias Börger tb79@st-andrews.ac.uk*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *20 June 2016* Accepted: *21 September 2016* Published: *07 October 2016*

#### Citation:

*Börger T, Broszeit S, Ahtiainen H, Atkins JP, Burdon D, Luisetti T, Murillas A, Oinonen S, Paltriguera L, Roberts L, Uyarra MC and Austen MC (2016) Assessing Costs and Benefits of Measures to Achieve Good Environmental Status in European Regional Seas: Challenges, Opportunities, and Lessons Learnt. Front. Mar. Sci. 3:192. doi: 10.3389/fmars.2016.00192* The EU Marine Strategy Framework Directive (MSFD) requires Member States to assess the costs and benefits of Programmes of Measures (PoMs) put in place to ensure that European marine waters achieve Good Environmental Status by 2020. An interdisciplinary approach is needed to carry out such an assessment whereby economic analysis is used to evaluate the outputs from ecological analysis that determines the expected effects of such management measures. This paper applies and tests an existing six-step approach to assess costs and benefits of management measures with potential to support the overall goal of the MSFD and discusses a range of ecological and economic analytical tools applicable to this task. Environmental cost-benefit analyses are considered for selected PoMs in three European case studies: Baltic Sea (Finland), East Coast Marine Plan area (UK), and the Bay of Biscay (Spain). These contrasting case studies are used to investigate the application of environmental cost-benefit analysis (CBA) including the challenges, opportunities and lessons learnt from using this approach. This paper demonstrates that there are opportunities in applying the six-step environmental CBA framework presented to assess the impact of PoMs. However, given demonstrated limitations of knowledge and data availability, application of other economic techniques should also be considered (although not applied here) to complement the more formal environmental CBA approach.

Keywords: MSFD, environmental cost-benefit analysis, benefit transfer, ecosystem services

# INTRODUCTION

The importance of biodiversity in providing ecosystem services and human wellbeing is globally recognized (MEA, 2005; Haines-Young and Potschin, 2010; Mace et al., 2012), with international agreements and legislation introduced to address biodiversity loss (e.g., Convention of Biological Diversity, 1992). For the 27 countries in the EU, there is legislation aimed at halting biodiversity loss and promoting the sustainable use of the ecosystem services that the natural environment supports (e.g., Common Fisheries Policy, Habitats and Birds Directives, Water Framework Directive). Regarding the marine environment, the EU Marine Strategy Framework Directive (MSFD, 2008/56/EC) was promulgated in 2008 as a means to ensure that European marine waters achieve Good Environmental Status (GES) by 2020. To this end, the MSFD requires Member States to define environmental targets and associated indicators and to develop and implement Programmes of Measures (PoMs) that will ensure the achievement of GES.

To comply with MSFD, all EU Member States undertook a baseline assessment of the current state of the marine environment in their jurisdictions in 2012. These assessments were carried out considering the 11 MSFD Descriptors for defining GES in marine waters (**Table 1**), with each Descriptor, having an associated set of criteria (29 in total) and indicators (56 in total; EC, 2010). During this initial assessment, several Member States defined reference conditions and targets that determine GES for these indicators (e.g., Anon, 2012). Prior to the implementation of the various PoMs the MSFD requires that each Member State undertakes an impact assessment, including environmental cost-benefit analysis (CBA), on any measure they are planning to implement to support the realization of GES (Article 13.3; EC, 2015).

In principle, environmental CBA can be used to quantify and compare all of the costs and benefits resulting from a particular policy measure in monetary terms (Boardman et al., 2006). To this end, positive and negative environmental, economic, and social impacts accruing to relevant stakeholders, including the general public, have to be assessed, quantified, and where possible valued monetarily. Even if certain effects cannot be valued in monetary terms environmental CBA requires at least a listing and acknowledgment of "all costs and benefits of a policy" (Hanley, 2001). The effects that can be monetized can then be compared in a partial CBA (EC, 2015). This analysis should include goods and services with market value, but also those which are not traded in markets and hence have no market prices (Hanley and Barbier, 2009; EC, 2015). Environmental CBA requires an interdisciplinary approach involving the collaboration of natural and social scientists (Hyytiäinen et al., 2015). Costs of environmental management measures may include administration and enforcement costs, income losses resulting from a specific policy measure or opportunity costs. The benefits are typically more diverse and include direct effects of environmental change on prices of marketable goods, household incomes and firms' profits, in addition to changes in the provision of those ecosystem services which are outside the market.

The requirement for socio-economic analysis by the MSFD, in particular the implementation of environmental CBA, has been discussed within the literature (COWI, 2010; Bertram and Rehdanz, 2013; Bertram et al., 2014) along with a focus on the appropriate methods for its implementation (Turner et al., 2010; WG-ESA, 2010; Reinhard et al., 2012; Interwies et al., 2013a,b). Bertram and Rehdanz (2013), Hanley et al. (2015), and Oinonen et al. (2016) discuss the limitations of



economic valuation in the context of the MSFD and point out that certain challenges threaten the effectiveness of this policy instrument. These latter studies highlight the challenges of translating changes in ecosystem service provision into welfare benefits changes. As such, existing valuation studies in the marine environment focus too much on direct ecosystem benefits (e.g., recreation), which are relatively easier to value, and often ignore less tangible effects on human welfare (e.g., aesthetic for spiritual wellbeing; Atkins et al., 2013).

It has been recognized that a major challenge of using environmental CBA in an MSFD framework is "the lack of knowledge on the links between potential measures, improvement of marine ecosystems and corresponding economic and social value" (EC, 2015, p. 29). One reason for this may be the potential non-linear form of links between changes in ecosystem properties and functions, ecosystem services and benefits. A second reason relates to the potential cumulative effects on services and benefits resulting from concomitant implementation of several measures. This latter issue also implies a risk of double counting when implementing an environmental CBA. While it has been suggested that ecosystem service classifications are a way to map and assess these links (Interwies et al., 2013b; Bertram et al., 2014; Maes et al., 2014), a comprehensive conceptual framework does not exist for the use of ecosystem service approaches in the assessments of benefits arising from the implementation of a PoMs under the MSFD. An additional challenge is the limited number of valuation studies related to marine ecosystems (Atkins et al., 2013), with most of such studies focussing on coastal ecosystems, such as beaches (e.g., Nunes et al., 2015), seagrass beds (e.g., Börger and Piwowarczyk, 2016), and fisheries (e.g., Crilly and Esteban, 2013).

Adding to existing guidance of applying environmental CBA to marine ecosystems for the MSFD (WG-ESA, 2010; EC, 2015), this paper presents applications of environmental CBA from three Member States: Finland, the United Kingdom (UK), and Spain. Each case study takes a different approach to this task, based on differing conditions at each site, data availability, and the nature of the descriptor(s) under study. To make case studies comparable an established six-step process of environmental CBA will be applied an tested (after Hanley and Barbier, 2009). In this way, strengths and challenges of each study can be interrogated in a systematic way and strengths and weaknesses highlighted. As a secondary objective, this paper also scrutinizes the applicability of ecosystem service approaches to facilitate the assessment of ecosystem benefits under the MSFD. It considers the challenges of the valuation of ecosystem benefits specific to the MSFD and the function of ecosystem services as a link between an impact assessment of environmental management measures and monetary valuation. Although, not explicitly required by the MSFD, the use of an ecosystem services approach for this task has been suggested (WG-ESA, 2010; Koss et al., 2011) because it can: (1) assess trade-offs between the provisions of different services; (2) mitigate the risk of double-counting by concentrating on final ecosystem services (Fisher et al., 2009); (3) support the mapping of changes in the provision of ecosystem services spatially; and (4) facilitate value transfer by offering established ecosystem services classifications. It will thereby extend and specify the analyses in Bertram and Rehdanz (2013) and Bertram et al. (2014) and propose an ecosystem services approach as a potential step forward. This paper is particularly relevant to six of the MSFD Descriptors: D1 Biological diversity; D2 Non-indigenous species; D3 Commercially exploited fish and shellfish; D4 Marine food webs; D6 Sea floor integrity; and D11 Energy including underwater noise.

#### MATERIALS AND METHODS

# The MSFD and Environmental Cost-Benefit Analysis

This section presents the steps required to assess both the costs and benefits of PoMs under the MSFD. A number of CBA frameworks have been cited within the literature (e.g., Boardman et al., 2006; Defra, 2007; Hanley and Barbier, 2009) and have been successfully applied within a marine context, for example in the case of seabed restoration following the cessation of marine aggregate extraction (Cooper et al., 2010). The present analysis is based on Hanley and Barbier's (2009) recommendations for environmental CBA which involves the following six steps:


#### Step 1

In the context of the MSFD, the definition of measures to be implemented may include technical, legislative, economic, and policy-driven actions (EC, 2015). Here, it is the PoMs to achieve GES which are the focus of any environmental CBA to be conducted in an MSFD framework. According to the MSFD, management measures can be classified as existing or new measures. Existing measures (Article 13.1 and 13.2) are those which are based on non-MSFD legislation and which have been fully or partially implemented (Categories 1.a and 1.b, respectively). New measures (Article 13.3) are those which are additional to measures based on existing legislation and build upon them or are completely new (Categories 2.a and 2.b, respectively).

The different types of measures available and/or implemented underpin the selection of case studies in this paper. In the Finnish case a catalog of new measures (Categories 2.a and 2.b) which are about to be administered by the relevant authorities are used for the analysis. In the UK case, focus is entirely on potential new measures (Category 2.b), which were not included in the package of measures for the first cycle of MSFD implementation in the UK. In the Spanish case study, the measures investigated fall into Category 1.b, which are based on the Common Fisheries Policy (CFP), but contribute to the achievement of GES under the MSFD. By selecting three contrasting case studies across Europe, this allows for a comparative study of approaches to environmental CBA based on different categories of MSFD PoMs.

#### Step 2

Once the management measures have been specified in detail, their potential impacts on ecosystems and human activities can be identified and where possible quantified. This is particularly challenging in the marine environment because of the open access, transboundary movement of resources and pollution, and combinations of different pressures with different and cumulative impacts arising from the general complexity of marine and coastal ecosystems and the activities undertaken therein. These factors increase the uncertainties in an assessment of the effectiveness and benefits of the management measures (EC, 2015). Regardless of which method is applied, it is always necessary to determine a baseline of the current status and future projections that would result without additional management measures. There are a number of tools that can be employed separately or in combination at this stage (Burdon et al., 2015), such as:


#### Step 3

The impact of PoMs measured quantitatively require valuation to arrive at monetary estimates of both costs and benefits. Cost quantification involves the identification of the opportunity costs that will be incurred due to the implementation of and compliance to the management measures. The costs often included are costs to the regulator and/or government, costs to businesses or industry for complying with the management measure including loss in income or Gross Value Added, potential environmental or damage costs and social costs. The quantification of benefits has been an area of active research by environmental economists for decades (Cummings et al., 1986; Mitchell and Carson, 1989; Bateman et al., 2002; Freeman et al., 2014), and a number of techniques to monetize impacts of environmental change have been developed (see for example Atkins et al., 2013). Marketed goods affected by such change (e.g., fish catches) can be valued using market prices, although these may not fully reflect the value of this provisioning service if market imperfections exist. Valuation of non-market effects can be undertaken by applying revealed and stated preference methods. Revealed preference methods, such as the travel cost method (Ward and Beal, 2000) or hedonic pricing (Palmquist, 1999) observe human behavior or economic outcomes and infer the value of non-market aspects of environmental change. Stated preference methods, such as contingent valuation (Carson and Hanemann, 2005) and discrete choice experiments (Louviere et al., 2000) are survey-based and directly elicit willingness to pay for improvement in environmental quality, which can be aggregated over the whole population affected to arrive at the total economic value (including use and non-use values) of any particular change. Stated preference methods can also assess benefits ex-ante, i.e., before a project is implemented and values are generated, in contrast to other valuation methods which can only assess values from an ex-post perspective. However, when primary valuation studies are not available for a particular study site, benefit transfer has been proposed as a tool to apply values estimated for other locations to the site in question (Richardson et al., 2015).

#### Steps 4 and 5

Once impacts have been valued in Step 3, positive impacts (i.e., benefits) and negative impacts (i.e., costs and explicit implementation costs) accruing at different points in time can be aggregated and compared. The streams of instantaneous costs C<sup>t</sup> and benefits B<sup>t</sup> over the time horizon of the PoMs (t = 1, . . . , T) have to be discounted to make them comparable. The present values (PV) of costs PV<sup>c</sup> and benefits PV<sup>B</sup> are

$$PV\_{\mathcal{E}} = \sum\_{t=1}^{T} \frac{C\_t}{(1+\delta)^t} \tag{1}$$

and

$$PV\_B = \sum\_{t=1}^{T} \frac{B\_t}{(1+\delta)^t},\tag{2}$$

respectively. δ is the discount rate, which might differ between case studies. MSFD related PoMs would pass the net present value (NPV) test if discounted benefits outweigh discounted costs, i.e., PV<sup>B</sup> > PV<sup>C</sup> (after conducting a sensitivity analysis, see Step 6). Put differently, the implementation of the proposed measures increase social welfare and hence should be implemented from a welfare economic perspective if the discounted net benefits (PV<sup>B</sup> − PV<sup>C</sup> > 0) are positive. There may be other perspectives that could call for the measures to be rejected, for example with respect to social acceptance or equity.

#### Step 6

It is often the case that scarce evidence and a lack of sufficient data requires assumptions to be made at different stages of the analysis. To test the dependency of the results on any one assumption, sensitivity analysis can be applied by altering any one assumption and investigating the implications on the results.

The particular challenges of applying environmental CBA to PoMs within the MSFD lie in Steps 2 and 3. This challenge will be further investigated below.

### APPLICATION OF CBA FOR PoMs TO ACHIEVE GES IN EUROPEAN REGIONAL SEAS

While following the general framework of environmental CBA laid out above, three case study analyses were conducted independently of each other, each responding to particular requirements and challenges in their own geographical locale. A summary of the three case studies is presented in **Table 2**. **Figure 1** indicates their location in Europe.

#### Finnish Marine Waters of the Baltic Sea

The Finnish case study describes the economic analyses undertaken to support the preparation of the national PoMs for the MSFD. The national PoMs Working Group led the process and prepared and planned the new measures. The Working Group members were environmental scientists and other related officials, researchers, and NGOs. The Finnish government approved the PoMs in December 2015 after a public hearing process.

#### Step 1. Development of the PoMs

There was a consensus within the national PoMs Working Group that a gap existed between the present status of the marine environment and GES, and thus a list of potential measures falling into Categories 2.a and 2.b was compiled (**Table 3**). A sub-group of economists was established and as requested by the MSFD (Article 13) its mandate was to conduct the costeffectiveness and CBA of the new measures.

#### Step 2. Identification of the Impacts of the PoMs

The impacts of the PoMs were assessed as part of the costeffectiveness analysis (Oinonen et al., 2016). The environmental effectiveness of a measure was defined as the probability of closing the gap between the present environmental status and GES. The joint effectiveness of two or several


measures was computed combining the distributions of the individual measures. This allowed the calculation of probability estimates regarding GES achievement by 2020 for each MSFD Descriptor. Due to the lack of comprehensive ecological-economic models applicable for MSFD-related analyses, the estimates of the effectiveness of measures were based on expert elicitation. The data were collected in six thematic workshops that followed a structured group interview format. Each workshop had 6–13 experts discussing 6–10 measures.

#### Step 3. Economic Valuation of the Impacts of PoMs

Similarly to the impact evaluation of the PoMs, the costs of measures were estimated using expert elicitation and conditional probability distributions. The cost estimation was conducted during the same workshops that estimated the environmental impacts of the measures. The expected total costs for the Finnish PoMs were estimated at €136.2 m (Oinonen et al., 2016). The cost-effectiveness analysis provided a ranking of new measures and proposed a set of cost-efficient candidate PoMs. The number of measures in the cost-efficient candidate PoMs ranged from 21 to 31 and the expected costs of the PoMs ranged from €20 to €136.2 m.

TABLE 3 | New measures in the Finnish PoMs (source: Oinonen et al., 2016).


The economic benefits of the PoMs were estimated based on existing valuation studies on the benefits of improving the state of the Baltic Sea. These studies elicit people's willingness to pay for specific (mainly cultural) ecosystem services, i.e., recreational and non-use values. Previous work revealed that it is relatively straightforward to link the valuation studies directly to the Descriptors of GES, instead of linking the Descriptors to ecosystem services and further to valuation studies (Hasler et al., 2016). Thus, the approach connected the benefit estimates directly to the change in the status of the GES Descriptors.

Three criteria were considered when choosing the economic valuation studies, to ensure the results would be reliable. Firstly, we followed the cost-effectiveness analysis (Oinonen et al., 2016) and focused on those GES Descriptors which were assessed as not achieving GES in the Initial Assessment in 2012, i.e., D1 (Biological diversity), D4 (Food webs), D5 (Eutrophication), D8 (Concentration of contaminants), D9 (Contaminants in fish and other seafood). Secondly, the search was limited primarily to valuation studies in Finnish marine waters. When Finnish waters' studies were not available, the suitability of valuation studies conducted in the other coastal countries of the Baltic Sea were considered. Thirdly, studies conducted within the last 5 years were used to provide up-to-date benefit estimates based on state-of-the-art valuation methodologies. In the estimation of benefits, the Descriptors for biological diversity (D1) and food webs (D4) were combined due to their partial overlap, but were treated separately from eutrophication (D5). This approach to the Descriptors was consistent with their treatment in the valuation studies; there was a valuation study focusing solely on eutrophication, and another focusing on characteristics pertaining to biodiversity and food webs, both using stated preference methods and estimating use and non-use values.

A recent contingent valuation study on eutrophication (D5) (Ahtiainen et al., 2014) provided a value estimate of improving the eutrophication level in the Baltic Sea from the business-asusual state to a (near) good state by the year 2050 (all other basins except the northern Baltic Proper would achieve good state). According to the study, the benefits of reaching GES to the Finnish population until 2050 would be €3580 m, of which €1022 m would accrue until 2020. The characteristics used to describe eutrophication in the valuation study were water clarity, blue-green algal blooms, fish species composition, underwater meadows, and oxygen conditions in sea bottoms. These are clearly linked to the Descriptor on eutrophication (D5) and its more detailed characterization, which mentions water clarity, algal blooms, ecosystem effects, and oxygen deficiency (Finnish Ministry of Environment, 2014). The differences between the timeframe of Ahtiainen et al.'s (2014) study and the timeframe of the PoMs to achieve GES (2050 vs. 2020), as well asthe differences between Ahtiainen et al.'s (2014) study area (the entire Baltic Sea) and the area of the case study (Finnish marine waters) were assumed to work in opposite directions. The longer timeframe may have led to lower benefit estimates, and the larger geographic area to higher estimates compared to the MSFD policy change.

The benefit estimate for the Descriptors for biological diversity (D1) and food webs (D4) was based on a choice experiment study that valued the preservation of pristine areas, increases in the amount of healthy vegetation (such as underwater meadows) and the size of fish stocks (Kosenius and Markku, 2015). The study indicated that the benefits to the Finnish population would be €363–1068 m, with the lower bound estimate including only the preservation of pristine areas and the upper bound including all three improvements in the marine environment. The attributes of the choice experiment are related to the Descriptors of biodiversity and food web and their specification (Finnish Ministry of Environment, 2014). Preservation of pristine areas and healthy vegetation can be linked to the area of distribution and status of species and biotopes in D1, whereas the condition of fish species is linked to healthy fish populations in D4. In Kosenius and Markku (2015), the timeframe coincided with the MSFD target year of 2020. Although, the benefits were estimated for the entire Finnish population, the study area was limited to the archipelago between Finland and Sweden. It is likely that the benefits would be larger if the environmental change were to take place in the entire Finnish marine area.

A challenge with both valuation studies was that the baseline and target scenarios specified in the studies do not necessarily correspond with those of the MSFD. As the value estimates are dependent on the extent of the change in the marine environment, this may cause some uncertainty in the benefit estimates. However, as no reliable correction for the differences was available, it was deemed better to use the original estimates than to apply some ad-hoc adjustment factors.

No benefits could be estimated for contaminants in the marine environment (D8) and contaminants in seafood (D9) due to a lack of site-specific evidence. The few existing valuation studies on contaminants in the Baltic Sea focus on individual substances, e.g., tributyltin (Noring et al., 2016) or oil (Ahtiainen, 2007; Juntunen et al., 2013). Moreover, the new measures targeting contaminants are related to research activities (measures 30 and 31 in **Table 3**) thus their contribution to achieving GES by 2020 was assessed to be very low (Oinonen et al., 2016, Table S2).

#### Step 4. Discounting of Flows of Benefits Occurring Over Time

Both the costs and benefits were discounted to the year 2014 using a discount rate of 3%. The net present value (NPV) of achieving GES for biological diversity (D1), food webs (D4) and eutrophication (D5) in 2020 is around €2000 m. However, the PoMs will not lead to GES in terms of these Descriptors in Finnish marine waters by 2020; based on the environmental effectiveness assessment as part of the cost-effectiveness analysis (Oinonen et al., 2016), the probability of reaching GES by 2020 is 0.77 for biodiversity and food webs, and 0.02 for eutrophication. Consequently, the benefits of this particular PoMs are lower than the benefits of achieving GES. To obtain the expected benefits from the PoMs the benefits were multiplied with the probability of reaching GES yielding benefits to the Finnish population of €300–894 m (**Table 4**). Most of the benefits result from improvements in D1 and D4, as the probability of achieving GES is relatively high for these Descriptors. Reducing eutrophication would also lead to significant benefits, but due to the low probability of reaching GES by 2020, the expected benefits are low.

#### Step 5. Application of the Present Value Test

Comparison of the estimated benefits (€300–894 m) to the costs (€140 m) of the Finnish PoMs indicates that despite the fact that the GES will not be achieved by 2020, the benefits of the PoMs exceed the costs by a factor of between 2 and 6.

TABLE 4 | Estimated benefits of implementing PoMs and achieving GES in Finnish marine waters.


*The benefit estimates are discounted to the year 2014 using a 3% interest rate and calculated for the Finnish adult population.*

#### Step 6. Sensitivity Analysis

As part of the sensitivity analysis, the benefits are presented as a range instead of point estimates. This range reflects different time frames and the extent of the environmental change. Although, there are interlinkages and overlaps between the Descriptors of eutrophication and biodiversity/food webs, there was very little overlap in the environmental change descriptions in the valuation studies. As the valuation studies covered only three of the 11 GES Descriptors, and other features of the valuation studies (time frame, study area) were considered to lower the value estimates compared to valuing the achievement of GES in the Finnish marine waters by 2020, the risk of double-counting and overestimating the benefits was considered low.

Using different budget constraints, Oinonen et al. (2016) provided a set of cost-efficient PoMs. The PoMs that included all measures and had the highest costs was selected and approved by the Finnish government. This might be due to the fact that it was impossible to achieve GES by 2020 with any of the proposed candidate PoMs, as it would take longer for most measures to take full effect. A candidate PoM, that would not significantly change the probability to achieve GES, would decrease the costs from €1362 to €90 m and thus the benefit-cost ratio would increase from 2–6 to 3–9.

# East Coast Marine Plan Area (UK)

The UK case study, the East Coast Marine Plan (ECMP), was selected as a case study because it is a defined area of management, being the first area in England where marine planning has been undertaken and a marine plan produced. The GES management measures considered in this paper for the ECMP are associated with reducing the impact of underwater noise (UWN) and invasive alien species (IAS) which are a subgroup of non-indigenous species. These pressures were selected for analysis so that only additional MSFD management measures can be studied, that is those which are not implemented based on existing legislation (Category 2.a and 2.b).The analyses of both sets of management measures use an ecosystem services approach to assess their benefits. It became apparent that the evidence base was limited regarding both the impacts of UWN and IAS on ecosystem services and more generally on ecosystem services within the ECMP area. The assessment of costs and benefits of PoMs with respect to these two pressures was therefore applied using a scenarios approach which facilitates the transparency of making assumptions for each scenario during the course of the analysis.

#### Step 1. Policy Analysis regarding UWN, IAS, and Ballast Water Management

Noise is addressed in the MSFD within D11 (underwater energy including noise), with effects upon D1 (biological diversity), D3 (commercial fisheries), and D4 (food webs) (**Figure 2**). In the ECMP area the main sources of underwater noise are likely to be associated with shipping and offshore construction such as marine energy development. Due to the scarcity of scientific data relating to the impacts of sound, management measures are limited, particularly for fish and invertebrates (Popper et al., 2014). This has led to uncertainty regarding how regulators, stakeholders and scientists should proceed when so many activities produce underwater sounds (Hawkins and Popper, 2014; Hawkins et al., 2014a).

Noise may affect behavior and physiology and may also elicit injury or damage in those exposed; there has been more research to date regarding marine mammals (e.g., Nowacek et al., 2007; Weilgart, 2007) than fish and invertebrates although this area is growing (e.g., invertebrates: Wale et al., 2013; Solan et al., 2016). Similarly there has been a focus upon short term behavioral changes, for example schooling variation in fish (Hawkins et al., 2014b) rather than on longer term impacts such as reproductive changes.

Pile driving of wind turbine foundations produces substantial impulsive noise which has potential effects on a number of marine species through the water and vibration through the sediment (reviewed in Roberts, 2015). Approaches to minimize the impacts of piling come at a cost to the wind farm developer, and encompass either engineering solutions (such as inflatable pile sleeves) or biological monitoring (such as employment of marine mammal observers) (Würsig et al., 2000; Nedwell J. et al., 2003; Nedwell J. R. et al., 2003; Thomsen et al., 2006; Nehls et al., 2007; Parsons et al., 2008). For shipping, which produces a continuous sound, mitigation examples include reduction of vessel speeds, exclusion from biologically sensitive areas or attempts to use "quieter" ships (De Robertis and Handegard, 2013).

Non-indigenous species are species, subspecies or lower taxa that occur outside of their natural range following intentional or unintentional introduction due to human activities (Ojaveer et al., 2014). Invasive alien species (IAS) are a subset of nonindigenous species that have been defined as having a "significant negative impact on biodiversity as well as serious economic and social consequences" (EC, 2014; Ojaveer et al., 2014). Their introduction has long been recognized as a key threat to marine ecosystems and the services these deliver. Nonindigenous species and therefore IAS are addressed by D2, but they also affect other Descriptors due to their potential impact on biological diversity (D1) and to food webs (D4) through changes in feeding relationships (**Figure 2**). In the marine environment, shipping is the key vector for species globally and the most efficient way to avoid the introduction of new species is successful ballast water management (Molnar et al., 2008; Ojaveer et al., 2014). As the North Sea has been described as one of the most invaded ecoregions of the world (Molnar et al., 2008), the following environmental CBA focuses on ballast water management as a measure to reduce the likelihood of introducing

IAS and thereby achieving GES in D2 and contributing to D1, D4, and D6. The analysis addresses the questions of what ecological, economic and social benefits effective ballast water management produces and which indicators are necessary to measure impacts of IAS on ecosystem services and benefits. To facilitate this analysis, for this study two species, the molluscan veined whelk (Rapana venosa) and the Japanese shore crab (Hemigrapsus sanguineus) were chosen. Both of these species do not presently occur in the ECMP but have the potential to arrive as they both already occur in European countries. They have a high potential to impact ecosystem service provision if introduced as shown in other areas, for example R. venosa in the Black Sea (Mann et al., 2004) and H. sanguineus along the French side of the English Channel (Dauvin and Dufossé, 2011).

#### Step 2. Bio-Physical Impacts of the Policy: A Scenario Analysis

For both management measures, secondary evidence and other information was gathered to assess their potential impacts and how these might affect the ecosystem services and benefits in the ECMP area. Conceptual models were developed which display the linkages between the respective management measures and ecosystem structure, processes, services and the resulting benefits (**Figure 2**) and serve as the basis for quantifying (and valuing) these impacts. However, these figures are most likely not comprehensive because the link between the properties of ecosystems such as biodiversity and ecosystem services are still a major scientific challenge (Pereira et al., 2010; Strong et al., 2015).

Insufficient (quantitative) evidence regarding the impact of both pressures on relevant ecosystems and their services required a scenarios analysis approach to be undertaken. Quantitative evidence was lacking on several levels: data on current ecosystem services provision in the ECMP was not available at the spatial scale necessary. Additionally, uncertainty in terms of effects of noise on ecosystems still exists. The effect of IAS on a naïve habitat is also not predictable. Finally, it is even more uncertain to predict how benefits are impacted for example, is bird abundance reduced when bivalve biomass is reduced, this cannot be predicted (Kendall et al., 2004). **Table 5** specifies two environmental scenarios, characterizing low and high impact of each pressure. For IAS scenarios were chosen based on high and low impact classifications as described by Ojaveer et al. (2015). Given the high uncertainty and scarcity of site-specific evidence, assumptions were made to characterize each scenario. With a low (high) impact scenario specifying the lowest (highest) possible impact, a separate environmental CBA can be conducted for each of the environmental scenarios, i.e., for each row in **Table 5**. Implementation costs of additional MSFD measures fall into the cost category; benefits are the avoided negative impacts of the

#### TABLE 5 | Development and definition of scenarios in the UK case study.


*UWN, underwater noise; IAS, invasive alien species. Impact is defined as potential impact, e.g., the likelihood of environmental damage if levels of UWN or IAS increase.*

pressure under study (underwater noise or IAS), i.e., from the difference between a future situation without management and hence with potential low or high negative impacts and a situation with MSFD management (and hence no adverse impacts).

For each scenario, an explicit list of assumptions can be formulated. These assumptions for the high and low impact scenarios should be detailed based on existing evidence. This is done for the installation of wind turbines, based on the implementation of development plans in areas leased for offshore wind farms in UK waters in Rounds 2 and 3 by the Crown Estate (Higgins and Foley, 2014) as an example:

Assumption L UWN C : Only UK Round 2 wind farm projects within the ECMP area will be completed, resulting in the need for pile driving for 374 turbines (4cOffshore, 2016).

Asumption H UWN 1 : All UK Round 2 and 3 wind farm projects within the ECMP area will be completed, resulting in the need for pile driving for 374 Round 2 and 1457 Round 3 turbines, totaling 1831 (4cOffshore, 2016).

As for the costs associated with noise reduction measures, Nehls et al. (2007) estimate the costs of using an inflatable sleeve during pile driving in Germany to be approximately €20,000–25,000 per turbine. Between 374 (Assumption L UWN C ) and 1831 (Assumption H UWN C ) turbines will be installed in the ECMP area until 2020. This means that the total present value cost of using inflatable sleeves to reduce underwater noise during construction ranges between €9.3 and €43.9 m. These figures have been adjusted for inflation. Aggregation over time assumes a 3.5% discount rate as suggested by HM Treasury (2003). Employing noise reducing technology during pile driving increases overall construction time by 3% (Nehls et al., 2007).

For considering the costs of installing and operating ballast water treatment systems on vessels going in and out of the ECMP area to reduce the risk of IAS introduction (Fernandes et al., 2016), the following assumptions are made: (1) All container ships and tankers and 30% of passenger vessels come from high seas (e.g., Asia); (2) 10% of the total operating costs of ballast water treatment systems is attributed to the case ECMP area; and (3) any type of vessel enters the ECMP area from an intercontinental origin five times per year on average. In the low and high impact scenarios, the assumptions with respect to the future development of shipping traffic are:

Assumption L IAS C : Intercontinental shipping traffic in and out of the ECMP area will decrease by 10% compared to average annual arrivals for 2011–2014 reflecting a downturn in international trade.

Assumption H IAS C : Intercontinental shipping traffic in and out of the ECMP area will increase by 25% compared to average annual arrivals for 2011–2014 reflecting a large increase in international trade.

#### Step 3. Economic Valuation: Costs and Benefits

With these assumptions, further cost monetization is possible. Fernandes et al. (2016) provide estimates of installation and operating costs of such systems on different vessel types. For the UK case study, the assumption is that the technologies examined in Fernandes et al. (2016) will be used by vessels traveling along the UK east coast. Based on average annual ship arrivals data from the Department for Transport and assumptions L IAS C and H IAS C the total discounted cost up to 2020 of installing and operating ballast water treatment systems on all relevant vessels in that area ranges between €3025 and €3929 m, following a change in intercontinental traffic into the area of −10 and +25%, respectively.

In combination with an explicit list of assumptions to describe the scenarios in **Table 5**, the environmental CBAs of the low and high impact scenarios produce upper and lower bounds of the net benefit of MSFD PoMs. If a sensitivity analysis were to be conducted, as required by Step 6 of the environmental CBA approach, any assumption can be modified to investigate effects resulting in net benefit figures.

In terms of benefits, the impacts of the aforementioned scenarios-response combinations on ecosystem service provision have been assessed qualitatively, based on literature and expert judgment, as there is insufficient site-specific evidence available at this time (**Table 6**). Benefits of these measures can be assessed through the improvement in ecosystem service provision or the avoided loss of the ecosystem service provision, but are not limited to these. Without quantitative data on the impacts on ecosystem services it was not possible to attribute changes in value of the benefits except in the same qualitative way. Indirect benefits can include any avoided costs (e.g., incurred by industry, local communities, or other specific interest groups) of having to correct or minimize adverse impacts. For example,



*bonly affected by UWN.*

the eradication of the invasive and non-indigenous carpet sea squirt Didemnum vexillum in Holyhead Marina (Wales) was estimated to cost approximately €150,000 per eradication attempt (Kleeman, 2009). If this cost is avoided due to the implementation of a cessation or reduction measure this can be considered an indirect benefit of that measure. Therefore, the benefit of the measure is a combination of the avoided loss of ecosystem service provision (direct benefit) and the avoided cost of having to address the issue at a latter (and potentially more problematic) stage (indirect benefit).

#### Steps 4, 5, and 6. Discounting, Present Value, and Sensitivity Analysis

Due to the high level of uncertainty in the bio-physical data and because of the novelty of this investigation, benefits arising from the measures under investigation could not be quantified in this case study. As a consequence, a quantitative environmental CBA based on an application of Steps 4, 5, and 6 cannot be reported for this area.

### Bay of Biscay (Spain)

This case study comprises the Spanish section of the Bay of Biscay (BoB). The most important maritime sectors in this area are: fisheries, ship building, maritime transport (including sea and coastal passenger and freight water transport, inland passenger freight water transport, renting, and leasing of water transport equipment), construction and coastal tourism (Fernández-Macho et al., 2015). Both "fully" and "partially" maritime sectors are taken into account (following terminology from Kalaydjian et al., 2010 and Foley et al., 2014). Of these, fisheries, maritime transport, construction, and coastal tourism are considered "fully" maritime sectors, and sport fishing a "partially" maritime sector.

#### Step 1. Management Measures Linked to Maritime Activities Development

The relevant management measures to reduce the pressures of these sectors on marine ecosystems were identified. These measures can potentially enhance the provision of ecosystem services. For this case study, the most relevant measures potentially contributing to achieving GES in the BoB come from the reformed Common Fisheries Policy (CFP), which came into force on 1 January 2014 (EC, 2013), rather than from the MSFD, thus falling into Categories 1.a and 1.b.

Fishing activities cause pressures on the marine environment. These pressures may directly affect Descriptors D1 (biological diversity), D3 (commercial fisheries), and D4 (food webs), and in turn the provision of several ecosystem services and benefits, such as "wild fish and shellfish for food." The reformed CFP introduces several new changes, such as: legally binding targets to achieve maximum sustainable yield (MSY) for all harvested stocks by 2015; progressive phasing out of discards of unwanted or overquota fish by 2019; and the establishment of biologically sensitive protected areas, in which fishing activities may be restricted or prohibited. Therefore, the implementation of management measures directly related to the reformed CFP are crucial to achieving GES.

Here the three above-mentioned measures have been established in the context of the reformed CFP, which directly affect GES (Category 1.b): the potential elimination of scrapping subsidies<sup>1</sup> , which ultimately reduce the fishing pressure on commercial fish species and sea-floor integrity; implementation of the new landing obligation which directly affects marine biodiversity and food webs; and the introduction of individual fishing rights which contribute to the new regionalization framework promoted by the CFP that aims to increase the profitability of regional fisheries and ultimately help to reduce fishing pressure on commercial fish species.

#### Step 2. Bio-Economic Modeling

Benefits associated with food provision were simulated using the bio-economic model FishRent (Salz et al., 2011). This model has been applied to the three management measures related to fisheries specified in Step 1. FishRent is a quantitative assessment model that allows for the evaluation of the bio-economic performance of fleets and therefore, the provision of fish as a food service over the medium (15 years) and long term (25 years). FishRent is composed of six modules: biological (stockgrowth relation and biomass function), economic (revenues, costs, cash flow, etc.), interface (production function, discards and landings), market (price of fish and fuel price), behavior (fleet size, effort and investment), and policy (level of landings and/or the effort involved).

#### Step 3. Impact Assessment (IA) Analysis

A set of scenarios regarding the different FishRent components (stocks, fleets, etc.) were identified for which medium- to longterm simulations were run. The scenario approach takes into account the baseline, the status quo and potential management measures identified in Step 1, for which different endogenous (simulated by FishRent as the fishing effort) and exogenous variables (e.g., first sales prices) associated with external factors (e.g., market prices) are considered. To develop Step 3, both private and public costs of the development and management of the fishing activity are considered. Private costs related to development of the fishing activity are included within the economic module of FishRent. In addition, the public cost programme that exists at the European level (i.e., the European Maritime and Fisheries Fund, EMFF) has also been considered to co-finance the national and regional public cost programmes. However, as most available information regarding the EMFF is aggregated (e.g., different stocks, management measures, countries, etc.) it is difficult to include those public costs within the economic module of FishRent for each proposed measure. Therefore, these public cost are not explicitly considered in the model.

To achieve the main objective of Step 3, an impact-analysis (IA) was undertaken to assess the quantitative impacts of new fishing management measures on food (fish) provision and the expected monetary benefits (after EC, 2009). Following Murillas et al. (2011) the value of food provision was assessed using the

<sup>1</sup>Physical scrapping of vessels implies a permanent removal of the vessels from fishing activities. Scrapping subsidies prevent vessels continuing with the activity as there are high costs associated with scrapping vessels.

gross value added, which is the difference between the revenue obtained from fisheries according to the market price and the private costs incurred in the production of the good. In addition, profit is also used when possible.

#### Steps 4 and 5. Net Present Value (NPV) of the Expected Value Added in Relation to the Public Costs

Through the application of FishRent, the NPV of gross value added (and profits) related to fish provision value and to the public costs (investments) from the EMFF for the Basque Country (EMFF hereafter) can be obtained. Expenditure of 6% of the total EMFF budget on scrapping subsidies results in a positive medium-term (15 years) impact on the BoB trawler fleet, leading to a 2.25% increase of NPV of gross value added over this period and a 52% increase of NPV of profits. There is no effect over the long term (i.e., 25 years). Of more importance than the temporal scope is the level of investment needed for this measure (scrapping subsidies). If the allocated investment was lower (<6%), the impact on the activity, and therefore on food provision, would be over-proportionally reduced. Furthermore, applying subsidies to different fleet segments causes different effects. Investing these subsidies in the management of purse seiners results in an increase of the NPV of gross value added. It would increase by 4.43% in the medium term (38% in the case of profits), and the impact would be extended over the long term.

For the analysis of the impact of the implementation of the management measures related to the landing obligation (discard ban) a total research-related public cost of 1% is applied. There is no direct relationship between public cost and economic benefits. This public investment will prevent a decrease of the NPV of gross value added of around 45% of BoB trawler activity over a 15-year period, which might happen when the landing obligation is implemented. From the biological perspective, this leads to a 33% increase in biomass, which might imply a positive impact on the value of fish provision in the long-term.

The introduction of transferable individual fishing rights may positively impact on food provision benefits with NPV increasing by around 33% making fishing activity more profitable, thereby supporting economic as well as biological sustainability.

An important cost, representing 4% of the total EMFF budget for the Basque Country, is assigned to control and enforcement activities for implementation of the CFP. Its main impact is assessed by assuming a high level of compliance in the application of the management measures. Thus, this additional general cost should also, although partially, be assigned to the environmental CBA of the above measures.

Finally, sport fishing is of great interest in the BoB. Input control measures, which limit the effort by controlling the number of vessels involved, are applied. Considering only the boats that are dedicated to sport fishing in the area (N = 376), the vast majority were licensed (98%) and enrolled in the Second Book of Ship Registration held by the Department of Agriculture and Fisheries (92.7%). Thus, only 29 vessels should be removed from this recreational activity. The direct impact on vessel investment, production value and the rent of decommissioning 29 sport fishing vessels is estimated based on Zarauz et al. (2013). Investment is reduced by €1.5 m, which implies a reduction of production and rent of around €2 and €0.5 m, respectively. Lastly, in Step 6 a sensitivity analysis was developed to assess the influence of varying the main external factors (fuel price, fish market prices, etc.) to check the robustness of the expected trend in relation with the NPV of the GVA. A summary of the results is presented in **Table 7**.

# DISCUSSION

# Challenges Applying a CBA Approach under the MSFD

While environmental CBA is an established analytical tool for the appraisal of environmental management measures (Boardman et al., 2006; Hanley and Barbier, 2009), its application within the marine environment, and particularly under the MSFD, is challenging. The main challenge in all three case studies presented here is the limited ecological evidence available for the analyses. This may lead to a focus on only a limited number of ecosystem services which are more easily quantified. **Table 8** provides an overview of experience within each case study. However, the Finnish case study, with its greater reliance on eliciting expert opinion, demonstrates an approach which can lead to a quantitative assessment and included a wide range of management measures. Moreover, the Finnish approach can be extended to include an ecosystem service assessment.

A second and related challenge is the scarcity of fit-forpurpose valuation studies that focus specifically on benefits arising from changes in all or specific MSFD Descriptors<sup>2</sup> . This challenge was highlighted by the limited use of existing valuation studies in the Finnish case and the total absence of such information that could be applied in the UK case study. Only the major cost components of the management measures in the UK study could be monetized. The effect on ecosystem benefits of reducing underwater noise and the likelihood of introducing IAS could only be established in a qualitative way. This is because of the lack of knowledge of the existing level of ecosystem services at the EMCP spatial scale, the effect that both UWN and IAS will have on the ecosystem and the associated services and benefits; and the uncertainty associated with the use of methods such as benefit transfer. For both measures it is therefore safer to do a qualitative assessment but in this way including those services benefits that cannot be valued on monetarily (such as bioremediation or bird watching). The Finnish study adopted a pragmatic alternative for estimating the economic value of marine protection when applicable data are available and conducting extensive new valuation studies is not feasible. Even though the existing studies did not explicitly assess the benefits of achieving GES, the results are suitable for indicating the benefits from the PoMs. Existing results were used as limited resources prevented undertaking new studies. The BoB case study highlighted the clear link between the investment (i.e., private and public costs) and ecosystem service

<sup>2</sup>Apart from the studies used in the Finnish case study to the best of our knowledge, the only valuation studies relating directly to MSFD Descriptors are Bertram and Rehdanz (2013) and Norton and Hynes (2014). However, Hanley et al. (2015) and Sagebiel et al. (2016) report valuation studies which could also be linked to some Descriptors.


#### TABLE 7 | Bio-socioeconomic impact on ecosystem services resulting from management measures on Bay of Biscay maritime activities.

*ITQ, Individual fishing quota; BFT, Atlantic bluefin tuna; GVA, Gross value added.*

benefits. However, important public costs attached to certain CFP-related management measures cannot be split between specific management measures, which may limit the application of an environmental CBA specific to MSFD Descriptors.

The third challenge in the context of practical MSFD implementation, is the lack of public resources to conduct fitfor-purpose valuation studies, such as Norton and Hynes (2014) for the case of Ireland. From a theoretical perspective, it is important in any type of valuation approach to focus on assessing additional benefits, i.e., what is the marginal change in the quantity and value of the benefits relative to what is present under a scenario without the management measure. However, in the case of ecosystem benefits, this has proven challenging due to the uncertainties regarding the marginal change in ecosystem service provision. The economic valuation of the benefits of management measures does not necessarily require new studies to estimate benefits since using estimates from existing studies with a similar context (i.e., benefit transfer) is acceptable practice in ecosystem service valuation or environmental CBA (Richardson et al., 2015). However, without knowing the marginal change in ecosystem service provision, it is difficult to apply these values with any degree of confidence. It is also possible that the direction of change of the economic value of the benefit and ecosystem service provision are not the same. The practice of assigning economic values to ecosystem services is inherently anthropocentric, and therefore benefit values (e.g., those that are measured by willingness to pay) are based on human perceptions. Given current limitation in human knowledge and understanding of the features and functioning of marine and coastal ecosystems, individuals are not always able to see how an improvement in biodiversity or in species populations could affect them (Duarte, 2000).

A final challenge highlighted by the case studies, relates to comparing the present values of costs and benefits (Pearce, 1998) for a specific period of time. The discount rate δ is crucial to make costs and benefits incurred at different points in time comparable in the present (Equations 1, 2). The discount rate reflects different levels of desirability between consumption and/or opportunity costs that occur at different points in time (Feldstein, 1964), and it is also an expression of concern regarding the distributional equity between current and future generations and among future generations (Arrow et al., 1995). A positive discount rate means that future values count less and hence are "penalized" and the higher the discount rate, the more future values are penalized. Depending on the discount rate used, benefits that are realized at a later point in time could have lower present values than the

#### TABLE 8 | Main findings from the three case studies.


\**The contribution of marine ecosystems to the maintenance of population dynamics, resilience through food web dynamics, disease, and pest control (definition from Hattam et al., 2015).*

costs that are incurred once the measure is implemented affecting the overall outcome of the CBA. Furthermore, CBAs for different management measures may not be comparable if they have not used the same discount rates. The length of the time period, over which costs and benefits are assessed, also affects the total NPV as it determines the temporal extent of the costs and benefits that are considered in the assessment. The MSFD states that GES should be achieved by 2020, however, it does not provide guidance on the time period for assessing the impacts of implementing new management measures to achieve GES. If the assessment covers only the time period from 2016 (when new measures are expected to be implemented) up to 2020 (when GES is supposed to be achieved), the short time span and the impact of discounting on benefits that materialize at the later date mean that there is a risk that costs of implementing new management measures will most often outweigh the benefits. For issues such as changes in the environment, biodiversity or climate change, which can only be detected over longer time periods, using a short time span for assessing the costs and benefits of any policy action is not appropriate. The choice to take action on these issues is a direct recognition that long time spans will be involved and several generations will be affected (Stern, 2006; HM Government, 2011). This means that for the assessment to be meaningful, the time period of assessment needs to be realistic and long enough to take into account any lags in the response of the environment to the implementation of MSFD PoMs.

# Opportunities for Applying CBA under the MSFD

If the management measures under study impact a range of ecosystem services which cannot be easily quantified, the qualitative approach adopted for the UK case study shows how the focus can be kept broad so as not to overlook important impacts of the management measures under consideration. This demonstrates the trade-off between a highly quantified environmental CBA which may only focus on a small number of specific ecosystem services and the broad approach taken in the UK case where quantification is currently problematic. For the latter, multi-criteria analysis (MCA) (Linkov et al., 2006) might be a potential way forward (DCLG, 2009), as it can incorporate cost and benefit measures reported in different units of account (e.g., non-monetized ecosystem service changes). Further research to test the applicability of MCA in this context is needed.

A further opportunity relates to the fact that environmental CBA allows for the examination of the trade-offs between different options to achieve GES within the parameters of economic efficiency (OECD, 2006), taking into account other constraints (e.g., discount rate, time period of assessment) that are used in the analysis and how these affect the values that are calculated. Therefore, results of an environmental CBA can help decision-makers to examine trade-offs, giving them the opportunity to develop a PoMs where the discounted net benefits (PV<sup>B</sup> − PVC) are maximized and which can effectively draw from existing policy actions already implemented to alleviate environmental pressures or even contribute to future policies. Additionally, sensitivity analyses which take into account different levels of the constraints used (e.g., different discount rates, different time periods for assessment) can be undertaken to show the variability of total costs and benefits in the face of different types of uncertainty.

# Lessons Learnt during Case Study Application of the Environmental CBAs under the MSFD

With respect to Finnish waters in the Baltic Sea, existing bioeconomic models were available for D2 (Non-indigenous species), D3 (Commercial fish and shellfish), and D5 (Eutrophication) but the models would need to have been updated to be applicable in the context of the MSFD. Expert elicitation was a successful alternative approach (to modeling) that provided comprehensive analysis covering all Descriptors. Benefit transfer was relatively straightforward to execute. Even though the original valuation studies only partially covered GES Descriptors, the estimated benefits were higher than the expected costs of the measures. The sensitivity analysis shows that the benefit-cost ratio would have been higher if another set of measures had been chosen. A candidate PoMs, that would not significantly change the probability of achieving GES would decrease the costs from €136.2 to €90 m and thus the benefit-cost ratio would increase from 2–6 to 3–9.

In the East Coast Marine Plan area, UK, scenarios analysis proved a useful tool where there is considerable uncertainty concerning the links between management measures and the ecosystem and the links to welfare impacts. Previous studies have demonstrated that scenarios can be used to "test" which policy actions are robust and sustainable, however it is recognized that the big challenge for using scenario analysis is communicating findings to stakeholders and policy-makers effectively (Burdon et al., 2015). The lack of, and high uncertainty associated with, data restricts the full application of an environmental CBA, in particular there were insufficient site-specific data to assess the potential ecological impact of the management measures (Step 2) and to quantify and value any changes in ecosystem services and welfare benefits (Step 3). As such, these two steps were only partially fulfilled precluding the completion of Steps 4, 5, and 6. In the context of high uncertainty about bio-physical and economic data on the impacts of a PoMs, the preliminary qualitative analysis undertaken in this case study proved valuable in identifying the main ecosystem services which may be affected under each management measure and thus identified areas for further site-specific research. A more in depth analysis, such as a MCA, could now be performed whilst waiting for reliable quantitative data to be made available.

In the Bay of Biscay, Spain, many different economic maritime activities operate. However, there has been little previous effort to develop any qualitative or quantitative assessments of impact (ecological or social) of those activities on the value of ecosystem services. One reason for this, amongst others, is that the "partial" maritime nature of most of the sectors involved hinders the extraction of the required data from the available statistics. Due to this limited knowledge only a few new measures are proposed for application to the different private economic activities, with the exception of the commercial fishing sector, which is mainly affected by the newly reformed CFP. The capacity for providing both qualitative and quantitative assessments related to the BoB management measures resulting from the CFP relies on existing bio-economic models that have been developed and applied in other areas of Europe and that are flexible enough to be applied in the BoB. While such models are well developed for commercial fishing activities, models that can be used for other maritime activities are still not sufficiently developed.

# CONCLUSIONS AND RECOMMENDATIONS

By showcasing and discussing three environmental CBA examples in the context of the EU MSFD, this paper highlights challenges and opportunities for the use and further development of this technique in the impact assessment of PoMs. The sixstep approach in Hanley and Barbier (2009) has been used as a structural framework to build an MSFD-specific environmental CBA set in different contexts across Europe. Both expert elicitation and the ecosystem services approach are shown to facilitate identification and quantification of physical impacts (Step 2). Challenges arise in valuing the physical impacts in economic terms (Step 3). While the Finnish and Spanish case studies monetize both costs and benefits, the UK case study could only express implementation costs of measures in monetary terms. Application of the step-by-step process for environmental CBA in contrasting case studies with differing levels of data availability has highlighted a number of issues. As such the following recommendations can be made:

(1) The environmental CBA approach needs to be further developed to better integrate the ecosystem services approach with established environmental valuation techniques (Börger et al., 2014). To aid this, further research into the linkages between MSFD Descriptors and established ecosystem service classifications is required so that the specific environmental CBA can then be linked to the suitable Descriptor via the affected ecosystem services. This would also help mitigate problems of adjusting existing valuation estimates to situations with slightly different types of environmental change, geographic area or time horizon. However, for pressure indicators (e.g., D5 eutrophication) where existing valuation studies show a reduction in ecological indicators and associated reduced economic values, an ecosystem service approach might be circumvented. In such cases it might be easier to link the existing value estimates directly to the Descriptors and assess the potential impacts on ecosystem services separately.


which can feed into the environmental CBA process. Modeling is also a valuable tool for projecting potential changes in ecosystems service provision in the future where real-time data is not available.

In conclusion, this paper has applied an established six-step framework for undertaking environmental CBA to assess PoMs chosen to achieve GES under the MSFD. The application of this framework to three contrasting European case studies has identified a number of challenges for undertaking such an approach. Despite these considerations, this paper has shown that there are opportunities in applying the six-step environmental CBA framework to assess the impact of PoMs under the MSFD.

### AUTHOR CONTRIBUTIONS

TB, SB, HA, DB, TL, AM, SO, LP, MU and MA conducted the analyses. TB, SB, HA, JA, DB, TL, AM, SO, LP, LR, MU and MA contributed to the writing of the manuscript.

# ACKNOWLEDGMENTS

This manuscript is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), http://www.devotes-project.eu. MU is partially funded through the Spanish programme for Talent and Employability in R + D + I "Torres Quevedo." MA and SB are partially funded by the Marine Ecosystems Research Programme, Natural Environment Research Council (NERC), and Department for Environment, Food, and Rural Affairs (DEFRA) (grant number NE/L003279/1). The authors wish to thank the two reviewers who's comments have significantly improved the paper.

# REFERENCES


sediment-dwelling invertebrates mediate ecosystem properties. Sci. Rep. 6, 1–9. doi: 10.1038/srep20540


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling Editor declared a collaboration with the authors and states that the process nevertheless met the standards of a fair and objective review.

Copyright © 2016 Börger, Broszeit, Ahtiainen, Atkins, Burdon, Luisetti, Murillas, Oinonen, Paltriguera, Roberts, Uyarra and Austen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Price vs. Value of Marine Monitoring

Henrik Nygård\*, Soile Oinonen, Heidi A. Hällfors, Maiju Lehtiniemi, Eija Rantajärvi and Laura Uusitalo

*Finnish Environment Institute (SYKE), Marine Research Centre, Helsinki, Finland*

Monitoring data facilitate the basic understanding of changes taking place in nature and provide information for making management decisions, but environmental monitoring is often considered expensive. Here, we apply the concept of value of information to evaluate the value of marine monitoring in the EU Marine Strategy Framework Directive context. We estimated the costs of the Finnish marine monitoring program and used the costs and economic benefits estimates of the Finnish marine strategy to assess the value of environmental monitoring. The numbers were applied to scenarios with different levels of information available prior to management decision-making. Monitoring costs were related to the value of perfect information prior to the management decision, assuming that managers will choose the management option that maximizes the benefits. The underlying assumptions of the conceptual model are that more accurate information about the status facilitates the selection of an optimal set of measures to achieve the environmental objectives and the related welfare gains from the improved environmental status. Our results emphasize the fact that monitoring is an essential part of effective marine management. Importantly, our study show that the value of marine monitoring data is an order of magnitude greater than the resources currently spent on monitoring and that an improved knowledge base can facilitate the planning of more cost-effective measures.

#### Edited by:

*Maria C. Uyarra, AZTI Tecnalia, Spain*

#### Reviewed by:

*Joana Patrício, Executive Agency for Small and Medium-sized Enterprises, Belgium Suzanne Jane Painting, Centre for Environment, Fisheries and Aquaculture Science, UK*

> \*Correspondence: *Henrik Nygård henrik.nygard@ymparisto.fi*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *15 June 2016* Accepted: *03 October 2016* Published: *20 October 2016*

#### Citation:

*Nygård H, Oinonen S, Hällfors HA, Lehtiniemi M, Rantajärvi E and Uusitalo L (2016) Price vs. Value of Marine Monitoring. Front. Mar. Sci. 3:205. doi: 10.3389/fmars.2016.00205* Keywords: environmental management, value of information, monitoring, MSFD, Marine biodiversity

# INTRODUCTION

In environmental management, monitoring activities constitute the foundation for understanding changes taking place in nature and provide information essential for decision making. However, monitoring is often looked upon as an expensive activity creating only costs, not considering the wide use of the data and the value of more informed decisions (Caughlan and Oakley, 2001). Considering environmental management, from monitoring to management programs, monitoring costs constitute only a small proportion (of the total costs) that becomes even smaller when adding the benefits achieved from efficient management (see Lovett et al., 2007 and references therein). Value of information (VoI) analysis is a tool for evaluating how much a rational decision-maker would be willing to pay for a new piece of information prior to making a decision (Stigler, 1961). Colyvan (2016) provides an overview of the concept and its application in conservation biology and Keisler et al. (2014) reviews the peer-reviewed literature from the years 1990–2011. Characteristic for the VoI analysis is that the value of information is in relation to the decision context. For example, Runting et al. (2013) found that when making decisions about where to locate a reserve system to preserve coastal biodiversity it is optimal to allocate a substantial proportion of the conservation budget in better data and models. In the fisheries management literature, VoI analysis has been recognized as a valuable tool in advising on the optimal fishing effort or quotas (Hilborn and Walters, 1992; Mäntyniemi et al., 2009). In this paper we apply the VoI concept to study marine environmental management and the optimal allocation of resources between monitoring and measures to improve the status of the marine ecosystem.

The EU Marine Strategy Framework Directive (MSFD; European Union, 2008) requires that Member States strive to obtain or maintain good environmental status (GES) in their marine waters by 2020. For management to be effective, indepth knowledge about the functioning of the marine ecosystem, changes in the system as well as the ability of monitoring to detect these changes is needed.

At the start of each MSFD cycle of 6 years the status of the environment is assessed and indicators and their relation to GES are set. Monitoring programs to ensure the collection of data needed for the indicators are then developed. Based on the status assessment, the distance from GES is evaluated and the descriptors not achieving GES are identified. To reduce the distance from GES and to remain in GES for descriptors already in GES, the program of measures (PoM) is set up where corrective measures need to be planned and implemented. Once the 6 year cycle is completed, the effect of the PoM is evaluated by a new status assessment, which starts the new MSFD cycle. Thus, assessment of GES is in the core of the MSFD and the assessment results will largely rely on the set of indicators used and their performance (Uusitalo et al., 2016a). In addition to fulfilling the quality requirements of an indicator (e.g., Queirós et al., 2016), indicator performance depends on the quality of the data used for calculating the indicator value as well as for setting the indicator GES boundary. Inadequate and/or insufficient monitoring will decrease the precision of the indicators, which can lead to erroneous assessment results; GES can be adjudged on false premises and needed corrective measures are omitted risking further degradation, or the indicators are unable to show a correct positive response leading to undertaking unnecessary measures.

The MSFD requires social and economic analysis when assessing the status of the marine ecosystem and when developing the PoM (e.g., Oinonen et al., 2016a), but cost-effectiveness analysis is not required for the monitoring programs. In this paper, our aim is to show the value of data and information produced by monitoring programs and how that value relates to the costs of the monitoring programs. We discuss how welldesigned monitoring programs can lead to cost savings in the marine management. As an example case, we illustrate the VoI concept with a hypothetical example and with data from the Finnish Marine Strategy.

#### MATERIALS AND METHODS

#### Data

In this study we used information from the Finnish national marine biodiversity monitoring program (Korpinen et al., 2014). The biodiversity monitoring program is divided into five monitoring themes (marine mammals, birds, fish, benthic habitats, and water column habitats), which are further divided into 19 sub-programs. For example, the water column habitat monitoring theme is split into phytoplankton and zooplankton sub-programs, among others. Data on the costs (year 2013) were collected from the institutes responsible for the monitoring and by interviewing involved experts. The cost data are based on Finnish prices. Flow charts were prepared to identify the different steps causing costs in monitoring (see **Figure 1** for an example). The biodiversity monitoring sub-programs are diverse and use multiple approaches and methodologies, but as a general frame the monitoring cost data were split into the following categories: research vessel, equipment, supplies, personnel, fixed costs, and other costs (following Veidemane and Pakalniete, 2015). Research vessel costs were based on the daily price for running the vessel (including crew, fuel and maintenance costs). When samples for several monitoring subprograms were collected during the same monitoring cruise (e.g., phytoplankton, zooplankton and benthos), the research vessel costs were divided with the total number of samples collected during the monitoring cruises to allocate specific research vessel costs per monitoring sub-program. Equipment costs (e.g., sampling gear, microscopes etc.) were calculated as the list price taking into account the expected lifetime of the equipment and a yearly discount rate. The costs of supplies (e.g., sample bottles, preservatives, petri dishes etc.) were calculated based on the yearly usage. The costs of both equipment and supplies were classified into sampling, analysis or data management expenses, to facilitate distinguishing the categories when adding up the costs. Personnel costs were likewise categorized into field, laboratory and data management expenses, and estimated based on the level of expertise and number of person-months needed per year for the various tasks. Overheads were applied to the personnel costs and included as fixed costs. Other costs included transport of equipment and personnel from the institute to the research vessel, costs for maintaining necessary professional skills, accreditations, participation in proficiency tests and sustaining continuity of expertise at the institute. The cost data were transformed to cost per sample, in order to facilitate estimating indicator costs and evaluations of cost-effectiveness with respect to the quality of data (e.g., how the number of samples or the spatial and temporal coverage of sampling affect the uncertainty of the indicator result).

For the costs of different management options we followed Oinonen et al. (2016b) who assessed the costs of the Finnish PoM (Laamanen, 2016), which were expected to be 136.2 million e. The economic benefit estimates are taken from the cost-benefit analysis of the Finnish PoM (Oinonen et al., 2015). Oinonen et al. (2015) followed Hasler et al. (2016) and linked existing valuation studies of Ahtiainen et al. (2014a) and Kosenius and Ollikainen (2015) with the GES descriptors and used a benefit transfer method (e.g., Richardson et al., 2015) to estimate the non-market value of reaching GES. As the management aim is to improve the environmental status, economic benefits arising only from an improvement in the environmental status are considered. The economic benefits of achieving GES for D1, D4, and D5 in 2020 were estimated to be around 2090 million e (Oinonen et al., 2015). The cost-effectiveness analysis of the Finnish PoM also provided knowledge on the probability of

achieving GES with different sets of measures; the probability of reaching GES by 2020 is 0.77 for biodiversity (D1) and food webs (D4), and 0.02 for eutrophication (D5) (Oinonen et al., 2016b). To obtain the expected benefits from the PoM, the benefits of reaching GES were multiplied with the probability of reaching GES. Thus, the economic benefits of the Finnish PoM were estimated to be 894 million e (Oinonen et al., 2015).

# Conceptual Model

To construct a model to evaluate the VoI gained through monitoring, the following components are needed (**Figure 2**):


For the computation of VoI, probabilities of the alternative possible states of the system (components 1, 4, and 7) are needed; for example, the status assessment in component 1 could be, simply, "based on what we know now (e.g., precision of the

indicator value or confidence of the indicator with regard to spatial and temporal coverage), we estimate that the probability of being in GES is 30% and the probability of not attaining GES is 70%." The classes (in the example, GES/sub-GES) can be defined according to the question at hand.

The VoI concept can be illustrated by a simple example (**Table 1**). In this example, the ecosystem status is divided into three classes (poor, moderate, and good), where the classes poor and moderate denote sub-GES (far from and close to the GES boundary, respectively) and good represents GES. Three management alternatives (do nothing, intermediate management, strict management) with different direct costs, and different benefits that they provide under the different environmental states, are applied. For illustration purposes, assume that good environmental status will bring benefits worth 1000 units and these benefits will not increase any further by added management. However, the net benefit will actually decrease because of the costs of the unnecessary management. The example shows that given the uncertainty about the environmental state, the optimal decision is to employ the intermediate management option, as it has the highest expected benefit. However, the best management action differs for between the three environmental states. This means that the decision maker might make different decisions if they knew the true state of the environment, and therefore, information about the true state has value. The value can be calculated by multiplying the maximum economic benefit that can be gained from each environmental state with the probability of each state, and summing up these figures. This number can be compared with the benefit that can be gained if the management scenario yielding the highest expected benefit is implemented. The difference between these figures is the value of information. In the example (**Table 1**), this value is 20. It must be noted that the value of information about the true state increases as the current uncertainty increases; and if the existing knowledge is already very certain, the value of perfect information may be very low.

The example in **Table 1** computes the value of perfect information, i.e., the value of knowing precisely the status of the ecosystem. In reality, perfect information is often unattainable. The value of imperfect information can, however, be estimated by comparing scenarios with different levels of knowledge. We illustrate this with an example of evaluating the expected value of biodiversity monitoring in the Finnish marine monitoring program in the Baltic Sea, using the best available estimates of monitoring costs, PoM, their effectiveness and costs.

# Scenarios to Assess Value of Information

Applying the VoI concept (**Table 1**), scenarios in which varying levels of knowledge were available for the status assessment were constructed in order to optimize the benefits of defined management options to achieve GES and estimate the value of perfect information. Perfect information is here defined as 100% certainty of the environmental status when choosing the management option. In the scenarios we applied three possibilities of initial environmental status: poor, moderate and good (as defined above).

Three hypothetical scenarios for monitoring were tested: (1) No prior knowledge of the environmental status, i.e., no monitoring takes place. In this situation the status assessment result was based on chance and all three status categories were equally probable (0.33). (2) Monitoring takes place, but it is insufficient to give a confident status assessment. In this scenario,


#### TABLE 1 | An example calculation of the value of information, based on hypothetical figures; for explanations and references to actual data see text.

*The shaded values highlight the maximum benefits in each ecosystem state and the highest net benefit given the uncertainty about the ecosystem state.*

the probability of the status to be correctly assessed was set to 0.5, with 0.25 and 0.25 probabilities for poor or good status when the true status is moderate. When the true status was poor or good, the probability for the status to be assessed as moderate was set to 0.3 with a 0.2 probability for assessing good or poor status, respectively. (3) Good monitoring, with a 0.8 probability of being correct in the status assessment. When the true status was moderate, 0.1 and 0.1 probabilities were set for assessing poor or good status. If the true status was poor or good, the probability for the status to be assessed as moderate was set to 0.15 with a 0.05 probability for assessing good or poor status, respectively. These probabilities are illustrative estimates based on the expected performance of ecological indicators. In ecological studies, indicators are often considered acceptable if they predict the status correctly more than 70% of the time, and excellent if more than 80% of the time (Hale and Heltshe, 2008).

Given the scenarios, three management options were applied: (1) no management, (2) intermediate management and (3) strict management. The "no management" option did not induce any costs and no change in the environmental status was expected. The "intermediate management" option was based on the current management scheme (Finnish PoM; Laamanen, 2016), which has been estimated to cost 136.2 million e (Oinonen et al., 2016b). Based on this management option, improvement from an initial poor status to moderate status was expected. However, if the initial status was moderate, this management option was not considered to reach GES within the management cycle (Oinonen et al., 2015). In the "strict management" option, we expected that the environmental status would improve from poor to moderate and from moderate to good, respectively, depending on the initial status. The costs for the "strict management" option were set to 500 million e (roughly the double of the expected maximum costs of the Finnish PoM Oinonen et al., 2015).

Since the benefits were considered as non-market benefits arising from improved environmental status, poor environmental status was not considered to yield any benefits in the scenarios. Moderate environmental status would bring 894 million e (the benefits achieved with the current Finnish PoM by 2020) and good environmental status was set to yield 2090 million e in benefits (Oinonen et al., 2015). The "no management" option would not bring any additional benefits. In the "intermediate management" option, the improvement from poor to moderate would yield 894 million e. Also, if the initial status was moderate, intermediate management was set to bring 894 million e, thus the benefits would be 1788 million e. Also in the "strict management" option and poor initial status, benefits were considered to be 1788 million e. If the initial status was moderate, the benefits with strict management would be 2090 million e.

#### RESULTS

#### Monitoring Costs

The yearly costs for the Finnish national marine biodiversity monitoring program were around 5.9 million e (**Table 2**). The largest costs were generated by the fish monitoring (2.58 million e), where the gathering of information for the Common Fisheries Policy accounted for 2.21 million e, as well as by the off-shore pelagic and benthic monitoring (2.20 million e), where running the research vessel constituted a major expense. The seal monitoring received administrative assistance from the Finnish Border Guard and thus all surveillance flights were not accounted for since the Border Guard would have flown anyway. The bird monitoring was partly based on voluntary work by ornithologists, thus reducing the costs.

Dividing the monitoring costs into the type of work and the categories from where the costs originated (see **Table 3** for an example of the zooplankton monitoring) allowed for a more critical evaluation of the monitoring expenses. Field work and laboratory work cost approximately the same, summing up to constitute almost 50% of the total expenses of the zooplankton monitoring sub-program. Although zooplankton monitoring

TABLE 2 | Yearly costs of the five marine biodiversity monitoring themes in Finland.


\**includes information for the Common Fisheries Policy (2.21 mill.* e*). The pelagic and benthic monitoring themes are here combined, and split in coastal and off-shore monitoring.*

TABLE 3 | Costs of the Finnish zooplankton monitoring sub-program itemized by the type of work and the categories from which the costs originate.


*Fixed costs include overheads of personnel costs and other costs include transport of equipment and personnel, maintenance of professional skills and accreditations (see text for full explanation).*

takes place off-shore and using a large research vessel, the research vessel cost was only 18% of the total costs when using the cost allocation of ship time per number of samples.

#### Value of Information

The scenarios showed that making the management decision based on better knowledge of the environmental status increased the expected net benefits (**Table 4**), with the exception of poor environmental status. In this case, strict management always brought the most benefits, regardless of the probability of correct status assessment. When no information was available for the environmental status assessment, the highest expected net benefits were achieved with strict management. If indicative information was available, strict management was the most beneficial option when the environmental status was poor or moderate, whereas intermediate management would yield the highest net benefits if the state was good. With good information available for the status assessment, the risk of making an erroneous management decision was smaller. In this case, strict management would be preferable if the environmental status was poor, and the intermediate management option would be the best choice if the initial status was moderate or good. Even in this case, the value of perfect information was 34–135 million e (**Table 4**).

The value of perfect information was the highest when no prior knowledge of the environmental status was available. In the scenarios where information was available for the status assessment (indicative or good information), the value of perfect information was highest when the state of the environment was good (**Table 4**). In these cases, the acquisition of additional information would help to distinguish between the possibility that the status is good and no management needs to be undertaken, and the possibility that the status is moderate or poor, and management measures are needed. Perfect information has the least value when the state is known (even with some uncertainty) to be poor, since strict management will be clearly the best option in that case.

Increasing the amount of knowledge available for making management decisions from no information to good information is worth 50–151 million e (the difference in the value of perfect information), depending on the environmental state. Thus, this sum could be invested in monitoring activities to increase the knowledge base and reduce the uncertainty of the made decisions. Given the assumptions, the net cost of this investment is zero, since the investment costs are covered by the increased benefits of the better decisions.

### DISCUSSION

The example presented in this paper shows that the value of improved information concerning the status of the sea can be an order of magnitude greater than the monitoring costs; in the case example up to more than a hundred million euros. While these numbers are indicative due to the simplified setup of the model, the calculation still illustrates the high value and tremendous significance of monitoring data and puts its costs into the perspective of the costs of the entire marine management framework (**Figure 3**).

Monitoring improves the quality and reliability of the environmental status assessment, but does not directly affect the environmental status. For effective management well-planned and effective measures are the key, and sufficient monitoring provides information to aid in the required decision-making. Because of this, monitoring can in many cases actually be the most efficient way to improve the status of the seas, since it facilitates targeting and scaling the management measures more accurately. For monitoring to be effective, links to the decision-making system and management strategies need to be clear. In the MSFD, monitoring data are used not only in the status assessment, but they also provide the fundamental understanding for linking pressures from human activities to changes in environmental status (**Figure 4**). Thus, monitoring data are utilized also to identify measures and scaling them properly to ensure an improved environmental status after their implementation.

If the environmental status is far from the GES boundary (the environmental status is either poor or excellent), this can usually be verified with less monitoring effort (e.g., with decreased frequency in monitoring): the whole confidence interval of the assessed indicator will be below/above the GES border even if the uncertainty is high. Moving closer to the GES


TABLE 4 | The results from the value of information analysis based on the three scenarios with varying amount of prior knowledge.

*The expected net benefits are based on the option maximizing the benefits (light blue* = *intermediate management, dark blue* = *strict management). The green line indicates the GES boundary. The "Do nothing" management option was not the best option in any of the cases. The pink cells mark the most probable status assessed.*

border, the indicator confidence interval needs to be narrower in order to correctly assess the status, meaning that a higher monitoring effort is required to attain a more precise estimate of the status. However, should the sampling frequency be reduced due to a high certainty of the current environmental status, the additional benefits obtained from monitoring data (scientific, educational, and cultural) may be compromised in a way that the net savings from the reduced monitoring will be dwarfed (Lovett et al., 2007). Monitoring data are also important for development and validation of ecological models. Ecological models have capabilities to evaluate ecosystem structure and function, involving impacts of human activities, and are potentially valuable aids in environmental management (Piroddi et al., 2015; Lynam et al., 2016; Tedesco et al., 2016). Moreover, in our scenarios, even good knowledge prior to the management decision indicated that additional information would be beneficial. Interestingly, additional information had the highest value when the environmental status was good, showing the savings made by avoiding unnecessary measures.

Status assessments indicate the situation of the state of the environment at a given moment. Although the MSFD integrates an assessment period of 6 years and thus incorporates natural variability to some extent, continuous monitoring is essential to place the assessed status in a long-term context. Long-term monitoring and data series provide baselines to detect changes in ecosystem structure and function, offer empirical data for mining when exploring new questions and for developing models, as well as identify ecological surprises (Lindenmayer and Likens, 2010). Continuous monitoring also allows for timely reactions when identifying changes. Such early-warning signals allow for less costly measures compared to reacting only at a more deteriorated stage and for avoiding a total ecosystem collapse (Hutchings and Myers, 1994).

As environmental status and biodiversity are by definition multifaceted concepts (e.g., Cochrane et al., 2010) often affected by a multitude of pressures acting through multiple pathways (Korpinen et al., 2012; Andersen et al., 2015; Uusitalo et al., 2016b), the information on numerous ecosystem components provided by monitoring is essential for informed decision-making. As a consequence, the link from any single monitoring sub-program to the management measures is less straightforward than with some other management targets. However, this is not taken into account in our model, where we assume that the pressure-status relationships are known and the uncertainty in the status assessment stem only from the quality (precision, temporal and spatial coverage etc.) of monitoring data feeding into the indicators. A well-known challenge in environmental management is that the pressurestate relationships of indicators are not always clear and that several pressures impact the environment simultaneously. Consequently, a careful development and selection of indicators is needed to reduce the uncertainty of the environmental assessment.

circles) are indicated. e indicate the steps where economic analyzes are needed.

Here, our main focus was the value of monitoring for management needs. When estimating the value of environmental monitoring, it is also important to consider benefits not directly associated with management. This aspect is seldom highlighted although monitoring is recognized as also contributing to science and to protecting resources (Griffith, 1998; Lovett et al., 2007). The scientific benefits, such as essential basic understanding of the natural processes and variability in the marine environment, are difficult to value in economic terms. The acquired scientific knowledge has uncertain, but potentially considerable, effects on the planning of future environmental management and use in ecological modeling, as well as on other parts of society such as education, culture, and other fields of science. The use of monitoring data to inform the public about changes in the environment can increase their interest for sustainable sea use and increased awareness can strengthen the commitment of citizens to facilitate and speed up the reaching of GES. Motivation of people to participate on marine protection in the Baltic Sea area has been studied for example by Söderqvist (1998) and Ahtiainen et al. (2014b).

An interesting observation and challenge was that the data on monitoring costs were not easily available. The information on costs usually consisted of lump sums from the monitoring program's accounting, and allocating them to indicator level to inform management decisions in the MSFD context was not trivial. As most monitoring sub-programs have been in place before MSFD coming into force and also before the development of indicators (which furthermore is still ongoing), none of the monitoring programs are aimed at producing only data for indicators. Thus, exact calculations of the cost of an indicator are complicated to perform. The biodiversity indicators are based on monitored parameters measured from samples. Often also other parameters are measured from the same sample and thus, not all information collected in the monitoring programs is used directly for indicators and management purposes, but this data contribute to the scientific understanding of processes taking place in nature. Additionally, an indicator may require data collected in other monitoring programs, if not for direct calculation, then at least for the interpretation of the indicator results. Since the use of research vessels, required for off-shore monitoring is expensive, ship time is used efficiently and the costs are shared by several monitoring programs and research projects. It was thus necessary to split the research vessel expenditures between the monitoring programs in order to allocate costs correctly. As the grounds for this division, we here used the number of samples collected for each monitoring program. This approach resulted in relatively low ship costs for monitoring programs relying on a low number of samples, e.g., zooplankton monitoring, compared to monitoring programs with more samples, e.g., physical and chemical monitoring of the water column, even though the data were collected during the same monitoring cruise and hence the days at sea and sea area covered were the same. The principles of gathering monitoring cost information and splitting it between indicators and/or monitoring programs need to be elaborated in order to better facilitate the use of this information for optimizing monitoring programs. Our approach, i.e., to estimate the cost per sample in the monitoring programs, is a useful approach when planning monitoring campaigns e.g., during revision of the spatial and temporal coverage of sampling.

In this study, we did not address the question of how much additional monitoring is needed in order to increase the precision of the environmental status assessment and how much resources this would require. Factors affecting the quality of the assessment


*The steps are exemplified by work needed in MSFD context as well as how the steps were done in this study.*

#### TABLE 5 | Steps for analyzing the value of information.

are measurement accuracy as well as the spatial and temporal scales of sampling. For example, Klais et al. (2016) showed that catching the population dynamics of zooplankton communities in the Baltic Sea requires sampling every 2 weeks. Compared to the present temporal resolution of the Finnish national zooplankton monitoring (sampling twice a year), a monitoring scheme fully covering the population dynamics of zooplankton would require considerably increased resources. However, the status assessment uses one zooplankton indicator (mean size versus total stock) and the twice a year sampling during the productive season fulfills the data requirements for this indicator (Gorokhova et al., 2016). Optimizing the sampling program needs to be considered carefully taking into account what the requirements for the indicator are and what would be gained by adding spatial or temporal coverage. The monitoring cost data collected in this study allow for such evaluations, since the data provide information on costs per sample.

The VoI concept has here been illustrated with an example that can be calculated easily on any spreadsheet program. The steps needed for a VoI analysis are summarized in **Table 5** with links to steps in the MSFD work. The same concept could be implemented as a Bayesian Network based influence diagram (e.g., Uusitalo, 2007) in a more refined form that would allow the direct comparison of different monitoring programs, their costs and the expected improvement in the level of knowledge about the ecosystem status.

Comparing the costs of the current monitoring with the value of making well-informed decisions highlights the unbalance in the present interpretation of monitoring expenses. Whereas, monitoring causes concrete costs for managers, the benefits of reliable information to more accurately scale measures are hard to trace and thus usually not considered. Further, the benefits achieved by an improved environmental status needs to be determined using economic valuation methods. Valuation of monitoring needs to have a broad approach that takes into

#### REFERENCES


account not only the immediate minimum knowledge needs but also the benefits gained through more efficient management and the scientific, cultural and societal value of the knowledge that is produced. Thus, the monitoring should not be priced according to its costs but according to the value it is creating to the society.

#### AUTHOR CONTRIBUTIONS

HN, LU, and SO: Conceived the paper; HN, HH, and ML: Collected data on monitoring costs; LU, SO, and HN: Developed the model and scenarios; All authors contributed to the interpretation of the results. ER: Made the figures; HN: Wrote the first draft; All authors contributed to and approved the final draft

### FUNDING

This study was supported by the DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing Good Environmental Status) project funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (Grant Agreement No. 308392), http://www.devotes-project.eu, the MARMONI (Innovative approaches for marine biodiversity monitoring and assessment of conservation status of nature values in the Baltic Sea) project funded by the European Union LIFE+ Nature and Biodiversity program (Project Nr. LIFE09 NAT/LV/000238), http://marmoni. balticseaportal.net and the BONUS BIO-C3 project that was supported by BONUS (Art 185), funded jointly by the EU, and Academy of Finland.

#### ACKNOWLEDGMENTS

We would like to acknowledge Joona Salojärvi for help collecting the monitoring cost data.

in the Field of Marine Environmental Policy (Marine Strategy Framework Directive). Available online at: http://eur-lex.europa.eu/legal-content/EN/ TXT/PDF/?uri=CELEX:32008L0056&from=en


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Nygård, Oinonen, Hällfors, Lehtiniemi, Rantajärvi and Uusitalo. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Mapping ecosystem services provided by benthic habitats in the European North Atlantic Ocean

#### *Ibon Galparsoro\*, Angel Borja and María C. Uyarra*

*Marine Research Division, AZTI-Tecnalia, Pasaia, Spain*

#### *Edited by:*

*Michael Arthur St. John, Danish Technical University, Denmark*

#### *Reviewed by:*

*Steve Whalan, Southern Cross University, Australia Jose M. Riascos, Universidad de Antofagasta, Chile Maria Salomidi, HCMR, Greece*

#### *\*Correspondence:*

*Ibon Galparsoro, Marine Research Division, AZTI-Tecnalia, Herrera kaia portualdea z/g, Pasaia 20110, Spain e-mail: igalparsoro@azti.es*

The mapping and assessment of the ecosystem services provided by benthic habitats is a highly valuable source of information for understanding their current and potential benefits to society. The main objective of this research is to assess and map the ecosystem services provided by benthic habitats in the European North Atlantic Ocean, in the context of the "Mapping and Assessment of Ecosystems and their Services" (MAES) programme, the European Biodiversity Strategy and the implementation of the Marine Strategy Framework Directive (MSFD). In total, 62 habitats have been analyzed in relation to 12 ecosystem services over 1.7 million km2. Results indicated that more than 90% of the mapped area provides biodiversity maintenance and food provision services; meanwhile, grounds providing reproduction and nursery services are limited to half of the mapped area. Benthic habitats generally provide more services closer to shore—rather than offshore—and in shallower waters. This gradient is likely to be explained by difficult access (i.e., distance and depth) and lack of scientific knowledge for most of the services provided by distant benthic habitats. This research has provided a first assessment of the benthic ecosystem services on the Atlantic-European scale, with the provision of ecosystem services maps and their general spatial distribution patterns. Regarding the objectives of this research, conclusions are: (i) benthic habitats provide a diverse set of ecosystem services, being the food provision, with biodiversity maintenance services more extensively represented. In addition, other regulating and cultural services are provided in a more limited area; and (ii) the ecosystem services assessment categories are significantly related to the distance to the coast and to depth (higher near the coast and in shallow waters).

**Keywords: ecosystem service, benthic habitat, Regional Seas, Marine Strategy Framework Directive, habitat classification**

#### **INTRODUCTION**

Functioning ecosystems are essential for maintaining the oceans in a healthy state (Tett et al., 2013). While being healthy, they provide numerous and diverse goods and services that contribute "for free" to the general well-being and health of humans (Van Den Belt and Costanza, 2012). The "ecosystem goods and services" term integrates two concepts: (i) the ecosystem goods, which represent marketable material products that are obtained from natural systems for human use, such as food and raw materials (De Groot et al., 2002); and (ii) ecosystem services, which refers to all "the conditions and processes through which natural ecosystems, and the species that make them up, sustain and fulfill human life" (Daily, 1997). The latter are not directly marketable services, and include nutrient recycling, biodiversity maintenance, climate regulation or cultural and esthetic services (Costanza et al., 1997). Ecosystem services occur at multiple spatial scales; from the global, such as climate regulation, primary production, and carbon sequestration, to a more regional or local scale, such as coastal protection and leisure.

Previous studies show that coastal ecosystem services provide an important portion of the total contribution of ecosystem services to human welfare (Pimm, 1997; Pearce, 1998). Costanza et al. (1997) showed that, while the coastal zone only covers 8% of the world's surface, the services that this zone provides are responsible for approximately 43% of the estimated total value of global ecosystem services. Despite our dependence on biodiversity and ecosystem services, population expansion and economic growth are leading to increasing anthropogenic pressures on coastal areas (Wilson et al., 2013) and consequently, to a decreasing supply of ecosystem services worldwide (Costanza et al., 2014). Recognizing that human pressures directly impact on ecosystem services and that in turn, ecosystem services directly benefit human wellbeing, they have sparked interest amongst coastal planners and have led to the integration of ecosystem services in conservation management measures (Cimon-Morin et al., 2013).

Due to the above-mentioned reasons, ecologists, social scientists, economists and environmental managers are increasingly interested in assessing the economic values associated with the ecosystem services of coastal and marine ecosystems (Bingham et al., 1995; Costanza et al., 1997; Daily, 1997; Farber et al., 2002; Liquete et al., 2013a). Different approaches and frameworks have been proposed to identify, define, classify and quantify services provided by marine biodiversity (MEA, 2003; Ten Brink et al., 2009; Cices, 2013; Liquete et al., 2013a). Neither of these approaches being a straight forward one; the accurate estimation of the values of services, and in particular their temporal and spatial variation, is relatively new and has not been extensively researched (Schägner et al., 2013).

Indeed, the complexity of the processes and functioning of marine ecosystems, and their highly dynamic nature, translates into the absence or low resolution of spatially explicit information. Furthermore, the deep sea, and in particular benthic habitats, is mostly lacking in ecosystem services assessments (Armstrong et al., 2012; Thurber et al., 2013). Due to these limiting factors, there are few published studies, and they mainly focus on food production, such as fisheries, with other services receiving minor attention (Murillas-Maza et al., 2011; Liquete et al., 2013a; Seitz et al., 2014). Mapping and assessing ecosystem services may help to overcome such hindrances. Maps not only enable the characterization of current benefits that services provide to society, but also the adoption management measures that guarantee their future provision and contribution to human welfare (Egoh et al., 2012).

To date, several habitat mapping efforts have been carried out at different spatial and temporal resolutions (Liquete et al., 2013a). Within Europe, Mapping and Assessment of Ecosystems and their Services (MAES) is one of the keystones of the EU Biodiversity Strategy to 2020 (Maes et al., 2013). This strategy demands Member States to map and assess the state of ecosystems and their services in their national territory (including their marine waters) with the assistance of the European Commission. The results of this mapping and assessment should support the maintenance and restoration of ecosystems and the services they provide (Maes et al., 2013). It will also contribute to the assessment of the economic value of ecosystem services, and promote the integration of these values into accounting and reporting systems at EU and national level by 2020. The results are expected to be used to inform policy decision makers and policy implementation in many fields, such as nature and biodiversity, territorial cohesion, agriculture, forestry, and fisheries. Outputs can also inform policy development and implementation in other domains, such as transport and energy (Maes et al., 2013). For example, the Marine Strategy Framework Directive (MSFD, 2008/56/EC) requires the availability of ecosystem services valuation for the assessment of the environmental status and to define the measures that make sustainable human activities at sea (Cardoso et al., 2010). Hence, according to the MSFD, the assessment of the environmental status should be undertaken for the Exclusive Economic Zone (EEZ) of the Member States within the four European Regional Seas: North Eastern Atlantic, Baltic, Mediterranean, and Black Seas.

In this context, the objectives of this research were: (i) the qualitative assessment and mapping of the ecosystem services provided by benthic habitats within the European North Atlantic Ocean; and (ii) to determine if ecosystem services assessment categories are related to the habitat distance to the coast and depth. The analysis was based on available cartographic information and ecosystem services assessment, focusing on the benefits that they provide in the Regional Seas and sub-regions defined by the MSFD.

#### **MATERIALS AND METHODS**

The implementation of ecosystem services valuation involves two dimensions: (i) a biophysical assessment of services supply; and (ii) a socio-economic assessment of the value per unit of services (Schägner et al., 2013). Within this investigation, we focused only on the first approach of trying to map and assess the ecosystem services provided by benthic habitats at the European North Atlantic Ocean scale. This is because the economic value of the services is still poorly known, needing comprehensive data supply, which the results from this investigation can provide.

#### **GEOGRAPHIC AREA**

For this investigation, the North Eastern Atlantic was selected. According to MSFD, the North Eastern Atlantic Ocean is divided into four sub-regions: Greater North Sea, Celtic Seas, Bay of Biscay and Iberian coasts, and Macaronesia (**Figure 1**). It should be noted that at the time of this investigation, no official geographical delimitations of the sub-regions were adopted, and therefore, they were defined according to the EEZs. The total area of the European North Atlantic Ocean covered by the MSFD is 4,540,025 km2, which corresponds to the EEZ of 10 European Member States and part of Norway (**Figure 1**).

**FIGURE 1 | European North Atlantic Ocean sub-regions.** Spatial limits are based on the Marine Strategy Framework Directive and Exclusive Economic Zone of the countries located in each sub-region. BE, Belgium; DK, Denmark; FR, France; DE, Germany; IE, Ireland; NL, Netherlands; NO, Norway; PT, Portugal (including Azores archipelago and Madeira archipelago); SP, Spain (including Canary archipelago); SE, Sweden; and UK, United Kingdom.

#### **BACKGROUND INFORMATION USED IN THE ANALYSIS**

In order to proceed with the mapping of ecosystem services, main bathymetric and habitat data were obtained from the following sources:


#### **DIGITAL ELEVATION MODEL**

To produce the digital elevation model information layer, bathymetric information from MeshAtlantic and EMODnet was mosaicked. The information on this layer enabled the investigation of the depth distribution of benthic habitats in the sub-regions of the mapped areas.

#### **BENTHIC HABITATS INFORMATION**

For practical purposes of mapping and assessment (i.e., data availability) this investigation focused on "benthic habitats," as a means to assess the provision of ecosystem goods and services.

Habitats were classified according to EUNIS (European Union Nature Information System) habitat classes (Davies et al., 2004). The EUNIS habitat classification aims to provide a common European reference set of habitat types to allow the reporting of habitat data in a comparable manner for use in nature conservation (e.g., inventories, monitoring, and assessments) (Davies and Moss, 2002; Davies et al., 2004; Galparsoro et al., 2012). The classification is organized into hierarchical levels (EUNIS habitat type hierarchical view is available at http://eunis*.* eea*.*europa*.*eu/habitats-code-browser*.*jsp). The present version of the classification starts at level 1, where "Marine habitats" are defined, up to level 6, by using different abiotic and biological criteria at each level of the classification. For seabed habitats for which EUNIS classes were not defined, underwater features defined under EUSeaMap (e.g., infralittoral seabed) were retained.

Habitat maps were transformed into raster format and mosaicked to obtain a total broad-scale habitat map. In overlapping cells, MeshAtlantic habitat classes were kept, according to the criteria that this represents the most recent information. The mapped area outside EEZ of Ireland was excluded from the later analysis, in order to make results comparable among different countries, in which only EEZ areas were included.

Finally, to analyse the spatial distribution of benthic habitats (in terms of their distance to shore) and therefore, that of the ecosystem services that they provide, the distance of each cell, assigned to each habitat type, to the nearest coastline point was estimated using Euclidean distance algorithm, in a Geographic Information System (GIS).

#### **ECOSYSTEM SERVICES ASSESSMENT**

In total, twelve ecosystem services were considered in this investigation: (i) Food provision; (ii) Raw materials (biological) (incl. biochemical, medicinal, and ornamental); (iii) Air quality and climate regulation; (iv) Disturbance and natural hazard prevention; (v) Photosynthesis, chemosynthesis, and primary production; (vi) Nutrient cycling; (vii) Reproduction and nursery; (viii) Maintenance of biodiversity; (ix) Water quality regulation and bioremediation of waste; (x) Cognitive value; (xi) Leisure, recreation and cultural inspiration; and (xii) Feel good or warm glow.

Ecosystem services were classified into: (i) Provisioning services (i.e., 1 and 2 from the above list); (ii) Regulating services (i.e., 3–9); and (iii) Cultural services (i.e., 10–12). The qualitative ecosystem services categories offered by each habitat were based on Table 1 from Salomidi et al. (2012), which, in turn, classified them based on an adaptation of the categories proposed by the Millennium Ecosystem Assessment (MEA, 2003) and Beaumont et al. (2007). Rather than using absolute metrics to classify services of each habitat, the assessment was based on the expert judgment of Salomidi et al. (2012), collated in the aforementioned **Table 1** of that manuscript, and the following guidelines: (i) when the provision of a specific service is well documented in the scientific literature and is widely accepted as important for the specific benthic habitat analyzed, it was considered as providing a "High" value for such ecosystem service (e.g., the role of seagrass beds in sediment retention and prevention of coastal erosion); (ii) when a service was or could be provided by a habitat but to a substantially lower magnitude than by other habitats and without being vital for the persistence of an important human activity, a "Low" value was assigned; and (iii) in all other cases, ecosystem services were classified as "Negligible/Irrelevant/Unknown." For the purpose of the present investigation, ecosystem services categories were rated into the following numerical values for further analysis: "High = 3," "Low = 1," "Negligible/Irrelevant/Unknown = 0." A similar classification and scores were successfully used in smaller areas (Potts et al., 2014) (see **Figures 3**, **4** in that manuscript).

The ecosystem services provisioning categories of each habitat type, was linked to the final habitat map. For those habitat classes that were included in the map, but not listed in Salomidi et al. (2012), the categories were assigned according to the knowledge of the authors, in a similar way to that of Potts et al. (2014).

To analyse the spatial distribution pattern of ecosystem services provisioning levels, the total area and its percentage cover of the total mapped area, mean depth, and mean distance to the coastline were calculated. The values of all cells encompassed within a polygon representing the extent of a habitat, were averaged to assign a unique value to each polygon for each variable (i.e., mean depth value within a polygon) To assess whether the distance to the coastline and depth had an effect on the categories at which the different ecosystem services are provided (i.e., high, low, and negligible values), Kruskal-Wallis non-parametric tests were applied using Statgraphics v.5.0. Then, differences in ecosystem services categories within the subregions were tested using Chi-Square tests. Finally, Friedman test, followed by *post-hoc* Wilcoxon tests, was undertaken to explore statistical differences between ecosystem services typologies (i.e., provision, regulation, and cultural).

#### **RESULTS**

The European North Atlantic Ocean (EEZ only) covers more than 4.5 million km2 (**Table 1**), of which 26% corresponds to continental shelf (up to 200 m depth) and 74% to deeper areas (**Figure 2**). To date, 88% of the continental shelf and 18% of the deeper areas have been mapped, accounting for 38.9 % of the total EEZ area of the European North Atlantic Ocean.

The Macaronesia accounts for the highest proportion of the European North Atlantic EEZ, followed by the Extended North Sea (**Table 1**). However, differences in the amount of mapped area can be found among sub-regions. Whereas countries located in the Celtic Sea and North Sea have already mapped almost all their EEZ seabed surface (i.e., 98 and 93%, respectively), countries located in Macaronesia, Bay of Biscay, and Iberian coasts (i.e., France, Portugal, and Spain) have still more than 80% of the seabed area without cartographic information (**Table 1** and

**Table 1 | Total spatial contribution of each sub-region to the Exclusive Economic Zone (EEZ) of the European North Atlantic Ocean, and their mapped area, represented in total and relative (%) terms.**


**Figure S1**). Indeed, habitat maps for the Canary and Madeira Archipelagos, in Macaronesia, are not available. It should be highlighted that these countries have some of the most extensive and deepest EEZs areas of the European North Atlantic Ocean.

The 1.7 million km2 covered by the integrated broad-scale habitat map encompassed 62 different benthic habitats and seabed seascape features (**Figure 3**). The North Sea and the Celtic Sea encompassed 58 and 55 habitats respectively, while the Bay of Biscay and Macaronesia only covered 42 and 20 habitats, respectively. Furthermore, very few habitats accounted for a large section of the mapped area (**Figure 4**). Ten habitats covered more than 75% of the total mapped area, of which deep sea mud (18.3%), deep circalittoral sand (16.2%), circalittoral fine sands, or circalittoral muddy sand (9.7%) were the most dominant ones. Opposite, a large number of habitats (i.e., 33) covered less than 10,000 km2 or 0.5% of the mapped seabed. The least dominant habitats in the European North Atlantic Ocean were the low energy infralittoral mixed hard sediments, Atlantic and Mediterranean low energy infralittoral rock and sponge communities on deep circalittoral rock, all of which cover less than 100 km2.

Of the 62 habitats identified in European North Atlantic Ocean, none of them provides the 12 ecosystem services considered in this study at the highest value (**Table 2**). However, four of these habitats (i.e., Infralittoral rock and other hard substrata, Atlantic and Mediterranean high energy infralittoral rock, High energy infralittoral seabed, and High energy infralittoral mixed hard sediments) provide high values for 11 services (excluding nutrient cycling). Another seven infralittoral habitats also provide high values for 10 of the services. On the other hand, 12 deep and bathyal habitats are considered as providing negligible values for 10 or more ecosystem services. The upper, mid, and lower bathyal seabed habitats provide the lowest number of ecosystem services and values.

Results also indicate that the highest provision of services is that of habitats located close to the coastline and in shallow waters (*p <* 0*.*001 for all services and in both cases—distance and depth; see **Tables 3**, **4**). Thus, there is a gradient on the level of services provision, from high to lower or negligible values, seawards and toward deeper areas. For example, areas providing high food provision services are located close to the coast (16 ± 35 km) and in shallow areas (47 ± 50 m). Furthermore, it is also observed that the level of service provision significantly varies across subregions (Chi-Square test: *p* always *<* 0.001), with the North Sea being the region generally providing services at the highest levels.

**Table 2** also suggests that none of the ecosystem services is provided by all the habitats. "Food," "biodiversity maintenance" and "nursery grounds" (i.e., "reproduction") are the ecosystem services most commonly provided by habitats (and to the highest level). Opposite, "photosynthesis," "disturbance prevention," "air quality" and "cultural services" are provided on a high level by a limited number of habitats. This pattern is also observed when considering not only the number of habitats providing specific ecosystem services, but also the area providing such

**FIGURE 3 | Benthic habitat map distribution within the European North Atlantic Ocean.** Habitats are listed in alphabetical order.

ecosystem services (**Table 3** and **Figures S2**–**S13**, in Supporting Information).

Indeed, 93% of the studied area provides food provision services, of which 62% corresponds with high food provision values. Similarly, a high proportion of the mapped area (99%) is considered as providing high (41%) and low (58%) biodiversity maintenance services.

The next ecosystem services, in terms of area coverage, are reproduction and nursery, which are provided by 53% of the mapped area. For the remaining ecosystem goods and services (i.e., air quality and climate regulation, water quality regulation and bioremediation, nutrient cycling, raw material provision, photosynthesis, chemosynthesis, and primary production), the area covered by habitats providing them at high values is much smaller. The disturbance and natural hazard prevention service has the smallest spatial coverage.

Finally, cultural services (i.e., cognitive value, leisure, recreation and cultural inspiration, and feel good and warm glow), showed similar patterns on their spatial distribution. The area covered by the habitats providing such type of services (both, at high and low levels) is very limited (around 11% of the total).

On the other hand, significant differences are observed in the spatial distribution of provision levels of aggregated ecosystem services (i.e., provisioning, regulating, and cultural), (Friedman test *<sup>χ</sup>*<sup>2</sup> <sup>=</sup> <sup>47</sup>*,* 858; *<sup>p</sup> <sup>&</sup>lt;* <sup>0</sup>*.*001) (**Figure 5**). The provisioning services are supplied at significantly higher levels than both regulating (Wilcoxon *post-hoc* test *z* = −154, *p <* 0*.*001) and cultural services (Wilcoxon *post-hoc* test *z* = −171, *p <* 0*.*001); and in turn, regulating services are also provided at significantly higher levels than cultural services (Wilcoxon *post-hoc* test *z* = −130, *p <* 0*.*001).

#### **DISCUSSION**

Seafloor maps are an essential source of information for resource exploitation and management purposes (Rice, 2010). Nevertheless, in Europe, it is worth noting that countries such as Spain, Portugal and France, with large EEZ areas have less mapped areas. This is probably due to the steepness of the seafloor, with large bathyal and abyssal areas, and the technical and economic challenge associated with mapping areas with such characteristics. Among others, marine shallow water areas support most of the human activities associated with the use and benefit of the ecosystem services provided by benthic habitats (Ramirez-Llodra et al., 2011; Korpinen et al., 2013), but accurate estimation of the values of services and their spatial distribution is not available for extensive areas. Within this research, the assessment and mapping of the ecosystem services provided by benthic habitats of the European North Atlantic Ocean has been undertaken for the first time.

#### **Table 2 | Ecosystem services assessment for each habitat and seabed feature type (H, high; L, low; and N, Negligible).**


*(Continued)*

#### **Table 2 | Continued**


*EUNIS habitat code is given for those habitats included in the classification; \* indicates that the assessment was based upon Salomidi et al. (2012).*

In the studied area, a clear gradient has been identified for the provision of ecosystem services, with significantly higher provision levels for habitats located in shallow waters and close to the shore. This is coherent with the fact that habitats provide more ecosystem services as people have easier access to them. In fact, accessibility is a crucial factor and it is typically included in the monetization of some services, especially for cultural services (Milcu et al., 2013). In the case of benthic habitats, access depends on depth, and generally, on the distance from the coastline. Therefore, deep-sea habitats and habitats located further away from the coast generally provide fewer ecosystem services and at lower degree due to limited access and lack of scientific knowledge for most of them. However, as exploration of the deepsea improves with recent technological advances, access to such habitats (Ramirez-Llodra et al., 2011) will become less difficult, increasing the ecosystem services that they provide in the near future (Thurber et al., 2013).

According to our estimations, between 93 and 99% (depending on the sub-regions) of the benthic habitats of the European North Atlantic Ocean deliver food provision and biodiversity maintenance services; meanwhile, reproduction and nursery services are provided by 53% of the area. We consider that the assessment of this last service could be underestimated due the fact that knowledge on life-cycles is mainly limited to commercially important species. But it should be taken into account that other non-commercial species, with unknown life cycles, also play an important role in food webs. Thus, the reproduction and nursery grounds are likely to cover a wider area than the one resulting from this investigation. In contrast, areas providing other services are smaller or have much more limited spatial distribution. For example, the area corresponding to habitats that supply raw materials is very limited, and the highest proportion of this area only provides low or negligible resources. To explain this pattern, it should be considered that few raw materials are exploited at present, and that their exploitation is regulated by national and international regulations as the impacts associated with such exploitation may be high. However, there may be high potential for habitats to provide higher provision of this service as new raw materials are discovered and exploited (i.e., pharmaceutical).

Another interesting pattern is that observed for the provision of coastal protection as an ecosystem service. Liquete et al. (2013b) propose the use of 14 biophysical and socio-economic



**Table 4 | Differences (Kruskal-Wallis test) between ecosystem services categories provided by benthic habitats, according to the distance to coastline, and depth (***N* = **55***,* **023).**


*\*\*\*Indicates significant results at 0.001 significance level. The superscripts within each service have been used to indicate significant (different superscripts) or non-significant (equal superscripts) differences on post-hoc tests between pairs of data, at 0.05 significance level.*

variables, from both terrestrial and marine datasets, in assessing coastal protection. In this investigation, we have only used benthic habitats, which may explain the relatively small area providing this service in the European North Atlantic Ocean. Furthermore, it is the limited distribution of biogenic structures and seagrass species within this ocean, considered as the main producer of this service, which may explain the limited provision to shallow and habitats located close to the

coast (Christianen et al., 2013; Cullen-Unsworth and Unsworth, 2013).

**and (D) Total ecosystem services.**

The remaining ecosystem services are provided in limited areas. This pattern is possibly explained by the fact that some of the services analyzed are provided by very specific, spatially limited benthic habitats (i.e., photic zones), or in a larger scale, by pelagic habitats, i.e., air quality and climate regulation, water quality regulation and bioremediation, nutrient cycling, photosynthesis, chemosynthesis, and primary production. For example, some of them, such as climate regulation or carbon sequestration, are very important in coastal margin habitats, rather than in subtidal habitats (Beaumont et al., 2014).

Very small areas (11%) have been identified as providing cultural services (i.e., cognitive, leisure, recreation and cultural inspiration, feel good, and warm glow). This result is likely to be a consequence of the dependence of these services on accessibility. Therefore, even if the current provision of these services is limited to few habitats and areas (which are probably heavily used), it is likely that over time, as access increases to certain areas, these services will increase their value and distribution (Ghermandi et al., 2012). The broad-scale spatial patterns of the ecosystem services assessment resulting from this investigation could be considered consistent for different spatial scales of analysis if the approach is implemented elsewhere.

When considering the approach and results obtained through this research, authors would like to highlight that, rather than getting a valuation of the ecosystem services provided by the benthic habitats of the European North Atlantic Ocean, in our investigation a pragmatic approach for benthic services mapping is applied, based on the best available knowledge (De Groot et al., 2010). We recognize that the reliability of the results obtained in this investigation depend on, among other things, two major aspects: (i) the quality and reliability of benthic habitat maps used, which is an important but insufficiently assessed issue (Schägner et al., 2013); and (ii) the valuation of the ecosystem services carried out by scientific expert judgment (extracted from Salomidi et al., 2012), which could be biased toward the knowledge of the experts who published that research; meanwhile, social and economic aspects could be under-rated.

Some of the aforementioned weaknesses could be overcome: (i) enhancing the scientific knowledge of marine ecosystem functioning by finalizing detailed benthic habitat maps of the complete study area (especially, for the EEZ of France, Spain, and Portugal and deeper benthic habitats; Liquete et al., 2013a); and (ii) improving the assessment of services valuation, promoting the multidisciplinary discussions among environmental and social scientists and economists, to achieve consensus on benthic habitat services values.

A more adequate ecosystem services assessment and valuation could be carried out following the steps below:


This process could result in the definition of proposals for management plans for different directives (e.g., MSFD, Habitats Directive) and instruments such as Marine Spatial Planning. Since oceans are facing an increasing number of human uses and threats, the inclusion of ecosystem services within management plans is growing in importance. In this context, the science of ecology must play a crucial role in bringing concepts like ecosystem goods and services to the forefront of the valuation debate (Bingham et al., 1995; Wilson and Carpenter, 1999; Liquete et al., 2013a).

The spatially explicit nature of the approach presented in this investigation is of special interest to support decision-making approaches and different aspects of the ecosystem-based marine spatial management *sensu* Katsanevakis et al. (2011). Among other things, the key to achieving a more comprehensive set of management mechanisms is, in the first instance, to know more about the ecosystem functions of benthic habitats (Martinez et al., 2011). In this way, there is a key goal of maintaining the delivery of ecosystem services, which must be based upon ecological principles that articulate the scientifically-recognized attributes of healthy functioning ecosystems (Foley et al., 2010), as required by the MSFD (Borja et al., 2013; Tett et al., 2013). This would require management measures for minimizing environmental impact and maximizing the socio-economic benefit of marine services (Salomidi et al., 2012); aspects that are basic to the Marine Spatial Planning.

This research has provided a first assessment of the benthic ecosystem services at Atlantic European scale, with the provision of ecosystem services maps and their general spatial distribution patterns. Related to the objectives of this research, the conclusions are: (i) benthic habitats provide a diverse set of ecosystem services, with the food provision and biodiversity maintenance services more extensively represented. In addition, other regulating and cultural services are provided in a more limited area; and (ii) the ecosystem services assessment categories are significantly related to the distance to the coast and with depth (higher near the coast and in shallow waters).

The results obtained in this investigation highlight the need for diverse, healthy and extensive benthic habitat areas to support the provision of important and valuable ecosystem services (i.e., food provisioning, disturbance prevention, nutrient cycling, etc.). Spatially explicit assessment and valuation of ecosystem services might be of crucial interest for future management measures adoption such as Marine Spatial Planning. The approach proposed here could be considered as a pragmatic way of getting a first snapshot of the distribution of ecosystem services based on the available information and we consider this as a promising starting point for further research and discussion on ecosystem services contribution of benthic habitats in Europe.

#### **ACKNOWLEDGMENTS**

This manuscript is a result of the projects MeshAtlantic (Atlantic Area Transnational Cooperation Programme 2007–2013 of the European Regional Development Fund) (www.meshatlantic.eu) and DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) funded by the European Union under the 7th Framework Program "The Ocean of Tomorrow" Theme (grant agreement no. 308392) (www.devotes-project.eu), and also supported by the Basque Water Agency (URA), through a Convention with AZTI-Tecnalia. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. We wish to thank Udane Martinez and Iñigo Muxika (AZTI-Tecnalia) for their significant contributions to the data analysis. This paper is contribution number 676 from AZTI-Tecnalia (Marine Research Division).

#### **SUPPLEMENTARY MATERIAL**

The Supplementary Material for this article can be found online at: http://www*.*frontiersin*.*org/journal/10*.*3389/fmars*.*2014*.* 00023/abstract

**Figure S1 | Depth distribution of the Exclusive Economic Zone (dark blue) and depth distribution of habitat-mapped areas (light blue), in the four subregions of the European North Atlantic Ocean; (A) Macaronesia; (B) Bay of Biscay and Iberian Coast; (C) Celtic Seas; and (D) Greater North Sea, including the Kattegat, the English Channel and Norway.**

**Figure S2 | Spatial distribution of food provision services.**

**Figure S3 | Spatial distribution of raw materials (biological, incl. biochemical, medicinal, and ornamental) services.**

**Figure S4 | Spatial distribution of air quality and climate regulation services.**

**Figure S5 | Spatial distribution of disturbance and natural hazard prevention services.**

**Figure S6 | Spatial distribution of photosynthesis, chemosynthesis, and primary production services.**

**Figure S7 | Spatial distribution of nutrient cycling services.**

**Figure S8 | Spatial distribution of reproduction and nursery services.**

**Figure S9 | Spatial distribution of maintenance of biodiversity services.**

**Figure S10 | Spatial distribution of water quality regulation and bioremediation of waste services.**

**Figure S11 | Spatial distribution of cognitive value services.**

**Figure S12 | Spatial distribution of leisure, recreation, and cultural inspiration services.**

**Figure S13 | Spatial distribution of feel good or warm glow services.**

#### **REFERENCES**


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

*Received: 14 April 2014; accepted: 29 June 2014; published online: 18 July 2014. Citation: Galparsoro I, Borja A and Uyarra MC (2014) Mapping ecosystem services provided by benthic habitats in the European North Atlantic Ocean. Front. Mar. Sci. 1:23. doi: 10.3389/fmars.2014.00023*

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science.*

*Copyright © 2014 Galparsoro, Borja and Uyarra. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.*

# **Lessons learnt**

# From Science to Policy and Society: Enhancing the Effectiveness of Communication

#### Marianna Mea<sup>1</sup> \*, Alice Newton2, 3, Maria C. Uyarra<sup>4</sup> , Carolina Alonso<sup>4</sup> and Angel Borja<sup>4</sup>

*<sup>1</sup> Ecoreach srl, Ancona, Italy, <sup>2</sup> Norwegian Institute for Air Research (NILU) - Department of Environmental Impacts and Economics (IMPEC), Kjeller, Norway, <sup>3</sup> Centre for Marine and Environmental Research (CIMA), Gambelas Campus, University of Algarve, Faro, Portugal, <sup>4</sup> AZTI, Marine Research Division, Pasaia, Spain*

#### Edited by:

*Michael Elliott, University of Hull, UK*

# Reviewed by:

*Mario Barletta, Federal University of Pernambuco, Brazil Joana Patrício, Executive Agency for Small and Medium-Sized Enterprises, Belgium Ricardo Serrão Santos, University of the Azores, Portugal*

> \*Correspondence: *Marianna Mea m.mea@ecoreach.it*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *17 June 2016* Accepted: *30 August 2016* Published: *14 September 2016*

#### Citation:

*Mea M, Newton A, Uyarra MC, Alonso C and Borja A (2016) From Science to Policy and Society: Enhancing the Effectiveness of Communication. Front. Mar. Sci. 3:168. doi: 10.3389/fmars.2016.00168* Dissemination is now acknowledged as an important component of the research process, in particular for European Union (EU) funded research projects. This article builds on the authors' experience during the EU project DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) and aims to assist other scientists to develop a successful dissemination strategy to communicate project achievements. We provide a critical review of the different tools used for outreach to our target audiences, from the academia to the policy makers, and the general public, and try to assess their impact. An effective dissemination strategy and plan should have a clear objective, be designed before the start of the project, identify the target groups and define the methods or tools to be used according to target groups and objectives. The DEVOTES dissemination strategy included two complementary approaches of communication with stakeholders: (i) traditional (e.g., peer reviewed publications, stakeholders workshops, and participation in scientific conferences), and (ii) new (e.g., social networks, smartphone applications) media tools. For each dissemination approach, we defined production targets (e.g., number of articles to be published, individual visitors on the website, etc.) to be achieved by the end of the project, and impact measurements (e.g., citation indices for peer reviewed articles) to monitor the successful implementation of DEVOTES Dissemination. This allowed us to identify which tools had been more (e.g., website) or less useful and relevant (e.g., Facebook) during the project. We conclude that impact measurements cannot be easily identified for all dissemination actions. However, for those that were possible, the DEVOTES dissemination targets were successfully achieved. Overall, the use of the tools and activities outlined in this article, combined with the constant evaluation of the dissemination goals throughout the project duration and the assessment of the effectiveness of the different tools, is essential for the achievement of an effective and timely communication of research results.

Keywords: dissemination strategy, media impact, media tools, ocean literacy networking, stakeholders, training

# IMPORTANCE OF DISSEMINATION/COMMUNICATION OF SCIENCE

# Common Techniques for Communication

Science communication has been defined as "the use of appropriate skills, media, activities, and dialogue to produce one or more of the following personal responses to science: Awareness, Enjoyment, Interest, Opinion-forming, and Understanding" (Burns et al., 2003).

Scientists are not only asked to communicate their findings inside and outside academia, but also to build bridges between research and the society at large and, more importantly, to engage the general public, developing a bi-directional and critical dialogue with the different categories of social actors, (i.e., stakeholders).

Dissemination of scientific results to different target groups is increasingly recognized as a responsibility of scientists (Brownell et al., 2013) that needs the support of other professionals, e.g., journalists, artists, Information Technology (IT) specialists and social networks managers (Uyarra and Borja, 2016). Awareness of the need for better science communication has grown enormously over the last 40 years. The communication of science to different target groups, including the society at large, and the transfer of knowledge is now required in research programmes. Science plays a central role in our life, so policy makers and the wide public are not be able to make informed decisions without understanding the scientific basis (Treise and Weigold, 2002; Fischhoff, 2013).

Science is mainly financed through public funds. Worldwide, numerous organizations (e.g., governments, agencies, foundations) and a large diversity of research programmes are in place to fund research and innovation [e.g., Horizon, 2020 European Union (EU) and National Science Foundation (US) programmes]. Both human and economic resources are being used to this end. Therefore, bridging the gap between science and policy through effective dissemination is a must for such funding programmes to be considered as useful and successful. Although some progress that has been made in disseminating health research output to bridge the gap between science and practitioners (Wilson et al., 2010; Neta et al., 2015), this does not apply to most fields of research. Whether research outputs reach the relevant target groups (e.g., society, consumers, specific economic sectors, decision makers, policy makers, etc.) is yet not well-studied, but it is crucial for societies to become more knowledgeable and reach a better capacity to make informed-decisions.

Indeed, until recent times, not much relevance was given to dissemination and a greater focus was placed on ensuring that scientific outputs were reflected in the scientific literature. The potential impact through the development, dissemination and use of project results was often neglected, both in the call for research proposals and the proposals themselves. Many calls for proposals clearly state the need for dissemination activities to increase impact. Science dissemination is now evaluated in research project assessments and constitutes an important criterion to achieve an outstanding and fundable project (Pohl et al., 2010). Furthermore, there is considerable pressure from the funding agencies for scientists to communicate with and to involve society in research through "citizen science." However, despite its importance, guidance on what it is expected from scientists in terms of dissemination is still weak, and little has been developed as to how the success of any dissemination strategy may be measured.

Taking this into account, the aim of this article is to provide guidance to scientists on planning and implementing an effective dissemination strategy. In order to do so, we first provide a brief overview of the EU approaches to the dissemination of science. We then review the most important dissemination approaches, tools and activities available to a science communicator, and report on their effectiveness and on the difficulties that could be encountered. We illustrate this using the experience gained during the EUfunded project DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status; http://www.devotes-project.eu). In this project, the consortium prepared a dissemination strategy during the planning phase of the project that aimed at maximizing the impacts of the research. We (the Dissemination Team of the DEVOTES project) have collated a number of theoretically and practically informed frameworks that could be used by other scientists as a guide for planning and accomplishing a fruitful dissemination of their project results and outputs, both at the European and the international level.

# The Importance of Science Dissemination for the EU

Over the last decades, the European Commission's economic policy has largely been based on the belief that progress and economic growth are achievable through techno-scientific knowledge and innovation (PotoCˇnik, 2007). Therefore, if society understands the critical role that science and technology plays, public support should follow naturally. The nature of the sciencesociety relationship has shifted since the 80's, but the idea still lies at the heart of Europe's strategy. Back in the late 1980's, sciencesociety issues were considered a problem that could be solved by increasing classic communication efforts. The paradigm "Public Understanding of Science" (Royal Society, 1985) regarded the communication model as a linear function, where dissemination efforts would fill the knowledge gap and would make citizens supportive of science and technology policies.

The 1990's and EU Framework Programme 5 (FP5) were oriented to "Raising Awareness," which stressed that researchers should increase their involvement in dissemination activities. Moreover, through the Marie Curie Actions and the launch of gender mainstreaming (European Commission, 2001), more effort was made to attract Early Career Scientists and women into research.

At the beginning of the millennium, the key concepts of "dialogue" and "participation" were introduced, anticipating new ways of governance in science and technology. The EU FP6 funded the "Citizen and Governance in a Knowledge-Based Society" and "Science and Society" calls. The latter was modified to "Science with Society" in FP7, with the aim of improving linkages between science and society. This stressed the idea of considering science and society as a single entity, increasing the role of the wider public and non-research actors in science policy making, and making the results of publicly funded research more accessible (Wilkinson et al., 2016).

The last step in the recent evolution of the European science communication strategy is constructed around "Innovation Union 2020," where innovation is seen as the key tool for strong and sustainable growth. In this framework, the Responsible Research and Innovation (RRI) concept implies that all societal actors (e.g., researchers, citizens, policy makers, third sector organizations, etc.) work together during the research and innovation process to align its outcomes with the needs, values and expectations of society. One of the key pillars of Horizon 2020 is "tackling societal challenges that are important to all EU citizens and can have a real impact benefitting the citizens." These benefits include:


In summary, the European view on science-society issues has evolved from considering science as a source of rarely questioned knowledge, to a practice deeply intertwined with society (ESF Science and Policy Briefing 50, 2013).

In 2013, the European Commission's launched Horizon 2020 (H2020), a research and innovation programme that will run from 2014 to 2020. H2020 supports scientific research and innovation with an overall budget of approximately €80 billion (European Commission, 2013). The H2020 Communication guidelines (European Commission, 2014) provide a checklist to guide the participants in building a communication strategy specific for their project. This includes guidelines for:


#### Communication Tools

There are various approaches to communicate scientific findings, ranging from more formal (e.g., academic activities, lectures, seminars, production of textbooks, SCI publications) to informal activities (e.g., exhibitions, documentaries, media programs, science clubs and societies, educational games, theater performance, open lectures, festivals, magazine articles, and internet-based tools such as websites, blogs, social media, podcasts, newsletters; Burns et al., 2003). Scientific journalism has traditionally been used as the main format for the communication between science and the public, with the aim of filling in the gaps in the knowledge of the society at large (Treise and Weigold, 2002). However, not all topics are equally covered, and around 70% of scientific journalism coverage is on medicine and health. Scientists used to communicate their results in two main ways: (i) publishing in peer-reviewed journals, and (ii) presenting their findings at conferences. Both these methods are mainly directed to other scientists as most of the scientific journals are accessible only through institutional subscriptions, and conferences are mostly attended by other researchers. More recently, scientists have started to use Internet and social media as means to directly communicate. Innovation in new technologies has led to the development of new approaches, which not only encourage the dialogue between scientists and the general public, but also stimulate people to have an active role in science. In this sense, social media has helped science communication to transform itself from a one-way to a two-way system, where users interact directly with the scientist (**Figure 1**). In addition, citizen science (i.e., the active engagement of general public in scientific research projects, often acting as collectors of data) and crowdfunding (i.e., the request by founders of for-profit, cultural, scientific, and social projects to request funding from many individuals, often in return for future products or equity; Mollick, 2014) are now becoming more and more important in research projects development.

# The Dissemination Experience of DEVOTES

DEVOTES is a EU FP7 collaborative project involving 22 partners distributed across 14 countries in the Atlantic Ocean,

and the Baltic, Mediterranean, Black, and Red Seas. DEVOTES was developed with the main objective of improving our understanding of the relationships between anthropogenic pressures, their influence on the climate and their effects on the marine environment. The project was funded for improving and/or enhancing the effectiveness of ecosystem based management (EBM) in order to fully achieve the Good Environmental Status (GES) of European marine waters, in the context of the European Marine Strategy Framework Directive (MSFD; 2008/56/EC). To achieve this goal, DEVOTES developed a wide set of innovative indicators, models and tools to assist in the characterization, quantification and assessment of marine biological diversity, non-indigenous species, food-webs and seafloor integrity status at an European scale.

The communication strategy of DEVOTES was developed during the preparation of the proposal, with the main aim to build a network with the stakeholders and to provide an effective dissemination of the project achievements. The dissemination activities included an interactive communication dialogue with stakeholders, policy makers and society at large, as well as a uni-directional communication of results. In addition to the traditional approach of dissemination, (e.g., publications, presentations in conferences, organization of workshops, documentaries, etc.), DEVOTES made an effort to define the use and development of new tools to actively involve the different target groups, through the development of apps and the use of social media.

All the planned dissemination activities were directed to achieve the main objectives of DEVOTES. These included building knowledge of the functioning of marine ecosystems (i.e., promoting Ocean literacy, see Uyarra and Borja, 2016), and raising the awareness of the implications of human activities on marine ecosystems. Without this solid understanding, the public cannot make informed decisions and respond in an efficient and timely manner to solve environmental issues.

The next two sections will describe the activities carried out during the lifetime of DEVOTES to disseminate results and progress, and will analyze the performance of each tool.

### DISSEMINATION APPROACHES

# Communication Strategy and Dissemination Plan

Effective communication enhances the impact of a project and the possible uptake of the results. Therefore, the communication strategy of a research project should be discussed in detail and the various phases of the communication strategy should be established during the development of the project proposal. These phases include capturing public interest about the topic, disseminating the project results and outcomes, and finally ensuring and communicating the legacy of the project. The chosen communication approaches should also be established at this stage, as should be the identification of the target audiences.

The different inter-related phases for an effective communication strategy in a research project were taken into account in DEVOTES: the development of a dissemination strategy and plan, and the identification of key reporting elements and of the cross-cutting issues (**Figure 2**).

The communication strategy should be developed by a small communication team that includes, at least, the project coordinator, the webmaster, the graphic designer, and one scientist in charge of the dissemination. The inclusion of additional professionals, such as a scientific journalists and artists would be beneficial to this team. In addition, and to ensure that all work carried out within the project has the potential for equal visibility, each work

package of the project should nominate their communication officer, who will be the contact point for the communication team.

A time-line must be established for the various phases of communication in line with the timing of deliverables, and considering the necessary time-lag to prepare for the dissemination product linked to the specific deliverable (objective) and target audiences, which should also be defined. Once the communication strategy has been discussed, the communication team should draft a dissemination plan. The dissemination plan is a document that is revised at 6 months intervals throughout the duration of the project. It serves as a guide to the communication team and other project members to outline the actions, product outputs and target audiences to be reached during the project. The lead partner(s) for the different actions are also identified. The dissemination plan is a "living document" that can be revised and adapted to accompany the project development. During the project, the details of the various actions that have been undertaken may be added so that the dissemination plan is slowly transformed into the dissemination report as the project is implemented.

The dissemination plan should be structured to include the following sections, although others may also be necessary: (i) an executive summary; (ii) the target audience(s); (iii) the messages; (iv) the tools and mechanisms; (v) the calendar including the post project legacy; (vi) the assessment and monitoring; (vii) the indicators for the evaluation of the dissemination goals, and (viii) the internal communication. Moreover, a SWOT (Strengths, Weaknesses, Opportunities, and Treats) analysis should be included and revised during the project (**Figure 3**). The SWOT analysis is a structured planning method that identifies the internal (strengths and weaknesses) and external (opportunities and threats) factors that are helpful or harmful to achieve a specific objective, and can be a useful tool to evaluate the dissemination strategy of a project. The results of the SWOT analysis determine what may assist the dissemination team in achieving its objectives, and in identifying what obstacles must be overcome or minimized to achieve foreseen results. Additionally,

annexes can be added in the dissemination plan containing tables with details about the venues, participants, link to the products and other pertinent information. Other annexes may include examples of posters, leaflets, and other materials.

Dissemination actions should be targeted at well-defined audiences. The results of a research project may be of interest to the general public, but also to specialists and high-level policy makers. Different means and media of dissemination, vocabulary, and message are appropriate for each of these categories. This audience needs to be informed about the project, its progress, its results, its outputs and its legacy.

In order to maximize the impact of a research project, it is important to engage with all interested parties and communicate the results of the research. "Interested parties" include a wide variety of stakeholders, as well as the "end-users," i.e., those who will be able to make use of the findings, outcomes, and products. For the results to be useful, they should be of interest and easily accessible. Ideally, the identified end-users engage with the project at the design stage. Co-design allows end-users to actively participate and communicate their interests, and help the scientists to co-develop the project so as to maximize its uptake and legacy.

Engaging with the stakeholders can be surprisingly difficult, due to insufficient funds to engage them dynamically resulting in "stakeholder fatigue," because of the multiple requirements both from the project and from other projects on similar topics. There are existing guidelines about stakeholder engagement, such as Durham et al. (2014). For a balanced viewpoint, it is important to engage with different types of stakeholders and to establish a solid discussion with end users and local stakeholders (Saint-Paul and Schneider, 2016).

#### DEVOTES Dissemination Strategy

The DEVOTES Dissemination Team developed its communication strategy during the negotiation phase of the grant and requested that each partner nominate a responsible for the dissemination. Dissemination influences the decisionmaking process, and therefore the first step is to identify the audience, listen to it, identify which decisions are required and therefore what information is necessary (Fischhoff, 2013). The DEVOTES Dissemination Team therefore first focused on building a stakeholder map, identifying the audience and the specific targeted messages, the mechanisms of communication and finally defining a specific timeline for the different activities.

Besides the general public, another six categories of stakeholders were identified as target groups of dissemination, through an analysis of the characteristics of the audience engaged with DEVOTES project: (i) scientists with interest in marine monitoring, biodiversity, and assessment, (ii) higher education institutions, (iii) environmental agencies and/or other institutions operating at the national and regional levels, (iv) decision making authorities, (v) environmental associations, NGOs, fishing, and aquaculture associations, maritime transport associations, port authorities, and (vi) private and industrial stakeholders, including Small and Medium Enterprises (SMEs). The dissemination approach included a strong web presence through a dedicated website, social network accounts, and e-newsletters, participation in conferences and fairs, publication of scientific papers, organization of training activities and networking with other EU funded projects.

The DEVOTES Dissemination Team, with the contribution of all partners, created the database of stakeholders, which now includes more than 1500 contacts in marine environment research and industry. All were contacted early on to introduce them to the project concept through unidirectional communication, emails and the distribution of the electronic newsletter.

#### Traditional Tools

The identification of the audience potentially interested in DEVOTES results and the categorization of the different stakeholder groups were fundamental for the dissemination planning: for each audience cluster identified in the stakeholder map we used dedicated dissemination tools (**Figure 4**). Statistic information about the use of these tools is discussed in Section Evaluation of the Dissemination Goals of this paper.

The Dissemination Team held regular meetings to revise the plan and adapt it to the progress of the project. This resulted, for example, in a deep revision of the homepage layout and website structure 2 years after the beginning of the project and on the participation in Regional Sea meetings rather than organization of workshops.

#### **The website**

Nowadays, the Internet is the primary medium of science communication (Kling and McKim, 2000), and web-based communication is crucial for engaging public audiences with science (Bultitude, 2011). The DEVOTES dissemination strategy included various Internet-based tools the foremost of which was a dedicated website, http://www.devotes-project.eu, used as the main communication channel for the project management, achievements, and progress. A special effort of the Dissemination Team was focused on developing an eye-catching layout and a user-friendly website map. The website, dedicated to all stakeholder categories, was developed by graphic designers, under the supervision of the project coordinator and in accordance with the EU guidelines. The website has been constantly and timely updated with news, promotional material and new project products. The site map included six main sections:



A full set of informative and promotional material, including factsheets, policy briefs, brochures, and posters, was produced during the lifetime of the project to promote the release of reports, software tools and deliverables. All the promotional products, the website and templates (for presentations, posters, reports, minutes of meetings) were developed using the corporate image of the project, always including the DEVOTES logo and using a consistent color code.

Special attention was dedicated to the early career researchers (ECR), within and outside the project: the Young Scientist Corner included a series of interviews "PhD students of the Month," as well as announcing job opportunities, post-graduate modules, summer schools, and training activities.

#### **The newsletter and email campaigns**

The dissemination campaign of DEVOTES was launched with the publication of press releases in the countries of the members of the consortia. This was followed by an email campaign presenting the project and launching the website to all the potential stakeholders. The mailing campaigns continued with a regular electronic newsletter (approximately every 6 months), brief news (every 3 months), and monthly updates on the project progress. All the issues of the newsletter have been made available for download on the project website and promoted via the project social networks.

To enhance the communication inside the consortium, distribution lists were created at Work Package and Task level, for the General Assembly, the Steering Committee members, and for Advisory Committee members. Moreover, in addition to the Partners' Area of the website, a sharing platform has been included among the e-media tools available for the participant to the project.

#### **Scientific publications**

In order to better communicate the scientific results, not only within the scientific community but also to decision and policy makers, all the scientific papers produced in DEVOTES have been made Open Access, either with the gold road, paying the fee for the open access, or with the green road, self-archiving the article. As indicated above, academic institutions subscribe to the different journals, but usually they can only afford the subscription to a small fraction of them. This situation decreases the potential usage and impacts of research, which would be maximized if all research papers were Open Access (Canessa and Zennaro, 2009). Open Access enhances the research cycle, improves the access to international research outputs and the impact of the research. There is a correlation between Open Access publication and citation-count, increasing this from 50 to 250% (Canessa and Zennaro, 2009). Additionally, articles in Open Access are immediately available for free consultation and download and, more importantly, permanently preserved in journals digital archives.

The Dissemination Team created a repository of scientific papers produced during DEVOTES life, named "FP7 EU DEVOTES Community" in Zenodo, the OpenAIRE "orphan repository" available under the link https://zenodo.org/ collection/user-devotes-project. With this repository DEVOTES is accomplishing one of the most important objectives of the FP7 Programme, which is the free access to all the research outcomes to scientists and public at large. In addition to Zenodo, the Dissemination Team created a Google Scholar profile for DEVOTES in which all papers are listed, (https://scholar.google. it/citations?user=oSH2JTkAAAAJ&hl=it&oi=pll). This allows scientists to easily obtain information on all the papers published by the project, consult the citations received by each paper, rank them, and obtain the Hindex of the project, as an index of the success of the project scientific outcome.

As Open Access publications lead to wider and more efficient dissemination of information, the dissemination strategy of DEVOTES included also the production of an ebook, reporting the scientific results and products developed during the project. The ebook, composed by the articles published in this Research Topic will be freely available for download from the website of the project. Moreover, the ebook will be part of one of the applications for smartphone, which will be available by the end of DEVOTES project (October 2016).

#### **Workshops and participation to conferences**

The engagement of stakeholders is crucial to reach the objective of generating improved interfacing mechanisms in the management process, among science, policy, and decision makers and the general public. This can be achieved through targeted workshops, conference sessions, and webinars. Once more, the dissemination has to be tailored to the audience. The scientists working in related fields and projects are more easily reached at special sessions in conferences. Practitioners working at environment agencies, either regional or national are best reached through specially organized workshops, if possible using locally relevant materials as examples. International practitioners, such as the Regional Seas Conventions, European Environment Agency and expert groups (e.g., "Good Environmental Status working group"), are best reached at workshops back-to-back with pre-organized meetings. This both increases the likelihood of participation and reduces travel expenses. It is essential to distribute targeted information that explains the workshop well in advance of the meeting, so that the attendees may register and prolong their stay to participate.

Companies and SMEs are more difficult to contact as a group. Environmental consultancy firms may be in competition with each other, and so reluctant to have a joint meeting, and it may be therefore necessary to have individual or small group meetings. However, it was easier to organize group workshops and meetings for other potential end users, for example aquaculture firms that rely on marine good environmental status.

#### **Documentaries**

Films and documentaries are one of the most powerful communication and educational tools (Barnett et al., 2006; Hooper et al., 2011), engaging the public in critical thinking and enhancing public awareness in environmental issues (e.g., climate change, pollution, acidification). The production of documentary films has grown significantly in the past decade, and the distribution of documentaries through the Internet created new opportunities to create societal impact (Karlin and Johnson, 2011). Platforms such as YouTube, iTunes, and Vimeo make online videos easier to be made available, accessed, used, and shared. With the aim of increasing the potential impact of DEVOTES, the dissemination strategy included the preparation of a documentary illustrating the background and the main results of the project. DEVOTES was selected by "Futuris," the award-winning program of EuroNews on European science, research and innovation, as a successful example of project studying the effects of human activities on marine ecosystems, to raise general interest about the environmental status of European seas. The episode "Improving our understanding of our seas" went on air for 1 week and was then made available on the programme EuroNews YouTube channel. The DEVOTES documentary prepared by the project team will be ready in October 2016. A professional company (partner of the project) worked on the details of the storyboard, collecting videos, interviews and images from the DEVOTES partners. It will be broadcast via Internet-based channels (YouTube, Vimeo), available from the project website and promoted via the project social network accounts. A wide audience will be reached by the use of e-media tools for the promotion of the film to increase the social impact.

#### **Training activities**

Training activities and summer schools are an important part of dissemination. They provide for the legacy of a project by disseminating the project results to end users, such as postgraduate students and practitioners. Whereas students enrolled in postgraduate courses may benefit from taught modules, practitioners usually do not have the time or professional freedom to enroll in long-term training courses. Focused and short summer schools therefore provide an important opportunity for practitioners to learn complementary skills. The uptake of scientific results published in scientific papers and text books into curricula usually has a long time lag, sometimes lasting several years. Hence, including the training into postgraduate and summer schools, which can be attended by practitioners, fast-tracks the information to current end-users and those about to enter the job market (postgraduates).

A successful training course should be disseminated to potential end-users in a timely manner. In this way, interested candidates can plan to attend, if they are fully employed, or plan to select the course if they are post-graduates. The information provided should include the necessary context so that the candidate understands what training will be on offer and why they would benefit from attending. The training programme should include the knowledge and skills that will be learned when completed.

In the DEVOTES project, the consortium organized four summer schools to disseminate current "hot topics" addressed throughout the life of the project by the different partners. The topics covered were: genomic tools applied to monitoring; new modeling applied to assess the status of marine systems; innovative, and integrative ecosystem quality assessment tools; and ecosystem services provided by seas. DEVOTES Summer Schools have attracted both early career and senior researchers alike. Keynote talks were given around the specific topics listed above. Unlike the classic symposium format, where attendants are exposed to many but very short presentations, the longer length of the talks in these Summer Schools allowed the speaker to extensively expose different aspects of the subject and disseminate the results of the project in detail.

In addition to the primary dissemination and training tasks, these summer schools had other important objectives: (i) networking with scientists not involved in the project, either as professors or attendees, to bring fresh ideas into the project tasks and deliverables; (ii) give the opportunity to managers, Ph.D. students, Post-Doc, and scientists attending the school to learn about emerging concepts that can be incorporated into their daily research; (iii) disseminate the findings among more ample communities, e.g., through the collaboration with organizations such as EuroMarine, an European marine research network (http://www.euromarinenetwork.eu); and (iv) publish position papers on the topics addressed, which can be a direct (e.g., Borja et al., 2016) or indirect (e.g., Bourlat et al., 2013; Piroddi et al., 2015) result of the school. The Summer Schools have spread the findings of the project to an ample audience, covering more than 30 countries from all continents. A qualitative analysis of the Summer Schools is reported in Section Impact Analysis.

In addition to summer schools, other ways of training have been explored and implemented in DEVOTES. The use of webinars (online live courses) has been used as means to train on specific topics. As indicated above, there is often interest for learning but difficulties in accessing such knowledge. In the case of physical courses, this might be difficult for those working full time or having limited time or economic resources. To overcome such issues, webinars can be a realistic solution. In DEVOTES, webinars have been used to train key stakeholders on the most relevant tool developed under the project. With a total participation of 76 relevant stakeholders, and feedback received, it can be considered a very cost-effective means for communicating and practical training. The webinars are also available on the website of the project, together with short, YouTube training videos, and guidelines.

#### New Tools

#### **Social media**

Internet platforms, mobile applications (Apps), and social media have now also become resources to share research progress and to learn. All these tools represent a unique opportunity for scientists to enhance ocean literacy, "understanding of the ocean's influence on you—and your influence on the ocean," (Carley et al., 2013), allowing citizens to take informed decisions and to be able to participate in public debate about ocean health (Fauville et al., 2014).

Generic and professional social media tools, such as ResearchGate, LinkedIn, Facebook, Twitter, or Instagram have exploded in popularity in the last decade, attracting more and more scientists to using them. As mentioned above, online presence is fundamental for science communication and, together with social media, offers a wide range of benefits for scientists: boost their professional profile, enhance professional network, improve research efficiency and scientific metrics (Bik and Goldstein, 2013; Jucan and Jucan, 2014). Using social networks to promote research results and paper publications has been proved to increase the number of citations of their articles and the Hindex (Liang et al., 2014). A strong presence on social media may result in papers having 11 times more possibility to be cited vs. articles lacking of social media presence (Eysenbach, 2011). Additionally, generic social networks offer the opportunity to reach a wide range of people with a more or less developed personal interest in science and to develop that interest (Fauville et al., 2014).

DEVOTES has been present on a few, carefully selected social media tools, both professional and generic, to take advantage of the specific features of each one (pros and cons of the different media tools will be discussed further in Section Comparison of Different Media Tools). The DEVOTES Dissemination Team created an account and a discussion group in LinkedIn, with 206 members, which served as tool to improve sharing knowledge with other scientists and industry professionals in the marine and environment fields, to enhance the ocean literacy among these two target groups. DEVOTES made its social debut early in 2013 (ca. 6 months after the beginning of the project), using the most popular platforms: Facebook, https://www.facebook.com/ Devotesproject/), Twitter (@DEVOTESproject), and YouTube. The social media campaign included publishing posts at least three times per week from the project and project coordinator accounts.

To make DEVOTES appealing for the general public and decision makers, the DEVOTES Dissemination Team published posts on the website and social networks on environmental days (e.g., the 22nd March World Water Day, 8th June World Oceans Day), linking the project activities with the topic of each day. For example, on the International Day of Biodiversity (22nd May) we linked its topic "Mainstreaming Biodiversity; Sustaining People and their Livelihoods" with the main message of the DEVOTES Final Conference: "Marine biodiversity is the key to healthy and productive seas."

Other messages were dedicated to different categories of stakeholders (e.g., environmental agencies, consulting companies) and therefore included more technical aspects, such as the production of the Catalogue of Monitoring Networks and the development of NEAT, the Nested Environmental status Assessment Tool.

#### **Mobile apps**

The innovation in mobile computing technologies and their affordability make the learning process possible using mobile applications ("apps" hereafter). Small devices, such as smartphones and tablets, are now part of our daily life, have strong computing power and they are potentially always connected. Applications for smartphones and tablets are considered useful communication tools, which are able to reach out further than our scientific reports and publications do, including society at large (Hsu and Ching, 2013). Therefore, mobile devices represent a great opportunity for education, science communication and ocean literacy. To this end, the DEVOTES dissemination strategy included the development of mobile applications. Two apps already available are "DevoMAP" and "MY-GES." Another two are planned to be released by October 2016. All apps will be available for iOS and Android devices and downloadable from the project website. DevoMAP and MY-GES aim to disseminate the results from innovative modeling to a wide audience, and to attract the attention of the public, including scientists involved in assessments of GES in European regional seas and those not involved in marine environmental assessments. "DevoMAP" focuses on people directly involved in research and policy, to support the implementation of the MSFD. "MY-GES" targets people interested in our achievements among the general public. By targeting the general public, we aim to make society aware about the Marine Strategy Framework Directive, its implementation and assessments of environmental status. The other two apps will focus on the dissemination of overall project findings: "DevoBook," as a result of this issue of Frontiers, and "DEVOTES," an interactive app for the general public, including key questions and findings from all DEVOTES Work Packages and promotional material produced during the project lifetime.

#### **Artistic Elements**

The use of arts in science communication is still poor but a study, conducted by Curtis et al. (2012), showed that ecologists are willing to use the arts in a scientific forum to promote their results. In particular, they think that the visual (e.g., painting) and performing (e.g., ballets, theater plays) arts can be very useful in communicating scientific information.

In 2015, DEVOTES decided to include a visual artistic element in its dissemination strategy. In collaboration with the EU project CoCoNet (Toward COast to COast NETworks of Marine Protected Areas coupled with sea-based wind energy potential), a calendar was produced to be distributed to the project stakeholders at the end of the year. The topic of the calendar was the MSFD implementation, including an artistic interpretation of the 11 MSFD descriptors of GES, which define how to assess the quality of EU marine systems. Each descriptor was represented in an evocative illustration, associated to each month, and briefly outlined in the explanatory text. December's plate describes an ideal observation system, to monitor environmental quality standards, and integrate the information to assess the status and achieve GES (**Figure 5**).

The Calendar, distributed to more than 800 relevant stakeholders, was also made available for download from the website, and in only 3 months the page received more than 600 visits.

# The Importance of Networking with Other EU Projects

Taking into account the integrative view of DEVOTES, it was necessary to collaborate with other international, European and regional projects, creating a strong network across Europe and overseas. The tasks and approaches have been multiple. These include:


These interactions have resulted in undertaking a real interand trans-disciplinary research (Lang et al., 2012), allowing DEVOTES to go farther beyond the state of the art. This could not have been possible with the resources of only one project.

# EVALUATION OF THE DISSEMINATION GOALS

# Impact Analysis

The key issue of success of a dissemination tool depends on the ability to supply information and to transfer knowledge to the stakeholders and the potential users (Vermeulen et al., 2009), and then for stakeholders and potential users to use this knowledge. In order to evaluate the success of DEVOTES in terms of public engagement, we present here the quantitative analysis of each dissemination tool discussed above. To assess the performance of the dissemination activities on the web, several analytical tools are being used. All statistical data were regularly analyzed and compared with the impact target identified during

FIGURE 5 | December's plate of the DEVOTES/CoCoNet calendar (Copyright: Alberto Gennari).


TABLE 1 | Impact targets of the main DEVOTES dissemination tools/mechanisms.

the preparation phase of the project (**Table 1**) in order to measure the success and usefulness of the different tools.

To record the accessibility of DEVOTES website, Advanced Web Statistic 7.0 (AWStats, 2010) is being used to analyze the DEVOTES server log files from October 2012 until 2 years after the end of the project. Here, we present the results from October 2012 to April 2016 (**Figure 6**). It can be seen that, besides predictable decreases during summer and holiday seasons, use of the website increased until January 2016. Between January and April 2016, a reduction of the DEVOTES social media presence due to other commitments, led to a decreased interest in the website. An average of 2600 visits have been registered per month, with peaks of up to 10,000 hits during the release of the newsletters (e.g., June, September, and November 2013), the annual meetings (e.g., December 2014 and 2015), the revision of the website (March 2015) and peaks in social network activity (e.g., October 2013). A large proportion of the visitors came from Europe, but the website received visitors also from USA, Africa and Asia. Most of them reached the website via direct link, search engine (i.e., Google) and from external pages (i.e., DEVOTES newsletter and LinkedIn).

been produced and the work of the documentary is running

In order to evaluate the scientific impact of the whole project, two analytical tools were used to monitor the citations: Google Scholar Citations on the Google Scholar DEVOTES profile, and Altmetric, on the Zenodo DEVOTES community. Google Scholar Citations provide the user with several citation metrics. The DEVOTES papers (139, as of 18th August 2016) have a cumulative Hindex of 18 and 1083 citations overall. The Altmetric Analytical tool shows the online attention and activity that have been found for each specific article, collecting relevant mentions from social media, newspapers, policy documents, blogs, Wikipedia, and other sources.


TABLE 2 | E-media users in DEVOTES and other EU projects in the framework of Ocean of Tomorrow initiative (FP7-OCEAN).

*N/A, not available; N/U, not used. AQUO, KILL-SPILL, SONIC, BIOCLEAN, and STAGES do not have any e-media tool (no social media presence).*

The E-media analytical tools and results to evaluate the social media impact of DEVOTES are reported in **Table 2**, together with the statistics from other "Ocean of Tomorrow" projects started the same year (2012). If we compare the number of social media users, it appears clear that, besides the Facebook page, DEVOTES was able to successfully build its own social community, both in generic (i.e., Twitter) and in professional social media (i.e., LinkedIn).

As the project progressed, there was a positive tendency as more followers (Twitter)/fan(Facebook)/professionallinks(LinkedIn) were registered. The traffic on social pages also followed from other dissemination activities, such as the DEVOTES presence in conferences, the organization of summer schools and special sessions, and the participation to global campaigns (i.e., Ocean Sampling Day) and citizen science projects (i.e., My Ocean Sampling Day).

The impact of a successful project dissemination may result in the reassessment and enhancement of the effectiveness of relevant policies, the use of the project results by stakeholders and decision makers, and the creation of business opportunity, as well as s sharing new science-based knowledge.

In order to evaluate the impact of DEVOTES results for policy and decision makers, we monitored the amount of downloads of reports and/or deliverables (**Table 3**). The number of people visiting and downloading some of the reports and deliverables was very high, going far beyond the amount of persons directly involved in the project (around 200).

In addition to these quantitative evaluations, the DEVOTES Dissemination Team carried out also a qualitative evaluation on the Summer Schools and the internal dissemination activities. Satisfaction surveys conducted after each Summer School indicate that attendants were satisfied with the event. From the 61 participants in the Summer School of 2015 who answered to the satisfaction questionnaire, 67% made at least one contact for future projects and general satisfaction was scored with 8.25/10 (±1.32). However, some of the comments show that attendants were expecting a more interactive format and more opportunities for networking. Therefore, Summer Schools willing to attract students should make an effort to schedule activities with different level of participation.

#### TABLE 3 | First five products most downloaded from the DEVOTES website (2012–2016).


# Comparison of Different Media Tools

The advancements in information and communication technology are leading to a rapid change in the world of science communication, which is now faster and more interactive. The abundance and diversity of online media sources led to an increased amount of content on offer (Porten-Cheé and Eilders, 2015). Scientists should be present in different arenas and make an effort to interact with the general public. DEVOTES took advantages of different new and traditional media tools (**Table 4**), with the aim of building a "DEVOTES community" which goes beyond the scientific community. If we compare the different dissemination methods used and their performances, it is clear that traditional (e.g., the website) and innovative (e.g., Twitter) tools are strongly related, and that an efficient use of the latter have a positive feedback on the performance of the former. In fact, after our experience in using the different tools during the DEVOTES project, we can rank the different media taking into account their usefulness and costbenefit: (i) very useful: website, open access publication, sessions at international conferences, stakeholders workshops, Twitter; (ii) useful: summer schools, LinkedIn groups, press releases; (iii) moderately useful: videos, newsletters; and (iv) not very useful: Facebook, smartphone apps.

All the innovative tools should be used as complementary outlet to the traditional tools for the dissemination of new



posts from the project website, to share articles, advertise job opportunities, and training events, promote meetings and circulate information about the project progress and results. This should include media that have been shown not to be very useful in the DEVOTES project such as Facebook and mobile apps, reaching audiences familiar with these media. In some cases, the lack of usefulness may be related with the longer time of maturation needed to reach a large audience, such as in the apps. However, not all media tools are necessary: the revision of the dissemination plan and the performance analyses should help to shape the social media strategy, also identifying which tools are redundant (e.g., Facebook and Google+), to avoid overlap. In the case of DEVOTES social media, we decided to focus our attention and efforts on Twitter campaigns, LinkedIn group discussions and website updates, although the Facebook account and the YouTube channel were still active.

#### Difficulties in Engaging the Stakeholders

Common difficulties encountered during dissemination to the different target group include sharing information between projects, engagement of local stakeholder, copyright, and open access. Researchers have often participated in previous, related projects but may face some constraints about sharing information. For example, contact details of stakeholders may be protected by privacy laws and therefore the effort of stakeholder mapping may have to be repeated. Conference organizers may also face constraints about distributing the contacts of participants. Another constraint is about data sharing. This may result from a number of issues. Often the data may have been previously collected by a team, of which only one member participates in the new project. This person may therefore not be able to share the data as they are not the sole owner of the data. Another typical example is about data format. Data may exist in a different format, and in the case of historical data, it may only be available in paper reports. The transcribing of such data into digital format can be a very onerous and thankless task. Other examples are obsolete storage such as floppy disks, or storage using obsolete software programmes. Trivial examples include different formats such as using a decimal point vs. a decimal comma or apostrophe. Units may also need to be converted, such as concentration in mass/volume instead of molar concentration.

Copyright and open access of information is another common problem. National or internationally funded research often requires that results be publically available or in "open access" format. While many publishers now offer that option, it comes at a price. The project participants may not have budgeted for such costs. A successful project that may publish about 200 articles may have open access costs of more than 500,000 Euros, a significant proportion of the budget. Making articles freely available without using open access, even for research and educational purposes, may infringe copyright laws.

The engagement of local stakeholders, and crucially of possible end-users, can also be problematic. First it is important to identify these potential stakeholders, and then be able to contact them. Once more, even if one project partner has this information, they may not be able to share it with the other project partners. Once the contact details are known, then the stakeholders are best approached personally, rather than through "mass" email messages. The dissemination team should communicate why the contact is considered to be an important stakeholder. How the stakeholder may participate in the co-design of the project at the onset and the project, how they may participate in the product development phase, and finally how the project information may be of use to the stakeholder, are also relevant points.

### Difficulties in Engaging the Wide Public

The health and state of our marine environment and the ecological changes being detected and predicted for the future are a global area of interest. No matter how far we live from the sea, the ocean has a strong influence on Human life, providing food energy, moderating climate, and playing an important role in the economic prosperity of many regions. Yet, the common knowledge and understanding of the oceans is not spread enough among the general public and decision makers.

A large part of the general public still obtains their science news from traditional media, such as television, and print newspapers, but internet-based tools are becoming more widely used among teenagers and young adults. Going online regularly and using Google searches now represent the standard approaches for discovering information about a topic (Bik and Goldstein, 2013). However, people feel overwhelmed by the amount of information available.

Another common problem in disseminating EU research project findings is the translation and cultural adaption of the dissemination tools/mechanisms. Most of the material is produced in English, and only selected products are translated into local languages. Moreover, although people think scientists and policy makers should be engaging in dialogue with the public about science, this is not always translated into a willingness to be personally involved. The general public tend to think that is the role of "experts" and not theirs to advise the governments on science issues. However, people show more interest in research and science when they can be directly involved in the project: citizens are more motivated if they can "actively" contribute to science advancements. If people do not see how they can make the difference or being actively involved, they may lose interest. To this end, we suggest that citizen science activities should be included in research project proposals.

#### CONCLUSIONS

An effective science communication allows people to make sound choices (Fischhoff, 2013) about environmental issues, and help key actors to improve processes and methodologies in marine environment management. From our perspective, the most useful media tools used to disseminate DEVOTES have been the

#### REFERENCES


There are several factors influencing the dissemination of European funded projects, such as the limited project duration (e.g., 2–4 years), which could threaten the dissemination of end products, (see "Threats" reported in **Figure 3**). This in turn could influence the assessment of the dissemination impact to the stakeholders and the general public. To solve these risks, we suggest to include periodic (at least every year) web-based and physical surveys to monitor the effectiveness of results. Additionally, recent studies reveal that, although having a positive view of science and technology, EU citizens think scientific research is difficult to understand and that scientists should be more effective in communicating scientific results (European Commission, 2007, 2010). Our suggestion is to include (where possible) a citizen science initiative in the communication strategy, in order to actively involve the general public, not only in the collection of data but also in the dissemination process (e.g., increasing the social media audience and presence). In fact, the lack of a citizen science initiative was the factor determining the low success of the DEVOTES Facebook page (see "Weakness" reported in **Figure 3**).

Therefore, it is fundamental to develop an effective dissemination strategy at the moment of writing a research project proposal, and to perform a constant evaluation of the dissemination results before, during and after the project lifetime, involving all the key actors, advisory board and partners (see "Strengths" reported in **Figure 3**). To achieve this, the use of different media tools, targeting them to the adequate audience, will ensure the success of the project, by making available all the outcomes and products to the end users.

# AUTHOR CONTRIBUTIONS

MM wrote a first draft of the manuscript, then AN, MU, CA, and AB contributed equally to the manuscript.

#### ACKNOWLEDGMENTS

This manuscript is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), http://www.devotes-project.eu. MU is partially funded through the Spanish programme for Talent and Employability in R+D+I "Torres Quevedo." The authors thank Ulisse Cardini for his support with the figures and valuable comments on the manuscript.

of science. J. Sci. Technol. 15, 179–191. doi: 10.1007/s10956-006- 9001-y


ecosystem approach in practice. Front. Mar. Sci. 3:20. doi: 10.3389/fmars.2016. 00020


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Mea, Newton, Uyarra, Alonso and Borja. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Bridging the Gap between Policy and Science in Assessing the Health Status of Marine Ecosystems

Angel Borja<sup>1</sup> \*, Michael Elliott <sup>2</sup> , Paul V. R. Snelgrove<sup>3</sup> , Melanie C. Austen<sup>4</sup> , Torsten Berg<sup>5</sup> , Sabine Cochrane<sup>6</sup> , Jacob Carstensen<sup>7</sup> , Roberto Danovaro<sup>8</sup> , Simon Greenstreet <sup>9</sup> , Anna-Stiina Heiskanen<sup>10</sup>, Christopher P. Lynam<sup>11</sup>, Marianna Mea<sup>12</sup>, Alice Newton13, <sup>14</sup> , Joana Patrício<sup>15</sup>, Laura Uusitalo<sup>10</sup>, María C. Uyarra<sup>1</sup> and Christian Wilson<sup>16</sup>

*<sup>1</sup> AZTI, Marine Research Division, Pasaia, Spain, <sup>2</sup> Institute of Estuarine & Coastal Studies, University of Hull, Hull, UK, <sup>3</sup> Departments of Ocean Sciences and Biology, Memorial University of Newfoundland. St. John's NL, Canada, <sup>4</sup> Plymouth Marine Laboratory, Plymouth, UK, <sup>5</sup> MariLim Aquatic Research GmbH, Schönkirchen, Germany, <sup>6</sup> SALT, Lofoten, Norway, <sup>7</sup> Bioscience, Aarhus University, Roskilde, Denmark, <sup>8</sup> Department of Life and Environmental Sciences, Polytechnic University of Marche, Ancona, and Stazione Zoologica Anton Dohrn, Naples, Italy, <sup>9</sup> Marine Scotland – Science, Marine Laboratory, Aberdeen, UK, <sup>10</sup> Marine Research Centre, Finnish Environment Institute, Helsinki, Finland, <sup>11</sup> Centre for Environment, Fisheries, and Aquaculture Science, Lowestoft, UK, <sup>12</sup> Ecoreach, Ancona, Italy, <sup>13</sup> NILU—IMPEC, Kjeller, Norway, <sup>14</sup> CIMA, University of Algarve, Faro, Portugal, <sup>15</sup> European Commission, Joint Research Centre, Directorate for Sustainable*

#### Edited by:

*Christos Dimitrios Arvanitidis, Hellenic Centre for Marine Research, Greece*

#### Reviewed by:

*Matt Terence Frost, Marine Biological Association of the United Kingdom, UK Marco Uttieri, Parthenope University of Naples, Italy*

> \*Correspondence: *Angel Borja aborja@azti.es*

#### Specialty section:

*This article was submitted to Marine Ecosystem Ecology, a section of the journal Frontiers in Marine Science*

Received: *15 June 2016* Accepted: *31 August 2016* Published: *12 September 2016*

#### Citation:

*Borja A, Elliott M, Snelgrove PVR, Austen MC, Berg T, Cochrane S, Carstensen J, Danovaro R, Greenstreet S, Heiskanen A-S, Lynam CP, Mea M, Newton A, Patrício J, Uusitalo L, Uyarra MC and Wilson C (2016) Bridging the Gap between Policy and Science in Assessing the Health Status of Marine Ecosystems. Front. Mar. Sci. 3:175. doi: 10.3389/fmars.2016.00175* *Resources, D.2 Water and Marine Resources Unit, Ispra, Italy, <sup>16</sup> Oceandtm Ltd. Lowestoft, UK* Human activities, both established and emerging, increasingly affect the provision of marine ecosystem services that deliver societal and economic benefits. Monitoring the status of marine ecosystems and determining how human activities change their capacity to sustain benefits for society requires an evidence-based Integrated Ecosystem Assessment approach that incorporates knowledge of ecosystem functioning and services). Although, there are diverse methods to assess the status of individual ecosystem components, none assesses the health of marine ecosystems holistically, integrating information from multiple ecosystem components. Similarly, while acknowledging the availability of several methods to measure single pressures and assess their impacts, evaluation of cumulative effects of multiple pressures remains scarce. Therefore, an integrative assessment requires us to first understand the response of marine ecosystems to human activities and their pressures and then develop innovative, cost-effective monitoring tools that enable collection of data to assess the health status of large marine areas. Conceptually, combining this knowledge of effective monitoring methods with cost-benefit analyses will help identify appropriate management measures to improve environmental status economically and efficiently. The European project DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing good Environmental Status) specifically addressed these topics in order to support policy makers and managers in implementing the European Marine Strategy Framework Directive. Here, we synthesize our main innovative findings, placing these within the context of recent wider research, and identifying gaps and the major future challenges.

Keywords: environmental status, marine health, status assessment, management, ecosystem approach, socioecology

# INTRODUCTION

A recent assessment of marine ecosystem ecology identified eight grand research challenges (Borja, 2014): (i) understanding the role of biodiversity in maintaining ecosystem functionality; (ii) understanding the relationships between human pressures and ecosystems; (iii) understanding the impacts of global change on marine ecosystems; (iv) developing integrative assessment of marine ecosystem health; (v) ensuring delivery of ecosystem services by conserving and protecting the seas; (vi) understanding the way in which ecosystem structure and functioning may recover through restoration; (vii) understanding the need for an ecosystem approach and integrated spatial planning in managing ocean use, and (viii) developing better ecosystem models to support more effective management.

These challenges reflect widespread recognition of clear effects of pressures from established and emerging human activities on marine ecosystems (Halpern et al., 2015) and, consequently, the potential of those pressures to alter the ability of ocean ecosystems to provide services that yield societal and economic benefits (Barbier et al., 2012; Turner and Schaafsma, 2015). Given the multiple pressures society places on marine ecosystems and the broad range of services they provide, a holistic assessment (Borja et al., 2016) of the status of marine ecosystems requires scientific evidence-based Integrated Ecosystem Assessments (IEA; Levin et al., 2009). Indeed, the former European Commissioner for Environment, Janez Potocnik, stated during the closing session of Euromares ˇ 2010, on the occasion of the European Maritime Day, that: "We are learning that the [Marine Strategy Framework] Directive has a weakness—and that weakness is the lack of knowledge." With a lack of knowledge "...these unknown variables pose a real problem for decision-makers. They need to be identified and addressed in a systematic way. And while we need to acknowledge the differences and diversity of our seas, there are some issues which can only be adequately addressed on a European scale." These statements capture the desire of policy-makers and managers worldwide to fulfill their moral mandate to conserve and protect the seas (Reker et al., 2015) using evidence-based decision-making. Hence, the vision for clean, healthy, biodiverse, and productive oceans and seas with sustainable resource use requires bridging the gap between policy and science in assessing the status of marine ecosystems by increasing scientific knowledge of marine ecosystems and their functioning, including humans and their role as part of the ecosystem (Borja et al., 2013). Indeed, recent European and national policies enshrine the vision of healthy and biologically diverse seas (e.g., DEFRA, 2002; European Marine Board, 2013). More recently, the European Union and United Nations have tried to address problems associated with exploitation of deep fishing resources and associated impacts on biodiversity (St. John et al., 2016).

The development and implementation of policy and legislation globally demonstrate a significant effort to improve the status of the seas, including an ecosystem approach to ocean use management (Browman et al., 2004; Nicholson and Jennings, 2004; Borja et al., 2008, 2016; Curtin and Prellezo, 2010). In the European Union (EU), the Marine Strategy Framework Directive (MSFD; European Commission, 2008) represents the most comprehensive marine environmental legislation. This Directive aims to achieve Good Environmental Status (GES) by 2020 in the four European Regional Seas (Baltic, North Eastern Atlantic, Mediterranean and Black Sea). The MSFD requires that Member States assess ecosystem characteristics, pressures, and impacts with respect to 11 descriptors related to: biological diversity, non-indigenous species, commercial fish and shellfish, food-webs, eutrophication, seafloor integrity, hydrographic conditions, concentration of contaminants in the environment and in fish and other seafood consumed by humans, marine litter, and introduction of energy including underwater noise. Within these 11 descriptors, the European Commission (2010) then defines 29 criteria and 56 indicators necessary in evaluating environmental status.

The assessment of environmental status, while scientifically challenging (Stanley, 1995), simultaneously offers many opportunities for European marine research to support an ecosystem approach to environmental management, which EU Member States have agreed to implement (Borja et al., 2013). The European project DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing GES, www.devotes-project.eu) was started in 2012 to facilitate MSFD implementation. This project considers these complex, inter-related scientific issues and management needs of the MSFD, as well as the challenges shared by the four regional seas identified within the MSFD. Its main objectives were:


We therefore set an overall goal of better understanding the relationships between pressures from human activities and climate change, and their effects on marine ecosystems, including biological diversity, in order to support ecosystembased management and attain GES of marine waters. Our harmonized approach to the four European regional seas tested and validated existing indicators, created new indicators when necessary, developed modeling tools for the assessment of biodiversity, tested new monitoring tools and established an integrative approach for assessing environmental status.

This overview describes how this research has contributed to advancing the state-of-the-art since 2012 in bridging the gap between science and policy in marine environmental status assessment. Specifically, this addresses elements such as human pressures, indicator development, model use, innovative monitoring, and integrative assessment tools), in order to achieve healthy and sustainable ocean use. Here we synthesize key responses to major environmental questions and the lessons learnt. This information will support managers and policymakers in making decisions for improved management of ocean use.

# WHY MUST WE UNDERSTAND IMPACTS OF HUMAN ACTIVITIES AT SEA?

### State-of-the-Art

Marine environmental managers primarily aim to protect and maintain natural structure and functioning while simultaneously ensuring that ecosystems provide services, which in turn deliver benefits for society (Atkins et al., 2011; Elliott, 2011). In the management of human activities in the marine environment, it is axiomatic that a regulatory body (i.e., an environmental protection agency, natural conservation body, fisheries body, or marine licensing body), does not have to prove that an activity or its developer (the "user," "polluter"—those undertaking the activity, such as a dredging company, industrial plan, or wind farm operator) causes an adverse impact (Gray and Elliott, 2009). In contrast, the developer must prove they will not cause an impact, hence creating the scientific and statistical challenge of "proving the negative." A second key feature, "the precautionary principle" (PP), assumes a deleterious effect resulting from a given activity in the system unless proven otherwise (O'Riordan and Jordan, 1995). However, detractors criticize the vague definition of PP, and balancing scientific uncertainty and appropriate management measures remains a challenge (Steel, 2014).

The third key feature states that any developer wishing to use the marine system must obtain permission from a regulatory body, hence the importance of sufficient administrative bodies (Boyes and Elliott, 2014, 2015; Elliott, 2014); this encompasses the whole of marine governance, defined as the net result of policies, politics, legislation, and administration (Barnard and Elliott, 2015). The fourth feature, the "polluter pays principle," requires a developer to pay for the costs associated with that use: the licensing of the activity, the monitoring, remediation and mitigation of any damage to the system and, if necessary, compensation. The latter requires integrating natural and economic sciences to enable sustainability within and across generations and it may require developers to compensate affected users, the affected resource (e.g., restocking affected fish), or the affected environment (e.g., by creating new environment; Elliott et al., 2016). However, all of these central features relate to how users use an area of the sea (e.g., dredging, wind farm, fishing, etc.) but superimposing a wider suite of natural and human influences, such as climate change, on all of these activities (Elliott et al., 2015). This complexity demands, as the fifth feature, assessing the anthropogenic change or pressure in question (a "signal") against a background of inherent variability and natural change or wider influences, i.e., the changes emanating externally to the area being managed (the "noise"; Gray and Elliott, 2009; Elliott, 2011). Finally, a sixth key feature requires quantitative and legally defendable detection of such change with a direct feedback into management.

#### Progress beyond the State-of-the-Art

These key features require a defendable, holistic, underlying framework, accepted, and communicable to marine managers and wider users. That framework must link causes of potential and actual changes to the marine environment, the types of changes experienced and societal responses to mediating or removing the drivers of change or at least accepting change for the benefits provided. Even in the recent past, stakeholders frequently used the DPSIR (Drivers, Pressures, State change, Impact and Response) interlinking framework (e.g., Atkins et al., 2011; Smith et al., 2014), without clearly defining each element. Hence, the wide use of DPSIR model (Gari et al., 2015; Lewison et al., 2016; Patrício et al., 2016a) not only introduced many variants and perpetuated confusion but also made it not-fit-forpurpose in providing management guidance.

Previous studies document the evolution of the DPSIR approach (Smith et al., 2014, 2016), and here we summarize and focus on the evolution from DPSIR to the most recent derivative DAPSI(W)R(M) (Patrício et al., 2016a; Scharin et al., 2016; Burdon et al., in press). This modified approach adds Activities, and relates the Impact to human Welfare and the Responses to the use of Measures (the term preferred by EU Directives). Drivers describe underlying basic human needs, such as for food, security, space, and well-being, which require Activities (fishing, building wind farms, creating navigation routes). These activities then create Pressures, such as scraping the seabed with bottom trawls or building infrastructure that removes space. Pressures are the mechanisms that change the system, potentially causing concern. Those changes encompass both the natural system, including its structure and functioning (the "State change"; Strong et al., 2015), and the human system [the Impact (on human Welfare)]. The term Welfare is used sensu stricto to include economic welfare and human and societal well-being (Oxford English Dictionary).

Furthermore, all of the activities and external changes could potentially adversely affect that main aim (the protection of the social and ecological systems), and may thus be considered hazards. If these hazards damage parts of the socio-ecological system we value, they may be termed risks, thus providing a hazard and risk typology used in the DEVOTES project (Elliott et al., 2014). Smith et al. (2016) illustrated DPSIR, using fishing activity and the pressure of trawling from abrasion on the seabed and its impacts on particular components as an example. The challenges were addressed in moving from conceptual models to actual assessments including: assessment methodologies (interactive matrices, Bayesian Belief Networks, ecosystem modeling, the Bow Tie approach, assessment tools), data availability, confidence, scaling, cumulative impacts, and multiple simultaneous pressures, which more often occur in multi-use and multi-user areas (Smith et al., 2016).

Society and environmental managers need to know not only the current status of a marine system, but also whether it has been altered, the cause of that alteration, its significance, and what can be done to reverse that change. Therefore, this requirement creates the need to consider how Pressures result in State change, in the natural system, and a societally relevant Impact of sea use (including the assessment of cumulative pressures and impacts, as shown by Korpinen and Andersen, 2016); hence the need to consider not just Welfare (sensu DPSWR in Cooper, 2013) but the Impact (on human Welfare). This need explicitly includes an economic approach and a human health and well-being approach to human-induced changes. Furthermore, while that State change may often relate to the physico-chemical and ecological structure of the marine system, it increasingly requires users to consider the ecological functioning (Strong et al., 2015) especially given that many MSFD descriptors relate to functioning aspects. This "biodiversity-ecosystem functioning debate," regarding the effect of functioning on biodiversity and vice versa, is an important and developing field (Zeppilli et al., 2016).

The detection or prediction of changes to the natural state and impacts on human welfare require action to minimize, mitigate, compensate, remove, or even accept changes through societal Responses (the R in DPSIR). However, based on terminology used in the EU Directives, environmental managers now refer to those Responses as Measures [hence Responses -using Measuresin DAPSI(W)R(M); Scharin et al., 2016]. During the past decade, management has recognized the need to include all measures which therefore, as referred to as the Programme of Measures in the MSFD, should consider aspects of ecology, technology, economy, legislation, and administration. They should also satisfy societal, cultural and moral imperatives while communicating decisions to stakeholders; hence the so-called "10 tenets" for sustainable and successful marine management (Elliott, 2013; Barnard and Elliott, 2015).

The prevailing governance system provides a central control on adverse effects of human activities. The EU arguably represents the pre-eminent proponent of marine environmental legislation and other aspects of governance (Boyes and Elliott, 2014), but the complexity of the marine system, the need for transboundary action and the joint implementation of different systems have produced anomalies, confusion, and a need for an inter-governmental transboundary approach (Cavallo et al., 2016).

Most of the above framework relates to activities and pressures emanating from within a system such as a sea region, under management, for example the Baltic or North Seas (Andersen et al., 2015; Scharin et al., 2016). These may be termed endogenic managed pressures in which the causes and consequences in the region are managed (Elliott, 2011) and under legislative control (Boyes and Elliott, 2014). Exogenic unmanaged pressures (i.e., those aspects emanating from outside a managed system; for example global climate change Elliott et al., 2015) represent the major current challenge; environmental managers cannot control the causes but must respond to the consequences. Climate change offers a primary example, in which human impacts (e.g., ocean acidification, increase in alien species, sea-level rise, temperature regime change; Danovaro et al., 2013; Katsanevakis et al., 2014, 2016) add to internal pressures in an area. Climate change therefore shifts baselines, complicating evaluation change associated with internal activities in a region, but also potentially nullifying the use of quantitative indicators or at least requiring the target values of those indicators to be continually revised. A Member State not meeting legislative controls, such as directives, may therefore cite climate change as a modifying factor but one outside of its control (Elliott et al., 2015). Targets that cannot be reached due to changes caused by climate change effects are not manageable and need to be revised as a part of the 6 years management cycle.

## Conclusions

Successful ocean use management relies on adequate and comprehensive monitoring, and identifying appropriate measurements of change. Management response requires a clear understanding of underlying causes and effects of change in the marine environment and their consequences. Hence, the use of conceptual models linking the marine drivers, activities, and pressures can provide that solid foundation to link to state changes, impacts on societal welfare, and the resulting management responses using programmes of measures. Similarly, management relies on the ability to predict and detect future responses of the system to changes with sufficient certainty; prediction requires conceptual, empirical, and deterministic models, whereas detection implies the presence of robust monitoring systems at appropriate spatial and temporal scales. However, the "paradox of environmental assessment" sets the backdrop for this framework whereby increasing national and European legislation (such as the MSFD) requires more understanding and better monitoring but monitoring organizations face reduced budgets (Borja and Elliott, 2013). Therefore, by expanding the concept of DPSIR into DAPSI(W)R(M), understanding the gaps and the Strengths, Weaknesses, Opportunities, and Threats in monitoring, and exploring how climate change could affect GES, DEVOTES has included human welfare in the modified approach, emphasizing the importance for future policy and management measures. Hence, an adequate assessment of marine status can only be achieved through fit-for-purpose monitoring based on sound scientific knowledge.

# WHY DO WE NEED BETTER INDICATORS TO ASSESS THE STATUS?

#### State-of-the-Art

The multifaceted concept of biodiversity encompasses everything from the genetic composition of species to the organization of habitats and ecosystems (CBD, 1992). Despite the widely recognized need to maintain biodiversity, its many interpretations make difficult any comprehensive evaluation and therefore it is necessary to use indicators, or simplified measures, that reflect or synthesize the status of important aspects of ecosystem structure or function. Marine assessments depend upon indicators to detect and evaluate changes in environmental status driven by either natural or human pressures, often in the context of implementing management targets for environmental objectives and measures. Therefore, scientists and managers worldwide seek accurate and reliable indicators that represent all relevant aspects of marine biodiversity either as individual aspects or as surrogates (proxies) for series of changes (for example the use of the breeding health of piscivorous seabirds as a proxy for the whole marine trophic system).

Although, many nations worldwide recognize the need for an ecosystem approach to ocean management, the EU has led in developing specific metrics toward that objective. The European Commission (2010) Decision specifies criteria and methodological standards to evaluate environmental status of marine waters, based upon a set of 56 MSFD indicators. Some indicators used in the assessment of coastal ecosystems under the Water Framework Directive (WFD; European Commission, 2000; Birk et al., 2012) also apply to the MSFD assessment beyond the narrow coastal strip where MSFD and WFD overlap (Borja et al., 2010; Boyes et al., 2016). In practice, during the first phase of the MSFD implementation, EU Member States used different methodological approaches to determine and assess ecosystem status (European Commission, 2014; Palialexis et al., 2014). Data availability, regional specificities, and potentially different interpretations of the EU Commission Decision led to discrepancies within methodologies reported by Member States, increasing the potential for non-harmonized approaches to status determination. Managers require further guidance on criteria for "good" indicators, and assessment of status (Patrício et al., 2014), and such a plan is currently being developed by the EU and its Member States, ICES (International Council for the Exploration of the Sea), EEA (European Environment Agency), and RSCs (The Regional Sea Conventions). Concurrently, the RSCs are developing indicators for holistic marine assessments (e.g., HELCOM, 2013; OSPAR, 2015; UNEP, 2016).

#### Progress Beyond the State-of-the-Art

#### Overview of Existing Indicators and Gaps in Relation to MSFD Requirements

To support the MSFD process, we completed a comprehensive overview of existing MSFD biodiversity-related indicators (MSFD descriptors: D1—biological diversity, D2—nonindigenous species, D4—food-webs, and D6—seafloor integrity), identified gaps, and developed/tested new indicators to assess the status in the marine environment (Patrício et al., 2014).

We created an inventory of current MSFD biodiversity indicators, which includes over 600 entries, and developed complementary software (DEVOTool; www.devotes-project.eu/devotool) to help users navigate the metadata. The DEVOTool includes instructions for its use as well as a description of the database contents. Developing the inventory demonstrated that, despite many available marine biodiversity indicators, obvious gaps remain regarding some biotic components and criteria required for the MSFD implementation (Teixeira et al., 2014). Furthermore, information regarding the quality and confidence of the indicators is currently insufficient. Most available operational indicators target coastal and shelf ecosystems and cover WFD biological quality elements, such as macroinvertebrates, fish, phytoplankton, macroalgae, and seagrasses. Major current gaps include ecosystem level and genetic population level indicators, as well as indicators for microbes, pelagic and planktonic invertebrates, reptiles, ice-associated species, and communities, and deep-sea habitats. Most indicators lack regional targets or GES threshold values, and few measure confidence levels or demonstrably link to pressures. Thus, although current indicators may be regarded as operational in the way that they have been used in marine assessments, their applicability to fulfill the criteria of MSFD indicators and to comply with indicator quality criteria (Queirós et al., 2016) has not been assessed.

#### Development of New Indicators

We developed 16 new indicators and refined another 13 indicators (Berg et al., 2016; **Table 1**) to address gaps in MSFD implementation (Teixeira et al., 2014). These indicators mainly relate to the biodiversity-related Descriptors (D1, D2, D4, and D6), and cover the full range of biological components (i.e., from microbes to seabirds and marine mammals). In addition, we developed indicator quality criteria, which were used to evaluate these indicators (Queirós et al., 2016). For example, we developed four new indicators for microbes (bacteria and cyanobacteria), but their poor score on pressure responsiveness and the potential to set targets indicated a need for further development and validation (Berg et al., 2016). Some phytoplankton biomass indicators, such as chlorophyll-a concentration from satellite measurements, provide valuable assessments of pressures leading to eutrophication, but linking changes in diverse and rapidly fluctuating phytoplankton composition with impacts of nutrient loading has proved challenging (Camp et al., 2015; Carstensen et al., 2015).

Also indicators were developed to address the environmental impacts of invasive non-indigenous species in European regional seas (Minchin and Zaiko, 2013; Zaiko et al., 2014; Katsanevakis et al., 2016). Moreover, the project developed new foodweb indicators focusing on primary and secondary producers, both for phytoplankton and fish. Of those, the novel foodweb indicator "Phytoplankton community composition as a food-web indicator" was a highly-evaluated indicator and hence it is currently a candidate HELCOM core indicator for holistic ecosystem assessment. An indicator for systematic highresolution habitat mapping and characterization scored high in the indicator-evaluation as it may be a proxy for many of the 56 MSFD indicators. We also recently developed and tested numerous promising indicators that capture effects of fishing on marine biodiversity, e.g., on the positive effects of fishing effort reduction on the increase of large fish indicator (Engelhard et al., 2015) and on the need of using biodiversity and conservation-based indicators complementarily to ecological indicators of fishing pressure to evaluate the overall impact of fishing on exploited marine ecosystems (Fu et al., 2015; Coll et al., 2016). Furthermore, a newly developed indicator based on DNA metabarcoding assesses genetic diversity of macroinvertebrates


#### TABLE 1 | Indicators developed or refined within DEVOTES project, in relation to some of the indicators proposed within the Marine Strategy Framework Directive (MSFD).

*(Continued)*

#### TABLE 1 | Continued


*Note the potential application of some indicators in assessing various MSFD indicators.*

and microorganisms (Aylagas et al., 2014, 2016; Carugati et al., 2015; Dell'Anno et al., 2015).

We also applied Signal Detection Theory (SDT) to assess the accuracy, sensitivity and specificity of refined benthic indicators (such as the Benthic Quality Index—BQI) and their response to eutrophication. In general, we found SDT to be a robust and scientifically sound strategy for setting threshold values for indicators (Chuševe et al., 2016 ˙ ). Finally, we introduced a new approach to set indicator targets in relation to ecosystem resilience (i.e., the ability to recover rapidly and predictably from pressures) and to select indicators and their target ranges (Rossberg et al., 2017). This approach is a specific, quantitative interpretation of the concepts of GES and sustainable use in terms of indicators and associated targets. Importantly, it distinguishes between current and future uses to satisfy societal needs and preferences.

#### Conclusions

Increasing legal challenges of marine and coastal management, both to the EU Member State implementation of Directives and industry compliance with national laws, which hinge upon detecting and demonstrating marine environmental change (Elliott et al., 2015), increases the need for scientifically defensible indicators. Those indicators must be comprehensive, either in covering all relevant aspects of the marine system or as conceptually defensible surrogates that represent a well-defined and well-accepted causal link (e.g., the health of breeding populations of top seabird and fish predators being dependent on the health of seabed populations).

We tested and refined 13 available biodiversity indicators, developed 16 new options for assessment, particularly for biological descriptors (considering species, habitat and ecosystem levels), identified gaps for future research, developed indicator performance criteria, and provided a user-friendly tool to select and rank indicators (**Table 1**). These publicly-available contributions (Berg et al., 2016), support the second phase of the MSFD implementation and assist marine management in Europe and elsewhere.

### WHY MODELS ARE NECESSARY IN MARINE STUDIES AND ASSESSMENT?

#### State-of-the-Art

Understanding how changes in biodiversity link to foodweb functioning, anthropogenic pressures, and climate changes requires novel, integrative modeling tools. Similarly, scaling determining change from small to large areas and from the present to future, also requires such modeling approaches. Once validated, modeling tools can elucidate expected risks and rewards for a range of management options, aimed at achieving or maintaining GES. The evidence base from such scenario testing thus provides a suitable platform to enable informed decision-making. Prior to 2012, the proposals for using models in the MSFD implementation were very limited (Cardoso et al., 2010) but now, in the context of using models in assessments, Pinnegar et al. (2014) for example have demonstrated the value of food-web models in assessing potential responses of ecosystems to invasions.

#### Progress beyond State-of-the-Art

We assessed the capabilities of state-of-art models to provide information about current and candidate indicators outlined in the MSFD, particularly on biological diversity, food-webs, nonindigenous species, and seafloor integrity descriptors (Piroddi et al., 2015; Tedesco et al., 2016). We demonstrated that models could explain food-webs and biological diversity, but poorlyaddressed non-indigenous (alien) species, habitats and seafloor integrity (Lynam et al., 2016).

#### Habitats and Non-Indigenous (Alien) Species

In order to address the key gap related to non-indigenous (alien) species, we developed a method to model the vulnerability of areas to invasions, using the Mediterranean Sea as a case study (Katsanevakis et al., 2016). This conservative additive model accounts for the Cumulative IMPacts of invasive ALien species (CIMPAL index) on marine ecosystems. It estimates cumulative impact scores based on distributions of invasive species and ecosystems, considering both the reported magnitude of ecological impacts and the strength of such evidence.

#### Theory and New Approaches to Model Ecosystem Function

The theory supporting advanced modeling of food-webs and biodiversity was extended (Rossberg, 2013; James et al., 2015). Through different projects, including DEVOTES, Fung et al. (2015) used this theory to explore links between Biodiversity-Ecosystem Functioning (BEF) in marine ecosystems to fill in a key knowledge gap. Strong et al. (2015), furthermore, showed the importance and potential of such functional indicators. The BEF relationship can change (Mora et al., 2014; Fung et al., 2015), but the protection of fish from predation provided the mechanism in this case, and BEF relationships depended upon species richness and fishing impacts. Previous studies by Danovaro et al. (2008) revealed that the BEF relationships can be exponential and thus extremely sensitive to changes in environmental conditions determining a biodiversity loss. Nagelkerke and Rossberg (2014) also developed a theoretical understanding whereby resource and consumer traits predict trophic space, such that empirical data can be used to determine trophic traits related to food-web functioning (James et al., 2015).

New modeling approaches using mass-balanced models were also developed to identify ecosystem structure, function (including Ecological Network Analyses) and reaction to disturbance (Lassalle et al., 2013, 2014a,b; Niquil et al., 2014; Chaalali et al., 2015; Guesnet et al., 2015).

#### Habitats and Function

To further understand the role of habitat in regulating function in marine food-webs, and thus link to other descriptors, such as seafloor integrity and biological diversity, we studied marine habitats at local [i.e., Basque coast, Galparsoro et al. (2015); Eastern Aegean Sea, Lynam et al. (2015b); Western Adriatic deep-sea, Zeppilli et al. (2016)], sub-regional (i.e., North Sea, Stephens and Diesing, 2015; van Leeuwen et al., 2015), and regional (i.e., Mediterranean, Katsanevakis et al., 2016) scales. For example, we developed a process-driven characterization of sedimentary habitats for the Basque continental shelf and demonstrated that species richness decreases rapidly with increased sediment resuspension (Galparsoro et al., 2013). Habitat modeling of elasmobranchs in the southern North Sea demonstrated the extirpation of some species such as common skate over time (Sguotti et al., 2016). Modeling spatial distribution of three common seabird species in the southern North Sea demonstrated the importance of habitat type and availability fish prey to seabird distributions (see Lynam et al., 2015a). Additionally, we demonstrated in Stephens and Diesing (2015) the feasibility of predicting substratum composition spatially across a large swath of seabed (North Sea) using legacy grain-size data and environmental predictors. We also demonstrated the suitability of such a quantitative prediction for further analyses of habitat suitability compared to traditional grid cell categorization (Stephens and Diesing, 2015).

We applied Benthic Traits Analysis (Alves et al., 2014; van der Linden et al., 2016a; Van der Linden et al., 2016b) specifically to understand benthic community function in relation to habitat. This analysis identified typological groups of benthic macroinvertebrates in the North Sea, based on response and effect traits, as potential ecological indicators for MSFD Descriptors 1 (Biological Diversity) and 6 (Seafloor integrity; Veríssimo et al., 2015). The creation and analysis of large data set on population genetics in species groups with different dispersal abilities linked genetic variation to constraints in movement within benthic habitats in macroinvertebrates. This finding appears consistent with a "neutral theory" explanation for marine biodiversity spatial patterns (Chust et al., 2013, 2016).

A major challenge in marine management and assessment relates to the ability to link the physico-chemical and ecological systems. For example, for pelagic habitats, we identified distinct physical regimes in the North Sea based on density stratification characteristics, and modeling identified five hydrodynamic regimes (van Leeuwen et al., 2015). These findings are valuable to support assessment at a sub-divisional scale within MSFD subregions. Effective marine management must consider these regimes and their likely biological interactions. These zones form the basis for the OSPAR biodiversity (pelagic habitat) assessment based on lifeforms, together with considering oxygen and eutrophication when assessing primary production for food webs.

#### Scenario Testing to Inform Management Decisions

Our research demonstrated that fisheries management may enhance biological diversity (such as the size-structure of the fish and elasmobranch community) but potentially produce unintended consequences for other ecosystem components (Lynam and Mackinson, 2015). For example, decreases in bentho-piscivores component. However, the system may nonetheless sustain economic yields with minimal risk of stock collapse (Lynam et al., 2015b) if managed through an ecosystem approach. In the long term, climate change may shift baselines for indicators (Lynam et al., 2015b) and so assessments of GES should recognize these effects (Elliott et al., 2015).

# Conclusions

Marine research and assessment require modeling studies that can support the use of indicators in fully encompass the functional linkages between ecosystem components and overwhelming pressures on the marine environment, such as climate change and ocean acidification. Such modeling provides the evidence for setting realistic targets and thus supporting better long-term marine planning. We used case studies to illustrate that modeling can assist in MSFD implementation, contributing to each step of the assessment and management cycle. Modeling can help to develop and refine novel indicators to support indicator-based assessment of GES. Modeling can incorporate indicator trends and responses, incorporating prevailing climatic conditions and anthropogenic pressures and, in this way, support the review of objectives, targets and indicators. Moreover, modeling can both inform adaptive monitoring programmes and be used in scenario testing to inform management decisions.

# MONITORING NETWORKS IN EUROPEAN REGIONAL SEAS: IS TRADITIONAL MONITORING SUFFICIENT TO ASSESS THE STATUS OF MARINE ECOSYSTEMS?

#### State-of-the-Art

Methods traditionally used in marine monitoring to investigate spatial and temporal variation in abiotic and biotic variables are time-consuming, costly and often limited in resolution (de Jonge et al., 2006; Borja and Elliott, 2013; Carstensen, 2014; Fraschetti et al., 2016). These constraints can severely limit our capacity to detect spatial and temporal changes in marine environmental health. In addition, most countries lack the tools to expand marine monitoring to the deep sea (Ramirez-Llodra et al., 2011), severely constraining the expected implementation of the MSFD in the open ocean and deep sea (Zeppilli et al., 2016). Moreover, marine monitoring methods currently limit analyses of some descriptors. For example, detecting cryptic and/or alien species (including those causing harmful algal blooms) will benefit from molecular approaches (Bourlat et al., 2013).

Activities that smoother, abrade or permanently-remove seabed habitat represent the greatest threats to seafloor integrity (Rice et al., 2012). Previous studies used benthic faunal analysis to indicate general seafloor integrity (Pearson and Rosenberg, 1978), drawing on an extensive catalog of methods and approaches for such a fundamental change (Gray and Elliott, 2009), but increasingly together with various visual assessment tools (Solan et al., 2003). Specific benthic faunal indicators exist for trawl abrasion (Jorgensen et al., 2016) but deriving these indicators is time-consuming and expensive to implement. Video inspection of seafloor smothering using Remotely Operated Vehicles (ROV), such as from seabed drilling activities, can visually map the environmental footprint (Gates and Jones, 2012), but we lack data to validate the uncertainty of the method compared to conventional biological sample collection.

The implementation of the assessments of marine environmental status required by the MSFD thus requires development and/or testing of innovative monitoring systems. Despite creating recent methodologies/technologies in DEVOTES, these are not yet used in routine monitoring of the MSFD descriptors. We encourage this through our summary analysis encompassing a catalog of monitoring networks and a wide array of potential tools, including: (i) molecular approaches (e.g., barcoding and metagenomic tools), (ii) remote sensing/acoustic methods, and (iii) in situ monitoring techniques.

# Progress beyond the State-of-the-Art

We have produced a catalog with the biodiversity monitoring networks, currently available in European Seas, with the aim to: (i) present a critical overview of the monitoring activities in Europe (i.e., the amount and reason for ongoing monitoring, whether it fulfills its objectives and to what pressures it is links), (ii) identify areas where no monitoring occurs, and (iii) recommend the further development and improvements for optimizing marine biodiversity monitoring in the context of the MSFD. Since the publication of the catalog (Patrício et al., 2014), new material has been added so that it currently identifies 865 monitoring activities corresponding to 298 monitoring programmes. A gap and SWOT (Strengths, Weaknesses, Opportunities, and Threat) analysis of the catalog (Patrício et al., 2016b), highlights uneven distributions of monitoring across regional seas (i.e., more monitoring activities in the North Eastern Atlantic and Mediterranean). Specifically, we note uneven monitoring effort between descriptors (e.g., more monitoring for Descriptor 1 on Biological diversity and Descriptor 4 on Food webs), between biological components (e.g., monitoring emphasis on fish and phytoplankton) and between pressures (e.g., high level of monitoring of organic matter enrichment across all regional sea). In addition, we consider whether monitoring networks are fit-for-purpose or sufficient for adequate implementation of the MSFD within the context of the need for better coordination, harmonization of methodologies, and cost-effectiveness considerations (Patrício et al., 2016b). This allowed us to explore different innovative monitoring approaches. Below we discuss these new approaches in terms of their potential applications to some of the 11 descriptors of the MSFD investigated by DEVOTES, in order to evaluate their broader applicability to future marine environmental monitoring.

#### Descriptors 1 (Biological Diversity) and 2 (Non-indigenous Species)

Future monitoring is increasingly likely to use molecular tools to complement classical taxonomic techniques in providing timely and inexpensive results (Bourlat et al., 2013). Classical biodiversity assessment is time-consuming and requires diverse taxonomic expertise. Metabarcoding could expedite biodiversity assessment, especially for microscopic organisms (either algae or animals) for which morphological identification is difficult (Carugati et al., 2015). For example, Dell'Anno et al. (2015) provided the first comparison of different DNA extraction procedures and their suitability for sequencing analyses of 18S rDNA of marine nematodes. They subsequently analyzed intragenomic variation in 18S rRNA gene repeats and reported that morphological identification of deep-sea nematodes matches the results obtained by metabarcoding analysis only at the order-family level. These results illustrate the importance of metabarcoding for exploring the diversity of benthic metazoans, but currently available databases have a limited coverage in quantifying the species encountered. Metabarcoding studies should therefore carefully consider these limitations in quantitative ecological research and monitoring programmes of marine biodiversity (Aylagas et al., 2016).

The routine use of microarrays for rapid detection of specific phytoplankton taxa, and particularly the presence of harmful algal blooms, requires further development to increase reliability and reduce associated time and expense. Nonetheless, monitoring strategies should include different molecular approaches [e.g., quantitative, in situ Polymerase Chain Reaction (PCR)] as these approaches offer far greater sensitivity to detect the presence, for example, of pathogenic bacteria compared to traditional approaches.

In addition to the above molecular tools, comparing biodiversity across different habitats and seas represents a critically important aspect of marine biodiversity monitoring, which metabarcoding can address. For example, in order to use metabarcoding to investigate the benthic biodiversity colonizing identical structures in different habitats, we deployed and later recovered Autonomous Reef Monitoring Structures (ARMS), initially developed by NOAA for coral reefs, after 12 months on hard bottoms at shallow depths at three sites (triplicates) within different regional seas (Baltic Sea, English Channel in the NE Atlantic, Adriatic Sea, Black Sea, and Red Sea). This highly reproducible approach allows a standardized comparison of colonizing biodiversity in different systems.

In parallel, molecular tools allowed us to identify aspects of biodiversity that classical tools could not, such as identifying microbial assemblages as indicators of biodiversity (Caruso et al., 2016), monitoring picoplankton (Ferrera et al., 2016), an early detection of invasive species (Ardura et al., 2015; Zaiko et al., 2015a,b), a census of meiofauna (Carugati et al., 2015), identifying functional gene diversity and plankton phylogeny (Reñé et al., 2013, 2015; Ferrera et al., 2015), revealing benthic eukaryotic diversity (Pearman et al., 2016a,b), or assessing the status of benthic macroinvertebrates (Aylagas et al., 2014).

The MSFD recognizes spatial changes in species and population distributions as key indicators. Numerous DEVOTES studies demonstrated the value of combining seabed geological information with biological variables (e.g., Galparsoro et al., 2013, 2014). However, whilst multiple needs drive the collection of such geological data (e.g., safety of navigation, renewable energy infrastructure, planning), mapping the entire marine area will require considerable time (although perhaps less than a decade with existing capabilities). Despite this potential, even after a comprehensive baseline survey, further monitoring for change will always be necessary. Existing monitoring programmes have enabled collection of high-resolution multibeam sonar data over a large area and extrapolation of these properties across 100,000 km<sup>2</sup> in the western English Channel. Only by addressing and interrogating environmental variables at scales and a resolution relevant to the biota will we understand the context of local ecosystem change and status.

#### Descriptor 3 (Commercial Fish Species and Shellfish)

At present, other than acoustic surveys that lack taxonomic resolution and exceed the science capability of developing nations, we lack novel approaches to replace traditional surveys and stock recruitment assessment in fish population studies. However, emerging molecular tools can identify connectivity among fish populations and help elucidate the role of connectivity in maintenance of fish stocks.

#### Descriptor 4 (Food-Webs)

Researchers can now cost-effectively monitor the functioning at the base of the food-web (i.e., primary and secondary production) using ferrybox systems [such as the Continuous Automated Litter and Plankton Sampler -CALPS-, developed on the RV Endeavor CONISMA, 2013] on research vessels and ships of opportunity. The zooplankton data collected by CALPS identifies broad geographic patterns in abundance and diversity and can be integrated within existing multidisciplinary surveys at minimal extra cost. As another example, semi-automated classification of zooplankton samples usefully provided data for a range of food web related indicators even in the northern Baltic Sea, where the generally small-bodied zooplankton is difficult to be classified using semi-automated methods (Uusitalo et al., 2016a). The OSPAR-led EU project "Applying an ecosystem approach to (sub) regional habitat assessments" (EcapRHA, www.ospar.org/work-areas/bdc/ecaprha) has further investigated this approach. Monitoring of phytoplankton community composition (i.e., ratio between diatoms and flagellates) by a combination of remote sensing, microscopy, and bio-optical methods can clarify food-web effects on higher trophic levels (Goela et al., 2015).

#### Descriptor 5 (Eutrophication)

Current instruments that can analyze chlorophyll-a from in situ sampling can ground-truth satellite image analysis for monitoring of phyto-pigments concentrations in surface waters (Cristina et al., 2014, 2016) or assess aquaculture impacts (Mirto et al., 2010, 2014; Luna et al., 2013; Bengil and Bizsel, 2014). In addition, pigment color analysis (particularly in situ flow cytometry) can provide insights on phytoplankton biodiversity (Goela et al., 2015), estimate and calculate time series of annual gross primary production, and support MSFD implementation (Cristina et al., 2015). We also investigated the influence of benthic trophic state on meiofaunal biodiversity and found that the benthic trophic status based on organic matter variables is not sufficient to provide a sound assessment of the environmental quality in marine coastal ecosystems. However, the integration of the meiofaunal variable allows providing robust assessments of the marine environmental status (Bianchelli et al., 2016).

#### Descriptor 8 (Contaminants)

Andrade et al. (accepted) developed a high frequency noninvasive (HFNI) bio-sensor as a potential tool for marine monitoring which uses the biorhythmic gaping behavior of clams (such as the Icelandic scallop Chlamys islandica and the Pacific oyster Crassostrea gigas) in response to environmental cues such as day length. These innovative microsensors measure the distance between the valves of bivalves held in underwater baskets at strategic locations, and can operate unattended for several years. Measurements every 1.6 s are telemetered from the field to the laboratory and further transferred to a "big-data" storage system for analysis. Minimal operational costs and online, real-time data availability offer major advantages of the system once installed.

Beyond biorhythm research (including growth and spawning behavior) in relation to climatic factors, the method has potential for monitoring marine contamination. Exposure to stressors such as sudden changes in water quality, temperature increases (e.g., around power plants), toxic algal blooms, or a plume of water-borne contaminants, interrupts otherwise regular gaping behavior. The automated, real-time detection could provide an early-warning system, with potential applications including monitoring of water quality at swimming beaches, harbors, petroleum installations (produced water and unintentional spillages), and aquaculture sites. This "talking clam" method can improve cost-efficiency by alerting users to periods of potential risk, narrowing the need for more labor-intensive physical sampling, as long as it is assumed that normal gaping behavior reflects good water quality status.

### Conclusions

As indicated above, we currently face a "paradox of environmental assessment"—with increasing monitoring requirements set against a backdrop of decreasing budgets. This paradox ensures the need for more cost-efficient and effective monitoring, and may eventually produce cheaper traditional monitoring, especially where monitoring requirements span large areas, as in the MSFD. The paradox requires wide-scale and rapid surveillance techniques, including innovative tools such as genomic approaches, remote sensing and acoustic sensors.

### WHY DO WE NEED AN INTEGRATIVE ASSESSMENT OF STATUS?

#### State-of-the-Art

The European Commission (2010) identified 56 indicators to consider when evaluating environmental status, but at least an order of magnitude more indicators already exist (Berg et al., 2015). Despite this, many of these indicators are variants on similar themes and hence measure related attributes, and are often geographical derivations, for example the health of seabed communities. The relevance and availability of indicators vary substantially among regional seas and their subdivisions; however, the MSFD provides no guidance on integration principles, despite multiple approaches to aggregating indicators whose selection may produce highly diverging results (Borja et al., 2014). These choices challenge the scientific community to develop harmonized approaches for integrating these indicators to compare across different assessment areas.

The Ocean Health Index (OHI; Halpern et al., 2012) was developed to assess the consequences of human impacts as well as societal benefits by calculating a weighted average of scores for pressure, status and resilience goals in different areas globally. Borja et al. (2011) were the first to address specifically the challenges of the MSFD, using weighting averaging principles for integrating indicator information. The MARMONI (Innovative approaches for MARine biodiversity MONItoring and assessment of conservation status of nature values in the Baltic Sea) assessment tool (Martin et al., 2015) then used an aggregation principle based on the hierarchical structure laid out by the European Commission (2010), rather than using aggregation approaches based on the structures of marine ecosystems. Nevertheless, all assessment methods standardize indicators to a common scale prior to aggregation (Borja et al., 2016). This standardization relies upon defining of targets or reference states, which MSFD describes as targets for GES (Borja et al., 2013). The OHI uses the relative deviation from a reference state, whereas the MARMONI tool uses a binary scoring system to determine whether GES has been achieved (score of 100) or not (score of 0). However, these standardization approaches do not always achieve translating indicator values to a common scale. A relative deviation from a reference state of 50% could indicate a minor human disturbance for one indicator but a major human disturbance for another. Similarly, a binary standardization approach does not differentiate between whether minimal attainment or high status level of GES was achieved.

#### Progress beyond the State-of-the-Art

We developed and released software for NEAT (Nested Environmental status Assessment Tool; freely available at: www.devotes-project.eu/neat), to overcome some of the deficiencies of current integrated assessment tools (e.g., aggregation of multiple indicators at multiple temporal and spatial scales; absence of uncertainty determination, etc.) NEAT is loosely based on previous tools (Andersen et al., 2014, 2016) and translates indicator values to a common scale ranging from 0 (worst possible status) to 1 (best possible status), with 0.6 defining GES or the good-moderate boundary according to the WFD. Similarly, NEAT also allows users to set boundaries representing high-good status (value of 0.8), moderate-poor status (value of 0.4), and poor-bad status (value of 0.2). It also employs stepwise linear interpolation between these fixed points to produce transformations with a high degree of flexibility spanning the entire scale (0–1) and in which 0.6 always represent GES. In comparison with the OHI and MARMONI tool, this transformation produces a more comparable scale for integrating standardized indicator values. NEAT also employs weighted averaging of standardized indicators, but bases averaging on ecosystem features to represent the whole ecosystem. The approach primarily divides the entire ecosystem into multiple Spatial Assessment Units (SAU) that

are nested to define a hierarchy of SAUs. Habitat information and relevant indicators according to organism groups are used to describe the environmental status which the given habitat may enter at different levels of the hierarchy, depending on the spatial representativity of the indicator and organism. First, averaging aggregate indicators at the organism level to produce a more even representation of relevant organism groups, i.e., to avoid an assessment biased by many indicators for the same organism group, before aggregating across habitats and SAU (Clark et al., 2011). Spatial information of the different SAUs, if provided, is used for weighting and habitats can be prioritized to weight complex habitats such as vegetated sea bottoms more heavily than deep, muddy sediments. In addition, NEAT indicators are associated with the different MSFD descriptors, supporting assessments based on various descriptor combinations (essentially from one to all).

Application of NEAT to 10 case studies across European marine waters with very different challenges, environmental conditions, and scales (Uusitalo et al., 2016b) highlights its flexibility adapting to these very different cases. This also highlighted the need for careful evaluation of the indicator set, their GES boundaries, and the selection of the SAUs, all of which can increase the accuracy of the GES assessment.

Finally, NEAT includes an uncertainty assessment at all levels of integration based on the propagation of errors (uncertainties) associated with the provided indicator information (Uusitalo et al., 2015). Therefore, assessing the confidence in the integrated assessment requires including an indicator value with an estimate of the standard error of that indicator value. Noting that few studies report or even determine the standard error of an indicator value, Carstensen and Lindegarth (2016) provide a framework for quantifying indicator uncertainty to enable such calculations. Knowing the distributions of the indicator estimates enables the calculation of the distribution of the standardized indicators as well as their aggregated values.

# Conclusions

A true ecosystem approach for ocean use management requires an integrative assessment of marine water status. In this way, NEAT provides a second-generation, integrated assessment tool that builds on the hierarchical structure of marine ecosystems and the organisms inhabiting different compartments within this structure, thereby improving upon previous tools. Such a hierarchical approach allows users to interrogate the results to understand the reasons for the failure or success at achieving GES. However, the integrated assessment is only as good as indicator information allows, and missing or omitting information on specific groups (e.g., biological components or descriptors relevant to the assessed area) can bias the assessment results. Therefore, managers should produce guidelines stipulating indicator minimum requirements [e.g., type, coverage (ecosystem components, area, etc.), number] and the integrated assessment tool should clearly indicate if there is non-compliance with such guidelines. Moreover, because NEAT includes a comprehensive uncertainty assessment, researchers should incorporate this information as part of their interpretation of outcomes and decision support, thus needing guidelines for confidence levels of decisions.

In conclusion, environmental managers must assess the status of marine waters, not only to comply with current legislation (i.e., MSFD, WFD), but also to determine how far from targets marine ecosystems may be. Such information will allow managers to make informed decisions on sustainable resource use and the adequate restoration of degraded systems.

# WHAT ECONOMIC AND SOCIAL DIMENSIONS AFFECT MARINE MANAGEMENT?

#### State-of-the-Art

Inevitably, new legislative framework directives bring about unforeseen challenges to the different stakeholders who need to be involved in their implementation, particularly when first applied. Managers already apply the MSFD, which is itself complex, to complex, heterogeneous, and dynamic environments. Furthermore, initiation of the MSFD coincided with a period of a growing, global economic crisis. The numerous objectives can potentially conflict with one another from the perspective of different government departments within the Member States and also between Member States sharing a regional sea. The MSFD legal status and implementation deadlines demand that scientists and decision makers ensure a collaborative and multidisciplinary approach to deliver multisectoral objectives that test the abilities of existing institutions. The rapid identification of the issues, and the problems that they can create, can help those responsible for MSFD implementation to consider best how to address such issues and ensure that the MSFD can provide the intended sustainable environmental benefits.

The introduction of complex and integrative environmental legislation such as the MSFD also inevitably incurs additional costs, such as establishing new monitoring and improving existing monitoring of multiple indicators across European seas. This demand can be economically challenging. Policy makers and regulators in all EU countries are obliged to manage their resources carefully and hence they will seek to comply with the MSFD in the most cost-effective way. Yet they have many choices on which types of monitoring to apply as they select the approaches that best comply with the legislative needs within the limits of their budgets (Veidemane and Pakalniete, 2015). Although the MSFD does not require consideration of the socioeconomic aspects of monitoring, Borja and Elliott (2013) noted that limited financial resources represent the most significant threat to ensuring adequate monitoring.

Furthermore, the law requires that EU countries determine whether they need new management measures and monitoring schemes to enable them to achieve GES and, if so, to implement them. Here, the socio-economic analysis of the use of marine waters, the cost of present-date degradation of the marine environment, and the cost-benefit analysis of implementing monitoring and new management measures required under the MSFD could motivate Member States to achieve GES. However, whilst a dominant tool of all governments, economic analysis approaches to achieve such analyses specifically for the MSFD in relation to the marine environment and its management were not developed at the start of the MSFD process.

#### Progress Beyond the State-of-the-Art Barriers to Achieving Good Environmental Status

A comprehensive review of the documented barriers to achieving GES indicated that Member States have encountered and reported legislative, governance, and socio-economic barriers during this first phase of implementing the MSFD (Boyes et al., 2015, 2016). Barriers include ambiguity in the text of the Directive resulting in different interpretations by Member States, creating uncertainty, and different levels of conformity and governance complications. For example, GES [Article 3(5)] is neither well defined nor quantitatively described (Boyes et al., 2016), not easily understandable, and requires specific guidance to achieve common understanding and to enable coherent practices between the Member States and across regional seas. The next revision of the European Commission (2010) Decision regarding MSFD implementation will provide more guidance on GES definition (for example the operational definition proposed by DEVOTES; Borja et al., 2013), and thus the input from different stakeholders, including the scientific community, will be extremely important. The effectiveness with which MSFD can achieve GES partially relates to the success of other EU legislation [e.g., the WFD, the reformed Common Fisheries Policy (CFP), Maritime Spatial Planning Directive (MSP), Integrated Maritime Policy (IMP)], acknowledging the ambiguity of the role and contribution of each individual piece of legislation. Despite limited reference to specific policies in the MSFD, it provides a framework that can incorporate earlier and future legislation to ensure that legislation provides spatially and temporally complete coverage for the protection of marine environment. The MSFD article 6 is quite clear on the purpose and role of RSC: "... Member States shall, where practical and appropriate, use existing regional institutional cooperation structures, including those under Regional Sea Conventions..." and "...Member States shall, as far as possible, build upon relevant existing programmes and activities developed in the framework of structures stemming from international agreements such as Regional Sea Conventions...." However, in the absence of clear guidance on how this objective should be implemented or the actual competence of the RSCs, Member States have not adopted the regional coordination and integration to achieve MSFD objectives. Boyes et al. (2015, 2016) offer recommendations to address these legislative and governance barriers, such as "continued clarification and harmonization of the definitions and methodologies within and between Member States and the different RSCs." The aims of other directives should be consistently included in considerations for GES together with clear reference to MSFD and other existing, forthcoming and amended directives. Systematic use of standards that already used within other EU legislation must be applied as minimum requirements. Implementation of the MSP Directive particularly provides measures that will support delivery of the goals of MSFD by facilitating a balance with blue growth objectives (Boyes and Elliott, 2014; Boyes et al., 2016). The RSC must have a mandate supported by their contracting parties in order to ensure that the measures implemented in EU countries are supported and complemented by respective measures also in non-EU countries. Achieving RSC aims requires continuous cooperation in regional seas between EU-Member and non-Member States in the context of RSCs (Cavallo et al., 2016).

Socio-economic barriers include a lack of appropriate biological, environmental, and socio-economic data, a limited application of the ecosystem-based approach and of economic impact analyses by Member States. Effective use of the findings of EU funded projects and pilot projects (involving both non-EU countries and Member States) can both boost the evidence and knowledge required. It can also provide a vehicle to improve and support regional coordination and encourage the coherent implementation of the MSFD in regional sea areas, and ensure engaging non-EU countries in programmes that enable measures to achieve true regional GES.

Discussions with stakeholders showed that often public and stakeholder consultations on the programmes of measures were only open for limited periods of 1–2 months, and stakeholders in most Member States, particularly NGOs, felt that they were not sufficiently involved in the MSFD process, with only limited integration of their feedback (Boyes et al., 2015). In recognizing the complexity of marine ecosystems, the existence of multiple stakeholders with imperfect and impartial knowledge, as well as resource constraints, we developed a workshop approach "to engage and share different perspectives, and develop models of the system under consideration that are seen to be valid and useful aids to decision making" (Boyes et al., 2015). This multi-stakeholder workshop based modeling approach, which focused on Causal Loop Diagrams (CLD) to describe and understand the case site, was developed and then trialed in a case study site in England (Boyes et al., 2015). Managers should consider this approach, which effectively engaged stakeholders in understanding the complex environment associated with GES and the barriers and opportunities for its achievement, is exemplary for moving forward in MSFD implementation.

#### Cost-Effectiveness of Monitoring

Building on the ecological criteria for monitoring developed in Queirós et al. (2016), we developed an approach that uses multi-criteria-decision-analysis (MCDA) for cost-effectiveness analysis incorporating both ecological and economic criteria as attributes of monitoring systems. This approach encompassed a standardized scoring system for each of the different attributes, readily adaptable to the analysis undertaken with the attributes and the scores used as input to the MCDA. The costeffectiveness of a given monitoring approach can be determined using the Rapfish software (www.rapfish.org), a non-parametric multivariate analysis tool, developed and tested in different contextual case studies of MSFD monitoring in Finland, Spain, and the UK. We also developed flow charts to help users identify the different elements of operational costs during monitoring. The tool can be applied to examine both the cost-effectiveness of the different monitoring elements and whether the monitoring programmes satisfy the requirements of the MSFD monitoring objectives. The tool has demonstrated, for example, a mixed ability of current monitoring programmes in Bay of Biscay to comply with the need to monitor changes in quality and quantity of different MSFD Descriptors. In addition, monitoring open sea areas in the Gulf of Finland becomes more cost-efficient when combining monitoring with research cruises on scientific vessels, which make up the largest single monitoring cost.

#### Cost-Benefit Analysis of New Management Measures to Achieve GES

By 2015, EU Member States had to define the Programme of Measures, including new measures if any, required to achieve GES. Oinonen et al. (2016a) stated that "the specific application of methods and uptake of resulting information are currently still evolving in the ecosystem-based and adaptive management framework that the Directive stipulates." They further recommend the use of environmental economics delivered through interdisciplinary research to support the needs of MSFD.

Three different case-studies showed interdisciplinary approaches to the cost-benefit analysis of management measures to achieve GES. In Finland, a quantitative cost-effectiveness analysis of implementing different management measures, based on opinion of interdisciplinary experts, identified the costs, and most cost-effective measures. Researchers estimated economic benefits of the management measures based on existing valuation studies (i.e., willingness to pay) on the benefits of improving the state of the Baltic Sea; these analyses connected the benefit estimates directly to the change in the status of the GES descriptors (Oinonen et al., 2016b). Extending from this analysis into a full cost-benefit analysis, the net value of achieving GES for indicators of biodiversity, food webs, and eutrophication alone in 2020 is placed at ∼2 bn e (although the planned management measures will not achieve GES of these Descriptors by 2020).

Alternative approaches to cost-benefit analysis of management measures were developed and applied in the Bay of Biscay and the East Coast of England Marine Plan Areas (ECE). These approaches built on research to determine changes in ecosystem services and the benefits that identified, mapped and modeled ecosystem services, and considered valuation of their benefits (e.g., Hattam et al., 2014, 2015; Galparsoro et al., 2014; Borja et al., 2015; Kleisner et al., 2015; Laurila-Pant et al., 2015).

The Bay of Biscay approach examined the links between ecosystem services and their benefits and management measures to control the development of maritime activities creating those benefits. We used the Fishrent bioeconomic model (Salz et al., 2011) to quantitatively assess the impacts, in terms of percentage changes in net present value, of implementing some of the measures under the European Common Fisheries Policy (CFP) expected to support attainment of GES.

The ECE approach used structured analysis of changes in ecosystem services and benefits arising from potential new management measures (ballast water treatment, underwater noise reduction) to identify the benefits of achieving GES alongside the costs of implementing the measures. Insufficient availability of valuation data, needed to quantify ecosystem benefit impacts in monetary terms, precluded the possibility of extending the analysis into a quantitative cost-benefit analysis.

# Conclusions

Ensuring that Member States implement sufficient and nonoverlapping measures to achieve GES will require the continuous review of legislation and policy, and the assessment of its implementation (Boyes et al., 2016). Furthermore, effective stakeholder engagement is likely to facilitate the acceptance of the measures and associated costs. The effective application of the MSFD requires knowledge and databases but these currently are limited by economics (Borja and Elliott, 2013). It remains to be seen how the different Member States identify the additional measures needed to improve the marine environment toward GES and close the gap between current status and GES in 2020. However, Member States must use existing budgets carefully to avoid further economic hardship resulting from financial penalties due to legal infraction proceedings in the European Court. For example, reduced funding for monitoring, if not guided toward more effective monitoring tools (see Section Monitoring Networks in European Regional Seas: Is Traditional Monitoring Sufficient to Assess the Status of Marine Ecosystems?), can reduce the quality of monitoring (e.g., by reducing spatial and/or temporal coverage). Such reduction can ultimately entail a greater cost than investment in monitoring as inaccurate evaluation could increase the risk of decision-making errors, potentially resulting in reduced ecosystem services and a devaluation of ecosystem benefits. Political decision makers may consider other aspects of monitoring as societally important, such as maintaining a bank of knowledge, technological development, professional skill and experience development and enhancing public engagement. Tools for determining the costeffectiveness of monitoring and of management measures, as well as use of the ecosystem service approach to determine ecosystem benefits in cost-benefit analysis can support decisions on activities undertaken to comply with the MSFD. However, many member states implementing the MSFD lack both the data required to underpin rigorous economic analysis of costs of monitoring and the valuation data for assessing changes in ecosystem benefits from improvements in ecosystem services. Furthermore, the relationship between MSFD indicators and ecosystem services still requires better understanding and the implementation of the MSFD still urgently requires such data and information.

# FILLING IN THE GAP BETWEEN SCIENCE AND POLICY

Some of the challenges in marine ecosystems ecology identified by Borja (2014) relate to socio-ecological topics, especially given the recognition of humans (and the activities they perform and pressures they pose in the oceans) as an integral part of the marine ecosystem in recent decades. The human dimension of marine systems remains poorly documented, and discussions


#### TABLE 2 | Progress beyond the state-of-the-art achieved by DEVOTES, within the different topics addressed by the project, and gaps bridged in science-policy.

*(Continued)*

#### TABLE 2 | Continued


*DPSIR and DAPSI(W)R(M): D, drivers; A, activities; P, pressures; S, change of state; I, impact; W, human wellbeing; R, responses; M, management; MSFD, Marine Strategy Framework Directive. GES, Good Environmental Status; NEAT, Nested Environmental status Assessment Tool; ARMS, Autonomous Reef Monitoring Structures; ASU, Artificial Substrate Unit.*

on ecosystem-based management of seas often minimize the importance of social sciences (Fréon et al., 2009), despite the explicit role of humans in implementing the Ecosystem Approach since its adoption in the Convention on Biological Diversity (CBD, 1992). Despite progress in recent years in connecting natural and social sciences, the gap between science (social and natural) and policy remains large in marine research (Nicholson et al., 2012). Europe has made efforts to close the research project-policy circuit in relation to the WFD (Oliver et al., 2005; Quevauviller et al., 2005; Hering et al., 2010), although until recently, few attempts have been made to close such a circuit for the MSFD (Borja et al., 2010).

Many challenges remain in bridging the gap between science and policy to support improved policy decisions in marine management (von Winterfeldt, 2013; Choi et al., 2015). As noted by Rodwell et al. (2014), improvement would require identifying: (i) the gap or perceived gap between marine science and policy; (ii) the obstacles that prevent us from bridging the gap, and (iii) the possible solutions.

More than 30 years ago, Sebek (1983) identified some of the reasons for marine public policy failure in incorporating scientific knowledge, but researchers since have removed some of the impediments (**Table 2**). Specific examples include:

(i) Lack of international regulation encompasses both EU and non-EU countries adjacent to the regional seas, which the different RSCs and, especially, the WFD and MSFD has overseen in recent years. DEVOTES brought together different pieces of legislation, identified gaps and overlaps and provided advice on future needs for the satisfactory implementation of the MSFD for the new Commission Decision expected in the coming months (Patrício et al., 2014);


ecosystems and provision of ecosystem services as well as between management measures and ecosystem services.

DEVOTES was conceived as an integrative project that aimed to expand and merge natural and social sciences, enable the natural scientists to understand the economic and legal requirements and economic and governance specialists to understand the limitations of natural science. **Figure 1** encapsulates the work done and the integration of pieces to assess the status and respond to multiple stakeholders and end-users (i.e., scientists, policy-makers, managers, industry, conservation organizations, and society). Hence, our outputs not only increased knowledge of marine assessment and assisted marine managers, but also communicated these findings to increase stakeholder uptake.

In connection with the development of the MSFD programme of measures, Oinonen et al. (2016b) developed a pragmatic approach to holistic cost-effectiveness analysis. This allows users to select a cost-effective set of candidate measures in order to reach the multidimensional environmental objectives of the MSFD. They concluded that the major challenge in applying cost-effectiveness analysis was in assessing the current state of the environment and the multiple effects of different measures in evaluating marine ecosystem components rather than in the concepts of economic analysis. Despite this, economics helped to determine socially optimal level of marine protection.

Many challenges remain despite the body of work undertaken over the last 4 years (**Table 2**). These challenges include, among others: (i) understanding how multiple pressures act in marine ecosystems, and managing those pressures in the context of climate change; (ii) identifying key indicators and setting targets and reference conditions for those indicators so they can be used in assessments, noting the need for comparability across regional seas, and recognizing scenarios of climate change that shift baselines; (iii) the need to develop models capable of operating at an ecosystem level, with powerful computational capacities able to handle big data; (iv) getting EU Members States to consider adopting new monitoring tools routinely, and coordinating monitoring activities within regional sea research activities; (v) the need for intercomparable and harmonized assessments across regional seas that include ecosystem services in the assessment, and (vi) the urgent need for a harmonized framework under which indicator development follows specific rules and aligns with specific criteria to readily use in an integrative assessment tool.

#### GENERAL CONCLUSIONS

Defining, attaining and maintaining the GES of the seas spans from the technical details of monitoring and indicator implementation to major social and economic issues of how to optimize the long-term delivery of ecosystem goods and services, and how to govern society fairly in relation to use of the sea. It requires integrating natural and social sciences, horizontal integration across stakeholders, and vertical integration through governance, and feedback between monitoring, measures, and management. The MSFD aspires to bring all these aspects under the same umbrella, an ambitious and highly relevant objective.

DEVOTES advanced the state-of-the-art and identified major gaps within various aspects of MSFD implementation, contributed to filling these gaps, and identified additional scientific and development needs. The further development and validation of marine biodiversity indicators requires improved data with better spatial and temporal coverage, based on novel and cost-efficient monitoring methods. Better ecological relevance and indicator responsiveness to pressures will require experimental research on different levels of biological organization from the cell to the ecosystem. Such research

#### REFERENCES


will also enable incorporation of indicators into models in order to extrapolate marine assessment results to larger spatial and temporal scales. Each of these aspects requires comprehensive and integrated natural and social sciences which cross international boundaries and regional seas.

#### AUTHOR CONTRIBUTIONS

AB conceived the paper and wrote the first draft; ME, AB contributed to the pressures section; TB, SC, MM, AH, JP, LU contributed to the indicators section; CL, CW contributed to the modeling section; RD, AB, AN, JP, SC, PS, MU contributed to the monitoring section; JC, AB, TB, SC, MU contributed to the integration section; MA, MU contributed to the socioeconomic area; all authors contributed equally to the discussion and conclusions, AB, ME, SG, PS, MU made a revision of the whole text.

#### ACKNOWLEDGMENTS

This manuscript is a result of DEVOTES (DEVelopment Of innovative Tools for understanding marine biodiversity and assessing GES) project, funded by the European Union under the 7th Framework Programme, "The Ocean of Tomorrow" Theme (grant agreement no. 308392), www.devotes-project.eu. MU is partially funded through the Spanish programme for Talent and Employability in R+D+I "Torres Quevedo". This is publication number 777 from AZTI's Marine Research Division.

and assessment. Front. Mar. Sci. 3:96. doi: 10.3389/fmars.2016. 00096


overview of biological methods to implement the Water Framework Directive. Ecol. Indic. 18, 31–41. doi: 10.1016/j.ecolind.2011.10.009


and Transitional Water Quality Monitoring. The Handbook of Environmental Chemistry, eds A. Munné, A. Ginebreda, and N. Prat (London: Springer), 135–160.

	- The European Commission's assessment and guidance. Available

online at: http://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX %3A52014DC0097


and valuation: mixed methods or mixed messages? Ecol. Econ. 120, 126–138. doi: 10.1016/j.ecolecon.2015.10.011


Available online at: http://www.devotes-project.eu/wp-content/uploads/2014/ 10/DEVOTES\_Deliverable-1-5.pdf


relationships in the deep sea. Sci. Rep. 6:26352. doi: 10.1038/srep 26352

**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2016 Borja, Elliott, Snelgrove, Austen, Berg, Cochrane, Carstensen, Danovaro, Greenstreet, Heiskanen, Lynam, Mea, Newton, Patrício, Uusitalo, Uyarra and Wilson. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.