Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Sea turtle conservation and ecosystem-based management with a focus on green turtles (Chelonia mydas).. 2010

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
ubc_2010_fall_wabnitz_colette.pdf [ 3.89MB ]
ubc_2010_fall_wabnitz_colette.pdf
Metadata
JSON: 1.0069963.json
JSON-LD: 1.0069963+ld.json
RDF/XML (Pretty): 1.0069963.xml
RDF/JSON: 1.0069963+rdf.json
Turtle: 1.0069963+rdf-turtle.txt
N-Triples: 1.0069963+rdf-ntriples.txt
Citation
1.0069963.ris

Full Text

i  Sea turtle conservation and ecosystem-based management with a focus on green turtles (Chelonia mydas) and seagrass beds  by  COLETTE CATHERINE CHANTAL WABNITZ B.Sc. (Biology and Environmental Sciences), McGill University, 1998 M.Sc. (Tropical Coastal Management), University of Newcastle-Upon-Tyne, 2000   A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Geography)   THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) April 2010  © Colette Catherine Chantal Wabnitz, 2010 ii  Abstract The design of effective species management and recovery plans for sea turtle populations requires targets that are informed by an understanding of knowledge gained at the level of individuals/populations in the context of the wider goal of protecting an ecosystem's structural and functional attributes. In this thesis I present the first detailed investigation of the multiple levels at which sea turtles, particularly green turtles (Chelonia mydas), interact with ecosystems.  I begin by developing a framework for an age-structured population-level assessment of food consumption for hawksbill (Eretmochelys imbricata), loggerhead (Caretta caretta), and green turtles. This entailed construction of species-specific growth models for the western Atlantic, and subsequent integration of results with morphometric, survival, abundance, and food conversion efficiency to derive consumption estimates.  At the ecosystem level, I developed models for the Caribbean and Hawai’í, where green turtles are present at very low abundances and reaching carrying capacity, respectively. In the Caribbean, results showed that green turtle grazing of seagrass substantially altered habitat complexity, reducing the refuge role of seagrass to reef fish and invertebrates, and leading to potentially dramatic changes in species biomass and composition. In Hawai‘i, by feeding on algae, green turtles were found to contribute to the resilience of reefs in the face of disturbance, a functional role that needs to be explicitly included in future studies of reef dynamics. Taken together, these findings highlight the need to consider trophic and indirect interactions in the evaluation of sea turtles’ role within ecosystems.  iii  Accurate and reliable estimates of foraging habitat extent are essential to inform realistic and sustainable turtle recovery targets, particularly given the current degraded state of coastal ecosystems. Using Landsat satellite imagery, I present a novel mapping approach for seagrass habitats at large scales. Such regional initiatives are also urgently needed if the international community is to meet aims to conserve 10 – 30 % of specific habitats. A comparison between reef extent, determined using remote sensing and existing data, showed that discrepancies ranged from + 1,316 % to - 64 %, underlining our limited ability to ascertain progress towards adopted global marine conservation targets based on current data. iv  Contents Abstract ................................................................................................................ii Contents .............................................................................................................. iv List of tables ........................................................................................................ vii List of figures ...................................................................................................... viii Acknowledgements..................................................................................................x Dedication........................................................................................................... xv Co-authorship statement ........................................................................................ xvi 1. Introduction ....................................................................................................... 1 1.1 Rationale.............................................................................................................................. 1 1.2 Context ................................................................................................................................ 3 1.2.1 Ecosystem-based management ....................................................................................... 5 1.2.2 Role of sea turtles and ecosystem models as support tools............................................... 7 1.2.3 Sea turtle habitat - large scale mapping using remote sensing .......................................... 8 1.3 Objectives............................................................................................................................. 9 1.4 Thesis outline ..................................................................................................................... 10 1.5 References .......................................................................................................................... 12 2. Population dynamics and food consumption of hawksbill (Eretmochelys imbricata), loggerhead (Caretta caretta), and green (Chelonia mydas) sea turtles in the western Atlantic ..................... 19 2.1 Introduction ....................................................................................................................... 19 2.2 Materials and methods ....................................................................................................... 24 2.2.1 Length-mass relationships............................................................................................ 24 2.2.2 Growth rates................................................................................................................ 25 2.2.3 Population biomass estimates and life cycle.................................................................. 27 2.2.4 Food conversion and estimates of food consumption.................................................... 28 2.3 Results ............................................................................................................................... 33 2.3.1 Length-mass relationships............................................................................................ 33 2.3.2 Growth rates................................................................................................................ 34 2.3.3 Population biomass estimates and life cycle.................................................................. 35 2.3.4 Food conversion and estimates of food consumption.................................................... 36 2.4 Discussion.......................................................................................................................... 37 2.4.1 Length-mass relationships............................................................................................ 38 v  2.4.2 Growth rates................................................................................................................ 38 2.4.3 Age at maturity............................................................................................................ 40 2.4.4 Population biomass estimates and life cycle.................................................................. 42 2.4.5 Food consumption....................................................................................................... 43 2.4.6 Conclusions................................................................................................................. 46 2.5 References .......................................................................................................................... 70 3. Restoration of Chelonia mydas in the Caribbean: Ecosystem changes resulting from a reduction in seagrass habitat complexity .................................................................................. 83 3.1 Introduction ....................................................................................................................... 83 3.2 Materials and methods ....................................................................................................... 87 3.2.1 Study area ................................................................................................................... 87 3.2.2 Modelling approach: Ecopath with Ecosim (EwE) ....................................................... 88 3.2.3 Functional groups and model parameters ..................................................................... 90 3.2.4 Dynamic simulations ................................................................................................... 93 3.3 Results and discussion ........................................................................................................ 95 3.3.1 Ecopath model of the Caribbean ecosystem.................................................................. 95 3.3.2 Ecosystem wide changes as a result of an increase in green turtle biomass..................... 97 3.3.3 Insights offered by the modelling exercise................................................................... 100 3.3.4 Conclusions............................................................................................................... 105 3.4 References ........................................................................................................................ 116 4. Carrying capacity of green sea turtles (Chelonia mydas) at Kaloko-Honokōhau, Hawai‘i, and their role in reef resilience ...................................................................................... 128 4.1 Introduction ..................................................................................................................... 128 4.2 Material and methods....................................................................................................... 132 4.2.1 Study area ................................................................................................................. 132 4.2.2 Modelling approach................................................................................................... 133 4.2.3 Model parameters and functional groups.................................................................... 135 4.3 Results ............................................................................................................................. 139 4.4 Discussion........................................................................................................................ 142 4.4.1 Description of the Kaloko-Honokōhau system............................................................ 142 4.4.2 Turtles at carrying capacity ........................................................................................ 144 4.4.3 Potential threats and impacts on natural resources as a result of urban development ... 147 4.4.4 Conclusions............................................................................................................... 151 4.5 References ........................................................................................................................ 161 vi  5. Regional-scale seagrass habitat mapping in the wider Caribbean using Landsat sensors: Applications to conservation and ecology................................................................... 172 5.1 Introduction ..................................................................................................................... 172 5.2 Data and methods ............................................................................................................ 176 5.2.1 Landsat data and Millennium Coral Reef Mapping Project (MCRMP)....................... 176 5.2.2 Study areas ................................................................................................................ 177 5.2.3 Image processing ....................................................................................................... 177 5.2.4 Accuracy assessment ................................................................................................. 179 5.3 Results ............................................................................................................................. 182 5.4 Discussion........................................................................................................................ 184 5.4.1 Remote sensing considerations................................................................................... 184 5.4.2 Implications for conservation ..................................................................................... 188 5.4.3 Conclusion and perspectives ...................................................................................... 190 5.5 References ........................................................................................................................ 202 6. Measuring progress towards global marine conservation targets................................... 209 6.1 Introduction ..................................................................................................................... 209 6.1.1 Need for accurate baselines on tropical marine habitat areas....................................... 210 6.2 Data and methods ............................................................................................................ 211 6.3 Results and discussion ...................................................................................................... 211 6.3.1 Need for functional information derived from habitat inventories ............................... 214 6.3.2 The way forward: Using remote sensing to establish accurate baselines....................... 215 6.3.3 Conclusion ................................................................................................................ 216 6.4 References ........................................................................................................................ 219 7. Conclusions..................................................................................................... 221 7.1 References ........................................................................................................................ 232 Appendices ......................................................................................................... 238 Appendix 1 - Clutch size for hawksbill, loggerhead, and green turtles ...................................... 238 Appendix 2 - Emergence success data for hawksbill, loggerhead, and green turtles .................. 247 Appendix 3 - Total estimate of adult hawksbill turtles ............................................................. 238 Appendix 4 - Estimate of total number of adult loggerhead turtles ........................................... 240 Appendix 5 - Total estimate of adult green turtles ................................................................... 242 Appendix 6 - Species included under each reef fish functional group ....................................... 250 Appendix 7 - Input data, references, and relevant observations for Kaloko model.. .................. 253  vii  List of tables Table 2.1 - Empirical equations used to convert curved carapace length (CCL-cm) into straight carapace length (SCL - cm) measurements for individual species....................................... 47 Table 2.2 - Life history parameters extracted from the published literature for all three species of sea turtle considered here ....................................................................................................... 48 Table 2.3 - Total number of individuals in the population (all life stages), percentage of adults, and overall population biomass (all life stages) for hawksbill, loggerhead, and green turtles ...... 49 Table 2.4 - Slope (β estimates) of -log(1-CE) [conversion efficiency] vs. log(W) [mass] plots for hawksbill, loggerhead, and green turtles. Regression lines were forced through three separate estimates of asymptotic mass (log(W∞))............................................................................. 50 Table 2.5 - Consumption to biomass ratios for the overall hawksbill, loggerhead, and green turtle populations in the western Atlantic................................................................................... 51 Table 2.6 - Estimates of food consumption for an age-structured population of hawksbill and green turtles (neritic life phases only).......................................................................................... 52 Table 3.1 - Trophic parameters for all functional groups of the balanced Caribbean reef model .... 107 Table 3.2 - Diet composition matrix for the functional groups in the balanced model for the Caribbean ...................................................................................................................... 108 Table 3.3 - Ecological indicators of the Caribbean model as pertaining to community energetics, community structure, and cycling of nutrients................................................................. 110 Table 4.1 - Trophic parameters for all functional groups of the balanced Kaloko-Honokōhau National Historical Park (Kaloko) model ........................................................................ 152 Table 4.2 - Diet composition matrix for the functional groups in the balanced model of Kaloko- Honokōhau National Historical Park (Kaloko) ............................................................... 153 Table 4.3 - Summary of outputs from the ecological network analysis ......................................... 155 Table 5.1 - List of sites for which thematic seagrass habitat maps were derived in this paper ........ 192 Table 5.2 - Accuracies achieved for two seagrass classes (* with the exception of San Blas and Akumal, one class only) and one generic 'other' class ...................................................... 193 Table 5.3 - Estimated seagrass surface areas (in km2) at each study site........................................ 194 Table 6.1 - Representative comparison of coral reef areas as estimated based on remote sensing (Landsat ETM+) with previously published estimates ..................................................... 217  viii  List of figures Figure 2.1 - Map of the western Atlantic region ............................................................................ 53 Figure 2.2 - Length-mass relationship for western Atlantic hawksbill turtles .................................. 54 Figure 2.3 - Length-mass relationship for western Atlantic loggerhead turtles ................................ 55 Figure 2.4 - Length-mass relationship for western Atlantic green turtles ........................................ 56 Figure 2.5 - Median sample size for growth rate measurements per 10-cm size class for hawksbill, loggerhead, and green turtles in the western Atlantic......................................................... 57 Figure 2.6 - Median growth rates in carapace length (cm·yr-1) according to 10-cm size bins for hawksbill, loggerhead, and green turtles in the western Atlantic ........................................ 58 Figure 2.7 - Relationship of growth rate (cm·yr-1) to carapace length for different populations of hawksbill, loggerhead, and green  turtles in the western Atlantic ....................................... 59 Figure 2.8 - Hawksbill growth rate (cm carapace length) for 10-cm size bins .................................. 60 Figure 2.9 - Loggerhead turtle growth rate (cm carapace length) for 10-cm size bins ...................... 61 Figure 2.10 - Green turtle growth rate (cm carapace length) for 10-cm size bins ............................. 62 Figure 2.11 - Generalized life history of western Atlantic hawksbill turtles .................................... 63 Figure 2.12 - Generalized life history of western Atlantic loggerhead turtles .................................. 64 Figure 2.13 - Generalized life history of western Atlantic green turtles........................................... 65 Figure 2.14 - Age-structured population biomass distribution for hawksbill, loggerhead, and green turtles in the western Atlantic ........................................................................................... 66 Figure 2.15 - Log-log plot of food conversion efficiency (CE) for hawksbill, loggerhead and green turtles............................................................................................................................... 67 Figure 2.16 - Mass-specific metabolic rates (MRs) (log; W·g-1) for loggerhead and green turtles...... 68 Figure 2.17 - Age-structured food consumption estimates (tonnes·year-1) for the western Atlantic hawksbill, loggerhead, and green sea turtle populations..................................................... 69 Figure 3.1 - Map of Puerto Rico and the US Virgin Islands ......................................................... 111 Figure 3.2 - Potential relationship between seagrass biomass (i.e., refuge capacity) with increasing green turtle abundance and vulnerability of prey to predators (modified after Canion and Heck (2009) and implemented in Ecopath as mediation scenarios ................................... 112 Figure 3.3 - Biomass trajectories (final relative to initial biomasses) of selected trophic guilds under a simulated increase in turtle biomass implemented concomitantly with mediation 1 ......... 113 Figure 3.4 - Biomass trajectories (final relative to initial biomasses) of selected trophic guilds under a simulated increase in turtle biomass implemented concomitantly with mediation 2 ......... 114 ix  Figure 3.5 - Biomass trajectories (final relative to initial biomasses) of selected trophic guilds under a simulated increase in turtle biomass implemented concomitantly with mediation 3 ......... 115 Figure 4.1 - Map of Kaloko-Honokōhau National Historical Park (Kaloko) ................................ 156 Figure 4.2 - Graphical representation of trophic flows within the Kaloko reef ecosystem ............. 157 Figure 4.3 - Proportions (%), in terms of biomass (t·km-2) of functional groups at Kaloko............. 158 Figure 4.4 - Mixed trophic impact analysis for herbivores at Kaloko............................................ 159 Figure 4.5 - Fate of total system throughput at Kaloko................................................................ 160 Figure 5.1 - Map of the Caribbean and geographic location of sites processed for this study ......... 195 Figure 5.2 - Flow chart of data processing steps involved in mapping seagrasses at the large scale of the Caribbean region ...................................................................................................... 196 Figure 5.3 - Examples of Millennium polygons for the Belizean focal area .................................. 197 Figure 5.4 - Accuracy assessment points selected for Lee Stocking Island (Exuma, Bahamas) using an IKONOS image and augmented with in situ observations .......................................... 198 Figure 5.5 - Thematic map for all of the Bahamas and for the Little Bahama Bank with two seagrass classes (green) and a single ‘other’ class (grey)................................................................. 199 Figure 5.6 - Examples of habitat maps displaying two seagrass classes and a single ‘other’ class that can be compared with previous published peer-reviewed studies...................................... 200 Figure 5.7 - Examples of temporal variation in vegetation cover, probably due to algal blooms, in two pairs of Landsat images ........................................................................................... 201 Figure 6.1 - Reef habitat connectivity.......................................................................................... 218 x  Acknowledgements There are many to whom I am indebted for their assistance, advice, and moral support in the completion of my doctoral work. I offer my sincerest gratitude to all who have made this effort possible. First and foremost, I wish to extend a most sincere thank you to all members of my committee, Dr Daniel Pauly, Dr Brian Klinkenberg, Dr Karen Bjorndal, Dr Alan Bolten, Dr Villy Christensen, and Dr Jackie Alder, for their support, and encouragement. ‘Merci’ also for affording me great independence in research, providing me with the freedom and opportunity to explore my potential during my time at UBC (in particular through the pursuit of a wide variety of contract opportunities), and for your responsiveness and the extra time and effort you put in at the 11th hour so I could squeeze in the extra chapter and still deliver.  Sections of this thesis were conducted as part of the Sea Around Us Project, a scientific cooperation between UBC and the Pew Environment Group. Funding towards completion of my studies was also provided by a tuition (and partial tuition) waiver UBC scholarship, a Mia Tegner Grant, and two research fellowships: one provided by the US National Marine Fisheries Service (Grant # WC 133F04 SE1183), the other by the Disney Wildlife Conservation Fund (Grant # UF-04-01) to Dr Karen Bjorndal, Dr. Alan Bolten and myself. Support was further provided by: the Institut de Recherche pour le Dévelopement (IRD) through Dr. Serge Andréfouët; the University of South Florida (USF) through Dr. Serge Andréfouët and Dr. Frank Müller-Karger and the University of Florida, through Dr. Alan Bolten and Dr. Karen Bjorndal, during my stay as a visiting scientist at both institutions; a variety of contract opportunities I took on while completing my dissertation; and in times of need the gracious support of my family, Savas, and Danaé.  I extend deep gratitude to my main research advisor, Dr Daniel Pauly. Thank you for your guidance, patience, support, and dedication, for always forcing me to go beyond the horizon, think outside of the box, and find insights by looking at old datasets in entirely new ways! You always will be a source of inspiration, ideas, wisdom, and good company. I am also indebted to my co- supervisors Dr Karen Bjorndal and Dr Alan Bolten. Thank you for your commitment, pragmatism, dedication, keen eye for detail, and the inherent trust you laid in my abilities - not an easy task given that I was so far away (physically!). For his enthusiasm, understanding, and unwavering support, I would also like to offer a deep thank you to my fourth co-supervisor, Dr Brian Klinkenberg. Tusind tak skal du have for al din hjælp Dr Villy Christensen - and for welcoming me so warmly always… xi  and letting me walk out of your office ‘not blue’!. Thanks are also due to Dr Jackie Alder for agreeing to serve on my committee and for your continued assistance and support throughout my time at UBC.  Un vraiment grand grand merci à Serge Andréfouët, for taking me under your wing and being an incredible mentor. I thank you for always making time for me, for the patience and guidance you have offered me along the way, for inspiring me, for always having faith in me, and thinking about me when opportunities for collaboration arose. It has been a privilege, and I am richer for having had the opportunity to work with you.  Along the way I have also had the incredible fortune to receive the much appreciated support from what is now a very long list of people. Chapter 2 - I would like to thank: E. Harrison and B. Krueger, for the provision of unpublished biometric data for nesting green turtles at Tortuguero, Costa Rica, and foraging hawksbill turtles in Barbados respectively; J. Blumenthal, for sharing raw published data on hawksbill and green sea turtles in the Cayman islands(Blumenthal et al. 2009a, Blumenthal et al. submitted); A. Harewood for providing nesting information for hawksbill turtles in Barbados; P. Eliazar for assistance in tracking down a number of references; M. Etienne and R. Ahrens for statistical advice; D. Varkey for her patience and assistance with coding in R; and T.T. Jones for discussions concerning certain methodological aspects of this chapter. B. Hunt is acknowledged for stimulating discussions and constructive comments on earlier drafts.  Chapter 3 - I would like to thank M. Coll for advice and guidance in the development of this chapter. I also wish to acknowledge P. Mumby for insightful discussions about seagrass beds and reef ecosystems. B. Hunt is acknowledged for constructive comments on earlier drafts.  Chapter 4 - Thanks are due to J. Polovina for his help in getting this project off the ground. R. Most and the National Park Service are acknowledged for facilitating many aspects of my work and valuable input throughout the study. M. Rice and students of the Hawai‘i Preparatory Academy (Kamuela, Hawai‘i) are acknowledged for their help in collecting information on green turtles. E. Brown, T. Clark, M. Deagle, H. Jessop, R. Hoover, M. Landry, L. Marrack, M. Weijerman, K. Uyehara, S. Waddington, and T. Wurth are acknowledged for assistance during the development of the model and/or provision of often unpublished data. I wish to thank C. Payri, J. Smith, C. Squair, and T. Sauvage for all questions related to algae; F. Parrish for discussions pertaining to model xii  development and valuable feedback on earlier drafts; and M. Bailey for drawing up such wonderful illustrations of species representative of functional groups; M. Coll for valuable feedback on analyses; and B. Hunt for assistance on all zooplankton and phytoplankton aspects of the model and constructive comments on earlier drafts.  Chapter 5 and 6 - I acknowledge the support provided by the NASA Data Buy Program led by F. Policelli and T. Frisbie who funded the acquisition of the IKONOS images between 2000 and 2002. The Millennium Coral Reef Mapping Project was funded by NASA grants NAG5-10908 to SA and grant CARBON-0000-0257 to J. Robinson. The NASA/Interdisciplinary Program grant NNG04GO90G to S. Anfréfouët and F. Müller-Karger provided additional support. Field data in Panama were collected thanks to H. Guzmán and J. Akl. C. Kranenburg at USF helped with all aspects of Landsat data management. Thanks are due to N. Davis and S. Morgan for constructive comments on earlier drafts.  I don’t think anyone can truly understand what a PhD thesis “does to you” unless you go through that very process. It is at once the most frustrating and self-deconstructing experience, yet all the while also representing the most remarkable and enlightening learning challenge (and test of endurance). If I traveled this path standing tall at the end of it, it is for I have been blessed, truly, with remarkable colleagues, peers, and the most amazing group of friends. I would particularly like to thank R. Ahrens, S. Guénette, M. Palomares and D. Zeller as well as C. Payri, for valuable input, suggestions at various times during my dissertation and most importantly for your support in the process! Janice and Ann - thank you for what you do! Thank you also to my fellow FC students and friends, current and past, I won’t list names, you know who you are :o) - for listening to my woes and rejoicing when celebrating was in order. A warm ☺ to ISTS folks.  My beautiful friends - my world quite truly would not be the same without you! You have helped me considerably throughout this venture, both logistically and emotionally, in your own ways. Your encouragement and boundless faith over the years have helped keep me together and on track. Thank you to all of you who stood by my side through thick and thin, when my patience was fading. For seeing me through the fog, for always reaching out and knowing how to touch me and make me smile, even when I was withdrawn, and for letting me lean on you. Know that I feel blessed for having you in my life, and never have, nor will, take your friendship for granted: Heather (for being so strong and tolerating my close company for 3 years … and some interesting parking practices!); Frederic (still there after all those years :o)); Nick W (for noodles, raw egg xiii  drinks, and doing the incredible task of reading my intro and conclusions! an out of space sized thank you :o)); Sian (for helping to put me back together when I fell apart once); Joanna (yay for being on my team); Emma; Leigh (for all those moments); Joe; Wilf (for all those favours, and particularly your patience and understanding these last few months!) Michael J (we will have that bottle of wine!); Suzanne; Michelle and the “Brady” bunch; Sonia B (te adoro pecesito); Chiara and Pablito; Zaz, Michelle, Isabel and Rebs - for keeping me in your hearts and thoughts during all these (sometimes withdrawn) years and for your warm welcomes and manic schedule rearrangements so that I could stay with you and/or see you whenever I came for a flying visit through the UK. TTodd - for showing me how it is done, helping a sister out and teaching me what really makes turtles go. Looking forward to that rum in Hawai‘i! Peter M. - for having such faith in me, always thinking of me and counting me in, and granting me opportunities others could only dream of. Barry - for the hugs, redirecting your flights via Vancouver all these years for an annual visit, for the laughs and those shared tough moments, and for restocking my rum cabinet! Divya - for your time, your patience, essential advice, for checking in on me, and always being there for me. You are a wonderful friend to me. Boris - for your enthusiasm, making the time for me, your kind words, and most importantly your warmth and support. Savas - for getting me into this mess to start with ;o), for your incredible generosity, and for keeping me in your heart the way you do. Marta - for your unwavering support, level-headedness, ecopath advice, and treasured friendship. Dr. Lou - for always thinking of me, for giving me what I needed most whenever I needed it most. Walking, at times running, down much of this road with you has been truly special; for your friendship and for all that we share BEAMS** and love. Kate - for always checking up on me, for being such a source of strength and guidance, and truly leading by example. You are a true star and I admire you a great deal! John P – for being the most wonderful you, defining what true friendships are, making me smile and sending beams of support, love and for understanding me the way you do. Danaé - ma soulsister, pour ta tendresse, ta patience, ta presence, ta confiance, tes pom-pom balls, et pour avoir toujours été á l’écoute et gardé ma main dans la tienne tout au long de cette course parfois éperdue. Neil - for your love, compassion and encouragement, for having shared such an important part of this journey and listened and held my hand, and for allowing us both to learn all that was needed in the spaces in between. Mike M. - for willing me on and to be STRONG :o), for sending strength and positivity, always, and for your continuous support despite the distance *HUG*! Julie - ma belle, quel plaisir de t’avoir rencontrée ….merci, merci merci pour avoir été à l’écoute, pour tes coups de pouces surtout dans la dernière ligne droite, pour m’avoir fait rire, et su me remettre d’aplomb quand les émotions prenaient le dessus parfois ;o) Karen - for your enthousiasm and courage, xiv  remembering the important moments, and for volunteering to proof read my thesis, even though you were traveling to Oz and jetlagged. May the future see our collaboration plots come to fruition!  I cannot end without thanking my family, on whose constant encouragement and never-ending love I have relied throughout my studies. Despite the distance you have always stood by my side, supported me in any - and all - the ways you could, and pushed me when I thought I couldn’t anymore. Pama in particular, thank you for the enormous amount of faith you’ve always had in me, for teaching me to never give up, strive for the best, and for believing in my dreams, even when at times they seemed far fetched, rather quirky, and unreal! Mon cici et poussin, pour avoir toujours été présents toutes ces années, malgré la distance et le tumulte quotidien de vos vies respectives. I feel blessed to have you as siblings and may we always look after ourselves and each other. …and Brian… I would not have seen the light at the end of this very long and at times dark tunnel if it wasn’t for you - thank you for your love and encouragement; for being there for me in my most doubtful hours, and pulling me through by showing me again how to walk one step at a time; for your patience and forcing me to slow down sometimes; for keeping me true to myself, helping me reconnect with my core and find my inner strength, and reminding me to see the forest for the trees; for letting me lean on you as hard as I have; for reading drafts after drafts after drafts after drafts after drafts; for your advice and unwavering patience. Your insights, focus, brilliance, discipline, support, faith in what is, and love have been, and will continue to be, an inspiration. I love you. Let the thesis-free life begin and the good times roll! :o) xv  Dedication   For Chantal and Hans-Werner Wabnitz xvi  Co-authorship statement With the exception of Chapters 1 and 7, all chapters have been prepared as stand-alone, peer- reviewed publications. Chapters 4 to 6 are either published or in review at refereed journals, while Chapters 2 and 3 are in preparation for submission. I am the senior author on all papers and assume primary responsibility for design, implementation, analysis, and writing of all co-authored chapters. Any errors or omissions are my own. Details of co-authorship contributions are detailed below.  Chapter 2: My co-author Dr. Daniel Pauly developed the original vision for this chapter. I embraced the opportunity to transform these ideas into reality and developed a much more complex methodology than originally envisaged. I expanded on the original idea, developed and carried out all analyses, and drafted the text. Dr. Daniel Pauly directed my thoughts at critical phases of methodological and analytical progress, which went through multiple redevelopments. Dr. Karen Bjorndal provided data, methodological guidance, and contributed to the idea of this chapter through stimulating discussions and comments on drafts of the manuscript. Dr. Andrew Cooper provided guidance on statistical aspects of the analyses and comments on drafts of the manuscript.  Chapter 3: My co-authors Dr. Karen Bjorndal and Dr. Alan Bolten had the original idea of using Ecopath with Ecosim to investigate the role of green turtles in Caribbean seagrass beds. I transformed this idea into testable hypotheses given the limitations of available data. I compiled all data for the manuscript, carried out model development and analyses, and drafted the text. Dr. Marta Coll was of guidance in model development and analysis. Dr. Villy Christensen and Dr. Daniel Pauly provided edits, comments, and suggestions on drafts of the paper.  xvii  Chapter 4: This chapter’s idea was put forward by researchers at the National Marine Fisheries Service in Hawai‘i, who saw it as an interesting comparison to Chapter 3. I expanded upon it, collected all information for the manuscript, carried out all data management, model development and analyses, and drafted the text. George Balazs and Stacy Hargrove provided green turtle data for the model, stimulating discussions, as well as edits and comments on multiple drafts of the manuscript. Sallie Beavers provided access to data, facilitated contacts to researchers in Hawai‘i and made constructive edits and comments to drafts of the paper. Dr. Villy Christensen contributed helpful suggestions during model development. Dr. Daniel Pauly, Dr. Karen Bjorndal, and Dr. Alan Bolten provided guidance on multiple stages of the work, and edits to several iterations of the paper.  Chapter 5: The original idea and its conceptualization for this paper are mine. I carried out all data management and analyses (with the exception of the image analysis for the Bahamas, which was performed by my co-author Damaris Torres-Pulliza), and drafted the text. Dr. Serge Andréfouët provided guidance on all aspects related to image processing and analysis and granted me access to the Millennium Coral Reef database. He also contributed suggestions to the interpretation of results and edits to drafts of the manuscript. Dr. Phil Kramer provided high resolution imagery and ground truth data. Dr. Frank Müller Karger indirectly financially supported this part of my research by hosting me for several months as a visiting scientist at the University of South Florida, granted me access to the NASA image database, and provided edits to the text.  Chapter 6: As with chapter 5 the idea and its conceptualization for this paper are mine. I analyzed all data and drafted the text. My co-author Dr. Serge Andréfouët provided guidance on, and the data required for the analyses, contributed to sections of the text, and offered constructive feedback on numerous iterations of the paper. Dr. Frank Müller Karger indirectly provided funding to support my research by hosting me for several months as a visiting scientist at the University of South xviii  Florida, granted me access to the NASA image database, and made suggestions on an earlier draft of the manuscript. 1  1.  Introduction "We have lived by the assumption that what was good for us would be good for the world. We have been wrong. We must change our lives so that it will be possible to live by the contrary assumption that what is good for the world will be good for us. And that requires that we make the effort to know the world and to learn what is good for it. We must learn to cooperate in its processes, and to yield to its limits"  - Wendell Berry   1.1  Rationale Sea turtles can have major impacts on the structure and function of the ecosystems that they inhabit and their ecological role at historical abundances has been described as essential to the maintenance of healthy marine ecosystems (Jackson 1997, Jackson 2001, Bjorndal 2003, McClenachan et al. 2006). The detailed investigation of the multiple levels at which sea turtles interact with ecosystems constitutes an important and first step towards determining the abundance at which sea turtles fulfil their ecological roles. Such a comprehensive analysis has never been undertaken for any sea turtle population and is essential for the design of effective future species management and recovery plans.  Sea turtle populations have suffered severe declines primarily as a result of overexploitation (Parsons 1962, Hirth 1997). In the Caribbean for example, estimates of historical green turtle (Chelonia mydas) abundance suggest that the present population represents only approximately 3 – 7 % of the abundance levels prior to European arrival (Jackson et al. 2001a, Moran and Bjorndal 2005, McClenachan et al. 2006). To date, these pre-settlement estimates of sea turtle populations have been considered the most appropriate baselines against which to assess current population status. 2  However, such ‘baseline population levels’ may not serve as appropriate recovery targets given the present degraded state of marine habitats and altered food web structures (Bjorndal and Bolten 2003, Pandolfi et al. 2003, Marsh et al. 2005). For example, documented accounts of permanent loss of seagrass beds (Waycott et al. 2009), the main foraging habitat of green turtles in the western Atlantic (Bjorndal 1980), mean that fewer individuals may now be supported by these ecosystems. Quantitative models, such as those developed using the software Ecopath with Ecosim (EwE) (Christensen and Walters 2004), are useful tools that can help clarify the changes in ecological interactions and ecosystem dynamics that would result from recovering numbers of sea turtles.  Given the degraded state of coastal ecosystems, an accurate and reliable estimate of current seagrass extent (in the case of Caribbean green turtles for example) is needed to adequately inform the development and implementation of successful recovery targets. Such estimates would also provide the baseline data necessary to respond to calls by the international policy-making community to conserve at least 10 % of the world’s marine habitats (CBD 2006). In the absence of comprehensive spatial data this goal remains elusive. Remote sensing is the only practical and most cost-effective means of making large scale synoptic assessments of coastal resources (Mumby et al. 1999, Andréfouët et al. 2000). Surprisingly, despite 30 years of tropical coastal habitat remote sensing, seagrass extent has never been mapped at large scales, although such maps are needed to assist in targeted species recovery plans, and more generally, inform transboundary coastal ecosystem-based management initiatives, including the development of comprehensive networks of Marine Protected Areas (MPAs) (Wood et al. 2008).  3  1.2  Context Sea turtles are remarkable creatures. They hatch on land, scurry over the sand at night into the sea, and on entering the water will swim vigorously and continuously for 24 hours (the “swimming frenzy”) (Salmon and Wyneken 1987, Wyneken and Salmon 1992), so as to be carried out, away from the coast (neritic zone) (Bolten 2003a) and the hungry mouths of predators (Gyuris 1994). In the western Atlantic, once offshore, they join up with major current systems (gyres) (Carr and Meylan 1980, Witham 1991), drift and swim while feeding on organisms within large seaweed rafts and other flotsam (Carr 1986a, Bolten 2003b). With the exception of Atlantic loggerhead turtles (Caretta caretta) we know relatively little about this oceanic phase of sea turtles’ life history (Witham 1980, Carr 1987, Zug and Glor 1998, Bolten 2003b, Reich et al. 2007). For hard shelled species (i.e., with the exception of the leatherback) in the Atlantic, after having spent from a few years to a decade out at sea, the small turtles return close to shore, presumably when the benefit of conditions in the neritic zone outweigh those of staying in the oceanic environment for longer (Snover 2008). They then spend decades at neritic foraging grounds until they reach maturity, at which point they undertake the often long migration back to their natal beaches, and if a female, crawl onto the sand and deposit several clutches of eggs, to start the cycle anew (Musick and Limpus 1997).  Humans have a long-standing fascination for sea turtles and they have figured prominently in the mythology of many cultures (Frazier 2003)… and in their diet (Wing and Wing 2001). Harvesting of sea turtles for the trade of their meat, oil, shells, and eggs has reduced populations that once numbered in the millions to the brink of extinction (Meylan 1999, Jackson et al. 2001a). Hawksbill turtles were particularly sought after for their richly patterned scutes, also referred to as tortoiseshell or bekko, and considered a precious material for millennia (Meylan 1999). Known for the delicacy of their meat, Caribbean peoples may have used green turtles for thousands of years without overexploiting the resource, but indigenous subsistence use was outpaced by the intensive capture of 4  sea turtles that began following the arrival of Columbus (Carr 1986b). Sea turtles could easily be caught while nesting, or harpooned on their foraging grounds. Most importantly sea turtles could be kept alive with a daily dousing of saltwater until needed (Nietschmann 1979), ensuring the provision of fresh meat on the long sailing journeys during the period of exploration, expansion, and settlement of the New World (Parsons 1962). While many countries now have regulations in place to limit this trade, or have banned it altogether, an increasing demand for subsistence and local markets, particularly of green and hawksbill turtles, and a suite of other threats contribute to the continued decline of a number of sea turtle populations. With a few notable exceptions (Bjorndal et al. 1999, Balazs and Chaloupka 2004a, Chaloupka et al. 2008), many populations are considered depleted or declining (e.g., Meylan 1999, Witherington et al. 2009). Besides directed take, sources of mortality include high volumes of incidental capture in fishing gear, debris ingestion, disease, and the widespread loss or degradation of coastal habitats (Eckert 1995). As a result of their marked reduction, sea turtles are classified as Endangered or Critically Endangered by IUCN (World Conservation Union) (IUCN 2009) and listed under the Endangered Species Act in the United States as either threatened or endangered.  In response to the dramatic declines, scientists and conservation practitioners throughout the world have accelerated their attempts to aid in the recovery of sea turtle populations. Current conservation practices include the protection of eggs and females on nesting beaches; reduced or no lighting on nesting beaches, chiefly to avoid the disorientation of hatchlings going to sea (Bjorndal et al. 1999); and the implementation of a number of measures to reduce bycatch of sea turtles in a variety of fisheries, such as spatial or temporal no-fishing closures (Lewison et al. 2003), turtle excluder devices (TEDs) in trawl fisheries (Crouse et al. 1987), and changes in hook design and bait type in pelagic longlines (Watson et al. 2005, Cox et al. 2007).  5  From lessons learned in fisheries management, we now know that initiatives focused on single species, with little consideration for the ecosystem in which the species occurs, are to be avoided (Pauly et al. 1998, Trites et al. 1998). Animals or populations do not occur in isolation, but depend on the environment in which they live, and successful long-term recovery of endangered species can only be achieved if conservation efforts consider the wider goal of protecting an ecosystem's structural (e.g., diversity) and functional (e.g., resilience) attributes. The explicit consideration of the ecosystem in developing management strategies for the recovery of protected species was recognised and embraced in the recommendations for a new, coordinated and comprehensive National U.S. Ocean Policy (U.S. Commission on Ocean Policy 2004), and re-emphasised at a National Oceanic and Atmospheric Administration (NOAA) protected-species stock-assessment implementation plan workshop (Merrick et al. 2007). Planning for the successful recovery of a species therefore, needs to include looking at how changes in a focal species (e.g., increased abundance) affect the complex interaction networks it is involved in at the scale of natural communities (Menge 1995) (i.e., ecosystem-based management).  1.2.1  Ecosystem-based management From the above, it is clear that the development of successful national and international sea turtle management regimes require detailed knowledge of species biology and population dynamics in order to promote population stabilization and species recovery. Yet these regimes need to be anchored in a solid understanding of the ecological roles sea turtles play in the ecosystems of which they are a part (Jones et al. 2004). To that end, a holistic approach is required whereby knowledge gained (i) at the level of individuals, including vital rates such as fecundity, mortality, and growth can be combined with information at the population level and (ii) further integrated at the ecosystem scale.  6  (i) Individual Individual growth rates are an important factor in setting the age structure and other demographics of a population, and understanding such biological processes has important implications for devising successful population management strategies. A number of studies have derived growth estimates based on mark-recapture programmes of wild animals (e.g., Boulon 1994, Bjorndal et al. 1995, Bresette and Gorham 2001, Balazs and Chaloupka 2004b) and skeletochronology on stranded animals (e.g., Zug and Glor 1998, Snover and Hohn 2004, Zug et al. 2006 ). To date, a meta-analysis of existing data has not been conducted. Such an analysis would provide a means to smooth out local variability and the data shortcomings of individual studies, and produce parameter estimates that are broadly applicable at the regional scale. By combining data from feeding experiments with mortality schedules and a generalised growth model, one can then derive food consumption estimates of an individual over its life cycle, which, together with an approximation of population size, can be scaled to an entire population (Pauly 1986, Palomares and Pauly 1989).  (ii) Ecosystem Once the ‘mechanics’ of growth, mortality, and food consumption have been understood at the scale of an individual and current populations, this understanding needs to be placed into an ecosystem context. Given the drastic reduction in population numbers of sea turtles, their importance, and likely impact, at historic abundance levels is difficult to imagine (and oft forgotten). Thus, to answer the question: “Are sea turtle species central to and essential for healthy ecosystem processes or are they relict species whose passing would have little effect on ecosystem function?” (Bjorndal 1999) 7  knowledge of their ecology needs to be integrated into the trophic matrix of the system within which they are found. Important aspects of their ecology to consider are their (a) direct impact as consumers, including their dietary preferences and food consumption rates; and (b) indirect impact resulting from foraging behaviour (e.g., changes to the structure of their foraging habitat and associated behavioural changes in other species). In the case of green turtles in the western Atlantic for example, which feed almost exclusively on turtlegrass (Thalassia testudinum) and crop blades down to 2 - 4 cm from the sediment layer (Bjorndal 1980), such indirect impacts include the potential reduction in the refuge capacity of seagrass beds for a variety of juvenile fish (Jackson et al. 2001a, Adams et al. 2006). Overall, prior to the work presented in this thesis, the integrative study of such ecosystem-level interactions has remained relatively poorly explored (but see Bjorndal (2003)).  1.2.2  Role of sea turtles and ecosystem models as support tools A very useful way to investigate biological interactions in natural communities is to develop ecosystem models. The primary role of models in ecosystem science is to permit controlled exploration of a complex reality and to help elucidate patterns and processes that are not apparent from empirical data alone, thereby guiding research and management efforts. Within pre-defined temporal and spatial resolutions, an ecosystem model must be able to describe the changes in a system based on the selected components that make up the system and generalities of how a system functions, including quantitative estimates of biomasses, flows, trophic and non-trophic interactions (see previous section) and functional dynamic relations. Such models should ideally contribute to a greater understanding of how the system behaves and the driving forces and interactions of the system, as well as allow a prediction in terms of future states of the system.  Probably the fore-runner in its field and the most widely used of such models is the Ecopath with Ecosim (EwE) family of mass balance ecosystem models (Christensen and Pauly 1992, Walters et al. 8  1997, Plaganyi 2007). Following the original work by Polovina (1984), EwE is a unique suite of ecological modelling software in that, unlike most traditional ecological models, it allows the user not only to present an analysis on model parameters and structure at time t (Ecopath) (Christensen and Pauly 1993), but also incorporates a time dynamic aspect of ecological systems (i.e., attempts to predict ecosystem structure and function at time t+1 (Ecosim)) (Walters et al. 1997). As such, EwE allows the integration of a wide variety of information and enables the user to make predictions over a wide range of possible states of nature (Walters and Martell 2004). Moreover, it has the unique advantage, over probably most other modelling methods, of explicitly incorporating indirect non- trophic effects such as mediation (Walters and Martell 2004, Christensen 2008). Such non-trophic effects have often been overlooked and/or ignored, yet are fundamental to fully understanding changes in marine food webs (Heithaus et al. 2008).  Following protection, a number of sea turtle populations have started to recover (Garduño-Andrade et al. 1999, Beggs et al. 2007) and some populations are reported as close to ‘full’ recovery (Chaloupka et al. 2008). Application of EwE then represents a useful and unique heuristic means to investigate the potential wider ecological changes brought about by an increase in turtle abundance, as well as to describe the importance of these animals within their ecosystem. While ecosystem models are unlikely to reach the stage where they can quantitatively and accurately predict all ecosystem dynamics, they may be useful for identifying robust management strategies and exposing trade-offs. Results can also help to identify current gaps in understanding and help to focus research programmes (Walters and Holling 1990).  1.2.3  Sea turtle habitat - large scale mapping using remote sensing Once a better understanding of the role of sea turtles within an ecosystem context has been obtained, the provision of reliable information on the extent of their foraging habitat is an important next step 9  towards determining sustainable recovery goals. This is particularly true given that anthropogenic impacts have led to significant reductions in the area and health of sea turtle foraging grounds (e.g., seagrass (Orth et al. 2006, Short et al. 2006, Waycott et al. 2009)), further highlighting that historical population estimates of sea turtles may be inappropriate recovery goals.  With a few exceptions, current available habitat data suffer from poor spatial representation and low spatial resolution. Given the geopolitical transboundary nature of sea turtle migrations, it appears sensible to develop habitat information at large, eco-regional scales (Lourie and Vincent 2004). Satellite remote sensing provides a tool to develop a reliable, methodologically consistent database of habitat extent over large regions, in a cost effective, objective, and timely fashion (Mumby et al. 1999, Balmford et al. 2005). But mapping at such a large scale will require overcoming a number of methodological challenges, including the heterogeneity of habitats and the lack of field data to be able to validate many of the products. The results of such an initiative, moreover, can provide the sea turtle community with important data from which to develop meaningful recovery targets and conservation plans (including, for example, representative MPA networks) (Gibson and Smith 1999). The products should also provide researchers and managers with a useful and much needed baseline to monitor changes registered in coastal habitats over time.  1.3  Objectives The research I present in this thesis is framed in the context of two main objectives. The first is to develop a better understanding of the ecological role of sea turtles in neritic habitats, focusing on green turtles as a case study. The second is to develop a reliable methodology for the consistent large scale mapping of neritic foraging habitats using remote sensing, and looking at seagrass beds in the 10  wider Caribbean region as a focal study. The ultimate goal of these two objectives is to inform sustainable and ecologically meaningful recovery targets for sea turtles.  1.4  Thesis outline In total there are seven chapters in this thesis, consisting of five research chapters, a general introductory chapter (this, Chapter 1) and concluding chapter. The first three research chapters relate to the first objective, to develop a better understanding of the role of (green) turtles in their neritic foraging environments. In Chapter 2, I summarise relevant ecological information for three species of sea turtles (hawksbill, loggerhead, and green turtles) at the scale of the western Atlantic. The first step includes developing broadly applicable length-mass relationships and a meta-analysis of available growth data. I then integrate these with mortality schedules over the different life stages of the species to present age-structured biomass estimates for the entire population. I finally combine these data with estimates of food conversion efficiency to determine food consumption for each age class. In Chapter 3, using the Ecopath with Ecosim software, I investigate the changes likely to accrue to a seagrass ecosystem through an increase in green turtle biomass. I particularly focus on the likely impacts brought about by the increased grazing activity of a large number of green turtles, and thus concomitant decline in habitat complexity at the ecosystem level (i.e., reduction in refuge capacity of seagrass for small fish and invertebrates). In Chapter 4, I present a detailed Ecopath model to investigate the role of green turtles at a Hawaiian foraging ground, where ancillary data suggest that green turtles are at carrying capacity. In Chapters 5 and 6, I address objective 2. In Chapter 5, I present maps for a number of seagrass areas in the Caribbean region and the methodology I developed to be able to (a) classify remote sensing images consistently over large scales, and (b) assess image accuracy based on a suite of data sets and types. In Chapter 6, I evaluate the accuracy of existing large scale habitat information and highlight the importance of developing 11  reliable databases at such scales to inform both species recovery plans and marine conservation initiatives more broadly. Finally, in Chapter 7, I provide a synthesis of the findings presented in this thesis, comment on strengths and weaknesses of the research, and discuss potential applications of the research findings, as well as future research directions. 12  1.5 References Adams, A. J., Dahlgren, C. P., Kellison, G. T., Kendall, M. S., Layman, C. A., Ley, J. A., Nagelkerken, I. and Serafy, J. E. (2006) Nursery function of tropical back-reef systems. Marine Ecology Progress Series, 318: 287-301. Andréfouët, S., Maritorena, S. and Loubersac, L. (2000) Un bilan de la télédétection appliquée aux milieux coralliens. Oceanis, 26 (3): 311-349. Balazs, G. and Chaloupka, M. (2004a) Thirty-year recovery trend in the once depleted Hawaiian green sea turtle stock. Biological Conservation, 117: 491-498. Balazs, G. H. and Chaloupka, M. (2004b) Spatial and temporal variability in somatic growth of green sea turtles (Chelonia mydas) resident in the Hawaiian Archipelago. Marine Biology, 145 (5): 1043-1059. Balmford, A., Crane, P., Dobson, A., Green, R. E. and Mace, G. M. (2005) The 2010 challenge: Data availability, information needs and extraterrestrial insights. Philosophical Transactions of the Royal Society B-Biological Sciences, 360 (1454): 221-228. Beggs, J. A., Horrocks, J. A. and Kruger, B. H. (2007) Increase in hawksbill sea turtle Eretmochelys imbricata nesting in Barbados, West Indies. Endangered Species Research, 3: 159-168. Bjorndal, K. A. (1980) Nutrition and grazing behavior of the green turtle, Chelonia mydas. Marine Biology, 56 (2): 147-154. Bjorndal, K. (1999) Priorities for research in foraging habitats. In Research and Management Techniques for the Conservation of Sea Turtles. IUCN/SSC Marine Turtle Specialist Group Publication No. 4 (eds K. L. Eckert, K. Bjorndal, F. A. Abreu-Grobois and M. Donnelly), pp. 12-14. Washington, D.C., USA. Bjorndal, K. A. (2003) Role of loggerhead sea turtles in marine ecosystems. In Loggerhead Sea Turtles. (ed A. B. Bolten), pp. 235-254. Smithsonian Books, Washington, D.C., USA. Bjorndal, K. A. and Bolten, A. B. (2003) From ghosts to key species: restoring sea turtle populations to fulfill their ecological roles. Marine Turtle Newsletter, 100: 16-21. Bjorndal, K. A., Bolten, A. B., Coan, A. L. and Kleiber, P. (1995) Estimation of green turtle (Chelonia mydas) growth rates from length frequency analysis. Copeia, 1995 (1): 71-77. Bjorndal, K., Wetherall, J., Bolten, A. and Mortimer, J. (1999) Twenty six years of green turtle nesting at Tortuguero, Costa Rica: An encouraging trend. Conservation Biology, 13 (1): 126-134. 13  Bolten, A. B. (2003a) Variation in sea turtle life history patterns: neritic vs. oceanic developmental stages. In The Biology of Sea Turtles. Volume II. (eds P. L. Lutz, J. A. Musick and J. Wyneken), pp. 243-257. CRC Press, Boca Raton (FL), USA. Bolten, A. B. (2003b) Active swimmers - passive drifters: The oceanic stage of loggerheads in the Atlantic system. In Loggerhead Sea Turtles. (ed A. B. Bolten), pp. 63-78. Smithsonian Books, Washington, D.C. (USA). Boulon, R. H. (1994) Growth rates of wild juvenile hawksbill turtles, Eretmochelys imbricata in St Thomas, United States Virgin Islands. Copeia, 1994 (3): 811-814. Bresette, M. and Gorham, J. (2001) Growth rates of juvenile green turtles (Chelonia mydas) from the Atlantic coastal waters of St Lucie county, Florida, USA. Marine Turtle Newsletter, 91: 5-6. Carr, A. (1986a) Rips, FADS, and little loggerheads. BioScience, 36 (2): 92-100. Carr, A. (1986b) The Sea Turtle: So Excellent a Fishe, University of Texas Publishers. 320 pp. Carr, A. (1987) New perspectives on the pelagic stage of sea turtle development. Conservation Biology, 1 (2): 103-121. Carr, A. and Meylan, A. B. (1980) Evidence of passive migration of green turtle hatchlings in Sargassum. Copeia, 1980 (2): 366-368. CBD (2006) Decisions adopted by the Conference of the Parties to the Convention on Biological Diversity at its eighth meeting (Decision VIII/15, Annex IV). Accessed January, 2009. http://www.cbd.int/doc/decisions/COP-08-dec-en.pdf Chaloupka, M., Bjorndal, K. A., Balazs, G. H., Bolten, A. B., Ehrhart, L. M., Limpus, C. J., Suganuma, H., Troeng, S. and Yamaguchi, M. (2008) Encouraging outlook for recovery of a once severely exploited marine megaherbivore. Global Ecology and Biogeography, 17 (2): 297-304. Christensen, V. (2008) Incorporating behavior in ecosystem modeling. Accessed July, 2009. http://www.ecopath.org/modules/Support/Forum/attachments/Incorporating%20behavior%2 0in%20ecosystem%20modeling.pdf Christensen, V. and Pauly, D. (1992) ECOPATH II - A software for balancing steady-state ecosystem models and calculating network characteristics. Ecological Modelling, 61: 169-185. Christensen, V. and Pauly, D. (eds.) (1993) Trophic Models of Aquatic Ecosystems, ICLARM, Manila, Philippines. 390 pp. Christensen, V. and Walters, C. (2004) Ecopath with Ecosim: methods, capabilities and limitations. Ecological Modelling, 72: 109-139. Cox, T. A., Lewison, R. L., Zydelis, R., Crowder, L. B., Safina, C. and Read, A. J. (2007) Comparing effectiveness of experimental and implemented bycatch reduction measures: The ideal and the real. Conservation Biology, 21: 1155-1164. 14  Crouse, D., Crowder, L. B. and Caswell, H. (1987) A stage based population model for loggerhead sea turtles and implications for conservation. Ecology, 68: 1412-1423. Eckert, K. L. (1995) Anthropogenic threats to sea turtles. In Biology and Conservation of Sea Turtles, revised edition. (ed K. Bjorndal), pp. 611-612. Smithsonian Institution Press, Washington D.C., USA. Frazier, J. (2003) Prehistoric and ancient historic interactions between humans and marine turtles. In The Biology of Sea Turtles. Volume II. (eds P. L. Lutz, J. A. Musick and J. Wyneken), pp. 1-38. CRC Press, Boca Raton (FL), USA. Garduño-Andrade, M., Guzman, V., Miranda, E., Briseno-Duenas, R. and Abreu-Grobois, F. A. (1999) Increases in hawksbill turtle (Eretmochelys imbricata) nestings in the Yucatan Peninsula, Mexico, 1977-1996: Data in support of successful conservation? Chelonian Conservation and Biology, 3 (2): 286-295. Gibson, J. and Smith, G. (1999) Reducing threats to foraging habitats. In Research and Management Techniques for the Conservation of Sea Turtles. IUCN/SSC Marine Turtle Specialist Group Publication No. 4. (eds K. L. Eckert, K. A. Bjorndal, F. A. Abreu-Grobois and M. Donnelly), pp. 184-188. Washington, D.C., USA. Gyuris, E. (1994) The rate of predation by fishes on hatchlings of the green turtle (Chelonia mydas). Coral Reefs, 13 (3): 137-144. Heithaus, M. R., Frid, A., Wirsing, A. J. and Worm, B. (2008) Predicting ecological consequences of marine top predator declines. Trends in Ecology & Evolution, 23 (4): 202-210. Hirth, H. F. (1997) Synopsis of the Biological Data on the Green Turtle Chelonia mydas (Linnaeus 1758). Fish and Wildlife Service, Washington D.C., USA. 120 pp. IUCN (2009) IUCN Red List of Threatened Species. Version 2009.2. Accessed January, 2010. http://www.iucnredlist.org Jackson, J. B. C. (1997) Reefs since Columbus. Coral Reefs, 16: S23-S32. Jackson, J. B. C. (2001) What was natural in the coastal oceans? Proceedings of the National Academy of Science, 98 (10): 5411-5418. Jackson, J. B. C., Kirby, M. X., Berger, W. H., Bjorndal, K. A., Botsford, L. W., Bourque, B. J., Bradbury, R. H., Cooke, R., Erlandson, J., Estes, J. A., Hughes, T. P., Kidwell, S., Lange, C. B., Lenihan, H. S., Pandolfi, J. M., Peterson, C. H., Steneck, R. S., Tegner, M. J. and Warner, R. R. (2001a) Historical overfishing and the recent collapse of coastal ecosystems. Science, 293 (5530): 629-638. 15  Jackson, E. L., Rowden, A. A., Attrill, M. J., Bossey, S. J. and Jones, M. B. (2001b) The importance of seagrass beds as a habitat for fishery species. Oceanography and Marine Biology 39: 269-303. Jones, D. R., Southwood, A. L. and Andrews, R. D. (2004) Energetics of leatherback sea turtles. A step toward conservation. In Experimental Approaches to Conservation Biology. (eds M. S. Gordon and S. M. Bartol), pp. 66-82. University of California Press, Los Angeles (CA), USA. Lewison, R., Crowder, L. and Shaver, D. (2003) The impact of turtle excluder devices and fisheries closures on loggerhead and Kemp's ridley strandings in the Western Gulf of Mexico. Conservation Biology, 17 (4): 1089-1097. Lourie, S. A. and Vincent, A. C. J. (2004) Using biogeography to help set priorities in marine conservation. Conservation Biology, 18 (4): 1004-1020. Marsh, H., De'Ath, G., Gribble, N. and Lane, B. (2005) Historical marine population estimates: Triggers or targets for conservation? The dugong case study. Ecological Applications, 15 (2): 481- 492. McClenachan, L., Jackson, J. B. C. and Newman, M. J. H. (2006) Conservation implications of historic sea turtle nesting beach loss. Frontiers in Ecology and the Environment, 4 (6): 290-296. Menge, B. A. (1995) Indirect effects in marine rocky intertidal interaction webs: Patterns and importance. Ecological Monographs, 65 (1): 21-74. Merrick, R., Kelty, R., Ragen, T., Rowles, T., Sandifer, P., Schroeder, B., Swartz, S. and Valette- Silver, N. (2007) Report of the Protected Species SAIP Tier III Workshop (7-10 March 2006). U.S. Dep. Commerce, NOAA Tech. Memo. NMFS-F/SPO-78, Silver Spring (MD), USA. 79 pp. Meylan, A. B. (1999) Status of the hawksbill turtle (Eretmochelys imbricata) in the Caribbean region. Chelonian Conservation and Biology, 3 (2): 177-184. Moran, K. L. and Bjorndal, K. A. (2005) Simulated green turtle grazing affects structure and productivity of seagrass pastures. Marine Ecology Progress Series, 305: 235-247. Mumby, P. J., Green, E. P., Edwards, A. J. and Clark, C. D. (1999) The cost-effectiveness of remote sensing for tropical coastal resources assessment and management. Journal of Environmental Management, 55 (3): 157-166. Musick, J. A. and Limpus, C. J. (1997) Habitat utilisation and migration in juvenile sea turtles. In The Biology of Sea Turtles. (eds P. L. Lutz and J. A. Musick), pp. 137-164. CRC Press, Boca Raton (FL), USA. Nietschmann, B. (1979) Ecological change, inflation, and migration in the far western Caribbean. Geographical Review, 69 (1): 1-24. 16  Orth, R. J., Carruthers, T. J. B., Dennison, W. C., Duarte, C. M., Fourqurean, J. W., Heck, K. L., Hughes, A. R., Kendrick, G. A., Kenworthy, W. J., Olyarnik, S., Short, F. T., Waycott, M. and Williams, S. L. (2006) A global crisis for seagrass ecosystems. BioScience, 56 (12): 987-996. Palomares, M. L. and Pauly, D. (1989) A multiple regression model for predicting the food consumption of marine fish populations. Australian Journal of Marine and Freshwater Research, 40 (3): 259-273. Pandolfi, J. M., Bradbury, R. H., Sala, E., Hughes, T. P., Bjorndal, K. A., Cooke, R. G., McArdle, D., McClenachan, L., Newman, M. J. H., Paredes, G., Warner, R. R. and Jackson, J. B. C. (2003) Global trajectories of the long-term decline of coral reef ecosystems. Science, 301 (5635): 955-958. Parsons, J. (1962) The Green Turtle and Man, University of Florida Press, Gainesville (FL), USA. 126 pp. Pauly, D. (1986) A simple method for estimating the food consumption of fish populations from growth data and food conversion experiments. Fishery Bulletin, 84 (4): 827-840. Pauly, D., Christensen, V., Dalsgaard, J., Froese, R. and Torres, F. (1998) Fishing down marine food webs. Science, 279 (5352): 860-863. Plaganyi, E. E. (2007) Models for an Ecosystem Approach to Fisheries. FAO Fisheries Technical Publication No. 477, Rome, Italy. 126 pp. Polovina, J. J. (1984) Model of a coral-reef ecosystem. I. The Ecopath model and its application to French Frigate Shoals. Coral Reefs, 3 (1): 1-11. Reich, K., Bjorndal, K. and Bolten, A. (2007) The ‘lost years’ of green turtles: using stable isotopes to study cryptic lifestages. Biology Letters, 3 (6): 712-714. Salmon, M. and Wyneken, J. (1987) Orientation and swimming behavior of hatchling loggerhead turtles Caretta caretta L during their offshore migration. Journal of Experimental Marine Biology and Ecology, 109 (2): 137-153. Short, F. T., Koch, E. W., Creed, J. C., Magalhaes, K. M., Fernandez, E. and Gaeckle, J. L. (2006) SeagrassNet monitoring across the Americas: Case studies of seagrass decline. Marine Ecology, 27 (4): 277-289. Snover, M. L. (2008) Ontogenetic habitat shifts in marine organisms: Influencing factors and the impact of climate variability. Bulletin of Marine Science, 83 (1): 53-67. Snover, M. L. and Hohn, A. A. (2004) Validation and interpretation of annual skeletal marks in loggerhead (Caretta caretta) and Kemp's ridley (Lepidochelys kempii) sea turtles. Fishery Bulletin, 102 (4): 682-692. 17  Trites, A. W., Livingston, P. A., Vasconcellos, M. C., Mackinson, S., Springe, A. M. and Pauly, D. (1998) Ecosystem considerations and the limitations of ecosystem models in fisheries management: Insights from the Bering Sea. In Ecosystem Approaches for Fisheries Management. Proceedings of the 16th Lowell Wakefiled Fisheries Symposium. AK-SG-99-01. pp. 609-620. University of Alaska Sea Grant, Fairbanks (AK), USA. U.S. Commission on Ocean Policy (2004) An ocean blueprint for the 21st century - Final report of the U.S. Commission on Ocean Policy. Accessed November, 2009. http://govinfo.library.unt.edu/oceancommission/documents/full_color_rpt/welcome.html Walters, C. J. and Holling, C. S. (1990) Large-scale management experiments and learning by doing. Ecology, 71 (6): 2060-2068. Walters, C. J. and Martell, S. J. D. (2004) Fisheries Ecology and Management, Princeton University Press, Princeton (NJ), USA. 399 pp. Walters, C., Christensen, V. and Pauly, D. (1997) Structuring dynamic models of exploited ecosystems from trophic mass-balance assessments. Reviews in Fish Biology and Fisheries, 7 (2): 139-172. Watson, J. W., Epperly, S. P., Shah, A. K. and Foster, D. G. (2005) Fishing methods to reduce sea turtle mortality associated with pelagic longlines. Canadian Journal of Fisheries and Aquatic Sciences, 62 (5): 965-981. Waycott, M., Duarte, C. M., Carruthers, T. J. B., Orth, R. J., Dennison, W. C., Olyarnik, S., Calladine, A., Fourqurean, J. W., Heck, K. L., Hughes, A. R., Kendrick, G. A., Kenworthy, W. J., Short, F. T. and Williams, S. L. (2009) Accelerating loss of seagrasses across the globe threatens coastal ecosystems. Proceedings of the National Academy of Sciences of the United States of America, 106 (30): 12377-12381. Wing, S. R. and Wing, E. S. (2001) Prehistoric fisheries in the Caribbean. Coral Reefs, 20 (1): 1-8. Witham, R. (1980) The "lost year" question in young sea turtles. American Zoologist, 20 (3): 525-530. Witham, R. (1991) On the ecology of young sea turtles. Florida Scientist, 54: 179. Witherington, B., Kubilis, P., Brost, B. and Meylan, A. (2009) Decreasing annual nest counts in a globally important loggerhead sea turtle population. Ecological Applications, 19 (1): 30-54. Wood, L. J., Fish, L., Laughren, J. and Pauly, D. (2008) Assessing progress towards global marine protection targets: Shortfalls in information and action. Oryx, 42 (3): 340-351. Wyneken, J. and Salmon, M. (1992) Frenzy and postfrenzy swimming activity in loggerhead, green, and leatherback hatchling sea turtles. Copeia, 1992 (2): 478-484. 18  Zug, G. R. and Glor, R. E. (1998) Estimates of age and growth in a population of green sea turtles (Chelonia mydas) from the Indian River lagoon system, Florida: A skeletochronological analysis. Canadian Journal of Zoology, 76 (8): 1497-1506. Zug, G. R., Chaloupka, M. and Balazs, G. H. (2006) Age and growth in olive ridley sea turtles (Lepidochelys olivacea) from the north-central Pacific: A skeletochronological analysis. Marine Ecology, 27 (3): 263-270.  19  2.  Population dynamics and food consumption of hawksbill (Eretmochelys imbricata), loggerhead (Caretta caretta), and green (Chelonia mydas) sea turtles in the western Atlantic1  2.1  Introduction Hawksbill (Eretmochelys imbricata), loggerhead (Caretta caretta), and green turtles (Chelonia mydas) are three of the six extant sea turtle species found in the Atlantic Ocean. Their populations have been overexploited for centuries (Bräutigam and Eckert 2006) and continue to be threatened throughout their range due to directed take, disease, incidental capture by fishers, destruction of critical nesting and foraging habitat, pollution, and climate change (e.g., Eckert 1995, Lutcavage et al. 1997, Meylan 1999, Mortimer et al. 2000, Lewison et al. 2004). Consequently, hawksbill turtles are classified by the          1 A version of this chapter will be submitted for publication. Wabnitz, C.C.C., Cooper, A., Bjorndal, K., and Pauly, D. Population dynamics and food consumption of hawksbill (Eretmochelys imbricata), loggerhead (Caretta caretta), and green (Chelonia mydas) sea turtles in the western Atlantic. 20  World Conservation Union (IUCN) as Critically Endangered and loggerhead and green turtles are both listed as Endangered (IUCN 2009).  Sea turtles can have major impacts on the structure and function of the ecosystems that they inhabit (Bjorndal and Jackson 2003). Given their current depleted abundances, sea turtles’ role as predators and grazers are likely substantially below historic levels. Conservation actions that facilitate increases in turtle abundance, therefore, have significant implications for trophic food webs. Hawksbill turtles are considered to be the largest spongivores on coral reefs (Meylan 1990, van Dam and Diez 1996). Notably, they selectively feed on Chondrilla nucula (León and Bjorndal 2002), a species recorded as an aggressive competitor in the struggle for space between sponges and corals on reefs (Vicente 1978, Suchanek et al. 1983, Hill 1998). By breaking into the tough exterior of sponges, hawksbill turtles also facilitate feeding on the internal tissue by other spongivores, such as fish (Meylan 1988, Blumenthal et al. 2009a). At natural abundance levels hawksbill turtles may therefore play a key role in maintaining the structure and dynamics, and thus resilience, of coral reef ecosystems. Green turtles are an herbivorous species whose diet, in the Caribbean region, consists primarily of the seagrass species Thalassia testudinum (Bjorndal 1980). Green turtle grazing has been shown to result in the compensatory growth of seagrass (Moran and Bjorndal 2005), to significantly increase seagrass blades’ energy and nitrogen content (Moran and Bjorndal 2007), and to increase the speed of nutrient recycling by reducing the decomposition times of Thalassia testudinum blades (Thayer et al. 1984). At pre-exploitation levels the intense grazing of several million green turtles would have led to a large scale reduction in the complexity of seagrass beds (Jackson 1997, Bjorndal and Bolten 2003, Bjorndal and Jackson 2003), in all likelihood affecting the densities of associated fauna and the nature of predator-prey relationships (Thayer et al. 1984, Chapter 3). The longer blades, and associated longer nutrient cycling times, that characterise today’s systems have been postulated to, at least partly, facilitate the development and spread of seagrass die offs due to disease (Jackson 2001). 21   As governments, researchers, and conservation advocates plan for effective management and conservation solutions to recover sea turtle populations, there is pressing need to understand the ecological role of sea turtles in order to quantify their current and future impacts on ecosystems. Two critical aspects of sea turtle life history are required in order to do so: (i)  Growth rates - Sea turtles are characterised by slow growth rates, late maturity, and high longevity (Chaloupka and Musick 1997). Their life histories are complex, comprising at least an oceanic and neritic (coastal) stage (Bolten 2003a), and spanning entire ocean basins and several decades (Musick and Limpus 1997). Individual turtles’ growth rates may be influenced by a variety of factors, including age/body size, sex, and source rookery (Bjorndal et al. 2000a); food quality and quantity (Diez and van Dam 2002, Kubis et al. 2009); competitors (Bjorndal et al. 2000a); predators; and ambient temperature (Kamezaki 2003). However, it remains unclear how the physiological and environmental variables highlighted above act, either singly or in concert, to influence the growth of sea turtles. (ii)  Food consumption - Upon emergence, hatchlings enter the Atlantic and remain in the oceanic environment until, at a range of sizes, depending on species and location, they recruit to the neritic zone. Little is known about the oceanic phase. But oceanic juveniles of all three of the species considered here are thought to be largely carnivorous, feeding primarily on epipelagic prey such as cnidaria, ctenophores, and salps (Booth and Peters 1972, Bjorndal 1997, Witherington and Witherington 2002, Bolten 2003b, Frick et al. 2009). The transition to the neritic environment is associated with a dietary shift, which differs appreciably among species. Hawksbill turtles in the western Atlantic preferentially consume sponges (Acevedo et al. 1984, Meylan 1988, van Dam and Diez 1997, Blumenthal et al. 2009a) and are fairly selective in their choice of prey species (Meylan 1988, León and Bjorndal 2002), despite also foraging on other benthic invertebrates (Carr and Stancyk 1975, Den Hartog 1980, Pemberton et al. 2000, Blumenthal et al. 2009a). Loggerhead turtles consume a variety of slow moving or sessile hard 22  shelled, benthic invertebrates with diets varying among geographic areas and seasons (Lutcavage and Musick 1985, Dodd 1988, Seney and Musick 2007). Green turtles are largely herbivorous, feeding, in the Caribbean, predominantly on the seagrass Thalassia testudinum and/or algae (Bjorndal 1980, Bjorndal 1982) - although some populations are known to consume small to considerable quantities of animal matter, particularly jellyfish, salps, sponges, and invertebrates (Bjorndal 1980).  With quantitative data on growth rates, diet, and food consumption rates, it is possible to model the population dynamics and trophic impact of turtles within an ecosystem. This will not only enhance our understanding of their ecological roles, but also facilitate the development of management scenarios. Simulation of populations’ response and their trophic impact to such initiatives will be informative prior to the implementation of conservation strategies in the field.  At the level of individual populations, sea turtles may show great variability in growth rates (e.g., Balazs 1982, Kubis et al., 2009) and food consumption. The continued study of growth patterns at local scales (e.g., of individual foraging aggregations) will undoubtedly be valuable in teasing apart some of the factors thought to impact sea turtle life-history trajectories. However, in many instances, the conclusions drawn from small scale studies are marred by their small sample sizes, short recapture intervals, and limited size range of animals encountered at a given site. Moreover, sea turtles’ migratory behaviour means that a single individual may be subject to variations in growth rates and food consumption associated with regionally different conditions. A synthesis of existing data for sea turtles in the western Atlantic (Figure 2.1) presents a means to smooth out regional variability and the data shortcomings of individual studies, and produce parameter estimates that are broadly applicable to the region. Given the encouraging outlook from a number of rookeries following protection from human pressures (e.g., Garduño-Andrade et al. 1999, Richardson et al. 2006, Beggs et al. 2007, Chaloupka et al. 2008), a meta-analysis of growth rate and food consumption 23  is necessary to answer key questions such as: what are the likely impacts of increased sea turtle abundances; what is the carrying capacity of today's ecosystems; and how do these numbers compare with historical population estimates? Such an understanding would also improve our ability to predict how environmental changes are likely to affect sea turtle populations and assist in the development of multilateral agreements and regional conservation strategies.  Here, I have undertaken a meta-analysis of the key facets of sea turtle life history required for deriving regional estimates of sea turtle biomass and food consumption. At the scale of the western Atlantic, using data compiled from a comprehensive review of the literature for hawksbill, loggerhead, and green turtles, this study presents: (i) length-mass relationships based on data sampled throughout the species’ size ranges; (ii) estimates of K (the curvature parameter of the von Bertalanffy growth function (VBGF)) based on estimates of asymptotic size and several sets of published growth rate data; (iii) estimates of population biomass based on the results of (i) and (ii), as well as published information on life history parameters (e.g., size at recruitment and sexual maturity, survival); (iv) consumption/biomass estimates (Q/B), that were derived by calculating conversion efficiencies for each of the three species, combining these values with growth rate information, and integrating these over the modelled populations’ age/size distribution to obtain total population food consumption. The latter calculations build on efforts by Bjorndal and Jackson (2003) and McClenachan et al. (2006), who present consumption estimates for historical abundances of adult Caribbean green and hawksbill turtles.  The method presented is intended as an example of how one might approach the question of how much sea turtle populations consume, to gain greater understanding of species’ impact on ecosystems as predators/grazers. The study’s objective was also to highlight some of the current knowledge gaps in order to help guide future research and improve upon the estimates presented here. 24  bLaW ⋅= 2.2  Materials and methods 2.2.1  Length-mass relationships The relationship between total length (L) and mass (W), for most animals, is generally represented by the equation:                         (1)  whose parameters (a, b) are estimated by the antilog of the intercept, and the slope, respectively, of a regression of the log(W) against log(L). The value of (b) is generally close to 3.  Sea turtles can be measured in a number of ways, thus often requiring standardisation before datasets can be compared. Straight carapace length (SCL) and curved carapace length (CCL) are the most commonly used measurements. Although variations exist in how these can be taken (e.g., there are three possible straight-line carapace lengths: maximum, notch-to-tip, and minimum (Pritchard et al. 1983, Bolten 1999)), most authors do not detail the specific technique used in measuring individuals beyond CCL or SCL. For all analyses, discrepancies were assumed to be minimal. CCL data were converted to SCL, based on published linear regressions of paired CCL and SCL data for the species in question (see equations listed in Table 2.1). Length-mass data pairs from different studies were compiled to cover the widest possible range of sizes to make the parameters of length- mass relationships as representative of all developmental stages (i.e., juveniles, subadults, and adults) as possible (Safran 1992). Data were fitted to the log of length and mass. In the case of hawksbill turtles, the samples were weighted by the mass of individuals, because the large sample size of juveniles would have otherwise biased the results.  25  )1( )( 0ttKt eLL −− ∞ −= 2.2.2  Growth rates I created a database of growth increments derived from mark-recapture studies on animals caught in the wild, or assessed using skeletochronology on stranded wild individuals. Published growth rates were often expressed as the difference between two sizes over the time passed and reported as the arithmetic mean of a given number of individuals within 10 cm-class size bins. In order to make all published data comparable, where growth measurements were given for individual animals, the arithmetic mean of these measurements for the same 10-cm size bins was calculated. Zero growth values and negative growth increments were included in the analysis. However, in a few instances, published 10-cm binned growth rates were derived from measurements that did not include zero and/or negative increments. Discrepancies in growth rates between SCL and CCL were assumed to be minimal (Bjorndal and Bolten 1989).  Short recapture time intervals are less reliable in estimating growth due to measurement errors (Chaloupka and Musick 1997); although individual studies often report no significant difference in results when short intervals are excluded from the analysis (e.g., Collazo et al. 1992). When findings were detailed enough to allow it, only measurements obtained from recapture intervals > 6 months (if possible > 12 months) were included.  For the purpose of this study, growth was assumed to follow a von Bertalanffy growth function (VBGF) (von Bertalanffy 1938), with growth rates declining linearly with length, reaching zero at asymptotic length, and which typically has the form:                 (2)  26  Kt t eLLLL − ∞∞ ⋅−−= )( 0 where Lt is the length at age t, L∞ the asymptotic length or the mean size an animal would reach if it were to grow indefinitely, K a growth coefficient of dimension time-1 and t0 is a theoretical age at which L = 0. In this form, emphasis is placed on t0, which can only be estimated when data regarding absolute age(s) are available (i.e., captive specimens or for species that can be aged reliably in the wild). Sea turtles recruit to the neritic environment after an unknown time as oceanic juveniles, and therefore age is rarely known. Equation (2) can be rewritten as (von Bertalanffy 1960):                   (3)  where L0 is the size of hatchlings as they emerge from the nest (which replaces t0) and the other parameters being as defined in equation (2) (see below for estimates of L∞). The assumption here is that growth from hatching onwards follows the VBGF, which is unlikely. It is more probable that hatchling growth is initially exponential before assuming a VBGF growth pattern. When related to biomass estimates this limitation is easily compensated for by the fact that overall mass attributed to these size classes is small compared to the remainder of the population.  Three estimates of asymptotic length (L∞) were derived for each species: (i) the mean size of nesting females in the western Atlantic, L∞(min), as sea turtles have near zero growth rates after first nesting (Frazer and Ehrhart 1985) and therefore ‘L∞’ can be seen to represent mean population nesting length; (ii) the mean size of nesting females / 0.9, L∞(mid), as some authors consider asymptotic size to be slightly above adult mean size, with older adults distributed normally about the mean (Frazer et al. 1990) ; and (iii) the size of the largest nesting female recorded in the region, L∞(max). These values were compared to the intercept of the weighted regression line derived from measured growth rate increments at y = 0 (L∞RL) to ensure that they fell within the confidence interval generated around L∞RL, and were therefore reasonable given data. 27  )( rttZ rt eNN −−⋅= To account for unequal sample sizes, all data were analysed using weighted general least squares linear regression models. Weights were set equal to the sample size, as standard errors were often unavailable. As datasets did not meet the assumptions of homogeneous variance about the regression line, an additional exponential term was used to correct for heteroskedasticity in the data. For each species, three K (growth curvature) parameters were derived, by ‘forcing’ the weighted regression line through each of the three calculated L∞ estimates. Age-at-length for all three species was then obtained through equation (3).  2.2.3  Population biomass estimates and life cycle Mass-at-age was determined using equation (3) and the species-specific length-mass relationships as defined in equation (1).  Individual species’ life cycles were split into four phases: yearlings, oceanic juveniles, neritic juveniles, and adults (Bolten 2003a). Mortality is usually modelled as:            (4)  where Nt and Nr are number of individuals of a given population at times t and tr respectively, with tr representing the age at recruitment into a given life phase, and where Z is the total instantaneous mortality rate applicable over each life phase. Size at hatching, size at which juveniles recruit from the oceanic into the neritic environment, and size at maturity (Table 2.2), as well as accepted ranges of survival rates from one phase to the next were all derived from the published or gray literature.  The size of individuals at transition between phases was held constant, based on literature estimates, and variation in age at transition was therefore a function of growth curves derived in this study. 28  ∫ ∫ ⋅⋅ ⋅ = max max )/( t t tt t t t t r r dtNW CE Ndtdw B Q Population estimation is complicated by the fact that females of all three sea turtle species nest several times within a breeding season, typically nest in non-annual breeding intervals that may vary in length, and may be reproductively active for decades (Miller 1997). I derived a range of annual number of hatchlings entering the population, based on published (i) number of annual nesters (Appendix 1 - 3), in so far as possible, averaged over a number of years to minimise bias in estimated rookery size caused by interannual variability in breeding numbers (Miller 1997, Broderick et al. 2003, Heppell et al. 2003a); (ii) mean number of total eggs laid per female (i.e., eggs per clutch (Appendix 4) x number of clutches); and (iii) mean emergence success (Appendix 5). Based on the range of hatchling number estimates, and assuming a stable population, I ran 1,000,000 Monte Carlo simulations to determine individual species’ age-structured abundance distribution. All solutions for which the total calculated number of adults fell within estimates of total adult population size were retained (3 - 16 % for all runs). Total adult population estimates were derived from published data on the number of nesters, mean remigration interval, and assuming a 1:1 sex ratio (see Appendix 1 - 3). Finally, each species’ age-structured population biomass distribution was calculated by multiplying the resulting matrix of number of individuals-at-age by the corresponding estimate of mass for each of the three estimates of K.  2.2.4  Food conversion and estimates of food consumption Pauly (1986) proposed the following model to estimate the consumption to biomass ratio Q/B:                                    (5)   29  ( ) F dtdWCE t= β ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛−= ∞W WCE t1 WWCE loglog)1log( ββ −=−− ∞ where Nt is the number of individuals at age t, Wt their mean individual mass, CEt their food conversion efficiency, tr the age at which individuals recruit into the population (or life phase), and tmax the maximum age in the population (or life stage) (Palomares and Pauly 1998). Key assumptions to this equation are (i) a stable age distribution, and (ii) that individuals in this population grow according to the VBGF.  Food conversion efficiency (CE), most often derived from feeding experiments, is defined as the ratio of an individual animal’s mass gain to food ingested (F) over a given time interval (Ivlev 1945):                (6)  As shown in Pauly (1986), CE can also be related to mass by:                                   (7)  where β is an exponent estimated as the slope of the linear regression                       (8)  An advantage of equation (7) over equation (6) is that CE approaches 0 as Wt approaches W∞. The model implies that CE = 1 when W = 0, irrespective of the values of β and W∞.  CE was determined using data from published feeding experiments conducted on animals in the western Atlantic region that reported intake and mass gain of captive individuals concurrently 30  ( ) t t t CE dtdWF = (Stickney et al. 1973, Kaufmann 1975, Buitrago 1987, Swingle et al. 1993, Gutiérrez Montero and Cabrera Peña 1995). Food conversion efficiencies derived from data presented in Gutiérrez Montero and Cabrera Peña (1995) did not appear realistic, and were excluded from further analysis, as they did not decline with increasing age and were much higher than any other published values, in one instance even exceeding 100 %. For all studies considered, temperature ranged between 25˚ C and 29˚ C. All animals were chiefly fed different species of fish, up to three times daily. Data from studies where diets consisted of, or included, dried pellets were excluded from analysis due to uncertainty associated with pellet composition and reliable conversion rates between wet and dry mass. For each species, three general least square regression models were computed based on equation (8) ‘forcing’ the regression line through individual W∞ values (W∞(min), W∞(mid), and W∞(max) corresponding to individual species’ L∞ values). Data for loggerhead turtles from Swingle et al. (1994) showed significantly higher conversion efficiency rates than data from other included studies. This effect is the result of ingesting a nutritionally balanced diet (i.e., much lower intake of artificial diets are required, compared to fish, to produce a gram of mass gain) (Bjorndal 2003). Based on current available information, we do not know the extent to which a turtle can ingest a nutritionally ‘optimal’ diet in the wild, and therefore, whether intake rates on balanced or unbalanced diets are better estimates for wild populations. To allow more direct comparison among the three species, a multiple regression analysis was conducted including a dummy variable for data from Swingle et al. (1994). Thus, the final regression line was drawn using the entire dataset, but removing the ‘artificial diet effect’.  The rate of food consumption as a function of age (Ft) can be determined by rearranging equation (6):                                  (9)  where CEt is an individual’s conversion efficiency as a function of age. 31  UMDSRPC e +++++= AECEFMR t ⋅−⋅= )1(  Validation of food consumption models To validate the food consumption models, metabolic rates (MRs) obtained from data used in this study were compared with published values for green and loggerhead sea turtles within their typical temperature range, and derived from respirometry and doubly labelled water studies. See the review by Wallace and Jones (2008) for a discussion of methods used to measure, infer, or model MRs, and their Table 2 for a summary of reported MRs used for validation purposes. The energy that an animal derives from its food is either used or stored (e.g., for somatic growth, or as adipose tissue respectively) (Nagy 1989, Speakman 1997, Sherwood 2005). The fate of ingested food energy (C) can be expressed by the following equation (Nagy 1989, Speakman 1997):               (10)  where P = production (i.e., somatic and reproductive growth), R = respiration (i.e., metabolic rate (MR)), S = storage (e.g., glycogen) (these terms are often listed together simply as P), D = defecation, Me = methane gas produced in the alimentary tract, and U = excretion (i.e., nitrogenous waste). The combined efficiency of P, S, R, and U (net absorbed energy) is known as assimilation efficiency (AE). Based on data for growth and food intake (Ft) as gross energy intake, conversion rates (1 - CE), and assimilation efficiency (AE), equation (10) can also be rewritten as (Jones 2009):                      (11)  Digestive efficiencies (relative percent of ingested energy that is absorbed across the gut wall (Kitchell and Windell 1972)) can be assumed to range between 60.5 % (mean weighted by the number of turtles) for Thalassia testudinum organic matter in Chelonia mydas (Bjorndal 1985) and 32  upwards of 85 % for most reptiles feeding on animal tissue (Zimmerman and Tracy 1989). As all captive animals were fed a carnivorous diet, assimilation efficiency was conservatively estimated at 80 % (where AE is digestive efficiency minus the absorbed energy that is lost through excretion of nitrogenous waste (Kitchell and Windell 1972)).  All intake rates from the published literature (in g) were converted into Watts (Joules·s-1) based on the diet composition detailed in the references used (see previous section), and energy contents provided for corresponding food items in Sidwell (1981). Obtained MR values were then transformed into mass-specific rates by dividing them by the midpoint of two consecutive weight measurements.  In the case of green turtles, the combined size-dependent intake levels (g dry mass (DM) T. testudinum) listed in Bjorndal (1980) and Williams (1988) were also converted into MRs so as to allow comparison to those derived above. Bjorndal (1980) calculated intake rates based on indigestible lignin ratio and daily faeces production, while Williams (1988) based her estimates on observed individuals’ daily bite counts and bite size. For MR calculations the different-sized turtles reported in the studies were assumed to represent only one growing individual. The length of time between individual mass measurements was estimated from green turtle-specific growth rate and length-mass relationship modelled in this study (see sections 2.2.1 and 2.2.2). Thalassia testudinum energy content was set at 14,000 KJ·kg-1 DM (Bjorndal 1990) and DM to wet mass (WM) ratio at 0.144 (Patriquin 1973). Mass-specific MR calculations were then performed as described above.  Unfortunately, to my knowledge, no directly measured MRs for hawksbill turtles are available to compare with rates estimated here from dietary intake values.  33  Population estimate of food consumption The overall (all life stages) food consumption for each species’ western Atlantic population was computed, as a function of age, for estimates of ‘low K’, ‘K’, and ‘high K’ based on (i) individual species’ mass gain as a function of age, (ii) estimates of CE as predicted from equation (7), and (iii) the size of the population as a function of age (see section 2.2.3 above). Consumption/biomass ratios (Q/B) (year-1) were computed for each species by dividing the total population’s intake by the population’s biomass according to equation (5). Food consumption estimates and Q/B ratios were also derived for each species’ neritic portion of the population only (neritic juveniles and adults).  All statistical analyses were performed in R (R Foundation for Statistical Computing, version 2.10.1).  2.3  Results 2.3.1  Length-mass relationships The meta-analysis resulted in the length-mass relationships shown in Figure 2.2 Figure 2.4. Data available for both green and hawksbill turtles had a bimodal size distribution: the vast majority of individuals were between 20 and 70 cm carapace length, with few individuals sampled in the 70 to 90 cm range, and a second grouping at > 90 cm carapace length. Hawksbill turtles were the smallest of all turtles, while green turtles registered the largest length and mass. Excluding hatchlings, carapace lengths ranged from 22.5 to 98.9 cm for hawksbill turtles, 41.1 to 104.8 cm for loggerhead turtles, and 26.8 to 121.9 cm for green turtles. Individuals’ mass ranged from 0.9 to 86.3 kg for hawksbill turtles, 10.0 to 154.0 kg for loggerhead turtles, and 2.4 to 224.0 kg for green turtles.  34  2.3.2  Growth rates A total of 138 mean growth rate estimates were obtained from the literature (47 for hawksbill turtles, 50 for loggerhead turtles, and 41 for green turtles), summarized per 10-cm size class bins. Each 10- cm published mean growth rate value was itself based on measurements recorded from a number of animals. Sample sizes used to obtain these means varied between 1 and 168 sea turtles. Overall, the sample sizes for mean calculations were highest for hawksbill and green turtles (Figure 2.5), with the greatest number of growth rate values recorded between 20 and 60 cm. Sample sizes were small for > 60 cm size classes. With a few exceptions, loggerhead turtle sample sizes were < 10 individuals (Figure 2.5, centre panel), with similar number of samples per 10-cm size class bins.  Median growth rate values for each size class are presented in Figure 2.6. Mean growth rates for individual 10-cm size class bins ranged from negligible (0.29 cm·year-1 for hawksbill turtles, 0.1 cm·year-1 for loggerhead turtles, and 1.18 cm·year-1 for green turtles), to ≥ 9 cm·year-1. Juvenile loggerhead turtles (45-cm size class) caught in Union Creek, Bahamas grew the fastest (15.7 cm·year-1, Figure 2.7, centre panel; and Figure 2.9) (Bjorndal and Bolten 1988a).  Variation in growth rates was evident both within and among species for any given 10-cm size class bin (Figures 2.8 - 2.10), with the greatest variation observed for the smallest size classes. In some cases individual population trajectories showed increased growth with increasing size (Figure 2.7). However, the overwhelming population trend for all three species was one of declining growth with increasing size, and this was well captured using a linear model (Figures 2.8 - 2.10). This linear trend was strongest for hawksbill turtles (Figures 2.6 and 2.8). Interestingly, both loggerhead and green turtles had a tendency to register a small increase in growth rate upon recruiting to neritic environments, though this ‘hump’ did not deviate appreciably from the linear model (Figure 2.6).  35  For hawksbill turtles, L∞RL was 94.4 cm (CI = ± 48.5 cm) (Figure 2.8). Thus, estimates of 86 cm, 95 cm, and 114 cm for L∞(min), L∞(mid), and L∞(max) respectively, fell within the confidence intervals generated by the weighted linear regression. Results were similar for loggerhead turtles, with L∞RL = 112.1 cm (CI = ± 79.8 cm), and L∞(min) = 92.0 cm, L∞(mid) = 102 cm, and L∞(max) = 115 cm (Figure 2.9), and for green turtles, with L∞RL = 123 cm (CI = ± 118.2 cm), and L∞(min) = 106 cm, L∞(mid) = 117 cm, and L∞(max) = 134 cm (Figure 2.10). For graphical purposes only, small, medium and large sample sizes, as indicated in Figures 2.8 - 2.10, corresponded to < 10, > 10 to < 30, and > 30 for hawksbill and loggerhead turtles, and < 15, > 15 to < 40, and > 40 for green turtles respectively.  Species-specific K estimates derived from weighted linear regression forced through each of the three L∞ estimates are presented in Table 2.2. K values ranged from 0.0359 year-1 for green turtles, to 0.0955 year-1 for loggerhead turtles.  2.3.3  Population biomass estimates and life cycle A summary of the life history parameters for each of the three turtle species in this study is presented in Figures 2.11 - 2.13. An estimated mean of 2,020,000 hawksbill, 5,337,000 loggerhead, and 10,178,000 green turtle hatchlings enter the western Atlantic each year. Survival rates for each life history stage were based on published estimates, and ranged between a minimum of 0.475 for yearlings and a maximum of 0.920 for adult loggerhead turtles (Figures 2.11 - 2.13).  Based on data presented in the literature, size at recruitment to the neritic environment was set at 20 cm, 45 cm, and 25 cm for hawksbill, loggerhead, and green turtles respectively, while size at maturity was set at 75 cm, 87 cm, and 92 cm (Table 2.2). Dependent on estimates of K, hawksbill turtles were found to recruit to the neritic environment at between 3 and 4 years of age, and become sexually mature between the ages of 22 and 24 years (Figure 2.11). Loggerhead turtles recruited to 36  the neritic environment at between 7 and 9 years of age, and reached sexual maturity between 25 and 30 years of age (Figure 2.12), while green turtles were 4 to 5 years old and 31 to 36 years old at neritic recruitment and sexual maturity respectively (Figure 2.13).  Total biomass estimates for hawksbill, loggerhead, and green turtle populations in the western Atlantic are presented in Table 2.3 and Figure 2.14. Not surprisingly, based on the annual number of nesters in the region, green turtles had the greatest, and hawksbill turtles the lowest, overall biomass. Figure 2.14 clearly highlights that for each species, the neritic juvenile stage encompassed the greatest proportion of total population biomass, while adults only represented a small fraction (Table 2.3).  2.3.4  Food conversion and estimates of food consumption Based on feeding regimens from captive studies, β values (see equations (7) and (8)) ranged between a low of 0.0286 for green turtles and a high of 0.0430 for hawksbill turtles (Table 2.4). At the individual species level, close congruence was found among β values derived using L∞(min), L∞(mid), and L∞(max).  Out of all species, the forced regression line derived for hawksbill turtles fitted the available data best (Figure 2.15, top left panel). As outlined in the materials and methods section, datasets for loggerhead turtles kept at the Virginia Marine Science Museum and Columbus Zoo, USA (Swingle et al. 1993) showed higher conversion efficiencies than other published estimates (Figure 2.15, top right panel). This is a result of animals being fed nutritionally balanced diets, compared to ‘fish muscle’ in all other studies. A multiple regression that included a dummy variable (‘artificial diet’ = 1, ‘fish muscle’ = 0) allowed the use of the whole dataset while excluding the ‘artificial diet factor’. Very limited data from the Atlantic (Georgia (USA) and Colombia) were available to calculate 37  conversion efficiencies for green turtles (Figure 2.15, bottom panel). Although the regression was drawn using these data only, data from the Pacific coast of Mexico are also presented to illustrate that calculated estimates approximated the empirically derived rates, inclusive of larger sizes.  Mass-specific metabolic rates (MRs) calculated from data used to derive the food consumption model showed relatively good congruence with resting and field MR measurements using doubly labelled water and respirometry (Figure 2.16). Note that in the case of green turtles, previously published values and data converted from direct food consumption estimates from wild animals provide independent validation of estimates that were derived in this study from captive data.  Age-structured food consumption estimates at the scale of the entire western Atlantic for hawksbill, loggerhead, and green turtle populations are shown in Figure 2.17 (all life stages). Integrating under the curves in Figure 2.17 for the neritic life stages only, total consumption by neritic juveniles and adults amounted to between 21,700 and 43,800 tonnes, 242,000 and 715,000 tonnes, and 720,000 and 1,400,000 tonnes for hawksbill, loggerhead, and green turtles respectively. Q/B estimates for the neritic life stages only of each species, ranged between 2.14 (low K) for green turtles and 3.86 (high K) for loggerhead turtles (Table 2.5).  2.4  Discussion A relatively large number of empirical studies, of growth trends in particular, on hawksbill, loggerhead, and green turtles have been carried out in the western Atlantic. The application of meta- analytical methods allowed the combination of such studies, conducted at different localities and of varying samples sizes, and to generate a suite of models that are generally applicable to the western Atlantic region. The identification of such commonalities is of critical importance in the 38  development of large scale conservation and management strategies for highly migratory and shared resources such as sea turtles.  2.4.1  Length-mass relationships Combining datasets of length and mass for hawksbill, loggerhead, and green turtles provided a robust relationship throughout the individual species’ size ranges. A number of published regressions to date, for logistical reasons, have focused on separate foraging (e.g., Bjorndal and Bolten 1988b, e.g., van Dam and Diez 1996) and nesting aggregations (e.g., Beggs et al. 2007, CCC unpublished), and thus tend to include only a subsample of a species’ size-spectrum at the regional scale. Monitoring of the body mass to carapace length relationships for local aggregations is important, and can give insights into changes in turtle condition factor for example, while further contributing to our knowledge of sea turtles as a whole. Conversely, relationships representative of the size- spectra of hawksbill, loggerhead, and green turtles at the scale of the western Atlantic provide a means to generate biomass estimates for data-poor locations, constitute a basis for comparison with other ocean basins, and are a necessary step in the calculation of population consumption estimates as presented below. Results from the meta-analysis revealed that there are only few length/mass measurements for hawksbill and green turtles in the 65 to 85 cm size range. Future initiatives should, therefore, target this size range, as well as oceanic phase individuals for which, with the exception of loggerhead turtles, measurements are rare.  2.4.2  Growth rates For all three species, considerable overlap was found in growth rates among individual aggregations in the western Atlantic (Figure 2.6). The growth model presented depends on two key assumptions: (i) that growth trajectories can be fitted according to the ‘simplest’ growth model, the VBGF, and (ii) 39  that reasonable estimates of L∞ are available. A wide range of studies with records of nesting female sizes provided the latter. Concerns over the application of the VBGF have been discussed (Chaloupka and Musick 1997, Day and Taylor 1997, Heppell et al. 2003a). However, combined at the regional level, regression of growth rates at age for all three species did not substantially deviate from the assumption of an overall monotonic declining growth curve with increasing carapace length (Figures 2.8 - 2.10). Such a pattern, therefore, suggests that the VBGF provides reasonable estimates of growth rates and size classes. Moreover, even in instances where growth patterns seemed to not entirely conform to the VGBF, with some aggregations of loggerhead and green turtles registering peak growth rates a few years following neritic recruitment (Figure 2.7), the extent of deviation will not result in substantial different estimates of population level intakes (one of this study’s key objectives). It is unclear whether the increased growth rates shortly following settlement may represent a continuation from the pelagic stages or a surge in growth following neritic settlement (Heppell et al., 2003a). In a recent study, Bjorndal and Bolten (2010) hypothesize that such dome-shaped patterns at small size classes in the Atlantic may be associated with delayed growth during the transition from oceanic to neritic diets. The authors further suggest that this ‘limited nutrition’ transition period could be followed by a time of compensatory growth.  In the context of results presented here, departure of individual data points from the overall relationship is expected. Efforts should focus on increasing sample sizes at small (especially for loggerhead turtles) and large size classes (all species), as some apparent trajectories may in part be an artefact of sample size (Figure 2.5). Although continued data sampling at the local scale and further analysis of the source and significance of this variation are important, so are analyses presenting a summary of current results and overarching patterns. As growth data accumulate for all stages of sea turtle species these will provide useful updates to models presented here.  40  2.4.3  Age at maturity Measuring age at maturity directly in wild sea turtles is difficult given the geographic scope and long time spans that any such effort would involve (but see Limpus (2009) for notching of hatchlings). Models based on the assembly of growth rate data, therefore, represent a valuable approach to determine age for any given size. Part of the difficulty in generating such estimates for sea turtles is that, currently, limited or no data are available on the growth and/or duration of the epipelagic phase of western Atlantic green turtles (Zug and Glor 1998, Reich et al. 2007) and hawksbill turtles, respectively. Estimates of age at maturity thus assume that (i) pelagic growth follows the same pattern as indicated by neritic growth measurements, a pattern that seems warranted for loggerhead turtles in the Atlantic (Parham and Zug 1997, Bjorndal et al. 2001) (Figure 2.7 and Figure 2.9); and (ii) growth rates are negligible post maturation. It is reasonable to assume that a similar pattern holds in the Atlantic, and data for loggerhead turtles nesting at Melbourne Beach, Florida (Bjorndal et al. 1983), and hawksbill turtles at Tortuguero, Costa Rica (Bjorndal et al. 1985), support the latter.  Hawksbill turtle (Eretmochelys imbricata) Results indicated that a hawksbill turtle spends on average between 18 and 21 years at neritic foraging grounds, prior to becoming sexually mature (Figure 2.11). This range is in line with the age at maturity estimate of 25 years adopted by Meylan and Donnelly (1999). Individual studies found hawksbill turtles to take 9.4 years to grow from 45.0 to 77.7 cm (at which size the turtle nested) (Bjorndal and Bolten 2010), < 14.7 years between neritic recruitment in Puerto Rico at 23.0 cm and average breeding size (Diez and van Dam 2002), and 16.5 years to grow from 21.3 to 78.8 cm, the minimum adult female size at Buck Island, USVI (Boulon 1994). For comparative purposes, age at maturity for Australian hawksbill turtles is thought to range between 20 and 50 years (Chaloupka and Limpus 1997, Limpus and Chaloupka 1997, Chaloupka et al. 2004). 41  Loggerhead turtle (Caretta caretta) Assuming ca. 42.4 to 59.5 cm SCL (converted from 46.0 to 64.0 cm CCL) as the range during which most loggerhead turtles leave the oceanic habitat, Bjorndal et al. (2000b) and Bjorndal et al. (2003a) estimated the duration of the oceanic stage for North Atlantic loggerhead turtles to last between 6.5 and 11.5 years. This range corroborates the estimates of 7 to 9 years provided by the model presented here - based on the assumption that loggerhead turtles recruited to neritic habitats at a size of 45 cm SCL, and a growth model that accounted for three separate estimates of L∞. The range of values presented for the neritic stage (16 - 23 years) also agrees with estimates derived by Braun- McNeil et al. (2008) (24 year long neritic stage). Similarly, the herein modelled age at maturity estimates (25 - 30 years) (Figure 2.12) are in line with results from earlier studies, which placed age at maturity between 12 and 30 years (Frazer and Ehrhart 1985, Zug et al. 1986, Klinger and Musick 1995), and estimated that it would take approximately 26.5 years for a loggerhead to grow to a size of ~81 cm SCL (converted from CCL) (Bjorndal et al. 2001). A more conservative age at maturity estimate (34 years) was chosen for a stage-based demographic model, presented in the recently completed recovery plan for the northwest Atlantic population of the loggerhead sea turtle (2009). The model assumed a total of 7 years in the oceanic stage, based on data presented in Bjorndal et al. (2000b).  Green turtle (Chelonia mydas) Of the species considered, green turtles took the longest to reach sexual maturity (31 - 36 years, Figure 2.13); a finding that is consistent with their slow growth rates. The long neritic juvenile stage (26 - 32 years) of green turtles is probably due to their herbivorous diet in post-pelagic habitats (Heppell et al. 2003a) and associated lower food assimilation efficiency (~ 60 %, whereas energy digestibility coefficients of 80 % are expected for higher quality diets). Estimates derived herein are 42  in line with results from earlier studies suggesting that green turtles would require 6 (Zug and Glor 1998) to 17 years (Bjorndal and Bolten 1988b, 1995, Bjorndal et al. 2000a) to grow from ~ 30 to 75 cm SCL, and reach mean nesting size between 18 and 36 years of age (Mendonca 1981, Frazer and Ehrhart 1985, Frazer and Ladner 1986). Similarly, Bjorndal et al. (2000a) noted that it takes ~ 30 years for green turtles at Union Creek, Bahamas, to grow from 30 to 82 cm SCL. For comparative purposes, in Hawai‘i, pelagic juveniles only recruit to neritic foraging grounds at a size of ca. 35 cm SCL and are thought to reach maturity between 35 and 50 years of age (Balazs and Chaloupka 2004). This range is consistent with age at maturity estimates derived for several foraging-ground populations comprising the southern Great Barrier Reef green turtle genetic stock (Limpus and Chaloupka 1997, Chaloupka et al. 2004). In contrast, green turtles in Bahía de los Angeles, central Gulf of California, are expected to attain maturity after 9 to 21 years (recruiting to neritic habitats at ~ 20 cm SCL (Seminoff et al. 2002)).  For all three species considered, it is encouraging to see that, by and large, estimates of time spent at neritic foraging grounds prior to sexual maturity derived from individual datasets, irrespective of the means with which these were analysed, corroborated each other.  2.4.4  Population biomass estimates and life cycle Biomass estimates for western Atlantic sea turtle populations were derived by combining disparate datasets on length-mass relationships, growth, and mortality. Of these three aspects of sea turtle’s life history, mortality, particularly for pelagic and neritic size-classes, remains one of the most data poor (Heppell et al. 2003b). This is unsurprising given the long time series required and inherent difficulties in estimating survival from field data (Bjorndal et al. 2003b). The majority of published estimates tend to fall into one of two categories. First, survival probabilities often reflect apparent survival (e.g., Campbell and Lagueux 2005) rather than true survival, due to mortality estimates 43  confounding mortality and permanent emigration (Frazer 1987, Heppell et al. 1996, Bjorndal et al. 2003c). Second, published data tend to represent inferred values rather than empirical estimates (e.g., NMFS 2001, Heppell et al. 2003b). The advantage to the approach presented here lies in the use of survival probability distributions to represent and incorporate some of the natural variability and uncertainty in these estimates. However, as longer time series and more data become available these should be used to revise and update the distributions presented. The synthesis of information conducted in this study has also underscored the need to address a number of existing information deficiencies, including the need for additional demographic information on annual reproductive output (e.g., hatching/emergence success and clutch frequency).  Each species’ model assumed equal adult sex ratio, as encountered for example in loggerheads at northern foraging ground aggregations (NMFS 2001). It is important to note however, that some authors have reported an apparent female dominance (e.g., immature loggerhead turtles along the southern Atlantic coast of the USA (Wibbels et al. 1991, Shoop et al. 1998, Braun-McNeil et al. 2009) and hawksbill aggregations in the Caribbean (León and Diez 1999, Geis et al. 2003)), while others have noted a slightly male-biased sex ratio (e.g., foraging hawksbill turtles at Mona Island, Puerto Rico (Diez and Dam 2003)). Based on current available evidence, a primary sex ratio of 1 : 1, therefore, remains a reasonable assumption.  2.4.5  Food consumption The food consumption model presented is the first to have been derived for any hard-shelled species of sea turtle over the course of its life history. Combined data, once standardised between the difference sources, were shown to behave coherently across gradients. For example, conversion efficiency values declined with increasing size and the resulting slopes, β, were generally consistent, despite the different species under study and variation in feed provided. 44   Comparison of MRs calculated from data used in this study with published MRs, and those converted from intake levels measured in the field (for green turtles), provided independent validation of the food consumption model. The meta-analysis demonstrated remarkable consistency in MR estimates, given the temporal, spatial, and dietary range of studies utilised. The agreement between datasets was particularly surprising in the case of green turtles, where data comparison included green turtles fed carnivorous diets and green turtles feeding on their natural seagrass diet.  The estimated total consumption for neritic hawksbill turtles (i.e., neritic juveniles and adults), ranging between 21,700 and 43,800 tonnes·year-1, was determined using data from captive turtles that were fed fish. These estimates were converted to a corresponding biomass of sponge that hawksbill turtles would consume in the wild. For illustrative purposes, I will use the sponges Chondrilla nucula and Geodia neptunia because their energy content has been studied (Bjorndal 1990, Meylan 1990), and both are selectively preyed upon by hawksbill turtles (Meylan 1988, León and Bjorndal 2002, Blumenthal et al. 2009a). From data presented in Buitrago (1987) and Kaufmann (1975), and from fish composition data in Sidwell (1981), the energy content of the average feed was estimated to average 4,530 kJ·kg-1. Assuming a DM to WM ratio of 0.17 (Rützler 1978, Trautman et al. 2000), the mean annual intake for the neritic portion of the hawksbill population was estimated at between 36,400 and 153,000 tonnes sponge WM·year-1 (Table 2.6). This would correspond to a Q/B estimate of 3.6 - 9.5. Based on a rough estimate of annual digestible energy intake of 11,000 kJ·kg-1, energy digestibility coefficients between 43 and 90 % and the energy content of C. nucula, Bjorndal and Jackson (2003) estimated that hawksbill turtles consume on average 0.77 - 1.61 kg sponge DM·kg-1 turtle·year-1. Using Geodia neptunia energy content published in Meylan (1990) neritic hawksbill turtles would have to consume slightly more to fulfil their energetic requirements (1.60 - 3.35 kg sponge DM·kg-1 turtle·year-1). Based on neritic hawksbill biomass estimates derived in this study (9,960 tonnes - 16,000 tonnes) (Table 2.3), the resulting consumption levels amount to 45  between 45,000 tonnes and 315,000 tonnes of sponge WM·year-1 (Table 2.6). The discrepancy in estimates partly reflects differences in applied energy digestibility coefficients: 43 % in the study by Bjorndal and Jackson (2003), which the authors highlight is likely to be too low because hawksbill turtles feed primarily on sponges; and 80 % in estimates derived in this study, which may well be too high given the refractive nature of much of the sponge organic matter.  Food consumption calculations for the neritic life stages of an age-structured green turtle population, based on data from captive studies, yielded estimates of between 719,000 and 1,400,000 tonnes of fish·year-1. Assuming a digestible energy content of 14,000 kJ·kg-1 Thalassia DM, this is equivalent to 2,140,000 - 4,160,000 tonnes Thalassia WM·year-1 (Table 2.6). This corresponds to Q/B estimates ranging between 6.2 and 7.5. Previous studies found green turtles to consume, on average, 0.74 to 1.77 kg DM·kg-1 turtle·year-1 (Bjorndal 1980, Bjorndal 1982, Williams 1988). Extrapolating to the modelled neritic green turtle population biomass estimates (i.e., 345,000 and 557,000 tonnes (Table 2.3)) yields total consumption estimates ranging between 1,770,000 and 6,840,000 tonnes Thalassia WM·year-1 (based on a DM to WM ratio of 0.144 (Patriquin 1973)) (Table 2.6). Overall, the derived consumption values compare favourably with modelled estimates.  Food consumption estimates derived for an age-structured population represent an important contribution towards our understanding of sea turtles’ roles within their ecosystems. The approach described here directly builds, while at the same time considerably expands, on studies by Bjorndal and Jackson (2003) and McClenachan et al. (2006), in which the authors present intake estimates for the adult portion of Caribbean hawksbill and green turtle populations. Nevertheless, the results described in this study are still necessarily rough as they are based on the small available intake/growth trials of small captive turtles on unnatural diets - some of which are nutritionally balanced and others not. As such, these calculations also illustrate the need, over a range of size 46  classes, for well-designed feeding trials with appropriate diets, and studies attempting to estimate intake rates of wild animals.  2.4.6  Conclusions This study provides an important step in furthering the evaluation of the impact sea turtles have on their ecosystems and, therefore, in improving our understanding of how these may change as a result of recovering populations of sea turtles. The estimates put forward represent an attempt at describing in quantitative terms the impact sea turtles have at their neritic foraging grounds, integrating data on growth, variable mortality schedules, and food consumption estimates for individual species’ life phases. Estimates presented here should be further validated and updated through initiatives aimed at determining the level of prey selectivity that loggerhead turtles, for example, exhibit in the wild (Bjorndal and Jackson 2003), and accurate intake levels of all species for different sizes at different sites in the wild.  A solid understanding of the role of sea turtles within their ecosystems will also improve our ability to predict how anthropogenic impacts, for example, will affect populations, and therefore our capacity to make more informed management decisions (Bjorndal 2003). In the context of the severely degraded state and accelerated loss of most coastal ecosystems (Gardner et al. 2003, Pandolfi et al. 2003, Hughes et al. 2009, Waycott et al. 2009), such knowledge would help elucidate the carrying capacity achievable by recovering populations of sea turtles (Chaloupka et al. 2008, Wabnitz et al. 2008). 47  Table 2.1 - Empirical equations used to convert curved carapace length (CCL - cm) into straight carapace length (SCL - cm) measurements for hawksbill (Eretmochelys imbricata), loggerhead (Caretta caretta), and green turtles (Chelonia mydas).  Species Equation R2 Reference Eretmochelys imbricata SCL= 0.939 * CCL - 0.154 n.a. CITES (2002) Caretta caretta SCL = 0.948 * CCL - 1.442 0.97 Teas (1993)  48  Table 2.2 - Life history parameters extracted from the published literature for all three species of sea turtle considered here. K values were not derived from the literature, but were estimated through weighted general least squares linear regression models of growth data (see methods). High K, K, and low K estimates were derived based on L∞(min), L∞(mid),  and L∞(max) respectively.  Species Hatchling (cm) Neritic recruitment (cm) Sexual maturity (cm) L∞(min)i (cm) L∞(mid)I (cm) L∞(max)I (cm) High K K Low K Eretmochelys imbricata 4.2ii 20iii 75iv 86 95 114 0.0824 0.0671 0.0467 Caretta caretta 4.5v 45vi 87vii 92 102 115 0.0955 0.0749 0.0543 Chelonia mydas 5.0viii 25ix 92x 106 117 134 0.0555 0.0457 0.0359  i See text ii Witzell (1983); van Buskirk et al. (1994); Barry Kruger (pers. comm.) iii Carr (1987); Boulon (1994); van Dam and Diez (1998); León and Diez (1999); Diez and van Dam (2002); Meylan (1999); Blumenthal et al. (2009a) iv “most nesting hawksbill turtles are at least 75 cm in carapace length” (CITES 2002); in Cuba, 50 % mature by 76 - 80 cm SCL and 100 % by 80 cm SCL (Moncada Gavilán et al. 1999); in Puerto Rico hawksbill turtles are assumed to mature at a mean size between 70 and 75 cm SCL (van Dam 1997) v Dodd (1988); van Buskirk and Crowder (1994); Snover et al. (2007); vi Bjorndal et al. (2000b) - converted from a CCL of 49 cm using CCL = 1.388 + 1.053 SCLnt vii Crouse et al. (1987) viii Hirth (1997); van Buskirk and Crowder (1994) ix Bjorndal et al. (2000a), Bagley et al. (2008) x Derived from a visual inspection of the histogram in Carr and Goodman (1970) 49  Table 2.3 - Total number of individuals in the western Atlantic (all life stages), percentage of adults, and overall population biomass (all life stages) for hawksbill (Eretmochelys imbricata), loggerhead (Caretta caretta), and green turtles (Chelonia mydas). Results presented are individual means for three separate values of K (in this case, the slope of the weighted regression line forced though the individual asymptotic sizes (L∞s)); values in brackets represent the 2.5 % and 97.5 % quantiles. Eretmochelys imbricata Low K K High K Total number of individuals in population 5,460,000 [2,760,000 - 8,520,000] 5,610,000 [2,950,000 - 8,600,000] 5,790,000 [3,080,000 - 8,820,000] % adults in population 0.46 [0.24 - 0.87] 0.44 [0.24 - 0.82] 0.43 [0.23 - 0.77] Total biomass of population (tonnes) 11,000 [6,700 - 16,100] 14,000 [8,610 - 20,300] 16,800 [10,200 - 24,400] Biomass of neritic stages only (tonnes) (neritic juveniles and adults) 9,960 [6,200 - 14,400] 13,400 [8,280 - 19,200] 16,000 [9,770 - 23,100] Caretta caretta Total number of individuals in population 20,700,000 [14,400,000 - 27,900,000] 20,300,000 [14,300,00 - 27,100,000] 22,200,000 [15,700,000 - 29,600,000] % adults in population 0.59 [0.39 - 0.86] 0.60 [0.41 - 0.87] 0.54 [0.37 - 0.79] Total biomass of population (tonnes) 144,000 [98,900 - 199,000] 166,000 [116,000 - 225,000] 242,000 [168,000 - 330,000] Biomass of neritic stages only (tonnes) (neritic juveniles and adults) 97,600 [69,000 - 132,000] 129,000 [91,000 - 173,000] 197,000 [137,000 - 266, 000] Chelonia mydas Total number of individuals in population 47,100,000 [29,500,000 - 67,300,000] 49,500,000 [31,400,000 - 70,000,000] 53,200,000 [34,300,000 - 75,100,000] % adults in population 0.47 [0.29 - 0.75] 0.44 [0.28 - 0.70] 0.41 [0.26 - 0.64] Total biomass of population (tonnes) 357,000 [255,000 - 477,000] 439,000 [316,000 - 583,000] 566,000 [406,400 - 753,200] Biomass of neritic stages only (tonnes) (neritic juveniles and adults) 345,000 [248,000 - 459,000] 439,000 [316,000 - 583,000] 557,000 [400,000 - 740,000] 50  Table 2.4 - Slope (β estimates) of -log(1-CE) [conversion efficiency] vs. log(W) [mass] plots for hawksbill (Eretmochelys imbricata), loggerhead (Caretta caretta), and green turtles (Chelonia mydas) and given three estimates of asymptotic mass (log(W∞)). W∞(min) is associated with an estimate of ‘high K’, whereas W∞(max) corresponds to values derived for ‘low K’.  Eretmochelys imbricata  Caretta caretta  Chelonia mydas W∞min = 59.39 kg y = -0.043x + 0.2052 W∞min = 96.23 kg y = -0.0315x + 0.1607 W∞min = 152.60 kg y = -0.0314x + 0.163 W∞ = 77.34 kg y = -0.0402x + 0.1968 W∞ = 126.85 kg y = -0.0328x + 0.1634 W∞ = 204.83 kg y = -0.0302x + 0.1604 W∞max = 125.42 kg y = -0.036x + 0.183 W∞max = 174.91 kg y = -0.0301x + 0.1579 W∞max = 306.92 kg y = -0.0286x + 0.1571  51  Table 2.5 - Consumption to biomass ratios for hawksbill (Eretmochelys imbricata), loggerhead (Caretta caretta), and green turtle (Chelonia mydas) populations (all life stages) in the western Atlantic. Standard deviations are presented in parentheses. Values presented in [square brackets] are for the neritic portion of the population only.  Q/B Species Low K K High K Eretmochelys imbricata  2.30 (±0.07) [2.17 (±0.05)] 2.60 (±0.05) [2.53 (±0.05)] 2.80 (±0.05) [2.73 (±0.04)] Caretta caretta  2.78 (±0.04) [2.48 (±0.02)] 3.22 (±0.04) [2.98 (±0.02)] 3.86 (±0.04) [3.63 (±0.02)] Chelonia mydas  2.14 (±0.03) [2.08 (±0.03)] 2.34 (±0.03) [2.31 (±0.03)] 2.54 (±0.03) [2.51 (±0.02)]  52  Table 2.6 - Estimates of food consumption for an age-structured population of hawksbill (Eretmochelys imbricata) and green turtles (Chelonia mydas) (neritic life phases only). Modelled estimates represent this study’s model outputs whereas the second set of values for each species represents estimates of food consumption based on published intake rates kg-1 turtle. Estimates are presented as dry mass (DM) and wet mass (WM) assuming a DM : WM ratio of 0.17 for sponge (Rützler 1978, Trautman et al. 2000) and 0.144 for seagrass (Patriquin 1973). Food consumption for all neritic life stages Eretmochelys imbricata  Low High  References and comments Modelled (extrapolated from results obtained based on a fish diet) 21,700 43,800 tonnes year-1  C. nucula                                    (DM) 6,180 12,500 tonnes year-1 Assuming 80 % digestibility (WM) 36,400 73,300 tonnes year-1 G neptuni                                   (DM) 12,900 25,900 tonnes year-1 Assuming 80 % digestibility (WM) 75,700 153,000 tonnes year-1 From published intake rates C. nucula                                    (DM) 0.77 1.61 kg kg-1 turtle year-1 Bjorndal and Jackson (2003) Assuming 43 – 90 % digestibility (DM)  7,660  25,700 tonnes year-1 (WM)  45,000  151,000 tonnes year-1 G neptuni                                    (DM) 1.60 3.35 kg DM kg-1 turtle year-1 Bjorndal and Jackson (2003); Meylan (1990) Assuming 43 – 90 % digestibility (DM)  15,900  53,510 tonnes year-1 (WM)  93,800  315,000 tonnes year-1 Chelonia mydas Modelled (extrapolated from results obtained based on a fish diet)  719,000  1,400,000 tonnes year-1  Thalassia testudinum                 (DM)  308,000  599,000 tonnes year-1 Assuming 60.5 % digestibility based on Bjorndal (1985) (WM)  2,140,000  4,160,000 tonnes year-1 From published intake rates  0.74 1.77 kg DM kg-1 turtle year-1  Bjorndal (1980) ; Bjorndal (1982); Williams (1988) (DM)  255,000  985,000 tonnes year-1 (WM)  1,770,000  6,840,000 tonnes year-1 53                   Figure 2.1 - Western Atlantic region. Map modified from www.reefbase.org United States Equator 54  0 20 40 60 80 100 0 20 40 60 80 Straight carapace length (cm) M as s (k g) Mass =0.00044*Length2.652 N= 2069 R2 = 0.974                       Figure 2.2 - Length-mass relationship for western Atlantic hawksbill (Eretmochelys imbricata) turtles. Data are from Barbados (Beggs et al. 2007, Krueger unpublished), the Bahamas (Bjorndal and Bolten unpublished), the Cayman Islands (Blumenthal et al. 2009a, Blumenthal et al. 2009b), Honduras (Dunbar et al. 2008), Puerto Rico (USA) (van Dam and Diez 1996), and Suriname (Pritchard 1969).  55  0 20 40 60 80 100 0 50 100 150 Straight carapace length (cm) M as s (k g) Mass =0.00053*Length2.678 N= 347 R2 = 0.928                         Figure 2.3 - Length-mass relationship for western Atlantic loggerhead (Caretta caretta) turtles. Data are from the Chesapeake (VA), USA (Byles 1988, Coles 1999); Florida, USA (Barichivich et al. 1997, Campbell and Sulak 1997); and Delaware Bay (NC), USA (Eggers et al. 1992).   56  0 20 40 60 80 100 120 0 50 100 150 200 250 Straight carapace length (cm) M as s (k g) Mass= 0.00014*Length2.981 N= 806 R2 = 0.995                         Figure 2.4 - Length-mass relationship for western Atlantic green (Chelonia mydas) turtles. Data are from the Cayman Islands (Blumenthal et al. submitted); Florida (USA) (Carr and Caldwell 1956, Barichivich et al. 1997, Campbell and Sulak 1997, Gilbert 2005); Suriname (Pritchard 1969); Tortuguero, Costa Rica (CCC unpublished); and Great Inagua, Bahamas (Bjorndal and Bolten 1988b).  57               Figure 2.5 - Median sample size (indicated by the horizontal black line in the middle of individual boxes) for growth rate measurements per 10-cm size class for hawksbill (Eretmochelys imbricata), loggerhead (Caretta caretta), and green turtles (Chelonia mydas) (left to right panel) in the western Atlantic. Open circles represent mild outliers. Note that the x-axis scale differs among species.  S a m p l e  S i z e  Carapace length (cm) 25 35 45 55 65 75 Chelonia mydas 25 35 45 55 65 75 85 0 50 100 150 p Eretmochelys imbricata 5 25 35 45 55 65 75 85 95 105 Caretta caretta 58               Figure 2.6 - Median growth rates in carapace length (cm·yr-1) according to 10-cm size bins for hawksbill (left panel), loggerhead (centre panel), and green turtles in the western Atlantic. Median values are indicated by the horizontal black line in the middle of individual boxes; open circles represent mild outliers. Note that the x-axis scale differs among species. 25 35 45 55 65 75 85 0 5 10 15 Eretmochelys imbricata 25 35 45 55 65 75 Chelonia mydas Carapace length (cm) G r o w t h  r a t e  ( c m · y e a r - 1 )  5 25 35 45 55 65 75 85 95 105 Caretta caretta 59                              Figure 2.7 - Relationship of growth rate (cm·yr-1) to carapace length for hawksbill (top panel), loggerhead (centre panel), and green turtles (bottom panel) at different sites in the western Atlantic. Lines connect mean values of 10-cm carapace length increments. Data points plotted as open circles were not linked as information to this end was lacking. Note that the x-axis scale differs among species. Carapace length (cm) G ro w th  ra te  (c m ·y ea r-1 ) 0 2 4 6 8 10 20 30 40 50 60 70 80 Chelonia mydas 0 4 8 12 16 0 20 40 60 80 100 120 Caretta caretta -2 0 2 4 6 8 10 20 30 40 50 60 70 80 90 Eretmochelys imbricata 60                  Figure 2.8 - Hawksbill turtle (Eretmochelys imbricata) growth rate (cm carapace length) for 10-cm size bins. Left panel: The black and dashed lines represent the weighted linear regression and the 95 % confidence intervals respectively. Right panel: Growth rate regression lines forced through L∞min = 86 cm, L∞(mid) = 95 cm, and L∞max = 114 cm. Open circles, grey circles, and solid circles represent data with small, medium, and large sample sizes respectively (see text for definition of sample sizes). Data are from Cabo Rojo, muelle oeste, playa oeste and Lanza Zo, in the Dominican Republic (León and Diez 1999 in CITES, 2002); Dolce Legas, Isla Pinos (CITES 2002), and one other site (Manolis et al. 2006) in Cuba; Mona cliff wall, Mona reef, and Monito cliff wall, in Puerto Rico (Diez and van Dam 2002); West coast bank reef in Barbados (Krueger unpublished data) (Krueger pers. comm. in CITES 2002); and the U.S. Virgin islands (Boulon 1994). Carapace length (cm) G r o w t h  r a t e  ( c m · y e a r - 1 )  -2 0 2 4 6 8 10 20 40 60 80 100 120 Linf(min) Linf Linf(max) 0 20 40 60 80 100 120 -2 0 2 4 6 8 10 (mid) 61                   Figure 2.9 - Loggerhead turtle (Caretta caretta) growth rate (cm carapace length) for 10-cm size bins. Left panel: The black and dashed lines represent the weighted linear regression and the 95 % confidence intervals respectively. Right panel: Growth rate regression lines forced through L∞min = 92 cm, L∞(mid) = 106 cm, and L∞max = 115 cm. Open circles, grey circles, and solid circles represent data with small, medium, and large sample sizes respectively (see text for definition of sample sizes). Data are from the Azores (Bjorndal et al. 2000b); Great Inagua, Bahamas (Bjorndal and Bolten 1988a); Chesapeake Bay (VA), USA (Klinger and Musick 1995); Mosquito lagoon (FL), USA (Mendonca 1981); Cape Canaveral (FL), USA (Eckert and Martins 1989); Cumberland island (GA), USA (Parham and Zug (1997) as in Bjorndal et al.(2003)); Melbourne Beach (FL), USA (Bjorndal et al. (1983) in Bjorndal et al. (2003)); and Core and Pamlico Sounds (NC), USA (Snover et al. 2007, Braun-McNeil et al. 2008). Carapace length (cm) G r o w t h  r a t e  ( c m · y e a r - 1 )  0 20 40 60 80 100 120 0 5 10 15 0 4 8 12 16 0 20 40 60 80 100 120 Linf(min) Linf Linf(max)Linf(mid) 62                   Figure 2.10 - Green turtle (Chelonia mydas) growth rate (cm carapace length) for 10-cm size bins. Left panel: The black and dashed lines represent the weighted linear regression and the 95 % confidence intervals respectively. Right panel: Growth rate regression lines forced through L∞min = 106 cm, L∞(mid) = 117 cm, and L∞max = 134 cm. Open circles, grey circles, and solid circles represent data with small, medium, and large sample sizes respectively (see text for definition of sample sizes). Data are from the Cayman Islands (Blumenthal et al. submitted); Culebra, Puerto Rico (Collazo et al. 1992); Great Inagua, Bahamas (Bjorndal and Bolten 1988b); Indian River Lagoon (FL), USA (Zug and Glor 1998); Mansfield Channel (TX), USA (Shaver 1994); Mosquito Lagoon (FL), USA (Mendonca 1981); St Lucie county (FL), USA (Bresette and Gorham 2001); Trident Submarine Basin, Indian River Lagoon, Sebastian Inlet and St Lucie (FL), USA (Kubis et al. 2009); and the U.S. Virgin Islands (Bjorndal and Bolten 1988b, Boulon and Frazer 1990). Carapace length (cm) G r o w t h  r a t e  ( c m · y e a r - 1 )  0 2 4 6 8 10 20 40 60 80 100 120 140 Linf(max)Linf(min) Linf 20 40 60 80 100 0 2 4 6 8 10 (mid) 63  4 clutches (NMFS, 2009) 120 - 160 eggs [see Appendix 4] Egg Hatchling Survival μ= 0.475 [based on loggerhead data (Heppell et al. 2003; Heppell et al., 2005)] Survival μ= 0.800 [based on loggerhead data (Heppell et al., 2003)] Number μ= 2,020,000 0 500000 1000000 1500000 2000000 2500000 3000000 3500000 0.0 0.2 0.4 0.6 0.8 1.0 Survival μ= 0.920 (Richardson, 1999) 0.85 0.90 0.95 1.00 55  - 90%  emergence success [See Appendix 5] 3,007 – 5,628 nesters [see Appendix 1] 2.5 – 2.7 remigration interval (Beggs et al., 2007; Richardson et al., 1999; Perez Castanada et al., 2007) 1:1 sex ratio 15,035 – 30,091 adults Survival μ= 0.800 [based on loggerhead data (Heppell et al. 2003)] 0.6 0.7 0.8 0.9 1.0 Oceanic Juvenile 0.6 0.7 0.8 0.9 1.0 Neritic Juvenile Yearling Adult 3 years (low K) 2 years (high K) 18 years (low K) 21 years (high K) Size at recruitment = 20 cm Size at maturity = 75 cm                      Figure 2.11 - Generalised life history diagram for western Atlantic hawksbill turtles (Eretmochelys imbricata). Number of years by arrows represents the time range hawksbill turtles are thought to spend as oceanic and neritic juveniles respectively, according to model calculations. All measurements are in SCL (cm). 64  4.1 clutches (NMFS, 2009) 100 - 128 eggs (Dodd, 1988) Egg Hatchling Survival μ= 0.475 (Heppell et al., 2003; Heppell et al., 2005) Survival μ= 0.810 (Heppell et al., 2003) Number μ= 5,340,000 0.0 0.2 0.4 0.6 0.8 1.0 Survival μ= 0.880 (Heppell et al., 2003; Hedges, 2007) 45 - 70%  emergence success (NMFS, 2009) 19,060 – 19,482 nesters [see Appendix 2] 2.5 – 3.7 remigration interval (Richardson et al., 1978; Bjorndal et al., 1983; Ehrhart, unpublished data in NMFS (2009)) 1:1 sex ratio 95,299 – 144,170 adults Survival μ= 0.860 (Heppell et al., 2003; Bjorndal et al., 2003) Oceanic Juvenile Neritic Juvenile Yearling Adult 3e+06 4e+06 5e+06 6e+06 7e+06 8e+06 0.70 0.75 0.80 0.85 0.90 0.95 1.00 b 0.6 0.7 0.8 0.9 1.0 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 b Size at recruitment = 45 cm Size at maturity = 87 cm 16 years (low K) 23 years (high K) 8 years (low K) 6 years (high K)                      Figure 2.12 - Generalised life history diagram for western Atlantic loggerhead (Caretta caretta) turtles. Number of years by arrows represents the time range loggerhead turtles are thought to spend as oceanic and neritic juveniles respectively, according to model calculations. All measurements are in SCL (cm). 65  3 clutches [NMFS (2009)] 110-140 eggs [see Appendix 4] Egg Hatchling Survival μ= 0.475 [based on loggerhead data (Heppell et al., 2003; Heppell et al., 2005)] Survival μ= 0.825 (Bjorndal et al., 2003) Number μ= 10,200,000 0.0 0.2 0.4 0.6 0.8 1.0 Survival μ= 0.825 (Solow et al., 2002; Campbell and Lagueux, 2005; Troëng et al., (2007) 35 – 85 %  emergence success [See Appendix 5] 27,595 – 48,091 nesters [see Appendix 3] 2.6 – 3.6 remigration interval [Hirth (1987)] 1:1 sex ratio 165,570 – 288,546 adults Survival μ= 0.860 (Heppell et al., 2003; Bjorndal et al., 2003) 0.6 0.7 0.8 0.9 1.0 Oceanic Juvenile Neritic Juvenile Yearling Adult 5.0e+06 1.0e+07 1.5e+07 2.0e+07 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 Size at maturity = 92 cm Size at recruitment = 25 cm 26 years (low K) 32 years (high K) 4 years (low K) 3 years (high K)                       Figure 2.13 - Generalised life history diagram for western Atlantic green turtles (Chelonia mydas). Number of years by arrows represents the time range green turtles are thought to spend as oceanic and neritic juveniles respectively, according to model calculations. All measurements are in SCL (cm). 66                Figure 2.14 - Age-structured population biomass distribution for hawksbill (left), loggerhead (centre), and green turtles in the western Atlantic based on survivorship (Figures 10 - 12) and assuming a von Bertalanffy length/age model and a stable age distribution. The solid lines represent results for the upper and lower quantiles of model runs for life history parameters associated with L∞(mid). The lower and upper sets of dashed lines represent results for the upper and lower quantiles of model runs using life history parameters associated with L∞(min) and L∞(max) respectively. Note that the y-axis scale differs among species. Age (years) M a s s  ( t o n n e s )  0 10 20 30 40 50 60 70 0 10000 20000 30000 40000 0 10 20 30 40 50 60 70 0 5000 10000 15000 20000 25000 0 10 20 30 40 50 60 70 0 500 1000 1500 2000 67                     Figure 2.15 - Log - log plot of food conversion efficiency ‘CE’ for hawksbill (top left panel), loggerhead (top right panel), and green turtles (bottom panel), showing the fitted curve through W∞(min), W∞(mid), and W∞(max). Data for hawksbill turtles are from Colombia (Kaufmann 1975) and Venezuela (Buitrago 1987); for loggerhead turtles from Colombia (Kaufmann 1975), two zoological facilities (Swingle et al. 1993) and Georgia, USA (Stickney et al. 1973); and green turtles from Colombia (Kaufmann 1975), and Georgia, USA (Stickney et al. 1973). Although not included in calculations of the regression line for green turtles, extra data points for the Pacific coast of Mexico (Godínez-Domínguez et al. 1993) (open circles) are shown here to illustrate the decline in ‘CE’ with increasing body size. For loggerhead turtles, inclusion of a dummy variable to account for ‘artificial diet’ versus ‘fish diet’ differences, allowed us to draw the regression line based on the entire dataset while removing the ‘artificial diet’ factor from relevant datapoints (see text for a more detailed explanation). Body mass (log; g) C on ve rs io n ef fic ie nc y (-l og (1 -C E )) Legend: ƒ Colombia • Venezuela x Columbus Zoo + VMSM ŠGeorgia o Mexico (Pacific) 0 0.1 0.2 0.3 1 2 3 4 5 6Winf(min) Winf Winf(max) Eretmochelys imbricata 0 0.1 0.2 0.3 1 2 3 4 5 6Winf(min Winf Winf(max) Caretta  caretta 0 0.1 0.2 0.3 1 2 3 4 5 6 Winf(min Winf Winf(max) Chelonia mydas 68                Figure 2.16 - Mass-specific metabolic rates (MR) (log; W g-1) for loggerhead (left panel) and green turtles determined from feeding experiment data used in this study ( ) plotted with MRs of values compiled in Wallace and Jones (2008) (). Data converted from field consumption estimates for green turtles are represented as (S). -5 -4 -3 -2 -1 0 0 1 2 3 4 5 6 -5 -4 -3 -2 -1 0 0 1 2 3 4 5 6 M a s s - s p e c i f i c  M R  ( l o g ;  W  g - 1 )  Mass (log; g) 69                 Figure 2.17 - Age-structured food consumption estimates (tonnes·year-1) for the western Atlantic hawksbill (top panel), loggerhead, (middle panel), and green sea turtle populations. The full lines represent results for the upper and lower quantiles of model runs for life history parameters associated with L∞(mid). The lower and upper sets of dashed lines represent results for the upper and lower quantiles of model runs using life history parameters associated with L∞(min) and L∞(max) respectively. Note that the y-axis scale differs among species.  Age (years) F o o d  c o n s u m p t i o n  ( t o n e s · y e a r - 1 )  0 10 20 30 40 50 60 70 0 1000 2000 3000 4000 5000 6000 0 10 20 30 40 50 60 70 0 20000 40000 60000 80000 100000 0 10 20 30 40 50 60 70 0 20000 40000 60000 80000 100000 70  2.5  References Acevedo, M., Gómez, O. and Berovides, V. (1984) Alimentación de tres especies de quelonios marinos en la plataforma suroccidental. Revista de investigaciónes marinas, 5 (3): 29-36. Bagley, D. A., Kubis, S. A., Bresette, M. J. and Ehrhart, L. M. (2008) Satellite tracking juvenile green turtles from Florida's east coast: the missing size classes found. In 27th Annual Symposium on Sea Turtle Biology and Conservation pp. 73. US Department of Commerce, NOAA, National Marine Fisheries Service, Southeast Fisheries Science Centre. NOAA Technical Memorandum NMFS-SEFSC-569, Myrtle Beach (SC), USA. Balazs, G. H. (1982) Growth rates of immature green turtles in the Hawaiian archipelago. In Biology and Conservation of Sea Turtles. (ed K. A. Bjorndal), pp. 117-125. Smithsonian Institute Press, Washington D.C., USA. Balazs, G. H. and Chaloupka, M. (2004) Spatial and temporal variability in somatic growth of green sea turtles (Chelonia mydas) resident in the Hawaiian archipelago. Marine Biology, 145 (5): 1043- 1059. Barichivich, W. J., Sulak, K. J. and Carthy, R. R. (1997) Characterisation of Kemp's ridley sea turtles in the Florida big bend area during 1997. Southeast Fisheries Science Center, National Marine Fisheries Service, Panama City (FL), USA. 12 pp. Beggs, J. A., Horrocks, J. A. and Kruger, B. H. (2007) Increase in hawksbill sea turtle Eretmochelys imbricata nesting in Barbados, West Indies. Endangered Species Research, 3: 159-168. Bjorndal, K. A. (1980) Nutrition and grazing behavior of the green turtle, Chelonia mydas. Marine Biology, 56 (2): 147-154. Bjorndal, K. (1982) The consequences of herbivory for the life history pattern of the Caribbean green turtle. In Biology and Conservation of Sea Turtles. (ed K. Bjorndal), pp. 111-116. Smithsonian Institution Press, Washington, D.C., USA. Bjorndal, K. A. (1985) Nutritional ecology of sea turtles. Copeia,  (3): 736-751. Bjorndal, K. A. (1990) Digestibility of the sponge Chondrilla nucula in the green turtle, Chelonia mydas. Bulletin of Marine Science, 47 (2): 567-570. Bjorndal, K. A. (1997) Foraging ecology and nutrition of sea turtles. In The Biology of Sea Turtles. (eds P. L. Lutz and J. A. Musick), pp. 199-231. CRC Press, New York (NY), USA. Bjorndal, K. A. (2003) Role of loggerhead sea turtles in marine ecosystems. In Loggerhead Sea Turtles. (ed A. B. Bolten), pp. 235-254. Smithsonian Books, Washington, D.C., USA. 71  Bjorndal, K. A. and Bolten, A. B. (1988a) Growth rates of juvenile loggerheads, Caretta caretta, in the southern Bahamas. Journal of Herpetology, 22 (4): 480-482. Bjorndal, K. A. and Bolten, A. B. (1988b) Growth rates of immature green turtles, Chelonia mydas, on feeding grounds in the southern Bahamas. Copeia,  (3): 555-564. Bjorndal, K. A. and Bolten, A. B. (1989) Comparison of straight-line and over-the-curve measurements for growth rates of green turtles, Chelonia mydas. Bulletin of Marine Science, 45 (1): 189-192. Bjorndal, K. A. and Bolten, A. B. (1995) Comparison of length-frequency analyses for estimation of growth parameters for a population of green turtles. Herpetologica, 51 (2): 160-167. Bjorndal, K. A. and Bolten, A. B. (2003) From ghosts to key species: Restoring sea turtle populations to fulfill their ecological roles. Marine Turtle Newsletter, 100: 16-21. Bjorndal, K. and Bolten, A. (2010) Hawksbill sea turtles in seagrass pastures: success in a peripheral habitat. Marine Biology, 157 (1): 135-145. Bjorndal, K. A. and Bolten, A. B. (unpublished) Morphometric data for hawksbill sea turtles at Inagua, Bahamas. Bjorndal, K. A. and Jackson, J. B. C. (2003) Roles of sea turtles in marine ecosystems: reconstructing the past. In The Biology of Sea Turtles. (eds P. L. Lutz, J. A. Musick and J. Wyneken), pp. 259-273. CRC Press, Boca Raton (FL), USA. Bjorndal, K. A., Meylan, A. B. and Turner, B. J. (1983) Sea turtles nesting at Melbourne Beach, Florida. 1. Size, growth and reproductive biology. Biological Conservation, 26 (1): 65-77. Bjorndal, K. A., Carr, A., Meylan, A. B. and Mortimer, J. A. (1985) Reproductive biology of the hawksbill Eretmochelys imbricata at Tortuguero, Costa Rica, with notes on the ecology of the species in the Caribbean. Biological Conservation, 34 (4): 353-368. Bjorndal, K. A., Bolten, A. B. and Chaloupka, M. Y. (2000a) Green turtle somatic growth model: Evidence for density dependence. Ecological Applications, 10 (1): 269-282. Bjorndal, K. A., Bolten, A. B. and Martins, H. R. (2000b) Somatic growth model of juvenile loggerhead sea turtles Caretta caretta: Duration of the pelagic stage. Marine Ecology Progress Series, 202: 265-272. Bjorndal, K. A., Bolten, A. B., Koike, B., Schroeder, B. A., Shaver, D. J., Teas, W. G. and Witzell, W. N. (2001) Somatic growth function for immature loggerhead sea turtles, Caretta caretta, in southeastern US waters. Fishery Bulletin, 99 (2): 240-246. . 72  Bjorndal, K. A., Bolten, A. B., Dellinger, T., Delgado, C. and Martins, H. R. (2003a) Compensatory growth in oceanic loggerhead sea turtles: Response to a stochastic environment. Ecology, 84 (5): 1237-1249. Bjorndal, K., Bolten, A. and Chaloupka, M. (2003b) Survival probability estimates for immature green turtles Chelonia mydas in the Bahamas. Marine Ecology Progress Series, 252: 273-281. Bjorndal, K. A., Bolten, A. B. and Martins, H. R. (2003c) Estimates of survival probabilities for oceanic-stage loggerhead sea turtles (Caretta caretta) in the North Atlantic. Fishery Bulletin, 101 (4): 732-736 Blumenthal, J. M., Austin, T. J., Bell, C. D. L., Bothwell, J. B., Broderick, A. C., Ebanks-Petrie, G., Gibb, J. A., Luke, K. E., Olynik, J. R., Orr, M. F., Solomon, J. L. and Godley, B. J. (2009a) Ecology of hawksbill turtles Eretmochelys imbricata on a western Caribbean foraging area. Chelonian Conservation and Biology, 8 (1): 1-10. Blumenthal, J. M., Austin, T. J., Bothwell, J. B., Broderick, A. C., Ebanks-Petrie, G., Olynik, J. R., Orr, M. F., Solomon, J. L., Witt, M. J. and Godley, B. J. (2009b) Diving behavior and movements of juvenile hawksbill turtles Eretmochelys imbricata on a Caribbean coral reef. Coral Reefs, 28 (1): 55-65. Blumenthal, J., Austin, T., Bothwell, J., Broderick, A., Ebanks-Petrie, G., Olynik, J., Orr, M., Solomon, J., Witt, M. and Godley, B. (submitted) Life in (and out of) the lagoon: insights into movements of green turtles using time-depth recorders. Journal of Experimental Biology and Ecology. Bolten, A. B. (1999) Management and research techniques for the conservation of sea turtles. (eds K. L. Eckert, K. A. Bjorndal, F. A. Abreu-Grobois and M. Donnelly), pp. 1-5. IUCN Marine Turtle Specialist Group. Publication No. 4. Bolten, A. B. (2003a) Variation in sea turtle life history patterns: Neritic vs. oceanic developmental stages. In The Biology of Sea Turtles. Volume II. (eds P. L. Lutz, J. A. Musick and J. Wyneken), pp. 243-257. CRC Press, Boca Raton (FL), USA. Bolten, A. B. (2003b) Active swimmers - passive drifters: The oceanic stage of loggerheads in the Atlantic system. In Loggerhead Sea Turtles. (ed A. B. Bolten), pp. 63-78. Smithsonian Books, Washington, D.C., USA. Booth, J. and Peters, J. A. (1972) Behavioural studies on the green turtle (Chelonia mydas) in the sea. Animal Behavior, 20 (4): 808-812. Boulon, R. H. (1994) Growth rates of wild juvenile hawksbill turtles, Eretmochelys imbricata in St Thomas, United States Virgin Islands. Copeia, 1994 (3): 811-814. 73  Boulon, R. H. and Frazer, N. B. (1990) Growth of wild juvenile Caribbean green turtles, Chelonia mydas. Journal of Herpetology, 24 (4): 441-445. Braun-McNeil, J., Epperly, S. P., Avens, L., Snover, M. L. and Taylor, J. C. (2008) Growth rates of loggerhead sea turtles (Caretta caretta) from the western North Atlantic. Herpetological Conservation and Biology, 3 (2): 273-281. Braun-McNeil, J., Epperly, S. P., Owens, D. W., Avens, L., Williams, E. and Harms, C. A. (2009) Seasonal reliability of testosterone radioimmunoassay (RIA) for predicting sex ratios of juvenile loggerhead (Caretta caretta) turtles. Herpetologica, 63 (3): 275-284. Bräutigam, A. and Eckert, K. L. (2006) Turning the tide: Exploitation, trade and management of marine turtles in the Lesser Antilles, Central America, Colombia and Venezuela. TRAFFIC International, Cambridge, UK. 551 pp. Bresette, M. and Gorham, J. (2001) Growth rates of juvenile green turtles (Chelonia mydas) from the Atlantic coastal waters of St Lucie county, Florida, USA. Marine Turtle Newsletter, 91: 5-6. Broderick, A. C., Glen, F., Godley, B. J. and Hays, G. C. (2003) Variation in reproductive output of marine turtles. Journal of Experimental Marine Biology and Ecology, 288 (1): 95-109. Buitrago, J. (1987) Cria con fines de repoblacíon de tres especies de tortugas marina en Los Roques, Venezuela. Memoria Sociedad de Ciencias Naturales la Salle, 47 (127-128): 169-200. Byles, R. A. (1988) Behaviour and ecology of sea turtles from Chesapeake Bay, Virginia. Doctoral thesis. College of William and Mary, Williamsburg (VA), USA. Campbell, C. L. and Lagueux, C. J. (2005) Survival probability estimates for large juvenile and adult green turtles (Chelonia mydas) exposed to an artisanal marine turtle fishery in the Western Caribbean. Herpetologica, 61 (2): 91-103. Campbell, C. L. and Sulak, K. J. (1997) Characterisation of Kemp's ridley sea turtles in the Florida big bend area during 1995 and 1996. Southeast Fisheries Science Center, National Marine Fisheries Service, Panama City (FL), USA. 17 pp. Carr, A. (1987) New perspectives on the pelagic stage of sea turtle development. Conservation Biology, 1 (2): 103-121. Carr, A. and Caldwell, D. (1956) The ecology and migrations of sea turtles, I. Results of field work in Florida, 1955 American Museum Novitates, 1793: 1-24. Carr, A. and Goodman, D. (1970) Ecologic implications of size and growth in Chelonia. Copeia, 1970 (4): 783-786. Carr, A. and Stancyk, S. (1975) Observations on the ecology and survival outlook of the hawksbill turtle. Biological Conservation, 8 (3): 161-172. 74  CCC (unpublished) Biometric data including carapace length, width, and weight for green turtles collected at Tortuguero, Costa Rica from 1986 - 1989. Caribbean Conservation Corporation. Chaloupka, M. Y. and Limpus, C. J. (1997) Robust statistical modelling of hawksbill sea turtle growth rates (southern Great Barrier Reef). Marine Ecology Progress Series, 146 (1-3): 1-8. Chaloupka, M. and Musick, J. A. (1997) Age, growth, and population dynamics. In The Biology of Sea Turtles. (eds P. L. Lutz and J. A. Musick), pp. 233-276. CRC Press, Boca Raton (FL), USA. Chaloupka, M., Limpus, C. and Miller, J. (2004) Green turtle somatic growth dynamics in a spatially disjunct Great Barrier Reef metapopulation. Coral Reefs 23 (3): 325-335. Chaloupka, M., Bjorndal, K. A., Balazs, G. H., Bolten, A. B., Ehrhart, L. M., Limpus, C. J., Suganuma, H., Troeng, S. and Yamaguchi, M. (2008) Encouraging outlook for recovery of a once severely exploited marine megaherbivore. Global Ecology and Biogeography, 17 (2): 297-304. CITES (2002) Hawksbill turtles in the Caribbean region: Basic biological characteristics and population status. Convention on International Trade in Endangered Species of Wild Fauna and Flora, 52 pp. Coles, W. C. (1999) Aspects of the biology of sea turtles in the Mid-Atlantic bight. Doctoral thesis. College of William and Mary in Virginia, Williamsburg (NC), USA. Collazo, J. A., Boulon, R., Jr. and Tallevast, T. L. (1992) Abundance and growth patterns of Chelonia mydas in Culebra, Puerto Rico. Journal of Herpetology, 26 (3): 293-300. Crouse, D., Crowder, L. B. and Caswell, H. (1987) A stage based population model for loggerhead sea turtles and implications for conservation. Ecology, 68: 1412-1423. Day, T. and Taylor, P. D. (1997) Von Bertalanffy's growth equation should not be used to model age and size at maturity. American Naturalist, 149: 381-393. Den Hartog, J. C. (1980) The food of sea turtles Eretmochelys imbricata and Dermochelys coriacea. Netherlands Journal of Zoology, 30 (4): 595-610. Diez, C. E. and van Dam, R. P. (2002) Habitat effect on hawksbill turtle growth rates on feeding grounds at Mona and Monito Islands, Puerto Rico. Marine Ecology Progress Series, 234: 301-309. Diez, C. E. and Dam, R. P. v. (2003) Sex ratio of an immature hawksbill sea turtle aggregation at Mona Island, Puerto Rico. Journal of Herpetology, 37 (3): 533-537. Dodd, C. K., Jr. (1988) Synopsis of the biological data on the loggerhead sea turtle Caretta caretta (Linnaeus 1758). U.S. Fish & Wildlife Service Washington, D.C., USA. 110 pp. Dunbar, S., Salinas, L. and Stevenson, L. (2008) In-water observations of recently released juvenile hawksbills (Eretmochelys imbricata). Marine Turtle Newsletter, 121: 5-9. 75  Eckert, K. L. (1995) Anthropogenic threats to sea turtles. In Biology and Conservation of Sea Turtles, revised edition. (ed K. Bjorndal), pp. 611-612. Smithsonian Institution Press, Washington D.C., USA. Eckert, S. A. and Martins, H. R. (1989) Transatlantic travel by juvenile loggerhead turtle. Marine Turtle Newsletter, 45: 15. Eggers, J. M., Haberland, M. W. and Griffin, J. C. (1992) Growth of juvenile loggerhead sea turtles near PSE&G's Salem Generating Station, Delaware Bay, New Jersey. Marine Turtle Newsletter, 59: 5-7. Frazer, N. B. (1987) Preliminary estimates of survivorship for wild juvenile loggerhead sea turtles (Caretta caretta). Journal of Herpetology, 21 (3): 232-235. Frazer, N. B. and Ehrhart, L. M. (1985) Preliminary growth models for green, Chelonia mydas, and loggerhead, Caretta caretta, turtles in the wild. Copeia, 1985 (1): 73-79. Frazer, N. B. and Ladner, R. C. (1986) A growth curve for green sea turtles, Chelonia mydas, in the U.S. Virgin Islands, 1913-14. Copeia, 1986 (3): 798-802. Frazer, N. B., Gibbons, J. W. and Greene, J. L. (1990) Exploring Fabens' growth interval model with data on a long-lived vertebrate, Trachemys scripta (Reptilia: Testudinata). Copeia, 1990 (1): 112-118. Frick, M. G., Williams, K. L., Bolten, A. B., Bjorndal, K. A. and Martins, H. R. (2009) Foraging ecology of oceanic-stage loggerhead turtles Caretta caretta. Endangered Species Research, 9 (2): 91- 97. Gardner, T. A., Cote, I. M., Gill, J. A., Grant, A. and Watkinson, A. R. (2003) Long-term region- wide declines in Caribbean corals. Science, 301 (5635): 958-960. Garduño-Andrade, M., Guzman, V., Miranda, E., Briseno-Duenas, R. and Abreu-Grobois, F. A. (1999) Increases in hawksbill turtle (Eretmochelys imbricata) nestings in the Yucatan Peninsula, Mexico, 1977-1996: Data in support of successful conservation? Chelonian Conservation and Biology, 3 (2): 286-295. Geis, A., Wibbels, T., Phillips, B., Hillis-Starr, Z., Meylan, A., Meylan, P., Diez, C. and Dam, R. V. (2003) Predicted sex ratio of juvenile hawksbill sea turtles inhabiting Buck Island Reef National Monument, U.S. Virgin Islands. Journal of Herpetology, 37 (2): 400-404. Gilbert, E. (2005) Juvenile green turtle (Chelonia mydas) foraging ecology: Feeding selectivity and forage nutrient analysis. Masters dissertation.University of Central Florida, Orlando (FL), USA. 76  Godínez-Domínguez E, R. E., Carretero-Montes, F. d. A., Silva-Bátiz, S. R. and Aguilar, B. (1993) Crecimiento de neonatos de Chelonia agassizii (Testudines: Chelonidae) en cautiverio. Revista De Biologia Tropical, 41 (2): 253-260. Gutiérrez Montero, W. and Cabrera Peña, J. (1995) Crecimiento, conversión de alimento y mortalidad de Eretmochelys imbricata (Reptilia: Chelonidae) en estanques artificiales en Costa Rica. Revista De Biologia Tropical, 44 (2B): 847-851. Heppell, S. S., Limpus, C. J., Crouse, D. T., Frazer, N. B. and Crowder, L. B. (1996) Population model analysis for the loggerhead sea turtle, Caretta caretta, in Queensland. Wildlife Research, 23 (2): 143-159. Heppell, S. S., Snover, M. L. and Crowder, L. B. (2003a) Sea turtle population ecology. In The Biology of Sea Turtles, Volume II. (eds P. L. Lutz, J. A. Musick and J. Wyneken), pp. 275-306. CRC Press, Boca Raton (FL), USA. Heppell, S. S., Crowder, L. B., Crouse, D. T., Epperly, S. P. and Frazer, N. B. (2003b) Population models for Atlantic loggerheads: Past, present and future. In Loggerhead Sea Turtles. (ed A. B. Bolten), pp. 255-273. Smithsonian Books, Washington, D.C., USA. Hill, M. S. (1998) Spongivory on Caribbean reefs releases corals from competition with sponges. Oecologia, 117 (1-2): 143-150. Hirth, H. F. (1997) Synopsis of the biological data on the green turtle Chelonia mydas (Linnaeus 1758). Fish and Wildlife Service, Washington D.C., USA. 120 pp. Hughes, A. R., Williams, S. L., Duarte, C. M., Heck Jr, K. L. and Waycott, M. (2009) Associations of concern: Declining seagrasses and threatened dependent species. Frontiers in Ecology and the Environment, 7 (5): 242-246. IUCN (2009) IUCN Red List of Threatened Species. Version 2009.2. Accessed January, 2010. http://www.iucnredlist.org Ivlev, V. S. (1945) The biological productivity of waters. Translated by W.E. Ricker, 1966. Journal of Fishery Research Board of Canada, 23 (11): 1727-1759. Jackson, J. B. C. (1997) Reefs since Columbus. Coral Reefs, 16: S23-S32. Jackson, J. B. C. (2001) What was natural in the coastal oceans? Proceedings of the National Academy of Science, 98 (10): 5411-5418. Jones, T. T. (2009) Energetics of the leatherback sea turtle, Dermochelys coriacea. Doctoral thesis. University of British Columbia, Vancouver (B.C.), Canada. Kamezaki, N. (2003) What is a loggerhead turtle? The morphological perspective. In Loggerhead Sea Turtles. (eds A. B. Bolten and B. Witherington), pp. 28-43. Smithsonian Books, Washington D.C., USA. 77  Kaufmann, R. (1975) Observaciones sobre el crecimiento de tortugas marinas en cautividad. Caldasia, 11 (53): 139-150. Kitchell, J. F. and Windell, J. T. (1972) Energy budget for the lizard, Anolis carolinensis. Physiological Zoology, 45 (2): 178-188. Klinger, R. C. and Musick, J. A. (1995) Age and growth of loggerhead turtles (Caretta caretta) from Chesapeake Bay. Copeia, 1995 (1): 204-209. Krueger, B. (unpublished) Morphometric data for foraging hawksbill sea turtles in Barbados. Kubis, S., Chaloupka, M., Ehrhart, L. M. and Bresette, M. (2009) Growth rates of juvenile green turtles Chelonia mydas from three ecologically distinct foraging habitats along the east central coast of Florida, USA. Marine Ecology Progress Series, 389: 257-269. León, Y. M. and Bjorndal, K. A. (2002) Selective feeding in the hawksbill turtle, an important predator in coral reef ecosystems. Marine Ecology Progress Series, 245: 249-258. León, Y. M. and Diez, C. E. (1999) Population structure of hawksbill turtles on a foraging ground in the Dominican Republic. Chelonian Conservation and Biology, 3 (2): 230-236. Lewison, R. L., Freeman, S. A. and Crowder, L. B. (2004) Quantifying the effects of fisheries on threatened species: The impact of pelagic longlines on loggerhead and leatherback sea turtles. Ecology Letters, 7: 221-231. Limpus, C. J. (2009) A biological review of australian marine turtle species. 3. Hawksbill turtle, Eretmochelys imbricata (Linnaeus). Queensland Environmental Protection Agency, Brisbane (QLD), Australia. 54 pp. Limpus, C. and Chaloupka, M. Y. (1997) Nonparametric regression modelling of green sea turtle growth rates (southern Great Barrier Reef). Marine Ecology Progress Series, 149: 23-34. Lutcavage, M. and Musick, J. A. (1985) Aspects of the biology of sea turtles in Virginia. Copeia, 1985 (2): 449-456. Lutcavage, M. E., Plotkin, P. T., Witherington, B. and Lutz, P. L. (1997) Human impacts on sea turtle survival. In The Biology of Sea Turtles. (eds P. L. Lutz and J. A. Musick), pp. 387-410. CRC Press, Boca Raton (FL), USA. Manolis, C., Moncada, F., Webb, G. J. W., Nodarse, G., Escobar, E. and Morales, E. (2006) Management of hawksbill turtles in Cuba: Lessons learned. In Twenty third annual sea turtle symposium (ed N. J. Pilcher), pp. 61-64. NOAA Technical Memorandum NMFS-SEFSC-536, Kuala Lumpur, Malaysia. McClenachan, L., Jackson, J. B. C. and Newman, M. J. H. (2006) Conservation implications of historic sea turtle nesting beach loss. Frontiers in Ecology and the Environment, 4 (6): 290-296. 78  Mendonca, M. T. (1981) Comparative growth rates of wild immature Chelonia mydas and Caretta caretta in Florida. Journal of Herpetology 15 (4): 447-451. Meylan, A. (1988) Spongivory in hawksbill turtles - a diet of glass. Science, 239 (4838): 393-395. Meylan, A. (1990) Nutritional characteristics of sponges in the diet of the hawksbill turtle Eretmochelys imbricata. In New Perspectives in Sponge Biology: Third International Conference on the Biology of Sponges (ed K. Ruetzler ), pp. 472-477. Smithsonian Institution Press, Washington D.C., USA. Meylan, A. B. (1999) Status of the hawksbill turtle (Eretmochelys imbricata) in the Caribbean region. Chelonian Conservation and Biology, 3 (2): 177-184. Meylan, A. B. and Donnelly, M. (1999) Status justification for listing the hawksbill turtle (Eretmochelys imbricata) as Critically Endangered on the 1996 IUCN Red List of Threatened Animals. Chelonian Conservation and Biology, 3 (2): 200-224. Miller, J. D. (1997) Reproduction in sea turtles. In The Biology of Sea Turtles. (eds P. L. Lutz and J. A. Musick), pp. 51-81. CRC Press, New York (NY), USA. Moncada Gavilán, F., Carrillo, E., Saenz, A. and Nodarse Andreu, G. (1999) Reproduction and nesting of the hawksbill turtle, Eretmochelys imbricata in the Cuban archipelago. Chelonian Conservation and Biology, 3 (2): 257-263. Moran, K. L. and Bjorndal, K. A. (2005) Simulated green turtle grazing affects structure and productivity of seagrass pastures. Marine Ecology Progress Series, 305: 235-247. Moran, K. L. and Bjorndal, K. A. (2007) Simulated green turtle grazing affects nutrient composition of the seagrass Thalassia testudinum. Marine Biology, 150 (6): 1083-1092. Mortimer, J. A., Donnelly, M. and Plotkin, P. (2000) Sea turtles. In Seas at the Millennium: An Environmental Evaluation. (ed C. R. C. Sheppard), pp. 59-71. Elsevier Science Ltd., The Netherlands. Musick, J. A. and Limpus, C. J. (1997) Habitat utilisation and migration in juvenile sea turtles. In The Biology of Sea Turtles. (eds P. L. Lutz and J. A. Musick), pp. 137-164. CRC Press, Boca Raton (FL), USA. Nagy, K. (1989) Field bioenergetics: accuracy of models and methods. Physiological Zoology, 62 (2): 237-252. NMFS (2001) Stock assessments of loggerhead and leatherback sea turtles and an assessment of the impact of the pelagic longline fishery on the loggerhead and leatherback sea turtles of the Western North Atlantic. U.S. Department of Commerce NOAA Technical Memorandum NMFSSEFSC-455, 343 pp. 79  NMFS and US FWS (2009) Recovery plan for the Northwest Atlantic population of the loggerhead sea turtle (Caretta caretta). Second revision. National Marine Fisheries Service, Silver Spring (MD), USA. 325 pp. Palomares, M. L. and Pauly, D. (1998) Predicting food consumption of fish populations as functions of mortality, food type, morphometrics, temperature and salinity. Marine and Freshwater Research, 49: 447-453. Pandolfi, J. M., Bradbury, R. H., Sala, E., Hughes, T. P., Bjorndal, K. A., Cooke, R. G., McArdle, D., McClenachan, L., Newman, M. J. H., Paredes, G., Warner, R. R. and Jackson, J. B. C. (2003) Global trajectories of the long-term decline of coral reef ecosystems. Science, 301 (5635): 955-958. Parham, J. F. and Zug, G. R. (1997) Age and growth of loggerhead sea turtles (Caretta caretta) of coastal Georgia: An assessment of skeletochronological age-estimates. Bulletin of Marine Science, 61 (2): 287-304. Patriquin, D. (1973) Estimation of growth rate, production and age of the marine angiosperm Thalassia testudinum König. Caribbean Journal of Science, 13: 111-123. Pauly, D. (1986) A simple method for estimating the food consumption of fish populations from growth data and food conversion experiments. Fishery Bulletin, 84 (4): 827-840. Pemberton, R. A., Coyne, M., Musick, J. A., Phillips, B. and Hillis-Starr, Z. (2000) Habitat utilization of hawksbill sea turtles at Buck Island Reef National Monument: The zoanthid question. In 20th Annual Symposium on Sea Turtle Biology and Conservation (eds A. Mosier, A. Foley and B. Brost), pp. 69-70. NOAA Technical Memorandum NMFS-SEFSC-477. Pritchard, P. C. H. (1969) Sea turtles of the Guianas. Bulletin of the Florida State Museum. Biological Sciences, 13 (2): 86-140. Pritchard, P. C. H., Bacon, P., Berry, F., Carr, A., Fletemeyer, J., Gallagher, R., Hopkins, S., Lankford, R., Márquez, M. R., Ogren, L., Pringle, W., Reichart, H. and Witham, R. (1983) Manual of sea turtle research and conservation techniques. Centre for Environmental Education, Washington D.C., USA. 126 pp. Reich, K., Bjorndal, K. and Bolten, A. (2007) The ‘lost years’ of green turtles: Using stable isotopes to study cryptic lifestages. Biology Letters, 3 (6): 712-714. Richardson, J. I., Hall, D. B., Mason, P. A., Andrews, K. M., Bjorkland, R., Cai, Y. and Bell, R. (2006) Eighteen years of saturation tagging data reveal a significant increase in nesting hawksbill sea turtles (Eretmochelys imbricata) on Long Island, Antigua. Animal Conservation, 9 (3): 302-307. 80  Rützler, K. (1978) Sponges in coral reefs. In Coral Reefs: Research Methods. (eds D. R. Stoddart and R. E. Johannes), pp. 299-313. Monographs on oceanographic methodology 5, UNESCO, Paris, France. Safran, P. (1992) Theoretical analysis of the weight-length relationship in fish juveniles. Marine Biology, 112 (4): 545-551. Seminoff, J. A., Resendiz, A., Nichols, W. J. and Jones, T. T. (2002) Growth rates of wild green turtles (Chelonia mydas) at a temperate foraging area in the Gulf of California, Mexico. Copeia, 3: 610-617. Seney, E. E. and Musick, J. A. (2007) Historical diet analysis of loggerhead sea turtles (Caretta caretta) in Virginia. Copeia, 2007 (2): 478-489. Shaver, D. J. (1994) Relative abundance, temporal patterns, and growth of sea turtles at the Mansfield Channel, Texas. Journal of Herpetology, 28 (4): 491-497. Sherwood, L. (2005) Fundamentals of Physiology: A Human Perspective, Thomson Brooks/Cole, Belmont (CA), USA. 736 pp. Shoop, C. R., Ruckdeschel, C. A. and Kenney, R. D. (1998) Female-biased sex ratio of juvenile loggerhead sea turtles in Georgia. Chelonian Conservation and Biology, 3 (1): 93-96. Sidwell, V. D. (1981) Chemical and nutritional composition of finfishes, whales, crustaceans, mollusks, and their products. National Marine Fisheries Service, Charleston (SC), USA. NOAA Technical Memorandum-NMFS-F/SEC-11. pp. Snover, M. L., Avens, L. and Hohn, A. A. (2007) Back-calculating length from skeletal growth marks in loggerhead sea turtles Caretta caretta. Endangered Species Research 3(1): 95-104. Speakman, J. (1997) Doubly Labelled Water: Theory and Practice, Chapman & Hall, London, UK. 399 pp. Stickney, R. R., White, D. B. and Perlmutter, D. (1973) Growth of green and loggerhead sea turtles in Georgia on natural and artificial diets. Bulletin of the Georgia Academy of Science, 31: 37-44. Suchanek, T. H., Carpenter, R. H., Witman, J. D. and Harvell, C. D. (1983) Sponges as important space competitors in deep Caribbean coral reef communities. In The Ecology of Deep and Shallow Coral Reefs. (ed M. L. Reaka), pp. 55-60. Symposia Series for Undersea Research, NOAA, Washington, D.C., USA. Swingle, W. M., Warmolts, D. I., Keinath, J. A. and Musick, J. A. (1993) Exceptional growth rates of captive loggerhead sea turtles, Caretta caretta. Zoo Biology, 12 (5): 491-497. Teas, W. G. (1993) Species composition and size class distribution of marine turtle strandings on the Gulf of Mexico and southeast United States coasts, 1985-1991. U.S. Department of Commerce, NOAA Technical Memorandum NMFS-SEFSC-315, 43 pp. 81  Thayer, G. W., Bjorndal, K. A., Ogden, J. C., Williams, S. L. and Zieman, J. C. (1984) Role of larger herbivores in seagrass communities. Estuaries, 7 (4A): 351-376. Trautman, D. A., Hinde, R. and Borowitzka, M. A. (2000) Population dynamics of an association between a coral reef sponge and a red macroalga. Journal of Experimental Marine Biology and Ecology, 244 (1): 87-105. van Buskirk, J. and Crowder, L. B. (1994) Life-history variation in marine turtles. Copeia, 1994 (1): 66-81. van Dam, R. P. (1997) Ecology of hawksbill turtles on feeding grounds at Mona and Monito Islands, Puerto Rico. Doctoral thesis. University of Amsterdam, The Netherlands. van Dam, R. P. and Diez, C. F. (1996) Diving behavior of immature hawksbills (Eretmochelys imbricata) in a Caribbean cliff-wall habitat. Marine Biology, 127 (1): 171-178. van Dam, R. and Diez, C. E. (1997) Predation by hawksbill turtles on sponges at Mona Island, Puerto Rico. In Proceedings of 8th International Coral Reef Symposium pp. 1421-1426. van Dam, R. P. and Diez, C. E. (1998) Caribbean hawksbill turtle morphometrics. Bulletin of Marine Science, 62 (1): 145-155. Vicente, V. P. (1978) An ecological evaluation of the West Indian demosponge Anthosigmella varians (Hadromerida: Spirastrellidae). Bulletin of Marine Science, 28: 771-779. von Bertalanffy, L. (1938) A quantitative theory of organic growth. Human Biology, 10: 181-213. von Bertalanffy, L. (1960) Principles and theory of growth. In Fundamental Aspects of Normal and Malignant Growth. (ed W. W. Wowinski), pp. 137-259. Elsevier, Amsterdam, The Netherlands. Wabnitz, C. C. C., Andréfouët, S., Torres-Pulliza, D., Müller-Karger, F. E. and Kramer, P. A. (2008) Regional-scale seagrass habitat mapping in the wider Caribbean region using Landsat sensors: Applications to conservation and ecology. Remote Sensing of Environment, 112: 3455- 3467. Wallace, B. P. and Jones, T. T. (2008) What makes marine turtles go: A review of metabolic rates and their consequences. Journal of Experimental Marine Biology and Ecology, 356 (1-2): 8-24. Waycott, M., Duarte, C. M., Carruthers, T. J. B., Orth, R. J., Dennison, W. C., Olyarnik, S., Calladine, A., Fourqurean, J. W., Heck, K. L., Hughes, A. R., Kendrick, G. A., Kenworthy, W. J., Short, F. T. and Williams, S. L. (2009) Accelerating loss of seagrasses across the globe threatens coastal ecosystems. Proceedings of the National Academy of Sciences of the United States of America, 106 (30): 12377-12381. 82  Wibbels, T., Martin, R. E., Owens, D. W. and Amoss, M. S. (1991) Female-biased sex-ratio of immature loggerhead sea-turtles inhabiting the Atlantic coastal waters of Florida. Canadian Journal of Zoology, 69 (12): 2973-2977. Williams, S. (1988) Thalassia testudinum productivity and grazing by green turtles in a highly disturbed seagrass bed. Marine Biology, 98: 447-455. Witherington and Witherington, B. (2002) Ecology of neonate loggerhead turtles inhabiting lines of downwelling near a Gulf Stream front. Marine Biology, 140 (4): 843-853. Witzell, W. N. (1983) Synopsis of biological data on the hawksbill turtle Eretmochelys imbricata (Linaeus, 1766). FAO Fisheries Synopsis No. 137, 78 pp. Zimmerman, L. C. and Tracy, C. R. (1989) Interactions between the environment and ectothermy and herbivory in reptiles. Physiological Zoology, 62 (2): 374-409. Zug, G. R. and Glor, R. E. (1998) Estimates of age and growth in a population of green sea turtles (Chelonia mydas) from the Indian River lagoon system, Florida: A skeletochronological analysis. Canadian Journal of Zoology, 76 (8): 1497-1506. Zug, G. R., Wynn, A. H. and Ruckdeschel, C. (1986) Age determination of loggerhead sea turtles, Caretta caretta, by incremental growth marks in the skeleton. Smithsonian contribution no. 427, 44 pp.  83  3.  Restoration of Chelonia mydas in the Caribbean: Ecosystem changes resulting from a reduction in seagrass habitat complexity 2  3.1 Introduction Sea turtles have suffered substantial reduction in numbers throughout the Caribbean (Parsons 1962, NMFS 1993, Seminoff 2004, Bräutigam and Eckert 2006, NMFS 2007a, 2007b). Overharvesting, in addition to threats resulting from disease, incidental capture by fishers, and destruction of critical nesting and foraging habitat, has resulted in a decline of over 90 % of many stocks compared with historic levels (Jackson et al. 2001). Over the last few decades, conservation and management programmes have been implemented throughout the region to help reverse these severe declines.          2 A version of this chapter will be submitted for publication. Wabnitz, C.C.C., Coll, M., Bjorndal, K., Bolten, A., Christensen., V. and Pauly, D. Restoration of Chelonia mydas in the Caribbean: Ecosystem changes resulting from a reduction in seagrass habitat complexity. 84  The green turtle (Chelonia mydas) nesting populations at Tortuguero (Costa Rica) (Bjorndal et al. 1999) and Florida (USA) (Chaloupka et al. 2008), for example, have responded to protection efforts and are now demonstrating increases in population size.  In the Caribbean, green turtles feed mainly on turtle grass Thalassia testudinum (Bjorndal 1980) and together with manatees are considered the largest seagrass herbivore in the region. Green turtles remove a large proportion of seagrass standing stock by cropping blades 2 - 4 cm above the base without disturbing the underground rhizome system (Lanyon et al. 1989, Brand-Gardner et al. 1999). Individuals commonly show a high degree of fidelity to feeding sites by repeatedly returning to the same grazing plots (Bjorndal 1980, Ogden 1980). At comparatively high turtle densities, these individual grazing plots may merge into one single-cropped seagrass bed (Williams 1988). Green turtles have had the strongest ecological and evolutionary impacts on seagrass since the extinction of the diverse dugongid fauna before the Pleistocene (Domning 2001). Before human exploitation, green turtles may have been as important in determining seagrass productivity as terrestrial herbivores were in grasslands (McNaughton 1979, Pandolfi et al. 2003). Hence, it seems valid to ask what ecosystem-wide changes are likely to be observed as a result of the relatively consistent population increases exhibited by Caribbean green turtles. In the context of ecosystem- based management, this means that actions taken to help green turtle populations recover need to consider aspects of this species’ population biology, as well as its interactions with other species in the ecosystem, at all stages of its life history. Therefore, seagrass, which constitutes the main habitat at foraging grounds for green turtles in the Caribbean, need to be explicitly considered in any species management and recovery plan.  Seagrass beds can form dense meadows in the coastal zone. They can typically be found in close proximity to coral reefs and/or mangroves, and represent important ‘nursery’ (Beck et al. 2001, Adams et al. 2006) and foraging grounds for a variety of organisms from reef habitats (Hemminga 85  and Duarte 2000). While many herbivorous fish shelter on reefs at night, they usually are found foraging in adjacent seagrass habitats during the day (see review by Valentine and Duffy (2006)). Conversely, many reef-associated carnivores (e.g., grunts (Haemulidae) and snappers (Lutjanidae)) will venture to feed in vegetated habitats at night (Ogden 1980, Baelde 1990, Burke 1995, Cox et al. 1997, Kopp et al. 2007, Luo et al. 2009), and many seagrass-associated fish and invertebrates figure prominently in the diets of reef predators (Randall 1967, Heck and Weinstein 1989). As essential ‘nursery’ areas for the juvenile stages of many species, often of commercial importance (Nagelkerken et al. 2000a, Beck et al. 2003, Heck et al. 2003, Mumby et al. 2004, Dahlgren et al. 2006), seagrass directly or indirectly support fisheries (Gillanders 2006). This nursery function is understood to be the provision of a habitat where juveniles have greater survival and growth rates than on nearby unvegetated substrate (Adams et al. 2006). It is assumed that such benefits are primarily the result of high food abundance (Cocheret de la Moriniere et al. 2003a, Cocheret de la Moriniere et al. 2003b), low predation pressure resulting from the high structural complexity of the habitat at fine scales (Parrish 1989, Heck and Orth 2006, Verweij et al. 2006), and increased space availability for settlement recruits (Shulman and Ogden 1987). All of these observations strongly point towards a reliance of reef fish on seagrass-based production (Nagelkerken et al. 2000a, Nagelkerken et al. 2001).  As major grazers of seagrass, it is likely that a recovery of green turtle populations will have a significant impact on the present structure of seagrass beds. Specifically, a shift from today’s ungrazed pastures, characterised by long fronds, to grazed pastures with very short blades can be expected. Changes are not only expected to be structural. Sustained grazing by green turtles has already been shown to lead to compensatory growth by seagrass (Moran and Bjorndal 2005), and increases in the nutritional quality of the blades (Moran and Bjorndal 2007). However, our knowledge of the community-wide changes that sustained high-level green turtle grazing would have on seagrass community dynamics are less clear.  86  In this study I sought to assess the theoretical impacts that an increasing green turtle biomass would have on a typical Caribbean seagrass/reef/mangrove ecosystem, as commonly found in the coastal landscape of the U.S. Virgin Islands (USVI) and Puerto Rico (Figure 3.1). This was done through the use of a model implemented in Ecopath with Ecosim (EwE) (Christensen and Walters 2004). Our model realisation considered the three ecosystems together as important exchanges typically occur between them (Nagelkerken et al. 2000a, Mumby et al. 2004).  Available evidence seems to indicate that decreasing habitat complexity will provide decreasing refuge to prey species by interfering with predator success (e.g., Gotceitas and Colgan 1989, Lipcius et al. 1998, Heck and Orth 2006, Horinouchi 2007). Repeated and increased grazing by recovering green turtle populations will lead to a dramatic reduction in the canopy height of seagrass. This loss of structural complexity (i.e., loss of refuge) is likely to be associated with significant changes in direct and indirect trophic interactions within coastal ecosystems. Therefore, under the overall scenario of increased turtle abundance, I contrasted model outputs resulting from the implementation of three alternative functions describing changes to prey vulnerability with an increasing loss of habitat complexity: 1. A ‘step’ function - Based on experimental studies with Zostera marina, amphipods as prey, and pinfish and grass shrimp as predators, Nelson (1979) found that predation rate may not be a simple linear function of seagrass density. Instead, a threshold seagrass density might be required before any significant reduction in predator foraging efficiency is observed. In other words, predator effectiveness increases with decreasing seagrass complexity according to a step function (see Figure 3.2, dotted line); 2. A ‘dome’ function - Results from experiments conducted with bluegills and a variety of invertebrates led Crowder and Cooper (1982) to propose that a predator’s foraging efficiency is maximized at intermediate levels of habitat complexity, and low at both high and low macrophyte biomass (see Figure 3.2, dashed line); 87  3. A flat line - Contrary to previous experiments that were conducted with constant numbers of prey and predators at differing densities of seagrass, Mattila et al. (2008) and Canion and Heck (2009) increased predators and prey with increasing seagrass density, thereby mimicking what is observed in nature. Results from outdoor tank and field (mesocosms) experiments showed that seagrass provides significantly more shelter than unvegetated substrate, but that no significant differences occur in predator success among seagrass densities (Mattila et al. 2008, Canion and Heck 2009). In other words, similar refuge levels can be expected from grazed and ungrazed beds (Figure 3.2, solid line).  The model presented here was derived from species-level data. However, in discussing outputs from simulation runs for each of the three scenarios, I sought to highlight overarching tendencies and patterns rather than arrive at specific numerical conclusions of declines or increases in given species groups. Key data gaps, brought to light by my analysis, are also outlined. I recommend that these be considered as critical research priorities to add a quantitative basis to our understanding of the role of green turtles in reducing habitat complexity in seagrass meadows.  3.2 Materials and methods 3.2.1  Study area The model developed for this study represents an average annual mid-1990s situation in the coastal waters, from the water’s edge to a depth of approximately 50 - 75 m, around the U.S. Virgin Islands (USVI) and Puerto Rico (Figure 3.1). This area is meant to be representative of a Caribbean region coastal system. The area below 75 m was not included as the model focused on system dynamics in seagrass beds/coral reefs. Analyses of the products of a recent mapping effort reveal that Puerto Rico harbours a total 1,599 km2 of coastal habitats (3 % of which are unconsolidated sediment, 45 % 88  submerged vegetation, 5 % mangroves, and 47 % coral reef and colonized hardbottom), while the USVI coastal habitats cover 487 km2 (5 % of which are unconsolidated sediment, 33 % submerged vegetation, and 62 % coral reef and hard bottom) (Kendall et al. 2001).  3.2.2  Modelling approach: Ecopath with Ecosim (EwE) The foundation of the Ecopath with Ecosim (EwE) suite is a trophic mass balance model (Polovina 1984, Christensen and Pauly 1992) that creates a static snapshot of the resources in an ecosystem and their interactions, represented by trophically linked biomass ‘pools’ (Christensen and Pauly 1992). The biomass pools, hereafter referred to as functional groups, consist of a single species, single size/age group of a given species, or species groups representing ecological guilds (sharing similar trophic behaviour, habitats, and other ecological traits). These may be further split into ontogenetic (juvenile/adult) groups that can then be linked in Ecosim (see stanzas below).  Ecopath operates under two main assumptions. The first is that biological production (P) of each functional group (i) in the ecosystem equals the sum of mortality due to fisheries (Yi) and predators (j) (Bj·M2ij), net migration (Ei), biomass accumulation (BAi), and other unexplained mortality (Pi·(1- EEi)) where EE is the ecotrophic efficiency: ∑ −++++= j iiiiiijji EE1PBAEY2MBP )·(·                                                                                   (1) Equation (1) can be re-expressed as: ( )i i ijiiij jj j i EE1 B PBBAEYDC B QB B PB −⋅⎟⎠ ⎞⎜⎝ ⎛⋅++++⋅⎟⎠ ⎞⎜⎝ ⎛⋅=⎟⎠ ⎞⎜⎝ ⎛⋅ ∑                                                         (2) where (P/B)i indicates the production of (i) per unit of biomass and is equivalent to total mortality, or Z, under steady-state conditions (Allen 1971); (Q/B)i is the consumption of (i) per unit of 89  biomass; and DCij indicates the proportion of (i) that is in the diet of predator (j) in terms of volume or weight units. EwE parameterises the model by describing a system of linear equations for all the functional groups of the model, where for each equation three of the basic parameters: Bi, (P/B)i, (Q/B)i, or EEi have to be known for each group (i). The principle behind this modelling approach is that, on an annual basis, biomass and energy in an ecosystem and each of its components are conserved (Walters et al. 1997, Walters and Martell 2004a). The second assumption is that consumption within a group equals the sum of production, respiration, and unassimilated foods. Flows are expressed in t·km-2·year-1 wet weight organic matter for flows, and biomasses in t·km-2. The Ecopath model was considered balanced when realistic estimates of the missing parameters were calculated (EE < 1) (Christensen et al. 2005).  Ecosim, the time-dynamic module of the software, was then used to investigate the impact that an increase in turtle biomass would have on the ecosystem. Using a system of time-dependent differential equations, biomass fluxes amongst functional groups are calculated as a function of time by accounting for changes in predation, consumption, and emigration rates, as well as fishing (Christensen and Pauly 2004). For each group, biomass growth rate is expressed as: ( ) iiiiiijji i i BeFMIQQ Q P dt dB ⋅++−+−⋅⎟⎟⎠ ⎞⎜⎜⎝ ⎛= ∑∑                                                           (3) where (P/Q)i is the gross efficiency; Mi is the natural non-predation mortality rate; Fi is the fishing mortality rate; ei is the emigration rate; Ii is the immigration rate; and Bi is the biomass of the functional group (i). Calculations of consumption rates Qij are based on “foraging arena” theory, where the biomass of prey (i) is divided into vulnerable and non-vulnerable pools to predation (Walters and Kitchell 2001, Walters and Martell 2004b). Low vulnerability rates to predators imply donor-driven (prey is limiting), density-dependent interactions. High rates, on the other hand, indicate predator-driven interactions where (a) the behaviour of both prey and predator have weaker 90  effects on limiting predation rates; (b) predation mortality is proportional to the product of prey and predator abundance; and (c) the predator’s initial biomass is low compared to its carrying capacity (Christensen et al. 2005). Vulnerability of seagrass to turtles was set high (i.e., turtles are far from carrying capacity). This appears to be a reasonable assumption, given that the current green turtle population in the Caribbean represents 3 - 7 % of pre-exploitation levels (Jackson et al. 2001) and that green turtles show specific physiological adaptations to a seagrass diet (Bjorndal 1980).  For species with complex life histories, individual life history stages can be incorporated into Ecosim through ‘stanzas’. This approach was used to model the juvenile and adult stages of snappers, grunts, parrotfish (Scaridae), and surgeonfish (Acanthuridae) (see below for rationale behind this choice of species). The stanzas are linked to each other and usually the adults are set as the leading group. (P/B)i and diet composition were provided for both juvenile and adult groups, while Bi and (Q/B)i were introduced for the leading stanza group only (Christensen et al. 2005).  For more detailed methods as well as model capabilities and limitations refer to Christensen et al. (2005), and Christensen and Walters (2004) and Coll et al. (2009) respectively.  3.2.3  Functional groups and model parameters Based in part on a species list drawn up by Delgado (2004), data were obtained for 282 species of fish known to occur in the area. These were then aggregated into 14 functional groups (for a total of 29 functional groups including non-fish groups and detritus) chiefly according to their feeding type and predators. An additional important consideration was the degree to which species are known to utilise seagrass as a nursery habitat during the juvenile portion of their life history, and/or whether they are often found foraging in seagrass beds (e.g., Randall 1963, Carr and Adams 1973, Kikuchi 91  and Peres 1977, Ogden and Zieman 1977, Robblee and Zieman 1984, Nagelkerken et al. 2000a, Nagelkerken et al. 2000b). For example, grunts, like many reef fish in the Caribbean exhibit ontogenetic shifts in habitat utilisation (Lindeman et al. 2000), and seagrass beds have been identified as particular important for their juvenile stages (Ogden and Zieman 1977, McFarland et al. 1985, Shulman 1985a, Adams and Ebersole 2002, Cocheret de la Moriniere et al. 2002). Adult grunts also often associate with coral reefs during the day, but migrate to forage primarily on benthic invertebrates in seagrass beds at night (Ogden 1980, Gaut and Munro 1983, Hammerschlag and Serafy 2009). Based on a review of the available literature, local variations notwithstanding, the most dominant species recorded as having juveniles using seagrass beds as refuge and adults foraging in seagrass included: snappers, grunts, surgeonfish, and parrotfish (Robblee and Zieman 1984, Nagelkerken et al. 2000a, Mateo and Tobias 2001, Layman and Silliman 2002, Chittaro et al. 2005, Aguilar-Perea and Appeldoorn 2007, Dorenbosch et al. 2007, Mateo and Tobias 2008). I therefore included multi-stanzas (i.e., separate juvenile and adult groups) for these trophic guilds. As these species also are known to substantially contribute to the local recreational and commercial fisheries (Hawkins and Roberts 2004), I maintained them as separate groups. Species that do not depend on seagrass beds during their juvenile stages, but nonetheless will regularly forage in this habitat were aggregated based on feeding preferences primarily (e.g., herbivores, piscivores, Mobile Invertebrate Feeders [MIF], and, for purposes of identification, they are listed separately from other species through the use of the prefix ‘SG’).  Species composition and biomass, and the extent to which individual habitats are likely to serve as ‘nurseries’, have been found to vary depending on the landscape configuration of these habitats (Baelde 1990, Bouchon-Navaro et al. 2004, Mateo and Tobias 2004, Mateo and Tobias 2007, Grober-Dunsmore et al. 2008, Nagelkerken et al. 2008). Therefore, in the context of model configuration, I assumed a landscape arrangement similar to that found in a number of coastal bays of Puerto Rico and the USVI, whereby shallow seagrass beds are located adjacent to reefs with 92  mangroves lining the shoreline. In other words, reefs would be located in relative close proximity to seagrass and mangrove habitats, and not separated by deep channels.  Data required to parameterize the model (B, P/B, Q/B, and diet) were informed by a model developed by Opitz (1996). Complementary data, specifically for fish groups, were obtained from the published literature for the USVI and Puerto Rico preferentially, and other regions of the Caribbean where necessary. Natural mortality (M) parameters were chiefly derived according to Pauly (1980). Catches and fishing mortality estimates were informed by data in Hawkins and Roberts (2004), Dunn et al. (2009), Rothenberger et al. (2008), and García-Sais et al. (2005). The food intake by a group during a certain period of time (here a year) divided by its biomass (Q/B), was estimated based on Palomares and Pauly (1989, 1998) for fish species, and adapted from Moran and Bjorndal (2005) and Bjorndal and Jackson (2003) for green turtles. For seagrass, biomass and P/B values were adjusted giving special attention to results in Gacia (1999) and Zieman et al. (1984), and Moran and Bjorndal (2005) and Bjorndal and Jackson (2003), respectively. Limited data were available for biomass estimates of ‘invertebrates’ on reef and in seagrass beds. Given that the majority of non-herbivorous reef fish consume and/or selectively feed on invertebrate prey (Randall 1967, Ogden 1980), I felt it reasonable to assume that predation mortality ‘used’ the largest proportion of invertebrate production in the system. Therefore, the biomass of all three invertebrate functional groups (Reef invertebrates, SG_small invertebrates, and SG_invertebrates) included in the model were based on demand and estimated by setting EE = 0.95. Once the model was balanced, the obtained biomasses were then transferred as input variables. B, P/B, and Q/B parameters for the remaining groups were informed from previously published models for the region including Opitz (1996), Manickchand-Heileman et al. (1998), Okey (2004), and Arias-González (1998). Diet information was compiled from the published literature (e.g., Randall 1967, Ogden 1976, Brook 1977, Ogden and Zieman 1977, Hay 1981, Tribble 1981, Zieman et al. 1984, Lewis and Wainwright 1985, Shulman 1985b, Sweatman and Robertson 1994, McGlathery 1995, McAfee and Morgan 93  1996, Kirsch et al. 2002, Layman and Silliman 2002, Cocheret de la Moriniere et al. 2003a, Cocheret de la Moriniere et al. 2003b) as well as FishBase (www.fishbase.org).  For all input data, parameters were set for each of the fish species in the model. Upon aggregation, each species contribution to the overall functional group value was weighted according to the individual species’ biomass contribution.  From the results of an Ecopath model, several indices can be generated that provide information about the overall structure and functioning of the ecosystem in the study area (Christensen 1995). Therefore, a few of the overall statistics of the system are presented (Christensen and Walters 2004).  3.2.4  Dynamic simulations I compared results from the implementation of three published models (see introduction for details), that describe the response of predator foraging efficiency to changes in habitat complexity. These models predict : (a) an increase in predator foraging efficiency with declining seagrass biomass according to a step function (Nelson 1979) (Figure 3.2, dotted line); (b) highest prey vulnerability to predators at intermediate levels of complexity (Crowder and Cooper 1982) (Figure 3.2, dashed line); and (c) no change in predator foraging efficiency with changing habitat complexity (Canion and Heck 2009) (Figure 3.2, solid line). Such non-trophic effects, whereby the biomass of a particular group affects the vulnerability of (i) prey to a given predator (j), can be captured in Ecosim through so-called ‘mediation’ functions. The three functions (see Figure 3.2) were then applied separately to all groups that benefit from the sheltering capacity of seagrass (i.e., our four juvenile fish groups and two seagrass invertebrate guilds). In applying mediation functions 1 and 2 in Ecosim, effects of changes in prey vulnerabilities are weighted so as to increase (or decrease) to all their predators as the biomass of seagrass declines due to the greater abundance of green turtles. Findings on coral 94  reefs in Hawai‘i (Hixon and Beets 1993) support the assumption that predators may non-selectively reduce all prey populations in proportion to their initial relative abundances (Hixon 1991).  In Ecosim, predator-prey interactions are each attributed a vulnerability term. This vulnerability parameter shows the maximum increase in predation mortality a given predator can cause on a given prey (see ‘Modelling approach: Ecopath with Ecosim’ section for details). Typically, models are calibrated in Ecosim using time series ‘reference’ data (e.g., relative and absolute biomass data of different species over a particular historical period), so that, where possible, model simulations match observed trends. This calibration is usually done by adjustment of the ‘vulnerability’ parameters (Walters et al. 1997, Walters et al. 2000). In the absence of time series data and other information that would justify adjusting vulnerabilities upwards or downwards, these were left at the default value; with the exception of the vulnerability term of seagrass to green turtles. The latter was set high based on evidence that green turtles have had the strongest ecological and evolutionary impacts on seagrass (Domning 2001) and their biomass is currently far from carrying capacity.  The starting seagrass biomass, with the lowest turtle grazing impact, was deemed representative of the maximum seagrass biomass attainable in the system (see grey block arrow, Figure 3.2). I initially ran the model to check start and end biomasses. This was done to ensure that the simulations spanned the range of seagrass biomass required to represent the full range of habitat complexity under the mediation 1 model.  Epiphyte biomass was assumed to decline steeply and linearly with declining seagrass biomass, as continuous grazing of seagrass blades by green turtles would not give epiphytic microalgae a chance to attach to seagrass fronds. This was implemented in Ecosim through another mediation function, whereby a decline in seagrass biomass was linked to a significant decline in the P/B of epiphytes.  95  The default value of maximum relative P/B for seagrass was raised from two to three, based on the comparison of mass-specific growth measurements recorded over the course of a 16-month seagrass clipping experiment (Moran and Bjorndal 2005).  To mimic an increase in turtle biomass two approaches were used: 1. Starting with an estimated current biomass of green turtles, it was then “forced” to increase at a constant rate through loading of increasing turtle biomass time series data in Ecosim; 2. I assumed a higher than observed starting biomass of green turtles, but with an intense fishery on the species that effectively kept starting biomass orders of magnitude below estimated carrying capacity levels (157 - 374.8 tonnes·km-2 (Moran and Bjorndal 2005)). Simulations were then run by removing the fishery and describing ecosystem changes as green turtle biomass increased. The advantage of scenario (2) over (1) is that an increase in green turtle biomass over time is calculated by Ecosim in light of entered input parameters for the group.  3.3 Results and discussion 3.3.1  Ecopath model of the Caribbean ecosystem Input and output parameters of the balanced model representing a Caribbean ecosystem are shown in Table 3.1. The diet composition matrix of the balanced model is presented in Table 3.2. Ecotrophic efficiencies (EE), respiration/assimilation ratios (R/A), production/respiration ratios (P/R), and net food conversion efficiencies were within the expected range for all functional groups (Christensen et al. 2005).  96  The total biomass (including detritus) encompassed within our model was approximately 994 g wet mass·m-2 (Table 3.3). Primary producers were the most important group, comprising 46 % of the total biomass, followed by invertebrates, which accounted for 32 % of total biomass. Fish only made up 5.5 % of total biomass, 82 % of which was represented by species that utilise the seagrass for forage and/or as refuge habitat. At current biomass levels, green turtles accounted for only a very small proportion of total biomass (< 1 %). Fish groups that do not venture much on to the seagrass were dominated by zooplanktivores and herbivores, while surgeonfish and snappers dominated the ‘seagrass’ (SG) trophic guilds.  Results show that a large component of the highest trophic levels’ production was being removed by fisheries rather than predators. This finding is not surprising given the history of exploitation of marine resources, particularly of higher level trophic groups, within the region (Wing and Wing 2001), and the conspicuous absence of large predatory fish. This in turn also explains the relatively low EE values obtained for some of the herbivorous groups (e.g., parrotfish and surgeonfish), and the generally large biomass registered by low trophic level groups relative to higher trophic level groups in the system. Mean trophic level of the catch was 3.2 and accounted only for 26.2 % of the primary production in the system (PPR %).  Results of the model further indicated that functional groups were mainly organised within four trophic levels (TL), with the highest TLs corresponding to the two piscivorous groups (TL 3.7 and 3.6, respectively), followed by snappers, grunts, and invertebrate feeders (Table 3.1). The lowest groups, by definition, were the primary producers and detritus groups (TL = 1).  97  3.3.2  Ecosystem wide changes as a result of an increase in green turtle biomass Mediation 1 Nelson (1979) proposed a step function (Figure 3.2, dotted line) to represent the inverse relationship he observed between prey vulnerability and seagrass biomass (analogous to complexity). Applying this function through Mediation 1, an increase in turtle biomass resulted in a dramatic decline in the biomass of seagrass, ‘SG_seagrass invertebrates’, seagrass MIF, and seagrass piscivores (Figure 3.3). Parrotfish were among the most affected groups, as a decline in seagrass biomass represented a decline in a significant source of forage, but also reduced the survivorship of juveniles that successfully recruit to adulthood. Adult snappers underwent dramatic declines as a result of (a) increased predation by seagrass piscivores that had lost a large component of their diet, including seagrass herbivores and juvenile fish (the latter group being negatively affected by a loss in seagrass); and (b) a decline in their own prey, mostly juvenile fish. As a result of the decline in predators, snapper juveniles first registered an increase in biomass, but with a further reduction in seagrass and concomitant increase in their own vulnerability, subsequently suffered a dramatic decline in biomass. The adults’ biomass decline due to lack of forage was therefore further exacerbated by only few juveniles making successful ontogenetic migrations to adulthood. However, a decline in snappers was associated with an increase in the biomass of grunts. Snappers were an important predator on grunt juveniles, due to their high initial biomass relative to other piscivorous groups. That release from predation, together with a relatively stable biomass of their main food source, small seagrass invertebrates, allowed for strong grunt recruitment to adulthood. The fact that adults themselves fed on small seagrass invertebrates, and were thus able to compensate for the decline in other diet items, further contributed to the observed overall increase in grunt biomass. Surgeonfish benefited from a similar increase in their biomass due to the release from predation by snappers and the fact that algae biomass, one of their main forage sources, increased, in part as a result of the decline in parrotfish. 98   Mediation 2 Experiments conducted with bluegills and a variety of invertebrates led Crowder and Cooper (1982) to propose a model whereby a predator’s foraging efficiency is maximised at intermediate levels of habitat complexity (Figure 3.2, dashed line). The effect of Mediation 2 under conditions of increasing turtle biomass was clearly divided into two phases. Until halfway through the simulations, corresponding to the peak in prey vulnerability, results were similar to those obtained under Mediation 1 (see above). The biomass of seagrass piscivores, particularly snappers, declined following the trend of a number of its diet components, including juvenile fish and parrotfish. The decline in seagrass piscivores and snapper biomass released grunts from predation, enabling their biomass to increase. Seagrass invertebrates and other groups to which the mediation function was applied to, also declined, a trend paralleled by the key predators of these groups. However, as turtle grazing reduced seagrass biomass beyond intermediate habitat complexity levels, prey vulnerability began to decrease. In response, surgeonfish juvenile (and later adult) and seagrass invertebrate biomass began to increase, a trend that was mirrored, with a slight delay, by their predators such as seagrass invertebrate feeders (Figure 3.4). Juvenile fish and parrotfish were similarly affected. As a consequence, seagrass piscivores began to recover, introducing a downward trend in grunt biomass from their peak abundance at intermediate levels of habitat complexity.  Mediation 3 The third mediation function was based on findings by Mattila et al. (2008) and Canion and Heck (2009) (Figure 3.2, solid line) that showed prey vulnerability to be independent of seagrass biomass. Increasing turtle biomass, while implementing Mediation 3, was therefore expected to primarily have an impact on species that are heavily dependent on seagrass for forage. This was indeed 99  matched by model simulations. Mediation 3 resulted in the decline of parrotfish, the only trophic guild composed of species (especially Sparisoma radians) with a large seagrass diet component (Figure 3.5). I assumed for both juveniles and adults to consume large quantities of seagrass and more minor quantities of algae, thus the trajectory of both groups mirrored that of declining seagrass quite closely.  Overall, our findings demonstrate that if allowed to recover to the range of carrying capacity biomasses that have been estimated from field-based seagrass productivity and green turtle food consumption studies (157 - 374.8 tonnes·km-2) (Moran and Bjorndal 2005), reef ecosystems would look different to what we see today. Results for all three mediations were observed whether turtle biomass was increased by ‘releasing’ turtles from predation, or whether biomass was ‘forced’ to increase over time (results not shown). The only difference was that vulnerabilities of seagrass to turtles had to be set much higher for the latter scenario. In the case of the “biomass forcing scenario”, green turtle biomass at the start of simulations was much lower, and thus much further removed from carrying capacity. Given the underlying Ecosim equations, the increase in turtle biomass over time is accompanied by a concomitant decline in Q/B. This results in simulations having no effect (i.e., no seagrass biomass decline with increasing turtle biomass), and is obviously aberrant. To compensate for this, the vulnerability of seagrass to turtles had to be set very high. This finding confirms that at current abundance levels, green turtles are functionally extinct. Once vulnerabilities were adjusted, however, the results were comparable to those obtained when initialising the model with a higher turtle biomass, subsequently removing the turtle fishery, and allowing green turtle populations to recover. As highlighted previously, the benefit to this approach is that it allows Ecosim to compute the rate at which turtle biomass should be recovering based on the parameters entered in the model (rather than forcing an incremental yearly biomass increase through time series data).  100  As emphasised in the introduction, model outputs are meant to highlight overarching patterns for each of the three scenarios, rather than represent specific numerical conclusions of declines or increases in given species groups. Of note however, is that the magnitude of predicted biomass trajectories in Ecosim is sensitive to the vulnerability parameter value. Future efforts should therefore target collection of data and design of experiments that can help set more realistic vulnerability parameters, and thus help lend more confidence to the magnitude of the biomass declines/increases. It is likely for example that functional groups seen to ‘collapse’ should have lower vulnerability parameters.  3.3.3  Insights offered by the modelling exercise Diet composition - behavioural and dietary adaptive flexibility Our findings indicate that based on currently available evidence and understanding of reef life cycle ecology, increased turtle grazing will result in direct and indirect changes in seagrass ecosystems. It is important to note that the observed changes are, to a large degree, influenced by the diet matrix of the starting Ecopath model (although, the diet composition can and does change in Ecosim as a function of the relative abundance of various groups within the system). It is clear from the dietary data available in the literature that our knowledge of species diets has yet to capture the full extent of their spatial and temporal variability (plasticity). Moreover, given the importance of seagrass invertebrate guilds in the diet of a number of especially juvenile, but also adult reef fish, more research is urgently needed to characterise these communities. Moreover, targeted surveys should be conducted to clarify what aspects of seagrass beds are most important in determining species composition and densities, and how these are likely to change under a scenario of increased turtle grazing.  101  Limited observations of juvenile parrotfish foraging behaviour from the Atlantic indicate that they are mostly herbivorous (Overholtzer and Motta 1999, 2000). However, stomach content analysis and other evidence from Australia show that juvenile parrotfish are either carnivorous or omnivorous (Bellwood 1988, Chen 2002). Therefore, more research is needed to resolve to what degree juvenile parrotfish in particular selectively feed on algae and/or seagrass (or possibly small invertebrates), and to what extent both juveniles and adults may switch to consuming a significantly larger proportion of algae than seagrass in the absence of their preferred diet item (which may also point towards more time spent on the reef to forage on algae for example). Fish are thought to exhibit adaptive flexibility in diet composition (Dill 1983, Gerking 1994) suggesting that such a response in foraging behaviour appears reasonable. Interestingly, however, correlative field studies have found a significant positive relationship between leaf nitrogen content and grazing by parrotfish, indicating preferential feeding on leaves rich in nitrogen (e.g., Goecker et al. 2005). Given that turtles have been shown to improve the nutritional content of their food by regular re-cropping of the same plots (Moran and Bjorndal 2005), it is likely that herbivorous fish will in turn, up to a point, preferentially feed on these turtle-grazed leaves. Further research is needed to elucidate this, and should include comparative studies between herbivore consumption rates of T. testudinum in ungrazed conditions and seagrass that has been continuously cropped by turtles. Such experiments should also include replicates under different vulnerability to predation scenarios.  The increase in grunt biomass under Mediation 1 and 2 was partly linked to a release from predation by declining snappers. Gut content data available for snappers often only indicate “small/juvenile” fish/fish in analysis results (Cocheret de la Moriniere et al. 2003a, Nagelkerken et al. 2006). It is therefore standard practice to allocate diet proportions to predators based on prey ratios. In light of our results, future studies should strive to determine (a) whether predators exhibit prey selectivity, or just feed in relative proportion to available prey items (this also influences a prey’s vulnerability to specific predators); (b) design field experiments to record potential changes in the diet composition 102  of these predators under scenarios of declining habitat complexity and thus prey availability; and (c) determine to what extent adults and juveniles may forage more heavily within habitats that previously may have represented peripheral foraging habitats (e.g., mangrove and/or reef). Cautiously extrapolating from results demonstrating that linkages between coastal ecosystems may be dependent on the spatial arrangement of habitats within the seascape (Nagelkerken et al. 2008), such a hypothesis could well be corroborated.  Habitat complexity and fish behaviour Results from the studies conducted on reefs where fish had limited access to seagrass nursery habitats show that some species may demonstrate behavioural plasticity in the habitats used by their juveniles. Dorenbosch et al. (2004), for example, demonstrated that when seagrass abundance was low, the juveniles of six out of 17 species known to use bays with seagrass and mangroves as nurseries were also observed on the reef. It is suggested that these species are able to use the reef as an alternative nursery and do not depend strictly on bays with seagrass beds and mangroves. For other, less ‘adaptable’ species, reefs with limited connectivity to specific nursery habitats resulted in significantly lower biomass of adults on the reef, when compared to sites that had greater availability of these nursery habitats (Dorenbosch et al. 2006, Mumby 2006, Grober-Dunsmore et al. 2007). For some species, behavioural plasticity in nursery habitat selection is further supported by the observation that juveniles may be obligate users of seagrass beds at one site, use additional habitats at another site (Nagelkerken et al. 2001, Adams and Ebersole 2002), or vary their habitat use inter- annually (Adams et al. 2006). This inherent variability indicates that caution is necessary when extrapolating findings from one location to another. Adams et al. (2006) have highlighted the general lack of baseline data on juvenile habitat use in tropical coastal systems, underscoring their overall poorly understood function as nursery sites. These data gaps make it difficult to reliably estimate the large scale impact of turtle grazing on the nursery function of seagrass beds. 103  Aside from increasing juvenile vulnerability to predation, competition and other interspecific interactions may significantly influence their abundance, and be affected by a reduced canopy height. However, given their demonstrated behavioural plasticity, juvenile fish may show an adaptive response to increased predation risk (and/or increased competition, and/or lowered food availability) and move to a different habitat (Mumby 2006). For example, it is not unreasonable to suggest that in the presence of large numbers of turtles, and thus closely cropped seagrass blades, species known to transition between mangrove and seagrass (e.g., bluestriped grunt (Haemulon sciurus) (Mumby et al. 2004, Faunce and Serafy 2007)) prior to their ontogenetic migration onto the reefs, may remain in the mangrove habitat during the entire portion of their early life cycle. Mangrove roots are typically considered key habitats, because they afford many species shelter from predators and support a diverse assemblage of epiphytic algae and invertebrates, providing an array of potential food sources for both predators and grazers (Farnsworth and Aaron 1996, Verweij et al. 2006). Our ability to predict the response of fish species to increased turtle grazing would be greatly improved if we had a better understanding of how the nursery function of different habitats was affected by their accessibility, availability, and relative structure. These therefore represent key areas for future research.  The prey ‘refuge’ concept is also likely to be dependent on the prey’s behaviour, the predator’s foraging strategy (Coen et al. 1981, Horinouchi 2007), the number of prey species of a given predator (Toscano et al. 2010), and the interaction of both predator and prey with the structure of the seagrass bed. Ambush predators, for example, utilise dense seagrass as camouflage, and may be more strongly affected than search and pursuit predators by the loss of habitat structure (Flynn and Ritz 2001). Changes in predator avoidance behaviour by prey in response to less structurally complex habitat may potentially be more important than physical interference of the habitat with predators while foraging (Main 1987).  104  Green turtles have been shown to introduce complexity within a seagrass bed through the maintenance of grazing plots (i.e., regular re-grazing of the same areas (Bjorndal 1980, Ogden et al. 1980)). At the landscape level, the patchwork-like pattern of grazed and ungrazed blades is likely to impact predator-prey spatial dynamics significantly. This type of heterogeneity is expected to be particularly important for juvenile fish that may prefer the open spaces for feeding, but are simultaneously in close proximity to effective protection from predators (Kerfoot and Sih 1987). Therefore, these edges or ‘ecotones’ may facilitate foraging, leading to a higher proportion of predators and fewer prey species than homogenous areas (Heck and Orth 1980, Irlandi 1994). One could hypothesise that intermediate turtle densities would create the most heterogeneous seagrass habitat and thus potentially sustain the greatest diversity (and/or abundance) of species. Moreover, juvenile fish densities in seagrass beds, although high relative to ‘reef’ habitats for example, are typically low overall (Christian 2003). Such a patchwork may lead to similar total abundance, but greater densities of juveniles within ‘better’ refuge patches. This in turn may increase, or alter, the outcome of potential competitive interactions between prey species and have unpredictable reciprocal effects on shared predators, given more limited refuge space.  Clearly, the behavioural response of different species to habitat structure may have a significant bearing on model outputs. The dramatic decline in snappers together with a visible increase in grunts resulting from Mediation 1 and 2 is an example of a relationship that may be substantially altered by behavioural adaptations. Although both trajectories can be explained from the changes accruing to the different groups based on prey availability and changes in vulnerability to predation, it is highly unlikely that structural changes brought about by turtle grazing would lead to the collapse of any one species. Future studies should, therefore, emphasise aspects that may guide the calibration of predator/prey vulnerability parameters. Field observations of grunts indicate that they react most strongly to spatial changes in habitat availability (P. Mumby, University of Exeter, pers. comm.), making their increase in biomass as a result of predation release surprising. This result 105  further underscores the need to more accurately quantify diet compositions and demonstrates the power of simulating changes in predator efficiency as a result of changes in habitat complexity at the ecosystem scale. It also highlights the need to conduct field experiments to accurately quantify response functions at the species level (Heck and Orth 1980, Bell and Westoby 1986) and in the presence of multiple predators (Toscano et al. 2010). This is further emphasized by the work of Young and Young (1978), who found that the variations in responses to seagrass blade clipping over a one year period did not consistently correspond to taxonomic groupings or feeding type, but instead resulted in species-specific changes.  3.3.4  Conclusions The model presented in this study, and simulated predator-prey interactions, highlight the potential changes that may be brought about by the recovery of a species whose grazing activity strongly alters habitat structure at the ecosystem level. Specifically, our results demonstrate that these changes are more complex than suggested by simple predator-prey experiments conducted under different seagrass densities. Small-scale studies performed in controlled conditions, and with a limited set of predator and prey species, are valuable to our understanding of how habitat complexity may mediate such interactions. However, under natural conditions, trophic pathways are governed by complex multi-species interactions. This is particularly true for species with ontogenetic migrations that will face different predators and feed upon different prey in different habitats. Ecological models help translate predictions made at experimental scales and consider changes that arise from trophic and non-trophic interactions (i.e., mediated through changes in habitat complexity) at the level of the ecosystem. They further highlight existing gaps in our current understanding of ecosystem processes and the need for field studies that mimic natural conditions. Field experiments in the form of green turtle exclosures, or experiments that mimic turtle grazing in dense seagrass beds, are required to add an empirical basis to some of the ecosystem-wide considerations highlighted in this 106  study (e.g., to what degree are given species associated with seagrass and/or to what degree do they exhibit adaptive flexibility to seagrass habitat use compared to mangrove/reef habitat).  The importance of seagrass, mangrove, and reef habitats as nursery areas and/or foraging grounds through the life history of single fish species (Nagelkerken et al. 2002, Christensen et al. 2003) highlights the need for a more holistic approach to conservation strategies (Sagarin and Crowder 2009). For example, the dramatic decline in mangrove extent worldwide (Valiela et al. 2001), including in the Caribbean, where local extinction of species that rely almost exclusively on mangrove habitats as juveniles has occurred (Mumby et al. 2004), needs to be considered in the context of the potential increased importance of mangroves with a decline in seagrass biomass due to recovering green turtle populations. While the design of Marine Protected Areas (MPA), one of a suite of oft advocated potential management tools (Agardy 1994, Allison et al. 1998, Jones 2001), involves numerous considerations (Carr et al. 2003), this study suggests that coral reef areas selected for conservation should focus on encompassing functionally linked essential habitat patches within the protected landscape. Model results further highlight the need for conservation measures that target sea turtle population recovery to consider ecosystem-level processes, and to develop such recovery initiatives within an ecosystem-based management framework. 107  Table 3.1 - Trophic parameters for all functional groups of the balanced Caribbean reef model. Outputs are presented in italics. B = biomass; TL = Trophic level; P/B = Productivity biomass ratio; Q/B = consumption rate; EE = Ecotrophic efficiency; P/Q = Production to consumption ratio or gross efficiency; P/R = Production to respiration ratio. MIF = Mobile Invertebrate Feeders; SIF = Sessile Invertebrate Feeders; SG = prefix for species known to forage in seagrass beds  Group name TL B P/B Q/B EE P/Q 1 Piscivores 3.59 1.640 0.458 4.500 0.556 0.102 2 SG_Piscivores 3.67 3.470 0.320 4.760 0.461 0.067 3 MIF 3.13 1.860 0.640 6.590 0.602 0.097 4 SIF 2.92 0.700 0.800 7.200 0.684 0.111 5 Zooplanktivores 3.02 2.500 1.590 13.400 0.599 0.119 6 Herbivores 2.08 2.900 1.940 35.010 0.693 0.055 7 Green turtles 2.00 0.020 0.150 10.420 0.114 0.014 8 SG_MIF 3.31 2.600 0.690 5.050 0.696 0.137 9 SG_Other 3.11 1.980 2.400 14.020 0.691 0.171 10 SG_Herbivores 2.01 2.700 0.480 17.490 0.811 0.027 11 Lutjanidae (adults) 3.39 9.517 0.710 5.740 0.946 0.124 12 Lutjanidae (juveniles) 3.07 1.325 0.700 13.591 0.955 0.052 13 Haemulidae (adults) 3.26 6.490 0.850 7.870 0.893 0.108 14 Haemulidae (juveniles) 3.07 0.167 1.300 28.975 0.926 0.045 15 Scaridae (adults) 2.01 5.175 1.410 43.200 0.633 0.033 16 Scaridae (juveniles) 2.01 1.239 1.580 85.258 0.925 0.019 17 Acanthuridae (adults) 2.01 9.651 0.580 27.040 0.559 0.021 18 Acanthuridae (juveniles) 2.22 0.066 1.200 142.112 0.915 0.008 19 SG_smallInvertebrates 2.07 39.980 7.010 125.000 0.950 0.056 20 SG_Invertebrates 2.44 27.943 3.700 18.000 0.950 0.206 21 Reef Invertebrates 2.13 24.854 2.800 15.000 0.950 0.187 22 Urchins 2.00 100.000 1.100 3.700 0.088 0.297 23 Seagrass 1.00 221.470 10.950 - 0.103 - 24 Algae 1.00 195.974 15.000 - 0.443 - 25 Epiphytes 1.00 37.553 25.000 - 0.276 - 26 Zooplankton 2.01 32.000 45.000 95.000 0.417 0.474 27 Phytoplankton 1.00 25.000 120.000 - 0.760 - 28 Corals/sponges 1.92 121.000 1.100 6.000 0.182 0.183 29 Detritus 1.00 100.000 - - 0.566 -  108  Table 3.2 - Diet composition matrix for the functional groups in the balanced model for the Caribbean. Shaded cells indicate diet proportion <0.0009.    Predator  Prey 1 2 3 4 5 6 7 8 9 10 11 1 Piscivores 0.001 - - - - - - - - - - 2 SG Piscivores 0.020  - - - - - - - - - 3 MIF 0.030 0.019 - - - - - - - - - 4 SIF 0.030 0.010 - - - - - - - - - 5 Zooplanktivores 0.101 0.099 - - - - - - - - - 6 Herbivores 0.200 0.140 0.009 - - - - - - - - 7 Green turtles    - - - - - - - - 8 SG MIF 0.010 0.050 0.005 - - - - -  - - 9 SG Other 0.051 0.110 0.005 - - - -  - - 0.015 10 SG Herbivores 0.007 0.020 0.005 - - - - - 0.007 - 0.005 11 Lutjanidae 0.114 0.090 0.005 - - - - - 0.007 - - 12 Lutjanidae (juv) 0.015 0.044  - - - -   - - 13 Haemulidae 0.070 0.120 0.005 - - - - - 0.007 - 0.016 14 Haemulidae (juv) 0.008 0.003  - - - -   - 0.002 15 Scaridae 0.219 0.107 0.005 - - - - - 0.009 - 0.012 16 Scaridae (juv) 0.010 0.042  - - - -   - 0.019 17 Acanthuridae 0.030 0.120 0.005 - - - - - 0.00100 - 0.007 18 Acanthuridae (juv) 0.001 0.001  - - - -   - 19 SG smallInvertebrates - - - - - - - 0.007 0.015 0.005 20 SG Invertebrates 0.001 0.024 0.010 - 0.001 - - 0.789 0.261 - 0.744 21 REEF Invertebrates 0.017  0.880 0.048 0.200 0.001 - 0.063 0.100 - 0.100 22 Urchins 0.015  0.040 - - - - 0.100 - - 0.015 23 Seagrass - - - - - 0.005 0.920 0.030  0.220 24 Algae - - 0.010 0.005 0.010 0.608 0.030  0.037 0.610 25 Epiphytes - - - -  0.005 - - - 0.155 0.001 26 Zooplankton - - - 0.001 0.739 0.001 - - 0.480 - 0.009 27 Phytoplankton - - - - - - - - - - - 28 Corals  sponges - - 0.015 0.946 0.050 0.090 - 0.00100 0.074 - - 29 Detritus - - 0.001   0.290 0.050 0.009  0.010 0.005 30 Import 0.050 - - - - - - - - - 0.050 109   Table 3.2. - cont.    Predator  Prey/Predator 12 13 14 15 16 17 18 19 20 21 22 26 27 1 Piscivores - - - - - - - - - - - - - 2 SG Piscivores - - - - - - - - - - - - - 3 MIF - - - - - - - - - - - - - 4 SIF - - - - - - - - - - - - - 5 Zooplanktivores - - - - - - - - - - - - - 6 Herbivores - - - - - - - - - - - - - 7 Green turtles - - - - - - - - - - - - - 8 SG_MIF - - - - - - - - - - - - - 9 SG Other - - - - - - - - - - - - - 10 SG Herbivores - - - - - - - - - - - - - 11 Lutjanidae - - - - - - - - - - - - - 12 Lutjanidae (juv) - - - - - - - - - - - - - 13 Haemulidae - - - - - - - - - - - - - 14 Haemulidae (juv) - - - - - - - - - - - - - 15 Scaridae - - - - - - - - - - - - - 16 Scaridae (juv) - - - - - - - - - - - - - 17 Acanthuridae - - - - - - - - - - - - - 18 Acanthuridae (juv) - - - - - - - - - - - - - 19 SG smallInvertebrates 0.979 0.400 0.979 - 0.010 0.010 0.200 0.008 0.350 0.001 - - 0.001 20 SG Invertebrates 0.010 0.563 0.010 - - - - - 0.020 0.001 - - - 21 REEF Invertebrates - 0.010 - - - - - - 0.010 0.090 - - - 22 Urchins - - - - - - - - 0.010 0.005 - - - 23 Seagrass - - - 0.700 0.200 0.048 0.001 - 0.001 0.008 0.120 - - 24 Algae 0.001 0.010 0.001 0.080 0.530 0.761 0.380 0.120 0.010 0.231 0.660 - - 25 Epiphytes  -  0.120 0.249 0.171 0.404 0.001 0.200 - 0.120 - - 26 Zooplankton 0.010 0.007 0.010 - - - 0.005 0.062 0.010 0.010 - 0.006 0.309 27 Phytoplankton - - - - - - - 0.010 - - - 0.650 0.350 28 Corals  sponges - - - 0.009 0.001 - -  0.001 0.010 - - - 29 Detritus - 0.010 - 0.091 0.010 0.010 0.010 0.799 0.388 0.644 0.100 0.344 0.34 30 Import - - - - - - - - - - - - - 110  Table 3.3 - Ecological indicators of the Caribbean model as pertaining to community energetics, community structure, and cycling of nutrients. The ‘cycling index’ is the fraction of an ecosystem's throughput that is recycled and is expressed here as percentage.  Parameter Value Unit Sum of all consumption 11,002 t/km²/year Sum of all exports 4,473 t/km²/year Sum of all respiratory flows 4,833 t/km²/year Sum of all flows into detritus 10,279 t/km²/year Total system throughput 30,587 t/km²/year Calculated total net primary production 9,303 t/km²/year Total primary production/total respiration 1.93 Total primary production/total biomass 10.6 Total biomass/total throughput 0.029 Total biomass (excluding detritus) 880 t/km² Mean trophic level of the catch 3.21 Mean transfer efficiency 8.9%  Network flow indices Finn's cycling index 13.62 % of total throughput System Omnivory Index 0.091 111  Puerto Rico US Virgin Islands    Figure 3.1 - Puerto Rico and the US Virgin Islands. The dashed line represents the approximate delineation between the US Virgin Islands and the British Virgin Islands. Map modified from http://reefgis.reefbase.org 112  Seagrass biomass ( g·m-2) HighLow P re y vu ln er ab ili ty Low High Simulation startSimulation end P re y vu ln er ab ili ty P re y vu ln er ab ili ty P re y vu ln er ab ili ty P re y vu ln er ab ili ty               Figure 3.2 - Potential relationship between seagrass biomass (i.e., refuge capacity) with increasing green turtle abundance and vulnerability of prey to predators (modified after Canion and Heck (2009) and implemented in Ecopath as mediation scenarios). Mediation 1 = dotted line; mediation 2 = dashed line; mediation 3 = solid line.  113  0.0 0.5 1.0 1.5 Total PP Invertebrates Fish Pelagic Demersal Fi na l B io m as s / I ni tia l B io m as s 0.0 0.5 1.0 1.5 2.0 2.5 0 20 40 60 80 100 Years Fi na l B io m as s / I ni tia l B io m as s Haemulids juv. Haemulids adult Acanthurids juv. Acanthurids adult Lutjanids juv. Lutjanids adult Scarids juv. Scarids adult SG piscivores SG MIF SG Herbivores SG others SG Small invert. SG invert. Seagrass Epiphytes (a)          (b)          Figure 3.3 – (a) Biomass trajectories (final relative to initial biomasses) of selected trophic guilds under a simulated increase in turtle biomass implemented concomitantly with mediation 1. Illustrated in panel (b), for the same time frame, is the ratio of all final relative to initial biomasses grouped by primary producers (PP), invertebrates, fish, and all pelagic versus demersal/benthic trophic guilds. SG = prefix for species known to forage in seagrass beds 114  0.0 0.5 1.0 1.5 Total PP Invertebrates Fish Pelagic Demersal Fi na l B io m as s / I ni tia l B io m as s 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0 10 20 30 40 50 60 70 80 90 100 Years Fi na l B io m as s / I ni tia l B io m as s Haemulids juv. Haemulids adult Acanthurids juv. Acanthurids adult Lutjanids juv. Lutjanids adult Scarids juv. Scarids adult SG piscivores SG MIF SG Herbivores SG others SG Small invert. SG invert. Seagrass Epiphytes (a)           (b)          Figure 3.4 – (a) Biomass trajectories (final relative to initial biomasses) of selected trophic guilds under a simulated increase in turtle biomass implemented concomitantly with mediation 2. Illustrated in panel (b), for the same time frame, is the ratio of all final relative to initial biomasses grouped by primary producers (PP), invertebrates, fish, and all pelagic versus demersal/benthic trophic guilds. SG = prefix for species known to forage in seagrass beds 115  0.0 0.5 1.0 1.5 Total PP Invertebrates Fish Pelagic Demersal Fi na l B io m as s / I ni tia l B io m as s 0.0 0.2 0.4 0.6 0.8 1.0 1.2 0 10 20 30 40 50 60 70 80 90 100 Years Fi na l B io m as s / I ni tia l B io m as s Haemulids juv. Haemulids adult Acanthurids juv. Acanthurids adult Lutjanids juv. Lutjanids adult Scarids juv. Scarids adult SG piscivores SG MIF SG Herbivores SG others SG Small invert. SG invert. Seagrass Epiphytes (a)           (b)          Figure 3.5 – (a) Biomass trajectories (final relative to initial biomasses) of selected trophic guilds under a simulated increase in turtle biomass implemented concomitantly with mediation 3. Illustrated in panel (b), for the same time frame, is the ratio of all final relative to initial biomasses grouped by primary producers (PP), invertebrates, fish, and all pelagic versus demersal/benthic trophic guilds. SG = prefix for species known to forage in seagrass beds 116  3.4  References Adams, A. J. and Ebersole, J. P. (2002) Use of back-reef and lagoon habitats by coral reef fishes. Marine Ecology Progress Series, 228: 213-226. Adams, A. J., Dahlgren, C. P., Kellison, G. T., Kendall, M. S., Layman, C. A., Ley, J. A., Nagelkerken, I. and Serafy, J. E. (2006) Nursery function of tropical back-reef systems. Marine Ecology Progress Series, 318: 287-301. Agardy, M. T. (1994) Advances in marine conservation - the role of marine protected areas. Trends in Ecology & Evolution, 9 (7): 267-270. Aguilar-Perea, A. and Appeldoorn, R. S. (2007) Variations in juvenile fish density along the mangrove-seagrass-coral reef continuum in southwestern Puerto Rico. Bulletin of Marine Science, 80 (3): 916-916. Allen, R. R. (1971) Relation between production and biomass. Journal of the Fisheries Research Board of Canada, 28: 1573- 1581. Allison, G. W., Lubchenco, J. and Carr, M. H. (1998) Marine reserves are necessary but not sufficient for marine conservation. Ecological Applications, 8 (1): S79-S92. Arias-González, J. E. (1998) Trophic models of semi-protected and unprotected coral reef ecosystems in the South of the Mexican Caribbean. Journal of Fish Biology, 53: 236-255. Baelde, P. (1990) Differences in the structures of fish assemblages in Thalassia testudinum beds in Guadeloupe, French West Indies, and their ecological significance. Marine Biology, 105: 163-173. Beck, M. W., Heck, K. L. J., Able, K. W., Childers, D. L., Eggleston, D. B., Gillanders, B. M., Halpern, B., Hays, C. G., Hoshino, K., Minello, T. J., Orth, R. J., Sheridan, P. F. and Weinstein, M. P. (2001) The identification, conservation, and management of estuarine and marine nurseries for fish and invertebrates. BioScience, 51 (8): 633-641. Beck, M. W., Heck Jr., K. L., Able, K. W., Childers, D. L., Eggleston, D. B., Gillanders, B. M., Halpern, B. S., Hays, C. G., Hoshino, K., Minello, T. J., Orth, R. J., Sheridan, P. F. and Weinstein, M. P. (2003) The role of nearshore ecosystems as fish and shellfish nurseries. Issues in ecology. Ecological Society of America, Washington, D.C., USA. 12 pp. Bell, J. D. and Westoby, M. (1986) Importance of local changes in leaf height and density to fish and decapods associated with seagrasses. Journal of Experimental Marine Biology and Ecology, 104 (1-3): 249-274. Bellwood, D. R. (1988) Ontogenetic changes in the diet of early post-settlement Scarus species (Pisces: Scaridae). Journal of Fish Biology, 33: 213-219. 117  Bjorndal, K. A. (1980) Nutrition and grazing behavior of the green turtle, Chelonia mydas. Marine Biology, 56 (2): 147-154. Bjorndal, K. A. and Jackson, J. B. C. (2003) Roles of sea turtles in marine ecosystems: Reconstructing the past. In The Biology of Sea Turtles. (eds P. L. Lutz, J. A. Musick and J. Wyneken), pp. 259-273. CRC Press, Boca Raton (FL), USA. Bjorndal, K., Wetherall, J., Bolten, A. and Mortimer, J. (1999) Twenty six years of green turtle nesting at Tortuguero, Costa Rica: An encouraging trend. Conservation Biology, 13 (1): 126-134. Bouchon-Navaro, Y., Bouchon, C. and Louis, M. (2004) Fish communities of seagrass beds in the Lesser Antilles: Interest of their protection. Revue d'Ecologie - La Terre et la Vie, 59 (1-2): 253-272. Brand-Gardner, S. J., Lanyon, J. M. and Limpus, C. J. (1999) Diet selection by immature green turtles, Chelonia mydas, in subtropical Moreton Bay, south-east Queensland. Australian Journal of Zoology, 47 (2): 181-191. Bräutigam, A. and Eckert, K. L. (2006) Turning the tide: Exploitation, trade and management of marine turtles in the Lesser Antilles, Central America, Colombia and Venezuela. TRAFFIC International, Cambridge, UK. 551 pp. Brook, I. M. (1977) Trophic relationships in a seagrass community (Thalassia testudinum), in Card Sound, Florida: Fish diets in relation to macrobenthic and cryptic faunal abundance. Transactions of the American Fisheries Society, 106 (3): 219-229. Burke, N. C. (1995) Nocturnal foraging habitats of French and bluestriped grunts, Haemulon flavolineatum and H. sciurus, at Tobacco Caye, Belize. Environmental Biology of Fishes, 42 (365-374). Canion, C. R. and Heck, K. L., Jr. (2009) Effect of habitat complexity on predation success: Re- evaluating the current paradigm in seagrass beds. Marine Ecology Progress Series, 393. Carr, S. and Adams, C. (1973) Food habits of juvenile marine fishes occupying seagrass beds in the estuarine zone near Crystal River, Florida. Transactions of the American Fisheries Society, 102: 511-- 540. Carr, M. H., Neigel, J. E., Estes, J. A., Andelman, S., Warner, R. R. and Largier, J. L. (2003) Comparing marine and terrestrial ecosystems: Implications for the design of coastal marine reserves. Ecological Applications, 13 (1): S90-S107. Chaloupka, M., Bjorndal, K. A., Balazs, G. H., Bolten, A. B., Ehrhart, L. M., Limpus, C. J., Suganuma, H., Troeng, S. and Yamaguchi, M. (2008) Encouraging outlook for recovery of a once severely exploited marine megaherbivore. Global Ecology and Biogeography, 17 (2): 297-304. Chen, L. S. (2002) Post-settlement diet shift of Chlorurus sordidus and Scarus schlegeli (Pisces: Scaridae). Zoological Studies, 41: 47-58. 118  Chittaro, P. M., Usseglio, P. and Sale, P. F. (2005) Variation in fish density, assemblage composition and relative rates of predation among mangrove, seagrass and coral reef habitats. Environmental Biology of Fishes, 72 (2): 175-187. Christensen, J. D., Jeffrey, C. F. G., Caldow, C., Monaco, M. E., Kendall, M. S. and Appeldoorn, R. S. (2003) Cross-shelf habitat utilization patterns of reef fishes in southwestern Puerto Rico. Gulf and Caribbean Research, 14 (2): 9-27. Christensen, V. (1995) Ecosystem maturity - towards quantification. Ecological Modelling, 77: 3-32. Christensen, V. and Pauly, D. (1992) ECOPATH II - A software for balancing steady-state ecosystem models and calculating network characteristics. Ecological Modelling, 61: 169-185. Christensen, V. and Pauly, D. (2004) Placing fisheries in their ecosystem context, an introduction. Ecological Modelling, 172 (2-4): 103-107. Christensen, V. and Walters, C. (2004) Ecopath with Ecosim: Methods, capabilities and limitations. Ecological Modelling, 72: 109-139. Christensen, V., Walters, C. J. and Pauly, D. (2005) Ecopath with Ecosim: A user's guide. Fisheries Centre, University of British Columbia, Vancouver (B.C.), Canada. 154 pp. Christian, J. (2003) Evaluating the energetic value and structural contributions of mangroves and seagrasses to local fish populations: the importance of habitat connectivity. Master thesis. University of Alabama, Tuscaloosa (AL), USA. Cocheret de la Moriniere, E., Pollux, B. J. A., Nagelkerken, I. and van der Velde, G. (2002) Post- settlement life cycle migration patterns and habitat preference of coral reef fish that use seagrass and mangrove habitats as nurseries. Estuarine Coastal and Shelf Science, 55 (2): 309-321. Cocheret de la Moriniere, E., Pollux, B. J. A., Nagelkerken, I., Hemminga, M. A., Huiskes, A. H. L. and van der Velde, G. (2003a) Ontogenetic dietary changes of coral reef fishes in the mangrove-seagrass-reef continuum: Stable isotopes and gut-content analysis. Marine Ecology Progress Series, 246: 279-289. Cocheret de la Moriniere, E., Pollux, B. J. A., Nagelkerken, I. and van der Velde, G. (2003b) Diet shifts of Caribbean grunts (Haemulidae) and snappers (Lutjanidae) and the relation with nursery- to-coral reef migrations. Estuarine Coastal and Shelf Science, 57 (5-6): 1079-1089. Coen, L. D., Heck, K. L., Jr. and Abele, L. G. (1981) Experiments on competition and predation among shrimps of seagrass meadows. Ecology, 62 (6): 1484-1493. Coll, M., Bundy, A. and Shannon, L. J. (2009) Ecosystem modelling using the Ecopath with Ecosim approach. In Computers in Fisheries Research. (eds B. A. Megrey and E. Moksness), pp. 225-291. Springer Science + Business Media. 119  Cox, C., Hunt, J. H., Lyons, W. G. and Davis, G. E. (1997) Nocturnal foraging of the Caribbean spiny lobster (Panulirus argus) on offshore reefs of Florida, USA. Marine and Freshwater Research, 48 (8): 671-679. Crowder, L. B. and Cooper, W. E. (1982) Habitat structural complexity and the interaction between bluegills and their prey. Ecology, 63 (6): 1802-1813. Dahlgren, C. P., Kellison, G. T., Adams, A. J., Gillanders, B. M., Kendall, M. S., Layman, C. A., Ley, J. A., Nagelkerken, I. and Serafy, J. E. (2006) Marine nurseries and effective juvenile habitats: concepts and applications. Marine Ecology Progress Series, 312: 291-295. Delgado, P. (2004) The US Caribbean region - wetlands and fish: a vital connection. Fish, crustaceans, and mollusks found in US Caribbean wetlands. National Oceanic and Atmospheric Administration, 93 pp. Dill, L. M. (1983) Adaptive flexibility in the foraging behavior of fishes. Canadian Journal of Fisheries and Aquatic Sciences, 40 (4): 398-408. Domning, D. P. (2001) Sirenians, seagrasses, and Cenozoic ecological change in the Caribbean. Palaeogeography Palaeoclimatology Palaeoecology, 166 (1-2): 27-50. Dorenbosch, M. (2004) The relationship of reef fish densities to the proximity of mangrove and seagrass nurseries. Estuarine Coastal and Shelf Science, 60 (1): 37-48. Dorenbosch, M., Grol, M. G. G., Nagelkerken, I. and van der Velde, G. (2006) Seagrass beds and mangroves as potential nurseries for the threatened Indo-Pacific humphead wrasse, Cheilinus undulatus and Caribbean rainbow parrotfish, Scarus guacamaia. Biological Conservation, 129 (2): 277-282. Dorenbosch, M., Verberk, W., Nagelkerken, I. and van der Veldel, G. (2007) Influence of habitat configuration on connectivity between fish assemblages of Caribbean seagrass beds, mangroves and coral reefs. Marine Ecology Progress Series, 334: 103-116. Dunn, D. C., Stewart, K., Bjorkland, R. H., Haughton, M., Singh-Renton, S., Lewison, R., Thorne, L. and Halpin, P. N. (2009) A regional analysis of coastal and domestic fishing effort in the wider Caribbean. Fisheries Research, 102 (1-2): 60-68. Farnsworth, E. J. and Aaron, M. E. (1996) Scale-dependent spatial and temporal variability in biogeography of mangrove root epibiont communities. Ecological Monographs, 66 (1): 45-66. Faunce, C. H. and Serafy, J. E. (2007) Nearshore habitat use by gray snapper (Lutjanus griseus) and bluestriped grunt (Haemulon sciurus): Environmental gradients and ontogenetic shifts. Bulletin of Marine Science, 80 (3): 473-495. 120  Flynn, A. J. and Ritz, D. A. (2001) Effect of habitat complexity and predatory style on the capture success of fish feeding on aggregated prey. Journal of the Marine Biological Association of the UK, 79 (3): 487-494. Gacia, E. (1999) Leaf dynamics and shoot production of the seagrass Thalassia testudinum in the Indian River Lagoon (Florida). Botanica Marina, 42 (1): 97-102. García-Sais, J. R., Appeldoorn, R., Bruckner, A., Caldow, C., Christensen, j. D., Lilyestrom, C., Monaco, M. E., Sabater, J., Williams, E. and Diaz, E. (2005) The State of Coral Reef Ecosystems of the Commonwealth of Puerto Rico. In The state of coral reef ecosystems of the United States and Pacific freely associated states: 2005. (ed J. E. Waddell), pp. 91-134. NOAA Technical Memorandum NOS NCCOS 11. NOAA/NCCOS Center for Coastal Monitoring and Assessment’s Biogeography Team, Silver Spring (MD), USA. Gaut, V. C. and Munro, R. L. (1983) The biology, ecology and bioeconomics of the grunts, Pomadasyidae. In Caribbean Coral Reef Fishery Resources. (ed R. L. Munro), pp. 110-141. ICLARM, Manila, Philippines. Gerking, S. D. (1994) Feeding Ecology of Fish, Academic Press, San Diego (CA), USA. Gillanders, B. M. (2006) Seagrasses, fish and fisheries. In Seagrasses: Biology, Ecology and Conservation. (eds A. W. D. Larkum, R. J. Orth and C. M. Duarte), pp. 503-536. Springer, Dordrecht, The Netherlands. Goecker, M. E., Heck, K. L. and Valentine, J. F. (2005) Effects of nitrogen concentrations in turtlegrass Thalassia testudinum on consumption by the bucktooth parrotfish Sparisoma radians. Marine Ecology Progress Series, 286: 239-248. Gotceitas, V. and Colgan, P. (1989) Predator foraging success and habitat complexity: Quantitative test of the threshold hypothesis. Oecologia, 80 (2): 158-166. Grober-Dunsmore, R., Frazer, T. K., Lindberg, W. J. and Beets, J. (2007) Reef fish and habitat relationships in a Caribbean seascape: The importance of reef context. Coral Reefs, 26 (1): 201- 216. Grober-Dunsmore, R., Frazer, T. K., Beets, J. P., Lindberg, W. J., Zwick, P. and Funicelli, N. A. (2008) Influence of landscape structure, on reef fish assemblages. Landscape Ecology, 23: 37-53. Hammerschlag, N. and Serafy, J. E. (2009) Nocturnal fish utilization of a subtropical mangrove- seagrass ecotone. Marine Ecology: doi:10.1111/j.1439-0485.2009.00337.x. Hawkins, J. P. and Roberts, C. M. (2004) Effects of artisanal fishing on Caribbean coral reefs. Conservation Biology, 18 (1): 215-226. Hay, M. E. (1981) Spatial patterns of grazing intensity on a Caribbean barrier reef: Herbivory and algal distribution. Aquatic Botany, 11 (2): 97-109. 121  Heck, K. L. and Orth, R. J. (1980) Seagrass habitats: The roles of habitat complexity, competition and predation in structuring associated fish and motile macroinvertebrates assemblages. In Estuarine Perspectives. (ed V. S. Kennedy), pp. 449-464. Academic Press, New York (NY), USA. Heck, K. L. and Orth, R. J. (2006) Predation in seagrass beds. In Seagrasses: Biology, Ecology and Conservation. (eds A. W. D. Larkum, R. J. Orth and C. M. Duarte), pp. 537-550. Springer, Dordrecht, The Netherlands. Heck, K. L. and Weinstein, M. P. (1989) Feeding habits of juvenile reef fishes associated with Panamanian seagrass meadows. Bulletin of Marine Science, 45 (3): 629-636. Heck, K. L., Hays, G. and Orth, R. J. (2003) Critical evaluation of the nursery role hypothesis for seagrass meadows. Marine Ecology Progress Series, 253: 123-136. Hemminga, M. A. and Duarte, C. M. (2000) Seagrass Ecology: An Introduction, Cambridge University Press, Cambridge, UK. 298 pp. Hixon, M. A. (1991) Predation as a process structuring coral-reef fish communities. In The Ecology of Fishes on Coral Reefs. (ed P. F. Sale), pp. 475-508. Academic Press, San Diego (CA), USA. Hixon, M. A. and Beets, J. P. (1993) Predation, prey refuges, and the structure of coral reef fish assemblages. Ecological Monographs, 63 (1): 77-101. Horinouchi, M. (2007) Review of the effects of within-patch scale structural complexity on seagrass fishes. Journal of Experimental Marine Biology and Ecology, 350 (1-2): 111-129. Irlandi, E. A. (1994) Large-scale and small-scale effects of habitat structure on rates of predation: How percent coverage of seagrass affects rates of predation and siphon nipping on an infaunal bivalve. Oecologia, 98 (2): 176-183. Jackson, J. B. C., Kirby, M. X., Berger, W. H., Bjorndal, K. A., Botsford, L. W., Bourque, B. J., Bradbury, R. H., Cooke, R., Erlandson, J., Estes, J. A., Hughes, T. P., Kidwell, S., Lange, C. B., Lenihan, H. S., Pandolfi, J. M., Peterson, C. H., Steneck, R. S., Tegner, M. J. and Warner, R. R. (2001) Historical overfishing and the recent collapse of coastal ecosystems. Science, 293 (5530): 629-638. Jones, P. J. S. (2001) Marine protected area strategies: Issues, divergences and the search for middle ground. Reviews in Fish Biology and Fisheries, 11 (3): 197-216. Kendall, M. S., Kruer, C. R., Buja, K. R., Christensen, J. D., Finkbeiner, M., Warner, R. A. and Monaco, M. E. (2001) Methods used to map the benthic habitats of Puerto Rico and the U.S. Virgin Islands. NOAA, NOS, NCCOS, Silver Spring (MD), USA. 45 pp. Kerfoot, W. C. and Sih, A. (1987) Predation: Direct and Indirect Impacts on Aquatic Communities, University Press of New England, Hanover (NH), USA. 386 pp. 122  Kikuchi, T. and Peres, J. M. (1977) Consumer ecology of seagrass beds. In Seagrass Ecosystems - A Scientific Perspective. (eds C. P. McRoy and C. Helfferich), pp. 147-194. Marcel Dekker, Inc., New York (NY), USA. Kirsch, K. D., Valentine, J. F. and Heck, K. L. (2002) Parrotfish grazing on turtlegrass Thalassia testudinum: Evidence for the importance of seagrass consumption in food web dynamics of the Florida Keys National Marine Sanctuary. Marine Ecology Progress Series, 227: 71-85. Kopp, D., Bouchon-Navaro, Y., Louis, M. and Bouchon, C. (2007) Diel differences in the seagrass fish assemblages of a Caribbean island in relation to adjacent habitat types. Aquatic Botany, 87 (1): 31-37. Lanyon, J. M., Limpus, C. J. and Marsh, H. (1989) Dugongs and turtles: Grazers in the seagrass system. In Biology of Seagrasses (eds A. W. D. Larkum, A. J. McComb and S. A. Shepherd), pp. 610-634. Elsevier, Amsterdam, The Netherlands. Layman, C. A. and Silliman, B. R. (2002) Preliminary survey and diet analysis of juvenile fishes of an estuarine creek on Andros Island, Bahamas. Bulletin of Marine Science, 70 (1): 199-210. Lewis, S. M. and Wainwright, P. C. (1985) Herbivore abundance and grazing intensity on a Caribbean coral reef. Journal of Experimental Marine Biology and Ecology, 87 (3): 215-228. Lindeman, K. C., Pugliese, R., Waugh, G. T. and Ault, J. S. (2000) Developmental patterns within a multispecies reef fishery: Management applications for essential fish habitats and protected areas. Bulletin of Marine Science, 66 (3): 929-956. Lipcius, R. N., Eggleston, D. B., Miller, D. L. and Luhrs, T. C. (1998) The habitat-survival function for Caribbean spiny lobster: An inverted size effect and non-linearity in mixed algal and seagrass habitats. Marine and Freshwater Research, 49 (8): 807-816. Luo, J. G., Serafy, J. E., Sponaugle, S., Teare, P. B. and Kieckbusch, D. (2009) Movement of gray snapper Lutjanus griseus among subtropical seagrass, mangrove, and coral reef habitats. Marine Ecology Progress Series, 380: 255-269. Main, K. L. (1987) Predator avoidance in seagrass meadows: Prey behavior, microhabitat selection, and cryptic coloration. Ecology, 68 (1): 170-180. Manickchand-Heileman, S., Soto, L. A. and Escobar, E. (1998) A preliminary trophic model of the continental shelf, south-western Gulf of Mexico. Estuarine, Coastal and Shelf Science, 46 (6): 885- 899. Mateo, I. and Tobias, J. (2001) Distribution of shallow water coral reef fishes on the Northeast Coast of St Croix, USVI. Caribbean Journal of Science, 37 (3-4): 210-226. Mateo, I. and Tobias, W. J. (2004) Survey of nearshore fish communities on tropical backreef lagoons on the southeastern coast of St. Croix. Caribbean Journal of Science, 40 (3): 327-342. 123  Mateo, I. and Tobias, W. J. (2007) A comparison of fish assemblages among five habitat types within a Caribbean lagoonal system. Gulf and Caribbean Research, 19: 21-31. Mateo, I. and Tobias, W. J. (2008) Seasonal patterns of juvenile fish abundance in seagrass meadows in Teague Bay bank barrier reef lagoon, St. Croix, US Virgin Islands. Gulf and Caribbean Research, 20: 59-65. Mattila, J., Heck, K. L., Millstein, E., Miller, E., Gustafsson, C., Williams, S. and Byron, D. (2008) Increased habitat structure does not always provide increased refuge from predation. Marine Ecology Progress Series, 361: 15-20. McAfee, S. T. and Morgan, S. G. (1996) Resource use by five sympatric parrotfishes in the San Blas Archipelago, Panama. Marine Biology, 125 (3): 427-437. McFarland, W. N., Brothers, E. B., Ogden, J. C., Shulman, M. J., Bermingham, E. L. and Kotchianprentiss, N. M. (1985) Recruitment patterns in young french grunts, Haemulon flavolineatum (Family Haemulidae), at St-Croix, Virgin Islands. Fishery Bulletin, 83 (3): 413-426. McGlathery, K. J. (1995) Nutrient and grazing influences on a subtropical seagrass community. Marine Ecology Progress Series, 122 (1-3): 239-252. McNaughton, S. J. (1979) Grazing as an optimization process - Grass ungulate relationships in the Serengeti. American Naturalist, 113 (5): 691-703. Moran, K. L. and Bjorndal, K. A. (2005) Simulated green turtle grazing affects structure and productivity of seagrass pastures. Marine Ecology Progress Series, 305: 235-247. Moran, K. L. and Bjorndal, K. A. (2007) Simulated green turtle grazing affects nutrient composition of the seagrass Thalassia testudinum. Marine Biology, 150 (6): 1083-1092. Mumby, P. J. (2006) Connectivity of reef fish between mangroves and coral reefs: Algorithms for the design of marine reserves at seascape scales. Biological Conservation, 128 (2): 215-222. Mumby, P. J., Edwards, A. J., Arias-Gonzalez, J. E., Lindeman, K. C., Blackwell, P. G., Gall, A., Gorczynska, M. I., Harborne, A. R., Pescod, C. L., Renken, H., Wabnitz, C. C. C. and Llewellyn, G. (2004) Mangroves enhance the biomass of coral reef fish communities in the Caribbean. Nature, 427 (6974): 533-536. Nagelkerken, I., Van der Velde, G., Verberk, W. and Dorenbosch, M. (2006) Segregation along multiple resource axes in a tropical seagrass fish community. Marine Ecology Progress Series, 308: 79-89. Nagelkerken, I., Bothwell, J., Nemeth, R. S., Pitt, J. M. and van der Velde, G. (2008) Interlinkage between Caribbean coral reefs and seagrass beds through feeding migrations by grunts (Haemulidae) depends on habitat accessibility. Marine Ecology Progress Series, 368: 155-164. 124  Nagelkerken, I., Dorenbosch, M., Verberk, W., de la Moriniere, E. C. and van der Velde, G. (2000a) Importance of shallow-water biotopes of a Caribbean bay for juvenile coral reef fishes: patterns in biotope association, community structure and spatial distribution. Marine Ecology Progress Series, 202: 175-192. Nagelkerken, I., van der Velde, G., Gorissen, M. W., Meijer, G. J., van't Hof, T. and den Hartog, C. (2000b) Importance of mangroves, seagrass beds and the shallow coral reef as a nursery for important coral reef fishes, using a visual census technique. Estuarine Coastal and Shelf Science, 51 (1): 31-44. Nagelkerken, I., Kleijnen, S., Klop, T., van den Brand, R., de la Moriniere, E. C. and van der Velde, G. (2001) Dependence of Caribbean reef fishes on mangroves and seagrass beds as nursery habitats: a comparison of fish faunas between bays with and without mangroves/seagrass beds. Marine Ecology Progress Series, 214: 225-235. Nagelkerken, I., Roberts, C. M., van der Velde, G., Dorenbosch, M., van Riel, M. C., de la Morinere, E. C. and Nienhuis, P. H. (2002) How important are mangroves and seagrass beds for coral-reef fish? The nursery hypothesis tested on an island scale. Marine Ecology Progress Series, 244: 299-305. Nelson, W. (1979) Experimental studies of selective predation on amphipods: Consequences for amphipod distribution and abundance. Journal of Experimental Biology and Ecology, 38: 225-245. NMFS (1993) Recovery plan for hawksbill turtles in the U.S. Caribbean Sea, Atlantic Ocean, and Gulf of Mexico. National Marine Fisheries Service, St Petersburg (FL), USA. 58 pp. NMFS (2007a) Green sea turtle (Chelonia mydas) 5 year review: Summary and evaluation. National Marine Fisheries Service, Office of Protected Resources, Silver Spring (MD), USA and US Fish and Wildlife Service Southeast Region, Jacksonville (FL), USA, 105 pp. NMFS (2007b) Hawksbill sea turtle (Eretmochelys imbricata) 5 year review: Summary and evaluation. National Marine Fisheries Service, Office of Protected Resources, Silver Spring (MD), USA and US Fish and Wildlife Service Southeast Region, Jacksonville (FL), USA, 93 pp. Ogden, J. C. (1976) Some aspects of herbivore-plant relationships on Caribbean reefs and seagrass beds. Aquatic Botany, 2: 103-116. Ogden, J. (1980) Faunal relationships in Caribbean seagrass beds. In A Handbook of Seagrass Biology: An Ecosystem Perspective. (eds R. C. Philipps and C. P. McRoy), pp. 173-198. Garland SPTM Press, New York (NY), USA. Ogden, J. and Zieman, J. (1977) Ecological aspects of coral reef-seagrass bed contacts in the Caribbean. In Proceedings of the third International Coral Reef Symposium pp. 377-382. University of Miami (FL), USA. 125  Ogden, J. C., Tighe, S. and Miller, S. (1980) Grazing of seagrasses by large herbivores in the Caribbean. American Zoologist, 20 (4): 949-949. Okey, T. A. (2004) Shifted community states in four marine ecosystems: some potential mechanisms. Doctoral thesis. University of British Columbia, Vancouver (B.C.), Canada. Opitz, S. (1996) Quantitative Models of Trophic Interactions in Caribbean Coral Reefs. ICLARM, Manila, Philippines. 343 pp. Overholtzer, K. L. and Motta, P. J. (1999) Comparative resource use by juvenile parrotfishes in the Florida Keys. Marine Ecology Progress Series, 177: 177-187. Overholtzer, K. L. and Motta, P. J. (2000) Effects of mixed-species foraging groups on the feeding and aggression of juvenile parrotfishes. Environmental Biology of Fishes, 58 (3): 345-354. Palomares, M. L. and Pauly, D. (1989) A multiple regression model for predicting the food consumption of marine fish populations. Australian Journal of Marine and Freshwater Research, 40 (3): 259-273. Palomares, M. L. and Pauly, D. (1998) Predicting food consumption of fish populations as functions of mortality, food type, morphometrics, temperature and salinity. Marine and Freshwater Research, 49: 447-453. Pandolfi, J. M., Bradbury, R. H., Sala, E., Hughes, T. P., Bjorndal, K. A., Cooke, R. G., McArdle, D., McClenachan, L., Newman, M. J. H., Paredes, G., Warner, R. R. and Jackson, J. B. C. (2003) Global trajectories of the long-term decline of coral reef ecosystems. Science, 301 (5635): 955-958. Parrish, J. D. (1989) Fish communities of interacting shallow-water habitats in tropical oceanic regions. Marine Ecology Progress Series, 58 (1-2): 143-160. Parsons, J. (1962) The Green Turtle and Man, University of Florida Press, Gainesville (FL), USA. 126 pp. Pauly, D. (1980) On the interrelationships between natural mortality, growth parameters, and mean environmental temperature in 175 fish stocks. Journal du Conseil, 39 (2): 175-192. Polovina, J. J. (1984) Model of a coral-reef ecosystem. I. The Ecopath model and its application to French Frigate Shoals. Coral Reefs, 3 (1): 1-11. Randall, J. E. (1963) An analysis of the fish populations of artificial and natural reefs in the Virgin Islands. Caribbean Journal of Science, 3: 31-47. Randall, J. (1967) Food habits of reef fishes of the West Indies. Studies in Tropical Oceanography, 5: 665-847. Robblee, M. B. and Zieman, J. C. (1984) Diel variation in the fish fauna of a tropical seagrass feeding ground. Bulletin of Marine Science, 34 (3): 335-345. 126  Rothenberger, P., Blondeau, J., Cox, C., Curtis, S., Fisher, W. S., Garrison, V., Hillis-Starr, Z., Jeffrey, C. F. G., Kadison, E., Lundgren, I., Miller, W. J., Muller, E., Nemeth, R., Paterson, S., Rogers, C., Smith, T., Spitzack, A., Taylor, M., Toller, W., Wright, J., Wusinich-Mendez, D. and Waddell, J. (2008) The State of Coral Reef Ecosystems of the Commonwealth of the U.S. Virgin Islands. In The state of coral reef ecosystems of the United States and Pacific Freely Associated States: 2008. (eds J. E. Waddell and A. M. Clarke), pp. 29-73. NOAA Technical Memorandum NOS NCCOS 11. NOAA/NCCOS Center for Coastal Monitoring and Assessment’s Biogeography Team, Silver Spring, MD (USA). Sagarin, R. D. and Crowder, L. B. (2009) Breaking through the crisis in marine conservation and management: Insights from the philosophies of Ed Ricketts. Conservation Biology, 23 (1): 24-30. Seminoff, J. A. (2004) 2004 global status assessment: Green turtle (Chelonia mydas). IUCN Marine Turtle Specialist Group, SSC, Red List Programme, Gainesville (FL), USA. 71 pp. Shulman, M. J. (1985a) Coral reef fish assemblages: Intraspecific and interspecific competition for shelter sites. Environmental Biology of Fishes, 13 (2): 81-92. Shulman, M. J. (1985b) Recruitment of coral reef fishes: Effects of distribution of predators and shelter. Ecology, 66 (3): 1056-1066. Shulman, M. J. and Ogden, J. C. (1987) What controls tropical reef fish populations - Recruitment or benthic mortality - an example in the Caribbean reef fish Haemulon flavolineatum. Marine Ecology Progress Series, 39 (3): 233-242. Sweatman, H. and Robertson, D. R. (1994) Grazing halos and predation on juvenile Caribbean surgeonfishes. Marine Ecology Progress Series, 111 (1-2): 1-6. Toscano, B. J., Fodrie, F. J., Madsen, S. L. and Powers, S. P. (2010) Multiple prey effects: Agonistic behaviors between prey species enhances consumption by their shared predator. Journal of Experimental Marine Biology and Ecology: doi:10.1016/j.jembe.2010.1001.1001. Tribble, G. W. (1981) Reef-based herbivores and the distribution of two seagrasses (Syringodium filiforme and Thalassia testudinum) in the San Blas Islands (Western Caribbean). Marine Biology, 65 (3): 277-281. Valentine, J. F. and Duffy, J. E. (2006) The central role of grazing in seagrass ecology. In Seagrasses: Biology, Ecology and Conservation. (eds A. W. D. Larkum, R. J. Orth and C. M. Duarte), pp. 463- 501. Springer, Dordrecht, The Netherlands. Valiela, I., Bowen, J. L. and York, J. K. (2001) Mangrove forests: One of the world's threatened major tropical environments. BioScience, 51 (10): 807-815. 127  Verweij, M. C., Nagelkerken, I., de Graaff, D., Peeters, M., Bakker, E. J. and van der Velde, G. (2006) Structure, food and shade attract juvenile coral reef fish to mangrove and seagrass habitats: a field experiment. Marine Ecology Progress Series, 306: 257-268. Walters, C. and Kitchell, J. F. (2001) Cultivation/depensation effects on juvenile survival and recruitment: implications for the theory of fishing. Canadian Journal of Fisheries and Aquatic Sciences, 58 (1): 39-50. Walters, C. J. and Martell, S. J. D. (2004a) Parameterization of ecosystem models. In Fisheries Ecology and Management. (eds C. J. Walters and S. J. D. Martell), pp. 286-309. Princeton University Press, Princeton (NJ), USA. Walters, C. J. and Martell, S. J. D. (2004b) Foraging arena theory (II). In Fisheries Ecology and Management. (eds C. J. Walters and S. J. D. Martell), pp. 229-255. Princeton University Press Princeton (NJ), USA. Walters, C., Christensen, V. and Pauly, D. (1997) Structuring dynamic models of exploited ecosystems from trophic mass-balance assessments. Reviews in Fish Biology and Fisheries, 7 (2): 139-172. Walters, C., Pauly, D., Christensen, V. and Kitchell, J. F. (2000) Representing density dependent consequences of life history strategies in aquatic ecosystems: Ecosim II. Ecosystems, 3 (1): 70-83. Williams, S. (1988) Thalassia testudinum productivity and grazing by green turtles in a highly disturbed seagrass bed. Marine Biology, 98: 447-455. Wing, S. R. and Wing, E. S. (2001) Prehistoric fisheries in the Caribbean. Coral Reefs, 20 (1): 1-8. Young, D. K. and Young, M. W. (1978) Regulation of species densities of seagrass-associated macrobenthos - Evidence from field experiments in Indian River estuary, Florida. Journal of Marine Research, 36 (4): 569-593. Zieman, J. C., Iverson, R. L. and Ogden, J. C. (1984) Herbivory effects on Thalassia testudinum leaf growth and nitrogen content. Marine Ecology Progress Series, 15 (1-2): 151-158.   128  4.  Carrying capacity of green sea turtles (Chelonia mydas) at Kaloko-Honokōhau, Hawai‘i, and their role in reef resilience3  4.1 Introduction Grazing by macroherbivores is one of the major processes structuring benthic coral reef communities. Studies highlighting the role of herbivores in promoting reef resilience and recovery to coral-dominated states, where disturbance has led to increased algal biomass (e.g., Bellwood et al. 2004), have focused almost exclusively on fish and sea urchins (Hay 1984a, Carreiro-Silva and McClanahan 2001, Mumby et al. 2006a, Paddack et al. 2006, Albert et al. 2008). Numerous Caribbean reefs have transitioned from coral to algal-dominated states (Gardner et al. 2003) following either a dramatic reduction in fish stocks that limited the distribution, abundance, and          3 A version of this chapter has been accepted for publication. Wabnitz, C.C.C., Balazs, G., Beavers, S., Bjorndal, K.A., Bolten, A.B., Christensen, V., Hargrove, S. and Pauly, D. Carrying capacity of green sea turtles (Chelonia mydas) at Kaloko-Honokōhau, Hawai‘i, and their role in reef resilience. Marine Ecology Progress Series (in press) 129  production of algae (Ogden and Lobel 1978, Hay 1981, Lewis 1986), and/or the region-wide loss to disease of an important echinoid herbivore (Diadema antillarum) in systems that had previously seen a shift from fish-dominated to echinoid-dominated herbivory (Lessios 1988, Mumby et al. 2006b). At some locations, increased anthropogenic nutrient loading has further increased the abundance of primary producers (Burkepile and Hay 2006, Littler and Littler 2007).  Few studies have focused on the role of green sea turtles (Chelonia mydas) in maintaining coral reef resilience. This is understandable in the Caribbean where green turtles mostly forage on seagrass (Bjorndal 1980, Mortimer 1981, Thayer et al. 1984) and thus, predominantly, have an indirect impact on the trophodynamics of reef systems (Valentine et al. 2002, Heck and Valentine 2006). In Hawai‘i, although the majority of reefs are not as severely impacted as those throughout the Caribbean region (Brainard et al. 2002, Waddell and Clarke 2008), several locations have shown increasing signs of stress as a result of mounting anthropogenic pressures on the coastal zone through development and runoff, tourism and recreation activities, and overfishing (Grigg 1994, Hunter and Evans 1995, Friedlander et al. 2008, Williams et al. 2008). Unlike in the Caribbean region, Hawaiian green turtles feed primarily on algal species that commonly occur on the reef (McCutcheon et al. 2003, McDermid et al. 2007, Arthur and Balazs 2008, Russell and Balazs 2009) and may therefore play a direct role in maintaining the resilience of coral ecosystems. The fact that green turtles feed on non-native algae, Acanthophora spp. and Hypnea spp. (Doty 1961, Russell 1992, Russell and Balazs 1994, Arthur and Balazs 2008, Russell and Balazs 2009) further highlights their contribution to the promotion of reef resilience.  Since turtle harvesting ended in the late 1970s (Witzell 1994), an approximately linear increase in abundance of nesting females has occurred at French Frigate Shoals (Balazs and Chaloupka 2004a, 2006), which accounts for > 90 % of all nesting within the Hawaiian Archipelago (Balazs 1980). This increase in abundance is interpreted as a recovery trend, because the Hawaiian nesting 130  population has been continuously monitored using dependable methodology for several decades (Balazs and Chaloupka 2004a, Chaloupka et al. 2008). Population trends at a number of foraging grounds that also have been subject to long-term monitoring seem to mirror this trajectory (Chaloupka and Balazs 2007).  An understanding of the role of turtles as grazers on reefs requires a process-oriented approach that assesses, preferably quantitatively, the relative contributions of all herbivore functional groups (i.e., fish, sea urchins and green turtles). Such an approach is currently lacking, though it may provide significant insights into the need for, and consequences of, improved turtle conservation and management.  Ecological modelling has developed ways to describe mathematically the complexity and nonlinear behavior of ecological systems. Ecopath with Ecosim (EwE) is a freely available, widely used, software for describing the structure of ecosystems and their food webs. It was recently named as one of the 10 major scientific breakthroughs in the 200-year history of the National Oceanic and Atmospheric Administration (see http://celebrating200years.noaa.gov/breakthroughs/welcome.html). Rather than providing outputs at the population level of biological organization, typical of many models, the EwE approach provides outputs at the ecosystem level, reflecting food-web linkages, energy cycling, and changes in biomass of each species group defined in the model (Christensen 2008). Although determining carrying capacity of a system has been highlighted as one of the uses of this software, few studies have investigated this aspect through the use of EwE (e.g., Christensen and Pauly (1998)).  I developed an Ecopath trophic model to investigate the role that green turtles play in the coral reef ecosystem of Kaloko-Honokōhau National Historical Park (hereafter referred to as Kaloko). Located on the west coast of Hawai‘i Island (the ‘Big Island’) (Figure 4.1), the park is characterised 131  by a healthy and relatively diverse coral habitat with little evidence of non-native species of macroalgae or diseased coral (Gibbs et al. 2007). Kaloko supports a low fish biomass yet high fish diversity (Parrish et al. 1990, Beets et al. 2006). The park also has a resident foraging population of immature green turtles that has been the subject of a mark-recapture study by the U.S. National Marine Fisheries Service (NMFS) and the U.S. National Park Service (NPS) since 1999. Three lines of evidence suggest that this foraging population has reached carrying capacity: (i) the significant increase in green turtle abundance within the archipelago over the last 10-20 years has been associated with a significant decrease in somatic growth rates (Balazs and Chaloupka 2004b), probably the result of density dependence (Bjorndal et al. 2000); (ii) field measurements of body volume and mass as an index of body condition show that turtles at foraging locations near Kaloko have lower body condition indices than green turtles at other sites in Hawai‘i (Kubis et al. 2008); and (iii) recent necropsy reports cite emaciation as a probable contributor in the death of a number of green turtles found stranded at foraging locations along the west coast of the Big Island (Work 2007, 2008a, 2008b).  The trophic model will also provide a ‘baseline’ of ecosystem state for Kaloko prior to major developments projected for areas around the park. Concern has been expressed over the future health of Kaloko’s coastal resources given proposed plans for development of lands adjacent to the south boundary of the park, including a 300 % expansion of a small-boat harbor, and construction of hotels, condominiums, and a light industrial park (Gibbs et al. 2007). Expected impacts include a reduction in groundwater flow, an important feature at Kaloko, with a concomitant increase in groundwater loads of sediment, nutrient, and chemical pollutants (Oki et al. 1999, Paytan et al. 2006, Johnson et al. 2008).    132  The goals of this study were, therefore, threefold: 1. to develop an ecosystem model of the marine portion of the park, synthesizing available data and describing the ecological structure and ecosystem processes of the reef system at Kaloko; 2. to ascertain whether Kaloko green turtles are at carrying capacity, by determining whether grazing by green turtles and other reef herbivores matches overall algal production; 3. to provide the management community with a tool that can simulate effects of increased urban development in the Kaloko area, as well as compare the outcomes of a range of potential management scenarios.  4.2  Material and methods 4.2.1  Study area Kaloko-Honokōhau National Historical Park was established in 1978 and spans 5.17 km2, of which 2.48 km2 is marine (Gibbs et al. 2007). The park is bordered to the south by the Honokōhau Small- Boat Harbor, and large scale industrial/commercial developments are located to the east (Figure 4.1). The terrestrial portion includes more than 180 anchialine (brackish) pools, two large ponds and wetlands modified for fish production by early Hawaiians, and a fish trap (Kaloko, ‘Aimakapā and ‘Ai’ōpio respectively; Figure 4.1). The coastal waters and reefs of Kaloko are within the West Hawai‘i Fisheries Management Area and are currently managed by the State of Hawai‘i. The legislated park boundary extends offshore for about 1,000 m at the widest point and to maximum depths of ca. 70 m (Parrish et al. 1990). Beyond this boundary the seabed drops off to depths > 180 m. Approximately 73 % of the marine section of the park is composed of hardbottom substrate. The remaining 27 % comprises unconsolidated sediment and artificial/historical features. Most of the hardbottom area has 10 % to < 50 % coral cover, and ~ 12 % has 50 % to 100 % coral cover (Gibbs et al. 2007). Benthic habitats are considered relatively healthy, with no signs of diseased corals or 133  non-native algal species (Marrack et al. 2009, Weijerman et al. 2009). Low sedimentation rates and the presence of relatively high coral cover in protected locations suggest that currently, the reef habitat in Kaloko is primarily controlled by natural wave-induced stresses (DeVerse 2006).  4.2.2  Modelling approach For model development, I used the Ecopath and Ecosim approach (EwE) (Christensen and Walters 2004, Christensen et al. 2005, http://www.ecopath.org). The Ecopath component of any EwE model provides a quantitative representation of an ecosystem for a defined time period as a snapshot of the resources in an ecosystem and their interactions, represented by trophically linked mass- balanced biomass ‘pools’ (Polovina 1984, Christensen and Pauly 1992). The biomass pools, hereafter referred to as functional groups, consist of a single species, or species groups representing ecological guilds. The idea behind the mass-balance approach is that “at any time within the system, and within the elements of that system, the amounts of matter that flow in must balance the amount that goes out, plus the change in biomass” (Pauly and Christensen 2002 p. 215). Ecopath, therefore, operates under two main assumptions: (a) That biological production within a functional group equals the sum of mortalities, i.e., on an annual basis, biomass and energy in an ecosystem are conserved (Walters et al. 1997, Walters and Martell 2004). This relationship can be expressed as follows:  )1()()/()/( 1 iiiiiiijj j jii EEBPBBAEYDCBQBBPB −⋅++++⋅⋅=⋅ ∑ =                     (1) where Bi and Bj are biomasses of prey (i) and predator (j) respectively; (P/B)i is the production to biomass ratio, equivalent to total mortality (Z) under most circumstances (Allen 1971); (Q/B)j is the food consumption per unit biomass of (j); DCij is the fraction of prey (i) in the average diet of predator (j); Yi is the total fishery catch rate of group (i); Ei is 134  the net migration rate (emigration - immigration); BAi is the biomass accumulation rate of group (i); and EEi is the ecotrophic efficiency, defined as the fraction of production that is consumed within the system or is removed by fishers. (b) That consumption within a group equals the sum of production, respiration, and unassimilated foods. This relationship can be expressed as follows:  GSBQBPTMQGSBPBBQB ⋅+⋅−−⋅−+⋅=⋅ )/()1()1()/()/(                             (2) where GS is the proportion of food unassimilated and TM is the trophic mode expressing the degree of heterotrophy of groups represented within the system - with 0 representing autotrophs, 1 heterotrophs, and intermediate values facultative consumers.  Ecopath then uses a set of algorithms to solve simultaneously n linear equations of the form in equation 1, where n is the number of functional groups. For each functional group, three of the basic parameters: Bi, (P/B)i, (Q/B)i  or EEi must be known, in addition to the fisheries yield (Yi), and the diet composition. Units of the model are expressed in t·km-2·year-1 wet weight organic matter for flows, and t·km-2 wet weight for biomasses. P/B and Q/B have the dimension year-1. For a review of Ecopath with Ecosim’s capabilities and limitations see Christensen and Walters (2004) and Plaganyi (2004, 2007).  To balance the model, changes were first made to the diet matrix, as diet compositions represent only snapshots of the feeding habits of individual species and are likely to be relatively variable based on location and time periods of data collection. The model required only minor adjustments and was considered balanced when: (a) the balanced model produced realistic ecotrophic efficiencies (EE < 1); and (b) values of the production to consumption ratio (P/Q) for functional groups ranged between 0.05 and 0.35, with the exception of groups with fast growth rates (higher ratios), and top predators with lower values (Christensen et al. 2005). 135   4.2.3  Model parameters and functional groups The model represented an annual average situation of ecosystem conditions in the marine portion of Kaloko-Honokōhau National Historical Park in 2005. I defined a total of 26 groups spanning the main trophic components of the ecosystem (including detritus) (Table 4.1, Figure 4.2). Eight groups were made up of fish, and represented the aggregation, based on ecological and biological similarities (e.g., diet, size, habitat, mortality), of 106 fish species recorded during underwater visual census studies (Beets et al. 2006) (see Appendix 6).  Biomass estimates for individual species were based on values from field studies, local expert opinion or from the literature. For species with data reported only from selected habitats, biomass values were extrapolated to the entire park by calculating an area-weighted biomass for each species relative to the proportion that each benthic habitat category covered within park waters (the latter based on Gibbs et al. 2007). For instances where P/B was equal to only natural mortality, estimates were taken directly from the literature or derived using the empirical formula of Pauly (1980). For exploited species, fishing mortality values were based on Friedlander and Parrish (1997) (see Fisheries section below). Where possible, the consumption of each group was obtained through field studies; otherwise consumption was estimated from empirical equations such as those available in Fishbase (www.fishbase.org) for all finfish. The diet matrix was constructed using data from field studies in Hawai‘i, preferentially the Kona Coast. Where no such data were available, the matrix was complemented with data obtained from the literature for the same species in similar ecosystems. Details of data sources and estimation methods for all parameters can be found in Appendix 7.  136  Fisheries Kaloko’s waters were designated as a Fish Replenishment Area (FRA) on December 31, 1999, effectively banning collection of fish for the aquarium trade in park waters, in response to declines of species targeted by collectors (Tissot et al. 2004). Biomass of targeted species may have responded to the closure, and because fishing pressure for ornamentals has been removed, P/B was set to natural mortality (M) for all formerly collected species (E. Brown, U.S. National Park Service, pers. comm.).  The park is a popular location for subsistence fishing and shoreline gathering, traditional activities that are permitted as long as they are consistent with state law and park mandates (i.e., with legal fishing gear for personal consumption) (DeVerse 2006). Harvesting is done primarily from shore using several methods, such as throw nets, spear, and pole fishing. Gill, or ‘lay’ netting, a serious threat to marine resources, including marine mammals and sea turtles, was restricted within park waters in August 2005 to locally constructed, handmade nets of natural fibres. The state of Hawai‘i does not have recreational and subsistence permitting or reporting requirements (Friedlander and Parrish 1997), despite surveys in the late 1980s that indicated that 19 - 35 % of residents fish (Smith 1993), and recent studies concluding that these fisheries dominate the catch of coral reef species (Zeller et al. 2005, Williams et al. 2008). Although several studies conducted throughout the Hawaiian Archipelago have assessed the importance of fishing impacts on coral reef fish assemblages (e.g., Tissot et al. 2004, Williams et al. 2008, Tissot et al. 2009, DAR unpublished data, in Weijerman et al. 2009), few data exist to quantify fishing mortality along the western shores of the Big Island. In the absence of catch or effort data specific to Kaloko, I assumed catch data/standing stock proportions in the park to be comparable to those in Hanalei Bay, Kauai (Friedlander and Parrish 1997), adjusting values to reflect species known to be targeted at Kaloko. As no data are available on macroinvertebrates in Kaloko (with the exception of sea urchins), nor information 137  regarding whether fishing occurs on those groups in the park, no catch was allocated to the benthic invertebrate group.  I divided existing fisheries into two ‘fleets’, one targeting ‘sharks and jacks’ specifically, and a recreational/subsistence fishery representing fishers operating mainly from shore using pole, spear, and line, and targeting small reef fish. I calculated the ratio of catch to standing stock for fish families targeted by fishers from Friedlander and Parrish (1997) and doubled them, given that fishing pressure is likely to be higher now than it was in 1997 (DAR unpublished data 2006 in Weijerman et al. 2009). I then applied these indices to fish families harvested in Kaloko (E. Brown, U.S. National Park Service, unpublished data 2008). For species groups targeted in Kaloko, but for which data were not available from the Friedlander and Parrish (1997) study, I applied the same value as for fish families in the same functional group. Fishing mortality for individual species/functional groups was calculated using F = catch / biomass.  Model analysis, indices, and uncertainty around input data I quantified total trophic flows within the ecosystem in terms of consumption, production, respiration, exports and imports, and flow to detritus (t·km-2·year-1). I then used ecological indices as obtained through Ecological Network Analysis (ENA) to evaluate the food web (Wulff et al. 1989, Christensen and Pauly 1993). ENA is a modelling technique used for understanding the structure and flow of material between components of an ecosystem (Ulanowicz 1997). As integrated in Ecopath, ENA allows for the calculation of ecosystem macro descriptors that quantify trophic structure, organic matter recycling, and ecosystem size and organization. These descriptors include total system Throughput (T), Ascendency (A), development Capacity (C), and the Relative Overhead (O/C). Throughput describes the size of a system and represents a measure of its metabolism (Christensen and Pauly 1993). Ascendency integrates both size and organization of a 138  given system (Christensen 1995). The development Capacity quantifies the upper limit to Ascendency while the system’s Overhead (O) is complementary to the Ascendency and measures the level of ‘redundancy’ of particular links (Heymans 2003). The O/C ratio was proposed by Heymans (2003) as an index of the resilience of the system (i.e., an index of the system’s ability to withstand disturbance (Ulanowicz 1997)). ENA further allows the derivation of Transfer Efficiencies (TE), summarizing the proportion of consumption that is passed up a food web. The TE is the ratio between the production of a given trophic level and the preceding trophic level (Pauly and Christensen 1995). Finally, I ran the Mixed Trophic Impact (MTI) analysis (Ulanowicz and Puccia 1990), which allows the theoretical quantification of direct and indirect interactions among functional groups within the food network. In other words, this analysis provides a first-order quantification of the beneficial and negative impacts of one group on another.  Any ecosystem model realization requires acknowledging the large amount of data required in its development, and the difficulty in quantifying the flows between the food web’s individual functional groups. Group dynamics can be verified by fitting model data to actual population trends over time. Such time series data do not exist for Kaloko. Given the primary goal of determining whether green turtles are at carrying capacity in Kaloko, I focused on sources of uncertainties associated with estimates of green turtle biomass, P/B, and Q/B; the proportion of sea urchins’ diet that was derived from turf growing on the nearshore lava bench (turfLB - green turtles’ primary forage base); and variability associated with turfLB biomass and P/B. To address these uncertainties, probability distributions for all input parameters were entered through the ‘pedigree’ (Funtowicz and Ravetz 1990) function of Ecopath (Pauly et al. 2000). Using a Monte Carlo resampling routine, the ‘Ecoranger’ module of Ecopath subsequently draws random input variables from within the confidence intervals defined for each parameter type in the pedigree tables and uses these as prior probability distributions for all input data. This approach leads to a large number of model realizations that are evaluated for their conformity to user-defined criteria as well as physiological 139  and mass-balance constraints. The results include probability distributions for the estimated parameters along with distributions of parameters in the accepted model realizations. This routine can be run for the model overall, as pedigrees are associated with all input parameters. However, I focused on uncertainty associated with input parameters that would affect the objective of estimating carrying capacity of green turtles.  4.3  Results Trophic parameters and the diet matrix for the 26 groups of the final balanced Ecopath model for Kaloko are presented in Table 4.1 and Table 4.2 respectively. Trophic flows between all functional groups are depicted in Figure 4.2. Sea urchins (Tripneustes gratilla, Echinometra mathaei, Heterocentrotus mammilatus, and Echinothrix spp.) accounted for the largest proportion of total living biomass in the system (40 %) (Figure 4.3). Green turtles and reef fish groups, including ‘sharks and jacks’, only represented 0.2 % and 5.5 % of total biomass respectively (Figure 4.3). Reef fish groups were dominated by herbivorous and species characterised as Mobile Invertebrate Feeders (‘MIF’). Not surprisingly, consumption by sea urchins had the biggest impact (45 %) on available resources at Kaloko; whereas fish accounted for 14.4 % and green turtles for 0.2 % of total consumption. Total fisheries catches represented < 1 % of the total fish biomass (Table 4.1). ‘Sharks and jacks’ were caught in large quantity compared with their relatively low biomass in the assemblage. Mean trophic level of the catch was 2.59 (Table 4.3).  Ecotrophic efficiency (EE) values (the proportion of a functional group’s production used within the system) were lowest for some of the highest trophic levels (e.g., ‘spinner dolphins’, ‘monk seals’, and ‘sea birds’), as well as for ‘crown-of-thorns starfish’, ‘hawksbill turtles’, ‘green turtles’, ‘sea urchins’, and ‘corals’ (Table 4.1). For the higher trophic level groups, this valuation is based on these species 140  deriving a significant proportion of their food from outside park waters. They were included chiefly to ensure that the model was representative of the system at Kaloko (i.e., to acknowledge that these species occur in the park, and may at some point in the future suffer from indirect effects of park development, even if they do not feed exclusively or primarily in park waters). For the five other groups, the low EE values are a result of low predation and fishing pressure being exerted on these species. Fish groups, overall, also had relatively low EE values, with higher values registered by those species pools that were subject to higher fishing mortality. This valuation suggests that the system generates a reasonable amount of surplus secondary production. Indeed, the largest component of the mortality coefficients within the system was due to predation mortalities, with the exception of ‘sharks and jacks’ for which the fishing mortality rate was greater than predation mortality. In contrast, most of the production by the macro- and turf algal groups, ‘zooplankton’, ‘benthic invertebrates’, and ‘phytoplankton’ was accounted for through consumption by other trophic groups within the model (EE > 90 %). In other words, sea urchins, herbivorous fish, and green turtles maintained all algae at Kaloko in a cropped state. Reef building corals, octocorals, benthic invertebrates, and zooplanktivorous fish were the main predators of zooplankton, while zooplankton consumed most of the phytoplankton production in the system. Benthic invertebrates were preyed on mostly by MIF reef fish and species within the benthic invertebrate group itself.  Throughput (total flows) values for each functional group highlighted the importance of turf algae, zooplankton, and phytoplankton, as well as sea urchins in system structure. The high EEs attained for all algal groups indicated that the system was at carrying capacity with respect to grazing, including for green turtles. The primary producers’ high EEs were also reflected in the low production/respiration ratio of 1.12 for the system overall (Table 4.3).  Focusing on mega- and macro-herbivores in the system only, the MTI routine highlighted the resource competition both within and between sea urchins, herbivorous fish, and green turtles 141  (Figure 4.4). Sea urchins were responsible for the largest between-group effect, impacting negatively on both fish and green turtles, and overall had the largest impact on algal and detrital resources. Given green turtles’ feeding preference on turfLB, they were the group most impacted by changes in turfLB (Figure 4.4).  When accounting for the uncertainty around parameters affecting consumption of turfLB, EE values of turfLB in successfully balanced model-runs ranged between 0.416 and 0.998.  The mean transfer efficiency in the ecosystem as a whole was 4.6 %, with 4.5 % for flows originating from primary producers and 4.7 % from the detritus. This low efficiency was due to consumption dominating total system throughput for the lower trophic levels, whereas respiration and flow to detritus dominated the higher trophic levels in the model (Figure 4.5). This result is to be expected in a system with low biomass at the highest trophic levels. Some upper trophic level species, such as trevally (also locally known as ‘ulua’ or ‘jacks’), are highly vagile. As a consequence, although not being heavily fished for in park waters, their abundance may still be depressed because of high fishing pressure along the remainder of the Kona Coast. Ascendancy was 31.5 % of capacity and relative overhead 68.5 % (Table 4.3), 46 % of which were attributable to internal flows, indicating that the system contains a number of ‘redundant’ trophic linkages. These observations are consistent with a system exhibiting relatively high resilience to perturbation with respect to energy flows, or a high system stability sensu Odum (1971).  Net primary productivity was 3,895 t·km-2·year-1 and was accompanied by a high flow to the detritus pool (Table 4.3). Turf algae (including turfLB) and phytoplankton accounted for 57 % and 24 % of total production in the system, or 65 % and 27 % of total primary production. Sea urchins (38 %) were the major contributor to the detrital pool, followed by zooplankton (28 %), while the combined fish groups (mainly herbivores) contributed 11 % of all flows to the detritus. The ratio between 142  primary production and respiration (PP/R) of Kaloko indicates that the system was at a low developmental stage, sensu Odum (1969, 1971). This finding was corroborated by the low Finn’s cycling index, showing that only a small fraction of the throughput (including detritus) was recycled (6 %) (Table 4.3). The proportion of flows originating from the detritus was 0.27 %.  The overall pedigree (i.e., ‘quality’) index of the model (0.539) was high in comparison to 50 other models for which pedigree values ranged between 0.164 and 0.676 (Morissette 2007).  4.4  Discussion 4.4.1  Description of the Kaloko-Honokōhau system Quantitative descriptions of the flux of matter and energy can provide significant insights into the fundamental structure of ecosystems. The Kaloko system is dominated by primary production (PP), 27 % of which is contributed by phytoplankton and 65 % by algal turfs, which is slightly lower than other published estimates for primary producers in tropical systems (e.g., Wanders 1976, Adey and Steneck 1985, Adey and Goertemiller 1987). The trophic networks at Kaloko were dominated by grazing, with herbivores accounting for 43 % of all living biomass within the system (Figure 4.3), of which 93 % were sea urchins. This dominance of grazing behaviour was further highlighted by the high EEs achieved for the main PP functional groups. Results from rapid assessment surveys throughout the state confirm that herbivorous fish, and sea urchins in particular, are commonly found on reefs in Hawai‘i (Rodgers et al. 2004). In contrast, relatively low EEs of higher trophic level functional groups indicate that the foraging activities of herbivores are not limited by predation pressure, as demonstrated by the analysis of total system throughput (Figure 4.5). Indeed, few predators commonly occur in the park, and fewer still have been observed feeding in the park. Therefore, as highlighted by the model’s average transfer efficiency of 4.6 %, which is lower than the 143  mean of 10 % obtained for various other ecosystems (Christensen and Pauly 1993), only a small proportion of production is being transferred up the food chain. Comparison between two sets of underwater visual census data (E. Brown and NPS, unpublished data 2007) showed a fivefold increase in piscivore biomass between 2005 and 2007. This inter annual difference may have been a result of real differences between the two datasets, possibly reflecting the restrictions on gillnetting in park waters implemented in August 2005. Differences in species biomass and composition may also reflect either variability in survey accuracy for species displaying vagile (e.g., Caranx melampygus) or cryptic behaviour (e.g., Gymnothorax flavimarginatus, Gymnothorax meleagris, Cephalopholis argus), or seasonal changes (Friedlander and Parrish 1998, Vitousek et al. 2009), as the 2005 and 2007 datasets were collected in April and October, respectively. Future surveys should, therefore, focus on good intra annual coverage to represent seasonal variation.  Phytoplankton and zooplankton had some of the highest EEs. The waters around Hawai‘i are generally oligotrophic (Bienfang et al. 2009) and, consequently, low biomass of phytoplankton and zooplankton groups is to be expected. Reef environments along the west coast of the Big Island are close to the deep slope of the Pacific and subject to strong wave action and currents along the shore (Presto et al. 2007). Therefore, I assumed that zooplanktivores on the reef derive a substantial portion of their diet from open-ocean plankton. I further assumed that ocean plankton also contributes to the energy intake of benthic invertebrates and other functional groups that feed on plankton. Experiments should be conducted to ascertain the proportion of ocean plankton that contributes to the diet of reef organisms at Kaloko or other similar systems in Hawai‘i.  At 42.54 t·km-2, the model’s benthic invertebrate biomass was similar to estimates provided for models in Raja Ampat, Indonesia (51.68 t·km-2) (Ainsworth et al., 2007), and the Great Barrier Reef, Australia (61.41 t·km-2) (Tudman, 2001), but substantially lower than estimates for Moorea, French Polynesia (barrier reefs = 198.26 t·km-2 and fringing reefs = 322.8 t·km-2) (Arias-González 1994, 144  Arias-González et al. 1997). Few studies have investigated the composition of benthic invertebrates on reefs, yet they are an important prey for a variety reef fish (e.g., Hobson 1974) and can represent an important fishery (Friedlander and Parrish 1997). Given that benthic invertebrates accounted for 6 % of total system biomass, species composition and their relative contribution to total group biomass of the benthic invertebrate community of Kaloko should be determined.  Sea urchin densities at Kaloko (~ 5 individual per m-2 for E. mathaei and < 1 individual per m-2 for all other species) were comparable to sea urchin densities recorded on reefs throughout the State of Hawai‘i (Rodgers et al. 2004), and similar to, or lower than, those reported from other reef locations (e.g., 5 to 6.8 individuals per m-2 (chiefly E. mathaei and T. gratilla) at La Réunion (Naim et al. 1997) and 2 to 4 individuals per m-2 (Echinometra) in Fiji (Appana and Vuki 2006)). Current sea urchin densities at Kaloko are also comparable to those recorded in the 1970s (Ebert 1971). These similarities suggest that high sea urchin densities are natural and not a release from predation pressure due to increased fishing pressure in the last 30 years, as noted at a number of locations along the coast of Kenya (Muthiga and McClanahan 1987, McClanahan 1998).  4.4.2  Turtles at carrying capacity Green turtles are at carrying capacity at Kaloko based on (a) their biomass estimates and consumption rates; (b) estimates of turf algae biomass growing on the lava bench (turfLB) as well as the primary production rates of these algae; and (c) the high degree to which sea urchins feed on the green turtles’ main food resource. Balanced model outputs under Ecoranger, taking into account the uncertainty surrounding these input parameters, showed that the EE of turfLB could range between 0.416 and 0.998. Specifically, it reflects the wide range in published biomass and mass-specific productivity rates for turf algae. For example, biomass estimates were found to vary between 230 g·m-2 (T. Sauvage, pers. comm. 2008) and 600 g·m-2 (Smith et al. 2001) in Hawai‘i alone, 78 g·m-2 in 145  the Solomon Islands (Albert et al. 2008), and 850 g·m-2 in the Caribbean (Bruggemann et al. 1994). Even greater ranges in productivity have been recorded for the different algal groups that constitute ‘turf’, particularly in relation to depth and seasonal cycles (Payri 2000). An evaluation of algal turf productivity and consumption rates by green turtles at Kaloko would help refine the estimates presented here.  Results from the MTI analysis highlighted the strong competition for the same resource between sea urchins, fish, and green turtles. Grazing pressure exerted by herbivorous fish does not occur uniformly on a reef. Grazing may be reduced in places where reef structure affords little refuge from predation (Lewis 1986). However, this may not be a significant factor at Kaloko where overall predation pressure is limited due to the low abundance of piscivores. Foraging efficiency may also be reduced on shallower portions of the reef because of high wave intensity. It was, therefore, assumed that fish at Kaloko chiefly grazed on reef turf algae rather than turfLB specifically (Table 4.2).  Recent evidence from the South Pacific demonstrates that few of the fish species previously classified as ‘herbivores’ are indeed exclusively herbivorous (Choat et al. 2002, Cvitanovic et al. 2007). Many appear to predominantly feed on detritus, only incidentally removing turf from the reef in the process (Crossman et al. 2001, 2005). Data on the extent of algae removed, but not consumed, are needed to correct the current assumption that all algal material removed by fish was ingested. MTI results reflected resource competition between turtles and fish primarily for macroalgae, which appeared ‘magnified’ as macroalgae biomass overall in the park is low. As sea urchins occur in the shallows (M. Weijerman, pers. comm. 2008, L. Marrack, U.S. National Park Service, pers. comm. 2008), probably emerging at night to feed (Mills et al. 2000, Vaitilingon et al. 2003), a small proportion of turfLB was included in the echinoids’ diet matrix (Table 4.2). Quantification in the field of this overlap and the extent to which relative densities among the three herbivore groups may 146  change with time would help refine carrying capacity estimates for green turtles. Clarification of the extent to which sea urchins feed on turfLB also has implications for their role in maintaining low algal biomass levels in the face of increased nutrient input associated with urban development activities around the park (see below).  Observations of Kaloko turtles appearing unusually ‘skinny’ (Kubis et al. 2008) and exhibiting low growth rates (Balazs and Chaloupka 2004b), compared to turtles at certain other Hawaiian foraging grounds, support input parameters within the range used in my model (Table 4.1). A key difference among individual foraging sites that may help explain variations in observed growth rates is that green turtles at Kaloko forage on turf, the dominant algal group on the central west coast of Hawai‘i, whereas other foraging grounds throughout the Hawaiian Archipelago have macroalgae available as forage. Recent observations have indicated a behavioural shift in the turtles’ foraging patterns. Fewer turtles are utilizing the shallow reef area in the summer months than they did up to about 2006 (S. B., unpublished data 2009). More interactions with humans as a result of recent increases in visitor use may be partly responsible for greater movement of green turtles than previously observed (S. B., unpublished data 2009). Greater turtle movement possibly away from Kaloko further strengthens the case for carrying capacity having been reached at the park because of limited available forage.  Turf algae growing on the lava bench close to shore and intertidal area were included in the model as a separate functional group, as resident turtles at Kaloko have been seen to concentrate their foraging to this portion of the park. Such focused foraging behaviour may occur for three reasons: (i) By focusing their feeding activities on shallow portions of the reef, individual turtles exert less effort foraging as the lava bench supports a highly productive algal mat that can grow unrestricted by space competition with live coral. If turtles were to forage on deeper sections of the reef, for a similar intake of algae, animals would have to roam over wider areas and pick 147  algae from in and around coral heads, resulting in greater energy expenditure. Results, corroborated by local observations of turf algae throughout the park being closely cropped (F. Parrish, NOAA, pers. comm. 2008), indicate that the system is also close to carrying capacity for overall turf algae, thus making foraging by green turtles in deeper reef sections even less energy efficient. ‘True’ carrying capacity of turf algae on the reef may actually be realized at lower EE values, because turf algae laden with sediment does not allow for much of the algae to be available to grazers. (ii) Green turtles may prefer to forage on the lava bench because of the protection afforded by the shallows against predation by tiger sharks. Sharks are regularly sighted at Kaloko (DLNR 2001, Thompson 2005), and a log of necropsies performed on turtles at Kaloko indicate that shark attacks were responsible for the death of 4 turtles since 1992 (obviously attacks that resulted in the complete ingestion of turtles by sharks cannot be monitored) (G. B., unpublished data 2009). Thus, the distribution of turtles, and the algae they consume, may be affected by differential vulnerability of turtles to shark predation in different habitats (Heithaus et al. 2002, 2006). From a predation perspective, nutritionally profitable microhabitats at Kaloko also appear to be low risk. (iii) The repeatedly grazed turfLB may be more palatable to green turtles than algae growing on deeper sections of the reef. The area where turtles aggregate to feed is located close to ‘Ai’ōpio fishtrap, an area of considerable groundwater discharge (Johnson et al. 2008). As groundwater is the only conduit for nutrients into the coastal zone, the nutritional content of algae in this specific location may be significantly different from algae on the remainder of the reef.  4.4.3  Potential threats and impacts on natural resources as a result of urban development On many coral reefs located near human population centres, benthic community composition has shifted from coral-dominated to macroalgae-dominated (Hughes 1994, Hunter and Evans 1995, 148  McClanahan and Mangi 2001). These shifts, often termed ‘phase shifts’ (Done 1992), have been attributed to increased anthropogenic nutrient input (e.g., Lapointe 1997) and to reductions in the abundance of herbivores (e.g., Hay 1984b). Studies investigating changes in grazing intensity at sites that have undergone phase shifts, such as Kāne‘ohe Bay, Hawai‘i, have focused primarily on the role of herbivorous fish, and to a lesser extent mesoherbivores (Cheroske et al. 2000) and sea urchins (Hunter and Evans 1995, Stimson et al. 2001, Smith et al. 2008). As results clearly demonstrate, future studies need to explicitly consider green turtles because of their important role as herbivores. Their consumption of non-native algae, including G. salicornia (Russell and Balazs 2009), abundant at a number of locations across Hawai‘i, including Kāne‘ohe Bay, further highlights their role in promoting resilience.  Ongoing and planned urban development activities and associated population growth around Kaloko will likely have significant and diverse impacts on the condition of the nearby reefs. Expected impacts include, but are not restricted to: (a) a reduction in groundwater discharge quality (i.e., increased nutrient loading) that could lead to increased macroalgal growth and subsequently a deterioration in the health of corals (e.g., Fabricius 2005, Smith and Smith 2006) particularly if combined with a reduction in the urchin population; and (b) an increase in fishing pressure and a resulting decrease in the herbivore fish population, which could lead to an increase in the abundance of macroalgae on the reef (Stimson et al. 2001, Williams et al. 2007). Based on this study’s findings, some key points emerge that should be taken into account if dynamic simulations of various ‘impact scenarios’ are run in the future.  Increased nutrient input and increased fishing pressure may lead to an increase in the proportion of macroalgal cover at Kaloko and potential forage for resident herbivores. In 2008, the NPS initiated a long-term, marine water quality monitoring program (quarterly sampling) that includes optical chlorophyll sampling. Additionally, plans are underway through a partnership between the NPS and 149  the U.S. Geological Survey to use stable isotope tracers to identify sources of nutrient inputs to park groundwater. The NPS, in partnership with the University of Hawai‘i, Hilo, recently also initiated a pilot fisheries harvest monitoring program to identify and quantify fishing pressure within the park; this should be a long-term initiative. Nutrient and fisheries’ catch time series in conjunction with regular monitoring of algal cover within the different habitats at Kaloko, and the herein developed Ecopath model, may help to determine the impacts of such changes at the ecosystem level. Based on model findings, I would expect the following responses to two scenarios:  (i) increased nutrient delivery to the reef system without changes in fishing pressure Under conditions of nutrient enrichment and intense herbivory, algal productivity may be high while biomass remains low (Hatcher and Larkum 1983). Our results demonstrate that current rates of herbivory by sea urchins, fish, and turtles maintain all algal groups closely cropped and may therefore be able to consume increases in algal production. This grazing disturbance, in turn, limits the establishment of macroalgae (Lewis 1986, Williams et al. 2001, Paddack et al. 2006), thereby implying that the current system is relatively resilient to change (as indicated by Kaloko’s O/C ratio). Indeed, it is the synergy of these three herbivorous groups that makes Kaloko an interesting case study. The ‘redundancy’ in herbivory at Kaloko stands in contrast to areas of the Caribbean, where sea urchin biomass remains low 20 years after the region-wide loss of the key echinoid herbivore to disease (Mumby et al. 2006b) and where, high fishing pressure may have limited the ability of herbivorous fish to maintain a high proportion of reef area free of macroalgae. Some evidence suggests that even robust herbivorous fish populations may be capable of maintaining only 50 % to 65 % of the substratum in a cropped state (Williams et al. 2001). Although herbivorous fish may be capable of increasing their grazing rates and/or population sizes in response to higher algal production, evidence for this in the published literature is limited (McClanahan et al. 1999, Williams et al. 2001, Garpe et al. 2006). Moreover, herbivorous fish, with the exception of Naso spp. (Choat et al. 2002), all prefer turf algae over other algal resources, and frequently avoid macroalgae (Bellwood 150  and Choat 1990, Bruggemann et al. 1994). Turtles and sea urchins, however, commonly feed on macroalgae, thus highlighting their respective roles in upholding reef resilience (Nystrom 2006).  (ii) increased nutrient delivery to the reef system with increase in fishing pressure Herbivorous fish and sea urchins both play different and complementary roles in maintaining low algal cover on reefs in the face of disturbance (e.g., Morrison 1988). Under a scenario of increased fishing pressure, model parameterization indicates that sea urchin abundance may increase as a direct response to more abundant forage on the reef and reduced predation pressure due to harvesting of the urchins’ predators. Sea urchins, when abundant, can have a considerable impact on benthic ecology. Healthy populations of sea urchins have been associated with reductions in macroalgal cover and increased coral recruitment (Edmunds and Carpenter 2001, Carpenter and Edmunds 2006). Typically, sea urchins will consume a greater diversity of algae than herbivorous fish and, as such, they will be more effective than the latter at controlling algal growth in the event of increased nutrient levels. Alternatively, as some species of echinoids can remove a large amount of calcium carbonate (CaCO3) while foraging (Hutchings 1986, Carreiro-Silva and McClanahan 2001), an overabundance of sea urchins can lead to rapid erosion of the reef framework (McClanahan and Kurtis 1991, Bak 1994, Mapstone et al. 2007). For example, Echinothrix diadema and Echinometra mathaei typically erode 20 % of the calcium carbonate accreted (Carreiro-Silva and McClanahan 2001). However, the relative contributions of grazing and bioerosion are species-dependent. The predominant browsing by Tripneustes gratilla predominantly suggests that this species would have less impact on the reef framework (Mills et al. 2000).  Currently, sea urchin densities are greatest in the shallows, while reef fish dominate grazing processes at depth. Higher fishing pressure may reduce fish grazing intensity on deeper reef sections. In response to reduced predation and increased forage availability, sea urchins could increase in abundance at greater depths. Similarly, green turtles may also displace some of their foraging 151  activities to deeper reef areas to take advantage of increased algal biomass (i.e., profitable, high-risk microhabitat sensu Heithaus). Increases in algal biomass would further be exacerbated if mixing carries nutrients to those deeper sections of the reef. 4.4.4  Conclusions The model presented allowed an evaluation of the trophic linkages and flows on a Hawaiian reef. This study is unique in its detailed analysis of herbivory levels and focus on the role of herbivores to determine whether a given functional group, in this instance green turtles, is at carrying capacity. Integration of additional quantitative field data on biomass, distribution, consumption rates, and diet selection of the three herbivore groups would clarify the proportion and type of algae consumed by sea urchins and green turtles versus herbivorous fish on a Hawaiian reef and further elucidate their respective roles in, and contribution to, reef resilience (Nystrom and Folke 2001, Bellwood et al. 2004). 152  Table 4.1 - Trophic parameters for all functional groups of the balanced Kaloko-Honokōhau National Historical Park (Kaloko) model. Outputs from the balanced model are presented in italics. B = biomass; TL= Trophic level; P/B= Productivity biomass ratio; Q/B = consumption rate; EE= Ecotrophic efficiency; P/Q Production to consumption ratio or gross efficiency; P/R= Production to respiration ratio. MIF = Mobile Invertebrate Feeders; SIF = Sessile Invertebrate Feeders; Zoo = Zooplanktivorous fish; CCA = Crustose Coralline Algae    TL B P/B Q/B EE P/Q P/R Catches   Group name   (t/km²)  (/year)  (/year)    (t/km²/year) 1 Spinner dolphins  3.21  2.740 0.151 11.519 0.007 0.013 0.017 2 Monk seals  3.89  0.179 0.121 11.508 0.033 0.011 0.013 3 Sea birds  3.17  0.002 0.127 76.515 0.012 0.002 0.002 4 Rays  3.15  4.233 0.200 3.100 0.002 0.065 0.065 5 Sharks and jacks  3.53  0.070 1.058 5.100 0.453 0.207 0.35  0.030 6 Hawksbill sea turtles  3.18  0.054 0.100 3.500 0.066 0.029 0.029 7 Green sea turtles  2.00  1.591 0.109 6.764 0.039 0.016 0.021 8 Reef fish - piscivores  3.39  1.730 0.615 6.121 0.527 0.100 0.144  0.003 9 Reef fish - herbivores  2.02  20.335 1.400 27.149 0.205 0.052 0.069  0.162 10 Reef fish - corallivores  2.60  0.542 2.100 12.918 0.547 0.163 0.255 11 Reef fish - detritivores  2.00  2.260 1.900 32.272 0.282 0.059 0.079  0.018 12 Reef fish - MIF  3.13  9.761 0.950 8.108 0.394 0.117 0.172  0.130 13 Reef fish - SIF  2.84  0.544 1.700 9.581 0.224 0.177 0.285 14 Reef fish - Zoo  2.85  3.046 1.450 13.378 0.585 0.108 0.157  0.004 15 Urchins  2.00  280.000 0.484 8.547 0.056 0.057 0.076 16 Crown-of-thorns  2.59  0.117 0.411 9.000 0.007 0.046 0.061 17 Benthic Invts  2.18  42.538 2.910 15.250 0.950 0.191 0.313 18 Corals  1.58  130.000 0.140 2.100 0.594 0.067 0.075 19 Octocoral  2.07  2.900 0.200 4.630 0.484 0.043 0.054 20 Macroalgae  1.00  22.691 9.824 - 0.925 - - 21 CCA  1.00  37.818 1.770 - 0.358 - - 22 Turf algae  1.00  128.780 19.000 - 0.942 - - 23 Turf algae_lava bench  1.00  3.065 25.000 - 0.921 - - 24 Zooplankton  2.02  1.240 219.000 949.000 0.979 0.231 0.625 25 Phytoplankton  1.00  3.290 325.458 - 0.984 - - 26 Detritus  1.00  100.000 - - 0.694 - - 153  Table 4.2 - Diet composition matrix for the functional groups in the balanced model for Kaloko. Shaded cells indicate diet proportion < 0.001     Predator  Prey 1 2 3 4 5 6 7 8 9 1 Spinner dolphins  -  -  -  - 0.008  -   -   -  - 2 Monk seals  -  -  -  -  0.002  -   -   -  - 3 Sea birds  -  -  -  -    -   -   -  - 4 Rays  -  -  -  -    -   -   -  - 5 Sharks and jacks  -  -  -  - 0.010  -   -   -  - 6 Hawksbill sea turtles  -  -  -  - 0.001  -   -   -  - 7 Green sea turtles  -  -  -  -  0.019  -   -   -  - 8 Reef fish - piscivores  -  0.200  -  - 0.110  -   -  0.010  - 9 Reef fish - herbivores  - 0.050  -  - 0.206  -   -  0.400   - 10 Reef fish - corallivores  -  -  -  - 0.040  -   -  0.050  - 11 Reef fish - detritivores  -  -  -  - 0.085  -   -  0.100  - 12 Reef fish - MIF  - 0.080  -  - 0.120  -   -  0.200  - 13 Reef fish - SIF  -  -  -  - 0.062  -   -  0.010  - 14 Reef fish - Zoo     0.010  - 0.010  - 0.051  -   -  0.100  - 15 Urchins  -  -   - 0.010  -   -   -  - 16 Crown-of-thorns  -  -   - 0.001  -   -   -  - 17 Benthic Invertebrates       0.190 0.107 0.578  - 0.070 0.894   -  0.129 0.004 18 Corals  -  -  -  -  - 0.001   -   0.001 19 Octocoral  -  -  -  -  -  -   -   -  - 20 Macroalgae  -  -  -  -  -  0.001   - 0.050 21 Crustose Coralline Algae (CCA)  -  -  -  -  -  -   -   -  - 22 Turf algae  -  - 0.010  -  -  -   -   - 0.782 23 Turf algae_lava bench  -  -  -  -  -  -  0.998   -  - 24 Zooplankton  -  - 0.001 0.010 0.050 0.005  0.001    0.013 25 Phytoplankton  -  - 0.001  -  -  -   -   -  - 26 Detritus  -  - 0.001  - 0.005  -   -   - 0.150 27 Import       0.800 0.563 0.399 0.990 0.150 0.100   -   -  - 154  Table 4.2 - cont.     Predator  Prey 10 11 12 13 14 15 16 17 18 19 24 1 Spinner dolphins  -  -  -  -  -  -  -   -   -  -  - 2 Monk seals  -  -  -  -  -  -  -   -   -  -  - 3 Sea birds  -  -  -  -  -  -  -   -   -  -  - 4 Rays  -  -  -  -  -  -  -   -   -  -  - 5 Sharks and jacks  -  -  -  -  -  -  -   -   -  -  - 6 Hawksbill sea turtles  -  -  -  -  -  -  -   -   -  -  - 7 Green sea turtles  -  -  -  -  -  -  -   -   -  -  - 8 Reef fish - piscivores  -  -  -  -  -  -  -    -  -  - 9 Reef fish - herbivores  -  - 0.010  -  -  -  -     -  -  - 10 Reef fish - corallivores  -  - 0.001  -  -  -  -    -  -  - 11 Reef fish - detritivores  -  -    -  -  -  -     -  -  - 12 Reef fish - MIF  -  - 0.010  - 0.010  -  -    -  -  - 13 Reef fish - SIF  -  - 0.001  -   -  -    -  -  - 14 Reef fish - Zoo  -  - 0.010  - 0.009  -  -     -  -  - 15 Urchins  -  - 0.085 0.032 -  -  -    -  -  - 16 Crown-of-thorns  -  -  -  -  -  -  -   -   -  -  - 17 Benthic Invertebrates       0.074  - 0.814 0.420 0.100  -  -  0.046   -  -  - 18 Corals       0.865  - 0.005 0.450  -   0.961     -  -  - 19 Octocoral  -  -  - 0.025  -  - 0.038     -  -  - 20 Macroalgae  -  - 0.001 0.001 0.001 0.070  -  0.017   -  -  - 21 Crustose Coralline Algae (CCA)  -  -  -  -  - 0.010 0.001   -   -  -  - 22 Turf algae       0.050 0.300 0.032 0.050 0.050 0.745  -  0.098   -  -  - 23 Turf algae_lava bench  -  -  -  -  - 0.025  -   -   -  -  - 24 Zooplankton       0.011  - 0.030 0.023 0.316  -  -  0.120  0.500 0.350 0.020 25 Phytoplankton  - 0.005  -  - 0.050  -  -  0.120  0.100 0.380 0.800 26 Detritus  - 0.695 0.001  0.014 0.150  -  0.598  0.330 0.250 0.175 27 Import  -  -  -  - 0.450  -  -   -  0.070 0.020 0.005 155  Table 4.3 - Summary of outputs from the Ecological Network Analysis  Parameter Value Unit Sum of all consumption  5,332.03 t/km²/year Sum of all exports  520.07 t/km²/year Sum of all respiratory flows  3,477.31 t/km²/year Sum of all flows into detritus  1,700.15 t/km²/year Total system throughput  11,030.00 t/km²/year Calculated total net primary production  3,895.09 t/km²/year Total primary production/total respiration  1.12 Total biomass (excluding detritus)  699.53 t/km² Total catches  0.35 t/km²/year Mean trophic level of the catch  2.59  Throughput cycled (excluding detritus) 54.52 t/km²/year Finn's cycling index 6.13 % of total throughput  Ascendency  31.50 % Relative overhead (O/C)  68.50 %       156    Figure 4.1 - Map of Kaloko-Honokōhau National Historical Park (Kaloko) (modified from www.nps.gov with permission from S. Beavers).   157  5 4 3 2 1 Spinner dolphins Monk seals Sea birds Rays Sharks and jacks Hawksbills Green sea turtles Piscivores Herbivores Corallivores Detritivores MIF SIF Zoo Urchins Crown of thorns Benthic Invts Corals Octocoral Macroalgae CCA Turf algae Turf algae_LB Zooplankton Phytoplankton Detritus              Figure 4.2 - Graphical representation of trophic flows within the Kaloko reef ecosystem. Each functional group is identified here by an illustration (© M. Bailey); where relevant, an image of a species representative of its guild is depicted. Images are not drawn to scale or proportional to the group’s biomass. The light grey horizontal lines and associated numbers represent trophic levels; lines connecting individual functional groups represent trophic links. Zoo = Zooplanktivorous fish; MIF = Mobile Invertebrate Feeding fish; SIF = Sessile Invertebrate Feeding fish; TurfLB = turf growing on the lava bench area; CCA = Crustose Coralline Algae; Benthic Invts = Benthic invertebrates. 158  Green sea turtles Reef fish - herbivores All other reef fish (incl. rays) Urchins Corals Algae Plankton Other invertebrates Mammals, hawksbill sea turtles and birds                 Figure 4.3 - Proportions (%), in terms of biomass (t·km-2) of aggregated functional groups at Kaloko.         159               Figure 4.4 - Mixed trophic impact analysis for herbivores at Kaloko. Impacts of increases in the biomass of a particular group (impacting) on another (impacted), resulting in an increase in the latter’s biomass, are recorded as a positive on the y-axis. Impacts resulting in a decline of the impacted group are recorded as a negative value. Group names along the upper x-axis represent the impacted groups. The impacting group is indicated on the lower right of individual graphs.   M T I Impacted functional groups -0.5 -0.3 -0.1 0.1 G re en  s ea  tu rtl es Re ef  fi sh es  - he rb iv or es Ur ch in s M ac ro al ga e Tu rf al ga e Tu rf al ga e_ la va  b en ch Reef fish - herbivores-0.5 -0.3 -0.1 0.1 G re en  s ea  tu rtl es R ee f f ish es  - he rb iv or es U rc hi ns M ac ro al ga e Tu rf al ga e Tu rf al ga e_ la va  b en ch Green turtles -0.5 -0.3 -0.1 0.1 G re en  s ea  tu rtl es R ee f f ish es  - he rb iv or es U rc hi ns M ac ro al ga e Tu rf al ga e Tu rf al ga e_ la va  b en ch Urchins -0.5 -0.3 -0.1 0.1 G re en  s ea  tu rtl es Re ef  fi sh es  - he rb iv or es Ur ch in s M ac ro al ga e Tu rf al ga e Tu rf al ga e_ la va  b en ch Macroalgae -0.5 -0.3 -0.1 0.1 0.3 0.5 G re en  s ea  tu rtl es Re ef  fi sh es  - he rb iv or es Ur ch in s M ac ro al ga e Tu rf al ga e Tu rf al ga e_ la va  b en ch Turf algae -0.5 -0.2 0.1 0.4 0.7 1 G re en  s ea  tu rtl es Re ef  fi sh es  - he rb iv or es Ur ch in s M ac ro al ga e Tu rf al ga e Tu rf al ga e_ la va  b en ch Turf algae_lava  160  0% 20% 40% 60% 80% 100% I II III IV Trophic levels to ta l s ys te m  th ro ug hp ut  (% ) A B C D                Figure 4.5 - Fate of total system throughput (A = Respiration, B = Flow to detritus, C = Export, D = Consumption by predator) in percentage-per-integer trophic level.        161  4.5  References Adey, W. and Goertemiller, T. (1987) Coral reef algal turfs: Master producers in nutrient poor seas. Phycologia, 26: 374-386. Adey, W. and Steneck, R. S. (1985) Highly productive eastern Caribbean reefs: Synergistic effects of biological and chemical, physical, and geological factors. In The ecology of Coral Reefs. Symposia Series for Undersea Research. (ed M. L. Reaka), pp. 163-187. NOAA, Rockville (MD), USA. Ainsworth, C., Varkey, D. and Pitcher, T. (2007) Ecosystem simulation models for the Bird's Head Seascape, Papua, fitted to data. In Ecological and economic analyses of marine ecosystems in the Birds Head Seascape, Papua, Indonesia. (eds. T. Pitcher, C. Ainsworth and M. Bailey), pp. 6-174. Fisheries Centre, Vancouver (B.C.), Canada. Albert, S., Udy, J. and Tibbetts, I. R. (2008) Responses of algal communities to gradients in herbivore biomass and water quality in Marovo Lagoon, Solomon Islands. Coral Reefs, 27 (1): 73- 82. Allen, R. R. (1971) Relation between production and biomass. Journal of the Fisheries Research Board of Canada, 28: 1573- 1581. Appana, S. D. and Vuki, V. C. (2006) Foraging behavior, substrate preference and influence of Echinometra sp A on the carbonate budget of Nukubuco Reef, Fiji Islands. Micronesica, 38 (2): 191-205. Arias-González, J. E. (1994) Trophic balance of a reef ecosystem (Tiahura-Moorea-French Polynesia). Comptes Rendus Académiques et Scientifiques Ser III-Sciences de la Vie, 317:1143-1150 Arias-González, J. E., Delesalle, B., Salvat, B. and Galzin, R. (1997) Trophic functioning of the Tiahura reef sector, Moorea Island, French Polynesia. Coral Reefs 16:231-246 Arthur, K. E. and Balazs, G. H. (2008) A comparison of immature green turtles (Chelonia mydas) diets among seven sites in the Main Hawaiian Islands. Pacific Science, 62 (2): 205-217. Bak, R. P. M. (1994) Sea urchin bioerosion on coral reefs: Place in the carbonate budget and relevant variables. Coral Reefs, 13 (2): 99-103. Balazs, G. H. (1980) Synopsis of biological data on the green turtle in the Hawaiian Islands. US Department of Commerce, 141 pp. Balazs, G. and Chaloupka, M. (2004a) Thirty-year recovery trend in the once depleted Hawaiian green sea turtle stock. Biological Conservation, 117: 491-498. 162  Balazs, G. H. and Chaloupka, M. (2004b) Spatial and temporal variability in somatic growth of green sea turtles (Chelonia mydas) resident in the Hawaiian archipelago. Marine Biology, 145 (5): 1043-1059. Balazs, G. H. and Chaloupka, M. (2006) Recovery trend over 32 years at the Hawaiian green turtle rookery of French frigate shoals. Atoll Research Bulletin, 543: 147-158. Beets, J., Brown, E. and Friedlander, A. (2006) Inventory of marine vertebrate species and fish-habitat utilization patterns in coastal waters off four National Parks in Hawai’i. USDI National Park Service, Inventory and Monitoring Program, Pacific Island Network, Hawaii Volcanoes National Park (HI), USA. 55 pp. Bellwood, D. R. and Choat, J. H. (1990) A functional analysis of grazing in parrotfishes (family Scaridae) - the ecological implications. Environmental Biology of Fishes, 28 (1-4): 189-214. Bellwood, D. R., Hughes, T. P., Folke, C. and Nystrom, M. (2004) Confronting the coral reef crisis. Nature, 429 (6994): 827-833. Bienfang, P., De Carlo, E. H., Christopher, S., DeFelice, S. and Moeller, P. (2009) Trace element concentrations in coastal Hawaiian waters. Marine Chemistry, 113 (3-4): 164-171. Bjorndal, K. A. (1980) Nutrition and grazing behavior of the green turtle, Chelonia mydas. Marine Biology, 56 (2): 147-154. Bjorndal, K. A., Bolten, A. B. and Chaloupka, M. Y. (2000) Green turtle somatic growth model: Evidence for density dependence. Ecological Applications, 10 (1): 269-282. Brainard, R., Friedlander, A., Gulko, D., Hunter, C., Kelty, R. and Maragos, J. (2002) Status of coral reefs in the Hawaiian archipelago. In Status of Coral Reefs (ed C. Wilkinson), pp. 238-250. Australian Institute of Marine Science, Townsville (QLD), Australia. Bruggemann, J. H., Vanoppen, M. J. H. and Breeman, A. M. (1994) Foraging by the stoplight parrotfish Sparisoma viride. I. Food selection in different socially determined habitats. Marine Ecology Progress Series, 106 (1-2): 41-55. Burkepile, D. E. and Hay, M. E. (2006) Herbivore vs. nutrient control of marine primary producers: Context-dependent effects. Ecology, 87 (12): 3128-3139. Carpenter, R. C. and Edmunds, P. J. (2006) Local and regional scale recovery of Diadema promotes recruitment of scleractinian corals. Ecology Letters, 9 (3): 271-280. Carreiro-Silva, M. and McClanahan, T. R. (2001) Echinoid bioerosion and herbivory on Kenyan coral reefs: The role of protection from fishing. Journal of Experimental Marine Biology and Ecology, 262 (2): 133-153. 163  Chaloupka, M. and Balazs, G. (2007) Using Bayesian state-space modelling to assess the recovery and harvest potential of the Hawaiian green sea turtle stock. Ecological Modelling, 205 (1-2): 93- 109. Chaloupka, M., Bjorndal, K. A., Balazs, G. H., Bolten, A. B., Ehrhart, L. M., Limpus, C. J., Suganuma, H., Troeng, S. and Yamaguchi, M. (2008) Encouraging outlook for recovery of a once severely exploited marine megaherbivore. Global Ecology and Biogeography, 17 (2): 297-304. Cheroske, A. G., Williams, S. L. and Carpenter, R. C. (2000) Effects of physical and biological disturbances on algal turfs in Kaneohe Bay, Hawaii. Journal of Experimental Marine Biology and Ecology, 248 (1): 1-34. Choat, J. H., Clements, K. D. and Robbins, W. D. (2002) The trophic status of herbivorous fishes on coral reefs. I. Dietary analyses. Marine Biology, 140 (3): 613-623. Christensen, V. (1995) Ecosystem maturity - towards quantification. Ecological Modelling, 77: 3-32. Christensen, V. (2008) Ecopath with Ecosim: Linking fisheries and ecology. In Handbook of Ecological Modelling, Network and Informatics. (eds S. E. Jørgensen, T. S. Chon and F. A. Recknagel), pp. 55- 70. WIT Press, Billerica (MA), USA. Christensen, V. and Pauly, D. (1992) ECOPATH II - A software for balancing steady-state ecosystem models and calculating network characteristics. Ecological Modelling, 61: 169-185. Christensen, V. and Pauly, D. (eds.) (1993) Trophic Models of Aquatic Ecosystems, ICLARM, Manila, Philippines. 390 pp. Christensen, V. and Walters, C. (2004) Ecopath with Ecosim: methods, capabilities and limitations. Ecological Modelling, 72: 109-139. Christensen, V., Walters, C. J. and Pauly, D. (2005) Ecopath with Ecosim: A user's guide. Fisheries Centre, University of British Columbia, Vancouver (B.C.), Canada. 154 pp. Crossman, D. J., Choat, J. H. and Clements, K. D. (2005) Nutritional ecology of nominally herbivorous fishes on coral reefs. Marine Ecology Progress Series, 296: 129-142. Crossman, D. J., Choat, J. H., Clements, K. D., Hardy, T. and McConochie, J. (2001) Detritus as food for grazing fishes on coral reefs. Limnology and Oceanography, 46 (7): 1596-1605. Cvitanovic, C., Fox, R. J. and Bellwood, D. R. (2007) Herbivory by fishes on the Great Barrier Reef: A review of knowledge and understanding. Unpublished report to the Marine and Tropical Sciences Research Facility. Reef and Rainforest Research Centre Limited, Cairns, Australia. 33 pp. DeVerse, K. (2006) Appendix A: Kaloko-Honokōhau national historical park resource overview. In Pacific Island Network vital signs monitoring plan. (eds L. HaySmith, F. L. Klasner, S. H. Stephens and G. H. Dicus), pp. 21. Natural resource report NPS/PACN/NRR-2006/003, National Park Service, Fort Collins (CO), USA. 164  DLNR (2001) DNLR urges water users near Kona's Honokōhau harbor to watch out for sharks. Accessed March 23, 2008. http://www.state.hi.us/dlnr/chair/pio/HtmlNR/01-46.htm Done, T. (1992) Phase shifts in coral reef communities and their ecological significance. Hydrobiologia, 247: 121-132. Doty, M. S. (1961) Acanthophora, a possible invader of the marine flora of Hawaii. Pacific Science, 15 (4): 547-552. Ebert, T. A. (1971) Preliminary quantitative survey of echinoid fauna of Kealakekua and Honaunau bays, Hawaii. Pacific Science, 25 (1): 112-131. Edmunds, P. J. and Carpenter, R. C. (2001) Recovery of Diadema antillarum reduces macroalgal cover and increases abundance of juvenile corals on a Caribbean reef. Proceedings of the National Academy of Sciences of the United States of America, 98 (9): 5067-5071. Fabricius, K. E. (2005) Effects of terrestrial runoff on the ecology of corals and coral reefs: review and synthesis. Marine Pollution Bulletin, 50 (2): 125-146. Friedlander, A. M. and Parrish, J. D. (1997) Fisheries harvest and standing stock in a Hawaiian bay. Fisheries Research, 32 (1): 33-50. Friedlander, A. M. and Parrish, J. D. (1998) Temporal dynamics of fish communities on an exposed shoreline in Hawaii. Environmental Biology of Fishes, 53 (1): 1-18. Friedlander, A., Aeby, G., Brainard, R., Brown, E., Chaston, K., Clark, A., McGowan, P., Montgomery, T., Walsh, W., Williams, I. and Wiltse, W. (2008) The state of coral reef ecosystems of the Main Hawaiian Islands. In State of coral reef ecosystems of the United States and Pacific Freely Associated States: 2008. (eds J. E. Waddell and A. M. Clarke), pp. 219-261. NOAA Technical Memorandum NOS NCCOS 73. NOAA/NCCOS Center for Coastal Monitoring and Assessment’s Biogeography Team, Silver Spring (MD), USA. Funtowicz, S. O. and Ravetz, J. R. (1990) Uncertainty and Quality in Science for Policy, Kluwer Academic Publishers, Dordrecht, The Netherlands. 229 pp. Gardner, T. A., Cote, I. M., Gill, J. A., Grant, A. and Watkinson, A. R. (2003) Long-term region- wide declines in Caribbean corals. Science, 301 (5635): 958-960. Garpe, K. C., Yahya, S. A. S., Lindahl, U. and Öhman, M. C. (2006) Long-term effects of the 1998 coral bleaching event on reef fish assemblages. Marine Ecology Progress Series, 315: 237-247. Gibbs, A. E., Cochran, S. A., Logan, J. B. and Grossman, E. E. (2007) Benthic habitats and offshore geological resources of Kaloko-Honokōhau national historical park, Hawai'i. U.S. Geological Survey. Scientific investigations report 2006-5256, 70 pp. Grigg, R. W. (1994) Effects of sewage discharge, fishing pressure and habitat complexity on coral ecosystems and reef fishes in Hawai'i. Marine Ecology Progress Series, 103 (1-2): 25-34. 165  Hatcher, B. G. and Larkum, A. W. D. (1983) An experimental analysis of factors controlling the standing crop of the epilithic algal community on a coral reef. Journal of Experimental Biology and Ecology, 69: 61-84. Hay, M. E. (1981) Spatial patterns of grazing intensity on a Caribbean barrier reef: Herbivory and algal distribution. Aquatic Botany, 11 (2): 97-109. Hay, M. E. (1984a) Patterns of fish and urchin grazing on Caribbean coral reefs: Are previous results typical? Ecology, 65 (2): 446-454. Hay, M. E. (1984b) Predictable spatial escapes from herbivory: How do these affect the evolution of herbivore resistance in tropical marine communities? Oecologia, 64: 396-407. Heck, K. L. and Valentine, J. F. (2006) Plant-herbivore interactions in seagrass meadows. Journal of Experimental Marine Biology and Ecology, 330 (1): 420-436. Heithaus, M. R., Frid, A. and Dill, L. M. (2002) Shark-inflicted injury frequencies, escape ability, and habitat use of green and loggerhead turtles. Marine Biology, 140 (2): 229-236. Heithaus, M. R., Hamilton, I. M., Wirsing, A. J. and Dill, L. M. (2006) Validation of a randomization procedure to assess animal habitat preferences: microhabitat use of tiger sharks in a seagrass ecosystem. Journal of Animal Ecology, 75 (3): 666-676. Heymans, J. J. (2003) Comparing the Newfoundland southern Labrador marine ecosystem models using information theory. In Ecosystem models of Newfoundland and southeastern Labrador: additional information and analyses for "back to the future". Vol. 11(5). (ed J. J. Heymans), pp. 62-71. Fisheries Centre Research Reports, Vancouver (B.C.), Canada. Hobson, E. S. (1974) Feeding relationships of teleostean fishes on coral reefs in Kona, Hawaii. Fishery Bulletin, 72 (4): 915-1031. Hughes, T. (1994) Catastrophes, phase shifts and large scale degradation of a Caribbean coral reef. Science, 265: 1547-1551. Hunter, C. and Evans, C. (1995) Coral reefs in Kaneohe Bay, Hawaii: Two centuries of western influence and two decades of data. Bulletin of Marine Science, 57: 501-515. Hutchings, P. A. (1986) Biological destruction of coral reefs - a review. Coral Reefs, 4 (4): 239-252. Johnson, A. G., Glenn, C., Burnett, W. C., Peterson, R. N. and Lucey, P. (2008) Aerial infrared imaging reveals large nutrient-rich groundwater inputs to the ocean. Geophysical Research Letters, 35: L15606, doi:15610.11029/12008GL034574. Kubis, S., Work, T., Murakawa, S. K. K. and Balazs, G. H. (2008) An alternative method for assessing body condition of Hawaiian green turtles. In 27th annual symposium on Sea Turtle Biology and Conservation pp. 21-22. US Department of Commerce, NOAA, National Marine Fisheries 166  Service, Southeast Fisheries Science Centre. NOAA Technical Memorandum NMFS-SEFSC- 569, Myrtle Beach (SC), USA. Lapointe, B. E. (1997) Nutrient thresholds for eutrophication and macroalgal blooms on coral reefs in Jamaica and southeast Florida. Limnology and Oceanography, 42: 1119-1131. Lessios, H. A. (1988) Mass mortality of Diadema antillarum in the Caribbean: What have we learned? Annual Review of Ecology and Systematics, 19: 371-393. Lewis, S. (1986) The role of herbivorous fishes in the organisation of a Caribbean reef community. Ecological Monographs, 56: 183-200. Littler, M. M. and Littler, D. S. (2007) Assessment of coral reefs using herbivory/nutrient assays and indicator groups of benthic primary producers: A critical synthesis, proposed protocols, and critique of management strategies. Aquatic Conservation-Marine and Freshwater Ecosystems, 17 (2): 195-215. Mapstone, B. D., Andrew, N. L., Chancerelle, Y. and Salvat, B. (2007) Mediating effects of sea urchins on interactions among corals, algae and herbivorous fish in the Moorea lagoon, French Polynesia. Marine Ecology Progress Series, 332: 143-153. Marrack, L., Beavers, S. and Weijerman, M. (2009) Baseline assessment of the coral reef habitat adjacent to the shores of Kohanaiki development in Kaloko-Honokōhau National Historical Park. Cooperative ecosystem studies unit, University of Hawai'i at Hilo, Hilo (HI), USA. 41 pp. McClanahan, T. R. (1998) Predation and the distribution and abundance of tropical sea urchin populations. Journal of Experimental Marine Biology and Ecology, 221 (2): 231-255. McClanahan, T. R. and Kurtis, J. D. (1991) Population regulation of the rock-boring sea-urchin Echinometra mathaei (Deblainville). Journal of Experimental Marine Biology and Ecology, 147 (1): 121-146. McClanahan, T. R. and Mangi, S. (2001) Coral and algal response to the 1998 El Niño coral bleaching and mortality on Kenya’s southern reef lagoons. Accessed March, 2009. https://iodeweb1.vliz.be/odin/handle/1834/482 McClanahan, T., Hendricks, V. and Polunin, N. (1999) Varying response of herbivorous and invertebrate feeding fishes to macroalgal reduction: A restoration experiment. In Proceedings of the international conference on scientific aspects of coral reef assessment, monitoring and restoration, NCRI pp. 133. Fort Lauderdale (FL), USA. McCutcheon, S., MCDermid, K. and Balazs, G. (2003) A nutritional analysis of the turf algal diet and fecal pellets of the green turtle, Chelonia mydas L Journal of Phycology, 39 (S1): 41. 167  McDermid, K. J., Stuercke, B. and Balazs, G. H. (2007) Nutritional composition of marine plants in the diet of the green sea turtle (Cholonia mydas) in the Hawaiian islands. Bulletin of Marine Science, 81 (1): 55-71. Mills, S. C., Peyrot-Clausade, M. and Fontaine, M. F. (2000) Ingestion and transformation of algal turf by Echinometra mathaei on Tiahura fringing reef (French Polynesia). Journal of Experimental Marine Biology and Ecology, 254 (1): 71-84. Morissette, L. (2007) Complexity, cost and quality of ecosystem models and their impact on resilience: a comparative analysis, with emphasis on marine mammals and the Gulf of St. Lawrence. Doctoral thesis. University of British Columbia, Vancouver (B.C.), Canada. Morrison, D. (1988) Comparing fish and urchin grazing in shallow and deeper coral reef algal communities. Ecology, 69 (5): 1367-1382. Mortimer, J. A. (1981) The feeding ecology of the west Caribbean green turtle (Chelonia mydas) in Nicaragua. Biotropica, 13 (1): 49-58. Mumby, P. J., Dahlgren, C. P., Harborne, A. R., Kappel, C. V., Micheli, F., Brumbaugh, D. R., Holmes, K. E., Mendes, J. M., Broad, K., Sanchirico, J. N., Buch, K., Box, S., Stoffle, R. W. and Gill, A. B. (2006a) Fishing, trophic cascades, and the process of grazing on coral reefs. Science, 311 (5757): 98-101. Mumby, P. J., Hedley, J. D., Zychaluk, K., Harborne, A. R. and Blackwell, P. G. (2006b) Revisiting the catastrophic die-off of the urchin Diadema antillarum on Caribbean coral reefs: Fresh insights on resilience from a simulation model. Ecological Modelling, 196 (1-2): 131-148. Muthiga, N. A. and McClanahan, T. R. (1987) Population changes of a sea urchin (Echinometra mathaei) on an exploited fringing reef. African Journal of Ecology, 25 (1): 1-8. Naim, O., Cuet, P. and Letourneur, Y. (1997) Experimental shift in benthic community structure. In Proceedings of the 8th International Coral reef Symposium pp. 1873-1878. Nystrom, M. (2006) Redundancy and response diversity of functional groups: Implications for the resilience of coral reefs. Ambio, 35 (1): 30-35. Nystrom, M. and Folke, C. (2001) Spatial resilience of coral reefs. Ecosystems, 4 (5): 406-417. Odum, E. P. (1969) The strategy of ecosystem development. Science, 104: 262-270. Odum, E. P. (1971) Fundamentals of Ecology, Saunders, Philadelphia, PA (USA). 574 pp. Ogden, J. C. and Lobel, P. S. (1978) The role of herbivorous fishes and urchins in coral reef communities. Environmental Biology of Fishes, 3 (1): 49-63. Oki, D. S., Tribble, G. W., Souza, W. R. and Bolke, E. L. (1999) Ground-water resources in Kaloko- Honokōhau National Historical Park, island of Hawaii, and numerical simulation of the effects of ground- 168  water withdrawals. U.S. Geological Survey, prepared in cooperation with the National Park Service; water-resources investigations report 99-4070, Honolulu (HI), USA. 55 pp. Paddack, M. J., Cowen, R. K. and Sponaugle, S. (2006) Grazing pressure of herbivorous coral reef fishes on low coral-cover reefs. Coral Reefs, 25 (3): 461-472. Parrish, J., Smith, G. and Norris, J. (1990) Resources of the marine waters of Kaloko-Honokōhau National Historical Park. Cooperative National Park Resources Study Unit, technical report 74, Department of Botany, University of Hawaii, Honolulu (HI), USA. 118 pp. Pauly, D. (1980) On the interrelationships between natural mortality, growth parameters, and mean environmental temperature in 175 fish stocks. Journal du Conseil, 39 (2): 175-192. Pauly, D. and Christensen, V. (1995) Primary production required to sustain global fisheries. Nature, 374 (6519): 255-257. Pauly, D. and Christensen, V. (2002) Ecosystem models. In Handbook of Fish Biology and Fisheries. (eds P. J. Hart and J. D. Reynolds), pp. 211-227. Blackwell Publishing, Oxford, UK. Pauly, D., Christensen, V. and Walters, C. (2000) Ecopath, Ecosim, and Ecospace as tools for evaluating ecosystem impact of fisheries. Ices Journal of Marine Science, 57 (3): 697-706. Payri, C. (2000) Production primaire et calcification des algues benthiques en milieu corallien. Oceanis, 26 (3): 427-463. Paytan, A., Shellenbarger, G. G., Street, J. H., Gonneea, M. E., Davis, K., Young, M. B. and Moore, W. S. (2006) Submarine groundwater discharge: An important source of new inorganic nitrogen to coral reef ecosystems. Limnology and Oceanography, 51 (1): 343-348. Plaganyi, E. E. (2007) Models for an ecosystem approach to fisheries. FAO Fisheries Technical Publication No. 477, Rome, Italy. 126 pp. Plaganyi, E. E. and Butterworth, D. S. (2004) A critical look at the potential of Ecopath with Ecosim to assist in practical fisheries management. African Journal of Marine Science, 26: 261-287. Polovina, J. J. (1984) Model of a coral-reef ecosystem. I. The Ecopath model and its application to French Frigate Shoals. Coral Reefs, 3 (1): 1-11. Presto, M. K., Storlazzi, C. D., Logan, J. B. and Grossman, E. E. (2007) Submarine groundwater discharge along the coast of Kaloko-Honokōhau National Historic Park, Hawai'i. Part I: Time series measurements of currents, waves and water properties. U.S. Geological Survey open file report. http://pubs.usgs.gov/of/2007/1310. Rodgers, K. u., Jokiel, P. L. and Brown, E. K. (2004) Rapid assessment of Kaloko-Honokōhau and Pu‘uhonua o Honaunau, west Hawai‘i. Hawaii Coral Reef Assessment and Monitoring Program (CRAMP), Hawaii Institute of Marine Biology, Kāne‘ohe (HI), USA. 55 pp. 169  Russell, D. F. and Balazs, G. H. (2009) Dietary shifts by green turtles (Chelonia mydas) in the Kane'ohe Bay region of the Hawaiian Islands: A 28-year study. Pacific Science, 63 (2): 181-192. Russell, D. J. (1992) The ecological invasion of Hawaiian reefs by two marine red algae, Acanthophora spicifera (Vahl) Boerg. and Hypnea musciformis (Wulfen) J. Ag. and their association with two native species Laurencia nidifica J. Ag. and Hypnea cervicornis J. Ag. ICES Marine Science Symposium, 194: 110-125. Russell, D. J. and Balazs, G. H. (1994) Colonization by the alien marine alga Hypnea musciformis (Wulfen) J Ag (Rhodophyta, Gigartinales) in the Hawaiian Islands and its utilization by the green turtle, Chelonia mydas L. Aquatic Botany, 47 (1): 53-60. Smith, C. M. and Smith, J. E. (2006) Algal blooms in north Kihei: An assessment of patterns and processes relating nutrient dynamics to algal abundance. Report for the city and county of Maui (HI), USA, 65 pp. Smith, J. E., Smith, C. M. and Hunter, C. L. (2001) An experimental analysis of the effects of herbivory and nutrient enrichment on benthic community dynamics on a Hawaiian reef. Coral Reefs, 19 (4): 332-342. Smith, J. E., Conklin, E. J., Smith, C. M. and Hunter, C. L. (2008) Fighting algae in Kaneohe Bay - Response. Science, 319 (5860): 157-158. Smith, M. K. (1993) An ecological perspective on inshore fisheries in the main Hawaiian islands. Marine Fisheries Review, 55 (2): 34. Stimson, J., Larned, S. T. and Conklin, E. (2001) Effects of herbivory, nutrient levels, and introduced algae on the distribution and abundance of the invasive macroalga Dictyosphaeria cavernosa in Kaneohe Bay, Hawaii. Coral Reefs, 19 (4): 343-357. Thayer, G. W., Bjorndal, K. A., Ogden, J. C., Williams, S. L. and Zieman, J. C. (1984) Role of larger herbivores in seagrass communities. Estuaries, 7 (4A): 351-376. Thompson, R. (2005) Kona officials fear shark being lured to harbor. Accessed Jan 15, 2008. http://starbulletin.com/2005/07/01/news/index5.html Tissot, B. N., Walsh, W. J. and Hallacher, L. E. (2004) Evaluating effectiveness of a marine protected area network in west Hawai‘i to increase productivity of an aquarium fishery. Pacific Science, 58 (2): 175-188. Tissot, B. N., Walsh, W. J. and Hixon, M. A. (2009) Hawaiian Islands marine ecosystem case study: Ecosystem- and community-based management in Hawaii. Coastal Management, 37 (3): 255-273. 170  Tudman, P.D. (2001) Modelling the trophic effects of fishing on a mid-shelf coral reef of the central Great Barrier Reef. Bachelor of Science with Honours thesis. James Cook University, Brisbane, Australia. Ulanowicz, R. E. (1997) Ecology: The Ascendent Perspective, Columbia University Press, New York (NY), USA. 201 pp. Ulanowicz, R. E. and Puccia, C. J. (1990) Mixed trophic impacts in ecosystems. Coenoses, 5: 7-16. Vaitilingon, D., Rasolofonirina, R. and Jangoux, M. (2003) Feeding preferences, seasonal gut repletion indices, and diel feeding patterns of the sea urchin Tripneustes gratilla (Echinodermata: Echinoidea) on a coastal habitat off Toliara (Madagascar). Marine Biology, 143 (3): 451-458. Valentine, J. F., Heck, K. L. and Cinkovich, A. M. (2002) Impacts of seagrass food webs on marine ecosystems: A need for a broader perspective. Bulletin of Marine Science, 71 (3): 1361-1368. Vitousek, S., Barbee, M. M., Fletcher, C. H., Richmond, B. M. and Genz, A. S. (2009) Pu’ukoholā Heiau National Historic Site and Kaloko-Honokōhau Historical Park, Big Island of Hawai'i. Coastal hazard analysis report. NPS Geologic Resources Division, 105 pp. Waddell, J. E. and Clarke, A. M. (eds.) (2008) The state of coral reef ecosystems of the United States and Pacific Freely Associated States: 2008, NOAA Technical Memorandum NOS NCCOS 73. NOAA/NCCOS Center for Coastal Monitoring and Assessment’s Biogeography Team, Silver Spring (MD), USA. 569 pp. Walters, C. J. and Martell, S. J. D. (2004) Fisheries Ecology and Management, Princeton University Press, Princeton, NJ (USA). 399 pp. Walters, C., Christensen, V. and Pauly, D. (1997) Structuring dynamic models of exploited ecosystems from trophic mass-balance assessments. Reviews in Fish Biology and Fisheries, 7 (2): 139-172. Wanders, J. B. W. (1976) The role of benthic algae in the shallow reef of Curaçao (Netherlands Antilles). I. Primary productivity in the coral reef. Aquatic Botany, 2: 235-270. Weijerman, M., Beavers, S., Marrack, L. and Most, R. (2009) Baseline assessment of the coral reef habitat in Kaloko-Honokōhau National Historical Park adjacent to the proposed harbor expansion and development, Kona Kai Ola. Cooperative ecosystem studies unit, University of Hawai'i at Manoa, Hilo (HI), USA. 57 pp. Williams, I. D., Polunin, N. V. C. and Hendrick, V. J. (2001) Limits to grazing by herbivorous fishes and the impact of low coral cover on macroalgal abundance on a coral reef in Belize. Marine Ecology Progress Series, 222: 187-196. Williams, I. D., Walsh, W. J., Sparks, R. T. and Polhemus, D. A. (2007) The slippery slope to slime: Invasive alien algae, coral to macroalgal shifts, and the scope for increasing herbivorous fish stocks as a 171  control mechanism. State of Hawai'i Division of Aquatic Resources, Honolulu (HI), Hawai'i. 11 pp. Williams, I. D., Walsh, W. J., Schroeder, R. E., Friedlander, A. M., Richards, B. L. and Stamoulis, K. A. (2008) Assessing the importance of fishing impacts on Hawaiian coral reef fish assemblages along regional-scale human population gradients. Environmental Conservation, 35 (3): 261-272. Witzell, W. N. (1994) The origin, evolution, and demise of the U.S. sea turtle fisheries. Marine Fisheries Review, 56 (4): 8-23. Work, T. (2007) Diagnostic case report, Hilton Waikaloa. U.S. Geological Survey, Biological Resources Division, National Wildlife Health Center, Honolulu field station, 1 pp. Work, T. (2008a) Diagnostic case report, Pu’uhonua O Honaunau NHP. U.S. Geological Survey, Biological Resources Division, National Wildlife Health Center, Honolulu field station, 1 pp. Work, T. (2008b) Diagnostic case report, Kaloko Honokōhau NHP. U.S. Geological Survey, Biological Resources Division, National Wildlife Health Center, Honolulu field station, 1 pp. Wulff, F., Field, J. G. and Mann, K. H. (eds.) (1989) Network Analysis in Marine Ecology: Methods and Applications Coastal and estuarine studies Vol. 32. Springer Verlag, Berlin, Germany. 284 pp. Zeller, D., Booth, S. and Pauly, D. (2005) Reconstruction of coral reef and bottom fisheries catches for US flag island areas in the western Pacific, 1950-2002. Final report to the Western Pacific Regional Fishery Management Council. Fisheries Centre, Vancouver (B.C.), Canada. 119 pp.   172  5.  Regional-scale seagrass habitat mapping in the wider Caribbean using Landsat sensors: Applications to conservation and ecology4  5.1 Introduction Seagrass are submerged flowering plants (angiosperms) that can form dense beds in shallow subtidal, mostly clear and sheltered, soft-bottomed marine and estuarine environments (Phillips and Menez 1988). These ‘seagrass meadows’ are important tropical, temperate, and subarctic coastal habitats (Hemminga and Duarte 2000, den Hartog and Kuo 2006), covering the equivalent of approximately 0.05 - 0.15 % of the surface area of the oceans globally (Spalding et al. 2003). By providing substratum for epiphytic algae, shelter for invertebrates and fish, and foraging areas for a variety of          4 A version of this chapter has been published. Wabnitz C.C.C., Andréfouët S., Torres-Pulliza D., Müller- Karger F.E., and Kramer P.A. (2008) Regional-scale seagrass habitat mapping in the wider Caribbean region using Landsat sensors: Applications to conservation and ecology Remote Sensing of Environment 112: 3455-67 173  organisms, they significantly contribute to the biodiversity of coastal waters (Williams and Heck 2001, Duffy 2006). The combined productivity of seagrass and epiphytic algae rank them among the most productive systems on Earth (Duarte and Cebrián 1996, Duarte and Chiscano 1999). These meadows also serve as critical breeding and nursery grounds for juvenile stages of many economically and ecologically important species (Beck et al. 2003, Heck et al. 2003, Dahlgren et al. 2006, Gillanders 2006).  Established in coastal zones, seagrass beds are inherently dynamic systems prone to natural physical disturbance, particularly in temperate regions (Fonseca et al. 2002). However, changes or losses in abundance, species composition, structure, and extent have commonly resulted from activities such as eutrophication, overfishing, and habitat alteration or destruction (Short and Wyllie-Echeverria 1996, Duarte 2002). Until recently, relatively little attention had been paid to the impacts of human activities on seagrass food webs (Jackson 2001, Duarte 2002), with most studies focusing on how physical disturbance alters the structure and function of the seagrass habitats themselves (Duarte 2002). The presence of green turtles (Chelonia mydas) may have had substantial ecological and evolutionary effects: increasing the productivity of seagrass in the same way as grazers in terrestrial grasslands (McNaughton 1979, Pandolfi et al. 2003, Moran and Bjorndal 2005, 2007). Changes in temperature, nutrient levels, and salinity, as well as a 93 - 97 % reduction in the Caribbean green turtle population compared to its size prior to human contact (Jackson et al. 2001), have been implicated in die-offs of seagrass throughout the region (Robblee et al. 1991, Jackson 1997, Fourqurean and Robblee 1999). Overall, anthropogenic impacts have contributed to seagrass now ranking among the most threatened of marine habitats (Green and Short 2003, Lotze et al. 2006, Orth et al. 2006).  Given ongoing coastal zone development around the globe, it is imperative to design and implement effective ways to protect coastal resources. Specifically, at the Fifth World Parks Congress (WPC) in 174  2003, the recommendation was made to develop extensive networks of Marine Protected Areas (MPAs) that “include strictly protected areas [amounting] to at least 20 - 30 % of each habitat” by 2012. However, exact predictions of the potential status of seagrass in the future and best ways to protect them are hampered - chiefly by the absence of consistent and reliable information concerning the present extent of this habitat. Similarly, current carrying capacity estimates of green turtles for the Caribbean (16 - 586 million) are based on only a very rough idea of seagrass extent thought to be available for foraging (Jackson et al. 2001).  A literature review conducted for this study suggests that there are many site specific studies and records of seagrass bed extent and distribution for the wider Caribbean Region (WCR). However, with few exceptions (e.g., Puerto Rico and the U.S. Virgin Islands), relevant documents are difficult to access and rarely, or poorly, document mapping methods or accuracies. Digital maps in GIS formats are often unavailable, or their use restricted. The only existing database generating a global overview was developed by the United Nations Environment Program-World Conservation Monitoring Centre (UNEP-WCMC) in 2003. The resulting “World Atlas of Seagrasses” was the first synthesis of the distribution and status of seagrass habitat at that scale (Green and Short 2003). However, direct habitat maps (i.e., chiefly derived from remotely sensed data), which provide the most accurate data on habitat distribution, were only available for a very limited subset of the world. The majority of geographic information thus falls into two main categories: (i) interpolation of expert knowledge and observations; and (ii) point-based samples, which are useful in a providing information regarding species presence, but give no information as to actual seagrass extent (Spalding et al. 2003). As a result, the worldwide UNEP-WCMC database, including the Caribbean section, suffers from substantial inaccuracy (vast commission or omission errors (i.e., including a seagrass pixel in a non-seagrass area and vice versa)), poor spatial representation, and limited spatial resolution.  175  Satellite remote sensing provides a tool to develop a reliable, methodologically consistent database of seagrass extent over large regions, in a cost-effective, objective, and timely fashion (Mumby et al. 1999, Krause-Jensen et al. 2004, Balmford et al. 2005). Habitat mapping on the scale of a region poses new environmental and methodological challenges rarely addressed in tropical initiatives to date (but see the Millennium Coral Reef Mapping Project (Andréfouët and Guzmán 2005, Andréfouët et al. 2006)). First, the diversity of environments (estuaries, cross-shelf areas, banks, atolls, and narrow continental reef terraces), the heterogeneity of habitats, and the vast extent of the targeted region imply that expertise and field data are seldom available with comparable quality, and often cannot be acquired due to their prohibitive costs. Second, high-resolution regional mapping requires hundreds of Landsat images to achieve complete cloud-free coverage. Working with such a large dataset presents substantial calibration problems in marine environments, and complicates or even prevents the use of standard analytical and statistical image processing approaches (Andréfouët et al. 2001, Thome 2001, Teillet et al. 2006).  This paper reports on the approach used to map seagrass beds using consistent methods, throughout the WCR, given the constraints associated with working at large spatial scales. First, I provide results obtained for well-documented sites throughout the region, where cross-comparison with ancillary data allowed for direct or indirect measures of accuracy. Second, I outline the implications of my findings for mapping seagrass beds in a cost effective fashion throughout the WCR. Finally, I discuss the relevance of the products presented and results for: (a) generally advancing future biodiversity research, conservation, and management in the region, and (b) specifically re-assessing carrying capacity estimates for green turtles.  176  5.2  Data and methods 5.2.1  Landsat data and Millennium Coral Reef Mapping Project (MCRMP) The archive of Landsat 7 Enhanced Thematic Mapper Plus (ETM+) images compiled by the MCRMP (Andréfouët et al. 2006) constituted the main data set used for seagrass mapping in this study. Most images in the database were acquired between 2000 and 2002. One to six images were available for each Landsat scene (path-row) intersecting coastlines. Where clouds were persistent, several Landsat Thematic Mapper 5 (TM) images, acquired mainly in the early 1990s, were also available (and often used). Images were all re-sampled to UTM WGS 84 projection, at 30-m spatial resolution. Only the first four bands were considered for the work presented here (blue, green, red, and near infra-red).  For each image path-row, MCRMP created polygons that characterise and delineate coral reef geomorphology according to a typology that is globally relevant (for more detailed information regarding the methodology applied see Andréfouët et al. (2006)). Individual Millennium classes are easily associated with a given set of environmental attributes (zonation, depth, hydrodynamic exposure), which favor the development of some habitats (e.g., seagrass), but not others. Although linkages between ‘geomorphology’ and ‘habitats’ have not been exhaustively quantified to date, preliminary analysis for several Caribbean sites where detailed habitat maps have been produced confirm that Millennium polygons can be used for a priori contextual editing (see Section 5.2.3 Image processing). By doing so, areas of the image unlikely to contain seagrasses are excluded, avoiding misclassification in subsequent analyses.  177  5.2.2  Study areas Table 5.1 lists the 19 focal areas of the WCR considered in this study (see also Figure 5.1). These sites were chosen based on the availability of independent information to: (a) compare obtained accuracies with published values and qualitatively contrast available thematic maps with products from this study; or (b) assess the accuracy of seagrass extent estimates derived from Landsat data using IKONOS imagery and/or field data points. The sites represent various levels of geomorphological complexity typical of the region.  5.2.3  Image processing For each site within a Landsat image, polygons were constructed that encompassed one, several, or all MCRMP classes with any likelihood of containing seagrass (see Figure 5.2 andFigure 5.3). For instance, one is unlikely to find seagrass within the deeper classes and on several exposed ‘reef’ areas (i.e., forereefs). The degree to which MCRMP classes were merged depended on the geomorphological complexity of the site.  A priori contextual editing has been reported by Andréfouët et al. (2003) and Andréfouët and Guzmán (2005) as a relatively simple and efficient way to enhance classification richness and improve accuracy of results. The method consists in applying a contextual decision rule throughout the image to sets of habitats which have similar spectra, but different yet predictable physical environments (Groom et al. 1996). By removing beforehand image areas that may create spectral confusion for the classes of interest, the classification process is much more accurate (Andréfouët et al. 2003). Given the scale of the area covered, this straight forward and reliable approach was favored here over the use of water column correction techniques, or the construction of depth- invariant indices. However, I am well aware of the benefits of the latter methodology (Andréfouët et 178  al. 2003), and in the few cases where polygons include a large number of habitats and significant depth range (e.g., wide cross-shelf areas), my assumption may not hold. In such instances, low map accuracies indicate Caribbean coastal configurations where bathymetric corrections would be required. Elsewhere, to quantify the validity of my assumption, I compared my results with those obtained from independent studies that corrected for the effects of the water column and published accuracy values.  For very large continuous areas spanning several Landsat scenes (e.g., Bahamas, Belize), images were inter-calibrated and normalized prior to classification (Figure 5.2). For this, digital counts (DN) were transformed into at-sensor reflectance using the gain/bias coefficients available for each image. A dark pixel correction scheme was then applied to remove part of the atmospheric effect. Specifically, subsequent to Rayleigh correction (Zhang et al. 1999), a deep-water reflectance value (average value of pixels sampled in ‘deep water’ in the short-wave infra-red band) was subtracted from pixels in all other bands, assuming a white aerosol signal (Zhang et al. 1999, Hu et al. 2001). Finally, using one arbitrary reference image, the next Landsat scene in the composition was adjusted using an empirical line calibration approach, based on training areas that have not changed between the different images (i.e., according to shapes of dense seagrass and sand patches, cf. Andréfouët et al. (2001)). This was possible since all Landsat scenes overlap neighboring scenes and I assumed stable conditions between the acquisitions of images (less than two years). All images were joined in a mosaic, which was expanded one image at a time. For all images, where necessary, clouds and shadows were masked out prior to classification.  Image classification training regions (e.g., sand, coral, and seagrass of different densities) were mostly selected by visual-interpretation, due to the lack of extensive field data for the majority of polygons. Generally, three to five classes were generated, but no more than eight, depending on intra-polygon spectral variability. Standard supervised classification, using ENVI ®’s Maximum 179  Likelihood classifier, was conducted for each image to produce a habitat map. Individual seagrass classes were subsequently merged to a maximum of three categories according to the density of the submerged vegetation: (i) dense (70 - 100 % cover); (ii) medium-dense (30 - 70 %); and (iii) sparse (<30 %). These closely correspond to categories derived for seagrass cover in other studies within the region (e.g., Mumby and Harborne 1999). All other habitat classes were merged into a generic ‘other’ class.  The study aimed to map overall seagrass presence/absence and density of cover throughout the region, without reference to particular seagrass species. However, turtle grass (Thalassia testudinum Banks ex König) is considered to be the most common species off the coast of Florida and throughout the Caribbean. Halodule wrightii Kütz. and Syringodium filiforme Aschers. are the other two species typically encountered, albeit at lower densities; and they are generally considered to be pioneer species (Gallegos et al. 1994).  5.2.4  Accuracy assessment The accuracy of classified images is generally assessed using a set of geo-referenced field data (Lillesand et al. 2004). However, ground-truth data that are adequately documented, of comparable quality, and that uniformly cover all areas of interest, present a substantial challenge for large scale mapping efforts. Such information is not uniformly available for seagrass beds throughout the WCR. Therefore, several strategies were adopted as proxies (Figure 5.2): (i) Data were gleaned from non peer-reviewed literature (e.g., consultancy reports, newspaper articles, electronic information, and government reports). The database developed from this information contains references of disparate quality and level of detail. Only a handful of these references provided detailed thematic maps or geographic coordinates of point observations associated with specific habitat classes. Only information gathered for Roatán Island 180  (Honduras) was considered of sufficient quality to enable accuracy assessment of our product (Porcher et al. 2001a, 2001b). (ii) Maps were extracted from peer-reviewed articles. Documented areas with published thematic maps allowing for qualitative comparisons with our products included Roatán (Maeder et al. 2002), Los Roques (Schweizer et al. 2005), Lee Stocking Island (Armstrong 1993, Call et al. 2003, Louchard et al. 2003), Martinique and Guadeloupe (Chauvaud 1997, Chauvaud et al. 1998, 2001), Glovers Atoll (Andréfouët et al. 2003), Alacranes (Bello-Pineda et al. 2005), Vieques Island (Hernández-Cruz et al. 2006, Shapiro and Rohmann 2006), Puerto Rico (Shapiro and Rohmann 2005), and Columbia (Díaz et al. 2003, Díaz and Gómez-López 2003). For these, accuracy values were generally reported. Although a number of other peer-review articles report on their seagrass mapping efforts in the Caribbean, the absence of available habitat maps precluded direct qualitative comparisons (e.g., Luczkovich et al. 1993, Mumby et al. 1998, Mumby and Edwards 2002, Garza-Perez et al. 2004). (iii) in situ data from mapping projects undertaken by the National Oceanic and Atmospheric Administration (NOAA) for coastal waters under US jurisdiction (NCCOS 2001) were downloaded from NOAA’s website. Under this initiative, habitat maps were created by visual interpretation of aerial photos. In situ data for Puerto Rico and the US Virgin Islands were available as: (a) ‘ground-truth’ points (large data set, geographically widespread for training); and (b) accuracy assessment data points (estimated at only two locations within the project area). The accuracy of our seagrass products was assessed using the larger ‘ground-truth’ dataset. (iv) Habitat survey points collected by S. Andréfouët and P. Kramer between 2001 and 2003 were collated into a single database (see Table 5.2). The dataset, providing seagrass presence/absence observations, spans San Blas (Panama), Los Roques (Venezuela), and Lee Stocking and Andros Islands (Bahamas). 181  (v) ‘Virtual ground truthing’ points were derived from high spatial resolution IKONOS imagery (Figure 5.4). High resolution images, such as aerial photographs, allow enhanced visual- interpretation of many benthic features, including seagrass beds (cf. NOAA’s approach). Although confusion with other submerged aquatic vegetation types remains possible, this method allows the practitioner to locate points (or polygons) within specific habitat types effectively and with a high level of confidence - even at depth. IKONOS images, collected between 2000 and 2003, were obtained from the National Aeronautics and Space Administration (NASA) Data Buy program archive. Several of these images had already been processed for habitat mapping purposes, with the resulting products and analyses published (e.g., Mumby and Edwards 2002, Andréfouët et al. 2003). To my knowledge, the use of other scenes is reported here for the first time. (vi) in situ points collected by contributors to the study by Andréfouët et al. (2003) were collated into a separate dataset and augmented for this study by visual-interpretation (Figure 5.4). Sites processed included: Boca Paila, Majahual, and Akumal (Mexico); Andros and Lee Stocking Island (Bahamas) (Figure 5.5); Roatán (Honduras); Glovers Atoll, Lighthouse Atoll, lagoonal patch reefs and two sites off the barrier reef (Belize).  In general, the time lag between the acquisition of ground observations, Landsat images, and IKONOS images ranged between a few months and a year. It was therefore assumed that, for the most part, the delineation of seagrass bed extent would not have dramatically changed during that timeframe.  The accuracy of habitat maps produced here was determined from confusion matrices elaborated using data obtained under (iii), (iv), (v), and (vi) above. Such a matrix allows for the calculation of specific accuracy measures including the overall accuracy and user’s and producer’s accuracies (Congalton 1991). Overall accuracy is computed by dividing the total number of correctly classified 182  pixels by the total number of pixels in the matrix (Congalton 1991). The ‘producer's accuracy’ refers to the probability of a reference pixel being correctly classified (i.e., measures the error of omission); whereas the ‘user's accuracy’ indicates the probability that a pixel classified on the map represents that category on the ground (i.e., measures the error of commission) (Congalton and Green 1999). An estimate of the Kappa coefficient, which quantifies the improvement of the classified map over a random class assignment, is also provided (Congalton and Green 1999, Foody 2002).  Locations used for the accuracy assessment spanned the entire WCR and included a large variety of habitat types, depth ranges, and water conditions inherent to the region. Therefore, these estimates were assumed to be representative of accuracies derived for thematic maps produced for the remainder of the Caribbean (ongoing project). NOAA outlined a similar approach to assess the accuracy of their Caribbean products (NCCOS 2001). However, for seagrass beds located in turbid estuaries and large coastal lagoons (e.g., Mexico and Venezuela) this assumption is unlikely to be verified and accuracy assessment of products derived for these areas will require the development of alternative strategies. Such sites were not included in this analysis.  5.3 Results Confusion matrices derived when using three seagrass classes of variable densities plus one ‘other’ (all non-seagrass areas) class, showed significant confusion between individual seagrass classes, resulting in low overall accuracy. However, as a whole, ‘seagrass’ was correctly classified. Consequently, the ‘medium’ and ‘sparse’ seagrass classes were merged to form a single ‘medium- sparse’ class (< 70 % cover) (see discussion for details). Associated ground-truth points were also merged for corresponding accuracy assessments. The ‘dense’ class (> 70 % cover) was left 183  unchanged. This improved accuracy, with overall values ranging between 46 % and 77 % (see Table 5.2).  Kappa values were found to vary widely between sites, spanning from 0.00 for a site west of La Parguera, Puerto Rico, to 0.64 for Los Roques, Venezuela (see Table 5.2). Overall accuracy averaged 68 %. Lowest overall accuracies were obtained for the classification of a site west of La Parguera, Puerto Rico (45 %). Highest accuracies were recorded for Akumal, Mexico (87 %) and San Blas, Panama (85 %). Given the local topographic complexity and patterns, thematic maps encompassed only one seagrass class for these two sites. Highest overall accuracy for sites with two distinct seagrass classes was registered at Los Roques, Venezuela (77 %). The ‘other’ class showed high producer accuracies, except for Puerto Rico. At all sites, most misclassifications still occurred between the two seagrass classes and some seagrass pixels that classified as “other”. My products can therefore be considered as conservative seagrass distribution maps (i.e., not overestimating seagrass areas) since very few “other” pixels were classified as seagrass (with the exception of Puerto Rico).  Figure 5.5 and Figure 5.6 show examples of thematic maps produced for two classes of seagrass, and one ‘other’ class for the Bahamas, Los Roques (Venezuela), and Alacranes (Mexico). Seagrass extent at each site is reported in Table 5.3. For comparative purposes, Figure 5.5 also includes current data available for seagrass extent in the Bahamas from UNEP-WCMC. Figure 5.6 includes sites for which qualitative comparisons were drawn with previously published seagrass maps. At these two sites and for areas of Martinique, Guadeloupe, and Roátan (not shown here) the two sets of products showed high consistency in distribution and extent of seagrass meadows.  184  As existing maps for Alacranes and Los Roques had also been derived from Landsat images, further analysis was undertaken for these two sites. At Alacranes, seagrass beds were predominantly found along the eastern rim of the bank. Minor discrepancies included seagrass beds mapped along the north part of the outer slope and areas within reef passes. Seagrass distribution and extent for my product at Los Roques closely matched those provided by Schweizer et al. (2005). Main differences included thinner seagrass margins in my product along the mangrove-lined northern and southeastern portions of the central atoll. While visual comparisons do not constitute quantitative accuracy assessments per se, the close agreement between my maps and those produced (a) with significant quantities of ground-truthed data, and (b) using depth correction techniques, provides confidence in the methodology developed for the purposes of this study.  5.4 Discussion 5.4.1  Remote sensing considerations Accuracies reported here span a broad range of values (46 - 88 %; Table 5.2), but they are comparable to those from previous Landsat-based seagrass mapping efforts. Despite the diverse datasets used to assess the accuracy of thematic maps, values do not reveal any positive or negative bias towards a ‘groundtruthing’ source. For example, two of the three poorest overall accuracies recorded (Majahual, Mexico, and Belize patch reef), and the best overall accuracy (Akumal, Mexico), were evaluated based on IKONOS image interpretation. For the two areas with highest accuracies, San Blas and Akumal, different sources were used in the assessment: in situ and IKONOS respectively.  Overall accuracies achieved in this study are well within the range of values reported in previous studies within the Caribbean, including results obtained using water column correction techniques 185  (e.g., Schweizer et al. 2005 for Los Roques). For most areas, the use of Millennium polygons to guide a priori contextual editing emerged as a useful approach to limit misclassifications in the absence of a formal water column correction step. However, the site west of La Parguera (Puerto Rico) stands out as a clear exception. This location is a wide cross-shelf area with a gentle slope from the shore down to a depth of 40 m, and is characterised by a variety of different habitats that are spectrally similar (deep diffuse seagrass, shallow dense seagrass, gorgonian plains, coral escarpment, and hard substrate covered with varying densities of algae). Yet, this mosaic of habitats is included in one single Millennium polygon labelled as ‘shelf slope’. In such instances, the practitioner is therefore faced with the usual challenges of benthic habitat mapping. Application of water column correction as an image pre-processing step to these sites is expected to yield improved accuracies. As a note, the very high accuracy reported by NOAA for Puerto Rico (100 % for submerged vegetation) is for the area of La Parguera itself (NCCOS 2001), characterised by much shallower seas and several distinct Millennium classes which facilitate classification.  Given the scale of the work presented here, it is necessary to put into perspective the different groundtruth data sets employed and the maps produced in previous studies. Although all authors mapped ‘dense’, ‘medium-dense’, and ‘sparse’ seagrass beds according to some comparable threshold of seagrass density, a closer look at individual studies reveals variations in definitions: ƒ NOAA has a fairly detailed scheme, with 5 seagrass classes, that needed to be interpreted and simplified into three classes of seagrass density (NCCOS 2001); ƒ in situ data for Lee Stocking Island and Andros were semi-quantitative, using an index of cover on a scale from 0 to 5 during the surveys (Andréfouët et al. 2003); ƒ in situ data for San Blas made use of a continuous measure of seagrass cover along boat tracks (Andréfouët, unpublished data). The data set thus needed to be broken down by Landsat pixels, resulting in only one class of seagrass density, as variations in cover were very patchy and tended to occur within one Landsat pixel; 186  ƒ data generated through visual-interpretation of IKONOS imagery, even when trained using well- know sites, may not always generate consistent density categories between sites. Consistency is difficult to maintain because of the variability in sediments and seagrass beds themselves, which may produce different signatures for similar benthic cover and densities. Variations such as leaves colonized by epibionts or calcareous growth, beds temporarily just below or even above the water surface at low tide, and dark background sediments, may all lower the spectral contrast of seagrass blades (Fyfe 2003), complicating the interpretation of densities. Similar challenges apply to categorizing in situ data despite measurable field experience (e.g., Lee Stocking Island, Andros, Glovers).  As algae can easily be mistaken for seagrass on Landsat images due to their highly similar spectral signatures (Green et al. 2000, Schweizer et al. 2005), a brief discussion of how I dealt with the issue follows. All ground-truth data that reported dominance (> 70 %) of algae were included in the ‘other’ class. Seagrass beds with varying levels of algae density were included in the appropriate corresponding seagrass class. Although very few ‘other’ pixels were classified as seagrass, seagrass pixels had a tendency to be inappropriately classified as ‘other’. In some instances, availability of several images acquired at different dates for individual Landsat scenes, recorded significant changes in the extent of submerged vegetation over short time periods (Figure 5.7). Since tropical seagrass beds are typically stable over the temporal scale of years, even in the advent of severe storms (but see also Fourqurean and Rutten 2004 and Byron and Heck 2006), these variations are most likely caused by ephemeral changes in algal biomass, and/or cyanobacterial blooms. Accounting for this factor during validation of final products allowed for large errors potentially associated with overestimating the extent of dense seagrass beds to be avoided. In some instances, such as across the shallow Bahama Banks, this type of image comparison constituted a critical aspect of the mapping effort.  187  High values registered for producer and user accuracies of the ‘other’ class indicate that the thematic maps do not overestimate seagrass extent. Moreover, confusion between individual seagrass classes explained most of the low accuracy. This is confirmed by the high accuracy results obtained in San Blas and Akumal where only one seagrass class was mapped. In these two focal areas, seagrass beds were primarily located on back reefs and terraces, and were typically small in size. Given the spatial resolution of Landsat images (30 m pixels), it was not possible to confidently discriminate between seagrass patches of different densities and, thus, only one seagrass class was mapped.  Uncertainty in pixel classification (e.g., differentiating between fuzzy ‘medium’ or ‘sparse’ classes given slight differences in class definitions) led me to collate the initial three-class seagrass typology into a two-class seagrass scheme. Although this approach may not be entirely satisfactory for all applications, given the constraints and goals of this study, the methodology adopted is a valuable first attempt at obtaining realistic figures of seagrass extent and distribution across the WCR.  Areas with dense seagrass cover are frequently patchy and narrow. These were generally correctly captured in shape and extent in classified images. Yet ground-truth points for dense seagrass often fell just onto the patch’s edge, or just outside the area classified as dense, leading to an incorrect classification assessment for that class. Misregistration of Landsat images, and/or ground-truth points, may partly explain this observation. Landsat image specifications may have up to 250 m geolocation errors (up to 7 pixels) (NASA 2007). My field experience further demonstrates that common errors are due to being off by one to two pixels (30 - 60 m); a distance still large enough to miss small targets defined on IKONOS images or in situ. It is usually possible, following personal data collection, to manually correct either images or GPS points to ensure their proper overlap. Here, such adjustments were impossible as data were collected from widely different sources. The problem was less common for medium and sparse seagrass beds, which generally cover wider areas. Areas of very low cover (< 5 %), which includes sandy areas, were generally classified as ‘other’. 188  Since one of my principle mapping objective was to re-estimate green turtle carrying capacity for the Caribbean, accurate mapping of low density seagrass beds was not of primary importance. Optimal turtle foraging grounds tend to consist of dense seagrass beds. I therefore assumed that accurate mapping of dense and medium-dense seagrass beds was an essential first step towards establishing turtle (and other) conservation targets.  Based on these observations, I suggest that thematic maps are both more useful and more valid for spatial analysis than the results of confusion matrices alone may indicate. Consistently higher values could have been achieved had images with high spectral and/or spatial resolution been used to map seagrass extent (Mumby and Edwards 2002, Andréfouët et al. 2003, Hochberg and Atkinson 2003). However, data availability and costs justify the use of Landsat images here. As of 2007, IKONOS and Quickbird licensed data cost 15 - 20 US$ per km2 depending on products, availability in archives, or needs for tasking an acquisition. In comparison, an archived copyright-free Landsat 7 image costs 600 US$, corresponding to a cost of ~ 0.02 US$ per km2. Therefore, although enhancement in accuracies and resolution may be substantial using IKONOS, Quickbird, or hyperspectral data - to date, given acquisition costs - none of these solutions present realistic approaches for large scale mapping efforts.  5.4.2  Implications for conservation While I recognize that my mapping efforts have limitations, results obtained to date are encouraging given the scale of the achievement, the rapid production of maps applying a consistent and uniform methodology, and the objectives to (a) regionally assist with management and conservation planning targets in a cost effective fashion; and (b) regionally re-assess carrying capacity estimates for green turtles.  189  Turtles and many fish species may move over large spatial scales and across different environments over the course of their life histories. Providing Caribbean-wide habitat information therefore significantly improves our ability to manage and conserve the goods and services provided by seagrass beds at biologically relevant scales (Olson and Dinerstein 2002, Stevens 2002). Indeed, it has been suggested that effective mapping for successful conservation should be carried out at transboundary (i.e., ecoregional) scale (Beck and Odaya 2001, Lourie and Vincent 2004), with a strong emphasis on methodological consistency. The products may also provide researchers and managers with a useful and much needed (Creed et al. 2003) baseline to monitor changes registered in seagrass ecosystems over time due to mounting human pressures on coastal ecosystems (Burke and Maidens 2004), and to develop spatially explicit models of impacts due to disturbances (e.g., Kelly et al. 2001). Even for management at a national level, the maps developed through this research can form the basis of efforts targeted at capturing important landscape patterns that may have significant management implications at a more local scale (Fonseca et al. 2002, Bell et al. 2006).  The average accuracy of 68 % across all sites does not adequately reflect the utility of organizing disaggregated spatial data into consolidated map products. Ultimately, which product is more environmentally relevant (i.e., higher classification accuracy but smaller spatial extent versus lower classification accuracy but larger spatial extent) will depend on the conservation task at hand. In the case of green turtle conservation for instance, deriving a more reliable estimate of seagrass extent represents a critical first step in re-assessing a regional estimate of carrying capacity. This in turn can inform the goals set for population recoveries. In doing so it is noteworthy that the theoretical number of green turtles that could be sustained by present seagrasses might not be consistent with turtle numbers that would assure optimal long-term seagrass productivity. This is clearly demonstrated in the diverse estimates of turtle carrying capacity for the WCR, derived from varying levels of seagrass productivity: from 586 million turtles for highly productive seagrass beds to 39 million in stressed meadows (Bjorndal et al. 2000). 190   5.4.3  Conclusion and perspectives Recognizing the urgency in establishing enhanced GIS data sets on the spatial distribution of habitats for large scale ecological and conservation applications, this study assessed the feasibility of using Landsat sensors to map seagrass beds effectively and subsequently apply consistent methods throughout the WCR. I processed a total of 40 Landsat 7 (ETM+) images covering regionally representative sites with contrasted sizes, structure, geomorphology, and seagrass bed extent. The results reported here are encouraging for the completion of a WCR-wide map of seagrass habitat using Landsat images, MCMRP coral reef products, and further IKONOS imagery. Such an effort is currently in progress, and builds on the discussion developed here. This is, to my knowledge, the largest such effort worldwide.  Accuracies obtained at individual sites are in agreement with local studies previously published using different methods. Comparison of the regional results also point to where some of the traditional image processing challenges in shallow coastal environments will be the most acute: wide cross shelf areas, such as those around Puerto Rico. Similar environments in the Bahamas, Cuba, and Florida, will require image depth-correction if useful accuracies are to be achieved.  Next steps include the processing of all remaining available images for the region and compiling the results into a GIS layer to facilitate use by the conservation and scientific community. In parallel, further map validation based on local expertise and high resolution images will help to highlight weaknesses and strengths of my products. There are mounting threats facing coastal areas (Burke and Maidens 2004), and significant gaps in existing habitat databases. Therefore, availability of consistent regional seagrass habitat maps for the WCR will assist governments and their partners in developing successful conservation plans. Furthermore, the design employed in this study should 191  also be useful in other tropical regions where reliable habitat data are critically needed for integrated coastal management purposes. 192  Table 5.1 - List of sites for which thematic seagrass habitat maps were derived in this paper. Available ancillary data and references on previous remote sensing and habitat mapping work at these same sites are also presented. Where applicable, the type of data used to assess accuracy of my products is indicated.   Focal area Landsat Path-Row Accuracy assessment data IKONOS data References Bahamas    13-41, 14-41, 14-42, 13-42, 12-42, 15-43, 14-43, 13-43, 12-43, 11-43, 15-44, 14-44, 13-44, 12-44, 11-44, 10-44, 12-45, 11-45, 10-45, 9-45 IKONOS & in situ    Lee Stocking Island Andros Island. (AUTEC) Armstrong (1993); Andréfouët et al. (2003); Call et al. (2003), and Louchard et al. (2003) Belize    19-48, 18-48, 19-49, 18-49 IKONOS    Lighthouse Atoll Glovers Atoll Barrier Reef section Patch Reef section Andréfouët et al. (2003) Mexico (Yucatán coast)   20-45, 19-45, 18-45, 19-46, 18-46, 19-47, 18-47 IKONOS   Akumal Boca Paila Mahahual Andréfouët et al. (2003); Garza-Perez et al. (2004) Roatán (Honduras) 17-49 IKONOS Roatán Maeder et al. (2002) St Croix (US Virgin Island)  4-48 NOAA   NOAA (2001) Puerto Rico (south coast) 5-48 NOAA  NOAA (2001) San Blas offshore banks and islands (Panama) 11-53 in situ  Andréfouët and Guzmán (2005) Los Roques (Venezuela) 4-52 in situ  Schweizer et al. (2005) Alacranes Bank (Mexico) 20-45 N/A (published value: 77 %)  Bello-Pineda et al. (2005) Guadeloupe 1-49 N/A (published value: 95.7 %)  Chauvaud et al. (2001) Bay du Robert (Martinique) 1-50 N/A (published value: 94 %)  Chauvaud et al. (1998) Providence Island (Colombia) 14-51 N/A  Díaz et al. (2003) San Andrés (Colombia) 14-51 N/A  Díaz et al. (2003) 193  Table 5.2 - Accuracy achieved for two seagrass classes (* with the exception of San Blas and Akumal) and one generic 'other' class using the following data for the assessment: IKONOS imagery, direct ground-truth data collected in situ, and NOAA ground-truth data. n is the total number of points used for each site.  Focal area Source n Accuracy Dense seagrass (> 70 %) Medium/Sparse Seagrass (< 70 %) Other  Overall Accuracy (%) Kappa coefficient Producer (%) User (%) Producer (%) User (%) Producer (%) User (%) Lee Stocking Island IKONOS & in situ 170 71.8 0.48 15.4 50.0 74.0 72.5 86.2 72.3 East Andros IKONOS & in situ 502 63.3 0.19 41.5 32.3 23.3 50.0 74.1 77.3 Roatán IKONOS 296 71.3 0.54 65.1 50.0 55.4 64.8 82.7 86.7 Lighthouse Atoll IKONOS 181 69.1 0.53 25.0 76.5 86.2 53.8 87.3 87.3 Glovers Atoll IKONOS 120 71.7 0.51 75.0 32.2 31.7 68.5 95.5 87.7 Belize patch reefs IKONOS 76 64.5 0.05 N/A N/A 0.0 0.0 96.1 66.2 Belize barrier reef IKONOS 102 55.9 0.34 33.3 29.6 42.9 85.7 86.1 57.4 Mahahual IKONOS 54 55.6 0.18 21.8 83.3 0.0 0.0 100 52.1 Boca Paila IKONOS 87 72.4 0.52 52.4 55.0 57.2 42.1 84.6 91.7 Akumal* IKONOS 57 87.7 0.46 N/A N/A 57.1 50.0 92.0 93.9 St Croix  NOAA 294 63.9 0.24 46.8 47.8 24.1 12.3 72.9 83.2 La Parguera NOAA 81 45.7 0.0 59.4 47.8 100.0 8,3 32.6 60.9 San Blas* in situ 293 85.3 0.57 66.2 67.2 N/A N/A 90.8 90.4 Los Roques in situ 279 77.1 0.64 90.5 79.2 49.4 88.9 86.8 71.2 194  Table 5.3 - Estimated seagrass surface areas (in km2) at each study site, with the exception of Puerto-Rico due to low accuracy achieved for the product.  Site Seagrass (km2) Bahamas 65,436 Belize 2,092 Mexico (Yucatán coast) 1,319 Roatán (Honduras) 64 St Croix (US Virgin Island) 62 San Blas offshore banks and islands (Panama) 11 Los Roques (Venezuela) 150 Alacranes Bank (Mexico) 58 Guadeloupe Island (France) 179 Providence Island (Colombia) 2.6 San Andrés (Colombia) 17  195               Figure 5.1 - Map of the Caribbean region and geographic location of sites processed for this study. Belize PR = Belize Patch Reef, Belize BR = Belize Barrier Reef (background map from www.reefbase.org). The lower panel shows data (both points and polygons) currently available on seagrass distribution for the region as displayed in “World Atlas of Seagrasses” (Green and Short 2003) and as taken here from the interactive IMAPS system (http://storp.unep-wcmc.org). Inset: Global location of region.   196  Calibration, mosaic Masking w/ Millennium polygons Supervised classification(s) In situ data IKONOS interpretation Literature search Accuracy assessment Landsat data More than one Path-row? yes no Acceptable? yes no Vectorization Transfer to GIS Bathymetric correction Millennium Coral Reef products Application no yes Stand-by (better data) Bathy correction already performed?                      Figure 5.2 - Flow chart of data processing steps involved in mapping seagrasses at the large scale of the Caribbean Region (WCR). ‘Path-Row’ > 1 = is more than one path-row covering the study site? Rounded boxes = objects; square boxes = processing steps; diamond shaped boxes = highlight specific tests or conditions that need to be fulfilled. Bathymetric correction is highlighted in grey as it was not utilised here, but is recommended for sites where depth presents a confounding factor (e.g., site west of La Parguera, Puerto Rico). 197                      Figure 5.3 - Examples of Millennium polygons for the Belizean focal area. Detailed geomorphological classes provided by the Millennium Coral Reef Mapping Project allow for a priori contextual editing prior to spectral supervised classification. The figure aims to highlight the complexity and details of the structures and products provided. This area includes close to 100 classes; thus for the purposes of readability and simplicity, no thematic legend has been provided. 198                   Figure 5.4 - Accuracy assessment points selected for Lee Stocking Island (Exuma, Bahamas) using an IKONOS image and augmented with in situ observations. Red: dense seagrass, yellow: medium-sparse seagrass, green: ‘other’.    199                      Figure 5.5 - Thematic map for all of the Bahamas and for the Little Bahama Bank with two seagrass classes (green) and a single ‘other’ class (white). Here, the ‘other’ class includes areas that were removed by a priori contextual editing for the actual seagrass mapping effort as well as processed areas classified as ‘non-seagrass’. The upper right panel shows the current seagrass product available from UNEP-WCMC, displayed with the interactive IMAPS service (http://storp.unep-wcmc.org). It is included here for comparative purposes and highlights commission and omission errors in the existing seagrass coverage for the Bahamas. 200                    Figure 5.6 - Examples of habitat maps displaying two seagrass classes and a single ‘other’ class that can be compared with previous published peer-reviewed studies: Los Roques (to use in comparison with Schweizer et al. 2005), and Alacranes (to be compared to Bello-Pineda et al. 2005). 201                    Figure 5.7 - Examples of temporal variation in vegetation cover, probably due to algal blooms, in two pairs of Landsat images taken less than 14 months apart (path-row 12-53, Bahamas). In the top pair, extensive darker areas appear on the bottom image. Note also (arrow most left) that dark areas have shifted to bright areas in the same period of time. In the second image pair, patches of dark water and darker bottoms appear on the bottom image (left arrow). Over the same time period (right arrow), vegetation has decreased on hard-bottom areas. These rapid changes from dark to bright and vice-versa are not typical of dense or medium-dense seagrass beds and are most likely to be associated with changes in algal cover. 202  5.5  References Andréfouët, S. and Guzmán, H. M. (2005) Coral reef distribution, status and geomorphology- biodiversity relationship in Kuna Yala (San Blas) archipelago, Caribbean Panama. Coral Reefs, 24 (1): 31-42. Andréfouët, S., Muller-Karger, F., Hochberg, E., Hu, C. and Carder, K. (2001) Change detection in shallow coral reef environments using Landsat 7 ETM+ data. Remote Sensing of Environment, 78 (1-2): 150-162. Andréfouët, S., Muller-Karger, F. E., Robinson, J. A., Kranenburg, C. J., Torres-Pulliza, D., Spraggins, S. A. and Murch, B. (2006) Global assessment of modern coral reef extent and diversity for regional science and management applications: A view from space. In 10th International Coral Reef Symposium (eds Y. Suzuki, T. Nakamori, M. Hidaka, H. Kayanne, B. E. Casareto, K. Nadaoka, H. Yamano, M. Tsuchiya and K. Yamazato), pp. 1732-1745. Japanese Coral Reef Society, Okinawa, Japan. Andréfouët, S., Kramer, P., Torres-Pulliza, D., Joyce, K. E., Hochberg, E. J., Garza-Perez, R., Mumby, P. J., Riegl, B., Yamano, H., White, W. H., Zubia, M., Brock, J. C., Phinn, S. R., Naseer, A., Hatcher, B. G. and Muller-Karger, F. E. (2003) Multi-site evaluation of IKONOS data for classification of tropical coral reef environments. Remote Sensing of Environment, 88 (1-2): 128-143. Armstrong, R. A. (1993) Remote-sensing of submerged vegetation canopies for biomass estimation. International Journal of Remote Sensing, 14 (3): 621-627. Balmford, A., Crane, P., Dobson, A., Green, R. E. and Mace, G. M. (2005) The 2010 challenge: Data availability, information needs and extraterrestrial insights. Philosophical Transactions of the Royal Society B-Biological Sciences, 360 (1454): 221-228. Beck, M. W. and Odaya, M. (2001) Ecoregional planning in marine environments: Identifying priority sites for conservation in the northern Gulf of Mexico. Aquatic Conservation: Marine and Freshwater Ecosystems, 11 (4): 235-242. Beck, M. W., Heck Jr., K. L., Able, K. W., Childers, D. L., Eggleston, D. B., Gillanders, B. M., Halpern, B. S., Hays, C. G., Hoshino, K., Minello, T. J., Orth, R. J., Sheridan, P. F. and Weinstein, M. P. (2003) The role of nearshore ecosystems as fish and shellfish nurseries. Issues in Ecology. Ecological Society of America, Washington, D.C., USA. 12 pp. 203  Bell, S. S., Fonseca, M. S. and Stafford, N. B. (2006) Seagrass ecology: New contributions from a landscape perspective. In Seagrasses: Biology, Ecology and Conservation. (eds A. W. D. Larkum, R. J. Orth and C. M. Duarte), pp. 625-645. Springer, Dordrecht, The Netherlands. Bello-Pineda, J., Liceaga-Correa, M., Hernandez-Nunez, H. H. and Ponce-Hernandez, R. (2005) Using aerial video to train the supervised classification of Landsat TM imagery for coral reef habitats mapping. Environmental Monitoring and Assessment, 105 (1-3): 145-164. Bjorndal, K. A., Bolten, A. B. and Chaloupka, M. Y. (2000) Green turtle somatic growth model: Evidence for density dependence. Ecological Applications, 10 (1): 269-282. Burke, L. and Maidens, J. (2004) Reefs at risk in the Caribbean. World Resources Institute, Washington, D.C., USA. 80 pp. Byron, D. and Heck, K. L. (2006) Hurricane effects on seagrasses along Alabama's Gulf coast. Estuaries and Coasts, 29 (6): 939-942. Call, K. A., Hardy, J. T. and Wallin, D. O. (2003) Coral reef habitat discrimination using multivariate spectral analysis and satellite remote sensing. International Journal of Remote Sensing, 24 (13): 2627-2639. Chauvaud, S. (1997) Cartographie par télédetection à haute résolution des biocénoses marine côtières de la Guadeloupe et de la Martinique. Estimation de la biomasse et de la production primaire des herbiers à Thalassia testudinum. Doctoral thesis. Université de Bretagne Occidentale, Brest, France. Chauvaud, S., Bouchon, C. and Manière, R. (1998) Remote sensing techniques adapted to high resolution mapping of tropical coastal marine ecosystems (coral reefs, seagrass beds and mangrove). International Journal of Remote Sensing, 19 (18): 3625-3639. Chauvaud, S., Bouchon, C. and Manière, R. (2001) Cartographie des biocénoses marines de Guadeloupe à partir de données SPOT (récifs coralliens, phanérogames marines, mangroves). Oceanologica Acta, 24 (Supplement): S3-S16. Congalton, R. G. (1991) A review of assessing the accuracy of classifications of remotely sensed data. Remote Sensing of Environment, 37: 35-46. Congalton, R. and Green, K. (1999) Assessing the Accuracy of Remotely Sensed Data: Principles and Practices, Lewis Publishers, New York (NY), USA. 137 pp. Creed, J. C., Phillips, R. C. and van Tussenbroek, B. I. (2003) The seagrasses of the Caribbean. In The World Atlas of Seagrasses. (eds E. P. Green and F. T. Short), pp. 234-240. University of California Press, Berkeley (CA), USA. Dahlgren, C. P., Kellison, G. T., Adams, A. J., Gillanders, B. M., Kendall, M. S., Layman, C. A., Ley, J. A., Nagelkerken, I. and Serafy, J. E. (2006) Marine nurseries and effective juvenile habitats: Concepts and applications. Marine Ecology Progress Series, 312: 291-295. 204  den Hartog, C. and Kuo, J. (2006) Taxonomy and biogeography of seagrasses. In Seagrasses: Biology, Ecology and Conservation. (eds A. W. D. Larkum, R. J. Orth and C. M. Duarte), pp. 1-23. Springer, Dordrecht, The Netherlands. Díaz, J. M. M. and Gómez-López, D. I. (2003) Cambios históricos en la distribución y abundancia de praderas de pastos marinos en la Bahía de Cartagena y áreas aledañas (Colombia). Boletin de Investigaciones Marinas y Costeras, 32: 57-74. Díaz, J. M., Barrios, L. M. and Gómez, D. I. (2003) Las praderas de pastos marinos en Colombia: Estructura y distribución de un ecosistema estratégico, Instituto de Investigaciones Marinas y Costeras (INVEMAR), Santa Marta, Colombia. 159 pp. Duarte, C. M. (2002) The future of seagrass meadows. Environmental Conservation, 29 (2): 192-206. Duarte, C. M. and Cebrián, J. (1996) The fate of marine autotrophic production. Limnology and Oceanography, 41 (8): 1758-1766. Duarte, C. M. and Chiscano, C. L. (1999) Seagrass biomass and production: A reassessment. Aquatic Botany, 65: 159-174. Duffy, J. E. (2006) Biodiversity and the functioning of seagrass ecosystems. Marine Ecology Progress Series, 311: 233-250. Fonseca, M., Whitfield, P. E., Kelly, N. M. and Bell, S. S. (2002) Modeling seagrass landscape pattern and associated ecological attributes. Ecological Applications, 12 (1): 218-237. Foody, G. M. (2002) Status of land cover classification accuracy assessment. Remote Sensing of Environment, 80: 185-201. Fourqurean, J. W. and Robblee, M. B. (1999) Florida Bay: A history of recent ecological changes. Estuaries, 22 (2B): 345-357. Fourqurean, J. W. and Rutten, L. M. (2004) The impact of hurricane Georges on soft-bottom, back reef communities: Site- and species-specific effects in south Florida seagrass beds. Bulletin of Marine Science, 75 (2): 239-257. Fyfe, S. K. (2003) Spatial and temporal variation in spectral reflectance: Are seagrass species spectrally distinct? Limnology and Oceanography, 48 (1): 464-479. Gallegos, M. E., Merino, M., Rodriguez, A., Màrba, N. and Duarte, C. D. (1994) Growth patterns and demography of pioneer Caribbean seagrasses Halodule wrightii and Syringodium filiforme. Marine Ecology Progress Series, 109: 99-104. Garza-Perez, J. R., Lehmann, A. and Arias-Gonzalez, J. E. (2004) Spatial prediction of coral reef habitats: integrating ecology with spatial modeling and remote sensing. Marine Ecology Progress Series, 269: 141-152. 205  Gillanders, B. M. (2006) Seagrasses, fish and fisheries. In Seagrasses: Biology, Ecology and Conservation. (eds A. W. D. Larkum, R. J. Orth and C. M. Duarte), pp. 503-536. Springer, Dordrecht, The Netherlands. Green, E. P. and Short, F. T. (2003) The World Atlas of Seagrasses, University of California Press, Berkeley (CA), USA. 310 pp. Green, E. P., Mumby, P. J., Edwards, A. J. and Clark, C. D. (2000) Remote sensing handbook for tropical coastal management, UNESCO Publishing, Paris, France. 316 pp. Groom, G. B., Fuller, R. M. and Jones, A. R. (1996) Contextual correction: Techniques for improving land cover mapping from remotely sensed data. International Journal of Remote Sensing, 17: 69-89. Heck, K. L., Hays, G. and Orth, R. J. (2003) Critical evaluation of the nursery role hypothesis for seagrass meadows. Marine Ecology Progress Series, 253: 123-136. Hemminga, M. A. and Duarte, C. M. (2000) Seagrass Ecology: An Introduction, Cambridge University Press, Cambridge, UK. 298 pp. Hernández-Cruz, L. R., Purkis, S. J. and Riegl, B. M. (2006) Documenting decadal spatial changes in seagrass and Acropora palmata cover by aerial photography analysis in Vieques, Puerto Rico: 1937-2000 Bulletin of Marine Science, 79 (2): 401-414. Hochberg, E. and Atkinson, M. (2003) Capabilities of remote sensors to classify coral, algae, and sand as pure and mixed spectra. Remote Sensing of Environment, 85 (2): 174-189. Hu, C., Muller-Karger, F., Andréfouët, S. and Carder, K. (2001) Atmospheric correction and cross- calibration of Landsat-7/ETM+ imagery over aquatic environments:  Multi-platform approach using SeaWiFS/MODIS. Remote Sensing of Environment, 78: 99-107. Jackson, J. B. C. (1997) Reefs since Columbus. Coral Reefs, 16: S23-S32. Jackson, J. B. C. (2001) What was natural in the coastal oceans? Proceedings of the National Academy of Science, 98 (10): 5411-5418. Jackson, J. B. C., Kirby, M. X., Berger, W. H., Bjorndal, K. A., Botsford, L. W., Bourque, B. J., Bradbury, R. H., Cooke, R., Erlandson, J., Estes, J. A., Hughes, T. P., Kidwell, S., Lange, C. B., Lenihan, H. S., Pandolfi, J. M., Peterson, C. H., Steneck, R. S., Tegner, M. J. and Warner, R. R. (2001) Historical overfishing and the recent collapse of coastal ecosystems. Science, 293 (5530): 629-638. Kelly, N. M., Fonseca, M. and Whitfield, P. (2001) Predictive mapping for management and conservation of seagrass beds in North Carolina. Aquatic Conservation: Marine and Freshwater Ecosystems, 11 (6): 437-451. 206  Krause-Jensen, D., Quaresma, A. L., Cunha, A. H. and Greve, T. M. (2004) How are seagrass distribution and abundance monitored? In European Seagrasses: An Introduction to Monitoring and Management. (eds J. Borum, C. M. Duarte, D. Krause-Jensen and T. M. Greve), pp. 45-53. The M&MS Project. Lillesand, T. M., Kiefer, R. W. and Chipman, J. W. (2004) Remote Sensing and Image Interpretation, John Wiley & Sons. 763 pp. Lotze, H. K., Lenihan, H. S., Bourque, B. J., Bradbury, R. H., Cooke, R. G., Kay, M. C., Kidwell, S. M., Kirby, M. X., Peterson, C. H. and Jackson, J. B. C. (2006) Depletion, degradation, and recovery potential of estuaries and coastal seas. Science, 312 (5781): 1806-1809. Louchard, E. M., Reid, R. P., Stephens, F. C., Davis, C. O., Leathers, R. A. and Downes, T. V. (2003) Optical remote sensing of benthic habitats and bathymetry in coastal environments at Lee Stocking Island, Bahamas: A comparative spectral classification approach. Limnology and Oceanography, 48 (1): 511-521. Lourie, S. A. and Vincent, A. C. J. (2004) Using biogeography to help set priorities in marine conservation. Conservation Biology, 18 (4): 1004-1020. Luczkovich, J., Wagner, T., Michalek, J. and Stoffle, R. (1993) Discrimination of coral reefs, seagrass meadows, and sand bottom types from space: A Dominican Republic case study. Photogrammetric Engineering and Remote Sensing, 59: 385-389. Maeder, J., Narumalani, S., Rundquist, D., Perl, R., Schalles, J., Hutchins, K. and Keck, J. (2002) Classifying and mapping general coral-reef structure using IKONOS data. Photogrammetric Engineering and Remote Sensing, 68: 1297-1305. McNaughton, S. J. (1979) Grazing as an optimization process - Grass ungulate relationships in the Serengeti. American Naturalist, 113 (5): 691-703. Moran, K. L. and Bjorndal, K. A. (2005) Simulated green turtle grazing affects structure and productivity of seagrass pastures. Marine Ecology Progress Series, 305: 235-247. Moran, K. L. and Bjorndal, K. A. (2007) Simulated green turtle grazing affects nutrient composition of the seagrass Thalassia testudinum. Marine Biology, 150 (6): 1083-1092. Mumby, P. J. and Edwards, A. J. (2002) Mapping marine environments with IKONOS imagery: Enhanced spatial resolution can deliver greater thematic accuracy. Remote Sensing of Environment, 82 (2-3): 248-257. Mumby, P. J. and Harborne, A. R. (1999) Development of a systematic classification scheme of marine habitats to facilitate regional management and mapping of Caribbean coral reefs. Biological Conservation, 88 (2): 155-163. 207  Mumby, P. J., Clark, C. D., Green, E. P. and Edwards, A. J. (1998) Benefits of water column correction and contextual editing for mapping coral reefs. International Journal of Remote Sensing, 19 (1): 203-210. Mumby, P. J., Green, E. P., Edwards, A. J. and Clark, C. D. (1999) The cost-effectiveness of remote sensing for tropical coastal resources assessment and management. Journal of Environmental Management, 55 (3): 157-166. NASA (2007) Landsat 7: Science data users handbook. Accessed