Canada Archives Canada Published Heritage Direction Du Branch Patrimoine De I'edition

Population- and Community-Level Consequences of the Exploitation of Large Predatory Marine Fishes

Julia K. Baum

Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Dalhousie University Halifax, Nova Scotia September 2007

395 Wellington Street 395, rue Wellington Ottawa ON K1A0N4 Ottawa ON K1A0N4 Canada Canada

Your file Votre reference ISBN: 978-0-494-39069-6 Our file Notre reference ISBN: 978-0-494-39069-6

NOTICE: AVIS: The author has granted a non L'auteur a accorde une licence non exclusive exclusive license allowing Library permettant a la Bibliotheque et Archives and Archives Canada to reproduce, Canada de reproduire, publier, archiver, publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public communicate to the public by par telecommunication ou par Plntemet, prefer, telecommunication or on the Internet, distribuer et vendre des theses partout dans loan, distribute and sell theses le monde, a des fins commerciales ou autres, worldwide, for commercial or non sur support microforme, papier, electronique commercial purposes, in microform, et/ou autres formats. paper, electronic and/or any other formats.

The author retains copyright L'auteur conserve la propriete du droit d'auteur ownership and moral rights in et des droits moraux qui protege cette these. this thesis. Neither the thesis Ni la these ni des extraits substantiels de nor substantial extracts from it celle-ci ne doivent etre imprimes ou autrement may be printed or otherwise reproduits sans son autorisation. reproduced without the author's permission.

In compliance with the Canadian Conformement a la loi canadienne Privacy Act some supporting sur la protection de la vie privee, forms may have been removed quelques formulaires secondaires from this thesis. ont ete enleves de cette these.

While these forms may be included Bien que ces formulaires in the document page count, aient inclus dans la pagination, their removal does not represent il n'y aura aucun contenu manquant. any loss of content from the thesis. Canada DALHOUSIE UNIVERSITY

To comply with the Canadian Privacy Act the National Library of Canada has requested that the following pages be removed from this copy of the thesis:

Preliminary Pages Examiners Signature Page (pii) Dalhousie Library Copyright Agreement (piii)

Appendices Copyright Releases (if applicable) For RAM

IV Table of Contents

List of Tables viii List of Figures xi Abstract xiv List of Abbreviations and Symbols Used xv Acknowledgements xvii CHAPTER 1 Introduction 1 CHAPTER 2 Shifting Baselines and the Decline of Pelagic Sharks in the Gulf of Mexico 11 ABSTRACT 12 INTRODUCTION 12 METHODS 14 Data 14 Shark Species 16 Modelling Change in Abundance 17 Estimating Change in Size 21 RESULTS 22 Unstandardized Catch Rates 22 Relative Abundance and Size Estimates 26 DISCUSSION 27 SUPPLEMENTARY MATERIAL 32

v Effect of Soak Time and Depth on Model Estimates 32 Model Robustness 33 Effects of Other Operational Factors 34 ACKNOWLEDGEMENTS 35

CHAPTER 3 Estimating Trends in Abundance from Fishery-Dependent Data: Recent Changes in Northwest Atlantic Pelagic Shark Populations 36 ABSTRACT 37 INTRODUCTION 37 METHODS 41 Data and Shark Species 41 Shark Catch Rates in U.S. Pelagic Longline Observer and Logbook Data 47 Observer Data Models 47 RESULTS 54 Shark Catch Rates in U.S. Pelagic Longline Observer and Logbook Data 54 Model Estimates of Recent Changes in Shark Abundance 63 DISCUSSION 70 Shark Catch Rates in U.S. Pelagic Longline Observer and Logbook Data 70 Estimates of Recent Changes in Shark Abundance 71 CHAPTER 4 Cascading Effects of the Loss of Apex Predatory Sharks from a Coastal Ocean 77 ABSTRACT 78 MAIN TEXT 78 SUPPLEMENTARY MATERIAL 94 Materials and Methods 94 Results and Discussion Ill ACKNOWLEDGEMENTS 115

CHAPTER 5 Top-Down Control in the Ocean: Methods, Evidence, Inferred Mechanisms, and Management Implications 116 ABSTRACT 117 1 INTRODUCTION 118 2 INSIGHTS FROM ECOLOGICAL THEORY AND EXPERIMENTS 121

vi 3 OVERCOMING METHODOLOGICAL & DATA LIMITATIONS 125 Improving Methods for Ecosystem-Scale Analyses 127 4 PREDATOR DEPLETIONS REVEAL TOP-DOWN CONTROL 134 Cascading Effects of a Massive Fisheries Collapse 134 Restructuring Oceanic Ecosystems: Mechanisms and Effects 139 5 MANAGEMENT & CONSERVATION IMPLICATIONS 148 6 SYNTHESIS & CONCLUSIONS 150

CHAPTER 6 Conclusions , 153

REFERENCES.. 159 APPENDIX 1 Models of 1992-2005 Observer Data 191 APPENDIX 2 Copyright Permissions 200

Vll List of Tables

Table 1.1. Shark species included in thesis analyses {/ at genus level indicates species within the genus were grouped for analysis) and their life history characteristics (age at maturity (for females), fecundity, size at maturity (for females) and maximum size (FL=fork length, TL=total length)) 8

Table 2.1. Comparison of 1950s and 1990s pelagic longline fishing data from the Gulf of Mexico used in analysis. Mean values ± 1 standard deviation 15

Table 2.2. Recorded oceanic and coastal shark species and sample sizes in 1950s and 1990s datasets, listed within each category in declining order of abundance in the 1950s 17

Table 2.3. Modelled species, coefficients for soak time (Ss) and depth (ysl, ys2, ys3) used to calculate the offset, and efficiency of hooks in the 1990s relative to the 1950s for the mean change in soak time and depth respectively 20

Table 2.4. Modelled species weights in kilograms ±1 standard deviation (sample size). 27

Table 2.5. Estimated declines (± 95% CI) between the 1950s and 1990s of oceanic whitetip, silky and dusky sharks in the Gulf of Mexico from the main model, from models examining the effect of the soak time offset and of the depth offset, and from robustness models 33

Table 3.1. Total of each elasmobranch species recorded in the U.S. Atlantic pelagic longline observer program between 1992 and 2005 43

Table 3.2. Total number of each oceanic and large coastal shark species (group) caught, and the total number of fishing sets monitored, each year between 1992 and 2005 as recorded in the U.S. Atlantic pelagic longline observer program (after data checks and removing shark targeted sets) 44

Table 3.3. Total number of each oceanic and large coastal shark species (group) caught, by fishing area, as recorded in the U.S. Atlantic pelagic longline

viii observer program from 1992 to 2005 (after data checks and removing shark targeted sets) 45

Table 3.4. Characteristics of observed sets in the U.S. pelagic longline fishery between 1992 and 2005 51

Table 3.5. Total number of monitored sets directed towards each target species between 1992 and 2005 in the US Atlantic pelagic longline observer program . 51

Table 3.6. Number of fishing sets monitored in the U.S. Atlantic pelagic longline fishery observer program between 1992 and 2005, in the full data set, in the data set with trips with only one fishing set removed (used in the GLMM-t models) and with vessels with only one trip also removed (used in the GLMM-vt models), in the data set with sets without hook type removed, with trips with only one fishing set also removed, and with vessels with only one trip also removed 52

Table 3.7. Final models (generalized estimating equations with trip marginal (GEE), generalized linear mixed models with trip random (GLMM-t), and with vessel random, trip residual (GLMM-vt)) for each species showing covariates included in the final model (shaded) and statistical significance level of each: * = <0.1; **=<0.01; ***=<0.001; ****=<0.0001, otherwise non-significant.... 69

Table 4.1. Data sources 81

Table 4.2. Survey, fisheries and landings data set descriptions, including area, gear type, season and years sampled, and total sample size 82

Table 4.3. Model results for each species of great shark from each of the research survey and fisheries data sources used in the meta-analysis shown in Figure 4.4, including the first and last year of capture in the data set, the number of years caught, the total number of the species caught, the model estimate (± 95% confidence intervals (CI)) of the instantaneous rate of change, for all years of data (All) and for only those years onwards from the baseline of 1970 (1970-).... 85

Table 4.4. Model results for each elasmobranch mesopredator species from each of the research surveys used in the meta-analysis shown in Figure 4.4, including the first and last year of capture in the data set, the number of years caught, the total number of the species caught, the model estimate (± 95% confidence intervals (CI)) of the instantaneous rate of change, for all years of data (All) and for only those years onwards from the baseline of 1970 (1970-) 89

Table 4.5. Taxa of elasmobranchs (sharks, skates, rays) consumed by the apex (or nearly apex) shark species included in the large shark group 96

Table 4.6. Summary of generalized linear models used to estimate trends in abundance for large sharks and elasmobranch mesopredators 105

IX Table 5.1. Studies identifying top-down control in oceanic ecosystems 128

Table Al. 1. Parameter estimates (including standard error and p-value) for fixed effects in final GLMM-vt model of 1992-2005 observer data for blue shark ... 192

Table A1.2. Parameter estimates (including standard error and p-value) for fixed effects in final GLMM-vt model of 1992-2005 observer data for mako sharks .193

Table A 1.3. Parameter estimates (including standard error and p-value) for fixed effects in final GLMM-vt model of 1992-2005 observer data for thresher sharks 194

Table A1.4. Parameter estimates (including standard error and p-value) for fixed effects in final GLMM-vt model of 1992-2005 observer data for oceanic whitetip shark 195

Table A 1.5. Parameter estimates (including standard error and p-value) for fixed effects in final GLMM-vt model of 1992-2005 observer data for hammerhead sharks 196

Table A 1.6. Parameter estimates (including standard error and p-value) for fixed effects in final GLMM-vt model of 1992-2005 observer data for tiger shark.... 197

Table A 1.7. Parameter estimates (including standard error and p-value) for fixed effects in final GLMM-vt model of 1992-2005 observer data for coastal group 1 (dusky, silky, night sharks) 198

Table A 1.8. Parameter estimates (including standard error and p-value) for fixed effects in final GLMM-vt model of 1992-2005 observer data for coastal group 2 (all Carcharhinus species and unidentified sharks) 199

x List of Figures

Figure 2.1. Map of the Gulf of Mexico showing unstandardized mean catches per 10,000 hooks on yellowfin tuna targeted sets during the day in the 1950's (a, c, e, g) and the 1990's (b, c, f, h) for oceanic whitetip (a, b), silky (c,d), dusky (e, f) andmako (g, h) sharks 23

Figure 2.2. Mean catch rates (± s.d.) of each shark species recorded in the 1950s data and/or 1990s data 25

Figure 2.3. The estimated change in abundance (± 95% CI) between the 1950s and 1990s of oceanic whitetip (OCS), silky (FAL), dusky (DUS), and mako (MAK) sharks 26

Figure 3.1. Map of the Northwest Atlantic Ocean showing the distribution of effort in the U.S. pelagic longline fishery's observer program between 1992 and 2005, categorized by number of sets (0 to 100) 54

Figure 3.2. Maps of the Northwest Atlantic Ocean showing unstandardized catch rates of oceanic shark species: blue shark (a,b), mako sharks (c,d), thresher sharks (e,f), and oceanic whitetip shark (g,h) as recorded in the U.S. pelagic longline logbook (a,c,e,g) and observer (b,d,f,h) data from 1992 to 2000 56

Figure 3.3. Maps of the Northwest Atlantic Ocean (with areas as in Figure 2) showing unstandardized catch rates of large coastal shark species: hammerhead sharks (a,b), tiger shark (c,d), and large coastal sharks of the genus Carcharhinus (e,f) as recorded in the U.S. pelagic longline logbook (a,c,e) and observer (b,d,f) data from 1992 to 2000 59

Figure 3.5. Catch per 1000 hooks (± 1 standard error) in all areas combined, by year for each species (group), as recorded in the observer data from 1992 to 2005 ... 62

xi Figure 3.6. Estimated change in relative abundance (standardized catch per 1000 hooks) between 1992 and 2005 based on the observer data for oceanic shark species: (A) blue, (B) mako, (C) thresher, (D) oceanic whitetip, and large coastal shark species: (E) hammerhead, (F) tiger, (G) coastal shark group 1, (H) coastal shark group 2 64

Figure 3.7. Estimated instantaneous rate of change in abundance in each area (• ± 95% CI) and in all areas combined (• ± 95% CI) between 1992 and 2005 based on the observer data for oceanic shark species: (A) blue, (B) mako, (C) thresher, (D) oceanic whitetip, and coastal shark species: (E) hammerhead, (F) tiger, (G) coastal shark group 1, (H) coastal shark group 2 66

Figure 4.1. Map of the U.S. Atlantic coast showing the location of each research survey, with 200m, 500m, and 1000m isobaths (dotted lines) given for reference 80

Figure 4.2. Change over time in species at each trophic level as estimated from individual data sources: trends in relative abundance (overall trend (solid line) and individual yearly estimates (•)) of great sharks (top row, UNC survey) and elasmobranch mesopredators (middle row, survey acronyms as in Table S3) estimated from GLMs, and the trend in North Carolina bay scallops (bottom row, NMFS landings) shown with loess curve from a generalized additive model 84

Figure 4.4. Instantaneous rates of change in relative abundance (±95% CIs) for (A) great sharks and (B) elasmobranch mesopredators, as estimated by random- effects meta-analyses of research survey (•) and fisheries (A) data 88

Figure 4.5. (A) Map of the southeastern U.S. indicating the study location (inset) and North Carolina bay scallop monitoring sites. (B) Mean scallop density measured in midsummer and mortality from early summer to early fall at Oscar Shoal for 10 years. (C) Scallop density trends at Oscar Shoal, based on 12 weekly surveys in 1998 and 8 in 2002 and 2003 92

Figure 4.6. Changes in landings (metric tons) by individual states of the U.S.A. plus east coast of Canada for a) oysters, b) bay scallops, c) hard clams, and d) soft- shell clams 101

xii Figure 5.1. Predatory effects observed among different levels of predators (blue circles), and on herbivores (white circles) and plants (black circle) 122

Figure 5.2 Meta-analysis of cod-shrimp interactions for nine regions in the North Atlantic Ocean 132

Figure 5.3. Simplified oceanic food webs showing only the species and the direct species interactions involved in the documented trophic cascades: significant predator (thick black arrow), minor predator (thin black arrow); for sharks diet data are often known only at taxonomic levels above species: predator on species within genus (medium grey arrow), predator on species within family (thin pale grey arrow) 135

Figure 5.4. Meta-analysis of large benthic fishes - small pelagic fish interactions for nine regions in the Northwest Atlantic Ocean (data from Frank et al. 2006).... 139

Figure 5.5. Temporal changes in abundance of demersal and pelagic marine fish from four regions in north-temperate oceans between 1978 and 2001, standardized to a value of one for 1978 141

xin Abstract

Predator removals may have far-reaching consequences for ecosystem structure and functioning because of the strong influence they can exert on other species. In the ocean, many top predators have experienced unprecedented declines in abundance, eliciting concern about their conservation and indirect ecosystem effects that might ensue from their losses. Large pelagic sharks are apex and near-apex predators of the ocean, but their life histories render them more vulnerable to overexploitation than most marine fishes. Exploitation of sharks has intensified globally in recent decades driven by high bycatch levels and increased demand for their meat and fins. Assessing the impacts of exploitation on sharks is complicated by the vast geographic scales over which they range and poor monitoring of these previously low-valued species.

This thesis sought to examine changes in abundance of large pelagic sharks in the Northwest Atlantic Ocean during the past half century, and to explore the evidence for community-level consequences of removing these and other top predators from the ocean. Indices of relative abundance for sharks and their prey species were derived by modelling catch rates in fishery-dependent and research survey data. In the Gulf of Mexico* oceanic whitetip (Carcharhinus longimanus) and silky (C. falciformis) sharks are estimated to have declined since the mid-1950s by >99% and 97% respectively, while the longest running shark-targeted survey on the U.S. east coast indicated precipitous declines in all large sharks between 1970 and 2003, ranging from 87% in sandbar shark (C plumbeus) to 97% or more in tiger (Galeocerdo cuvier), scalloped hammerhead (Sphyrna lewini), bull (C. leucas), and dusky (C. obscurus) sharks. Recent data suggest that declines have continued for some species whereas others like tiger shark have now stabilized. Concomitant with shark declines, 12 of 14 elasmobranch (shark, skate, ray) prey species increased significantly in abundance, and an apparent trophic cascade was induced from sharks through cownose ray (Rhinoptera bonasus) to bay scallops (Argopecten irradians). Evidence from recent studies of top-down control in open ocean, continental shelf, and coastal ecosystems highlights the need for improved understanding of oceanic predator-prey interactions and to maintain predator abundances above thresholds of ecological extinction.

xiv List of Abbreviations and Symbols Used

AIC Akaike's information criterion AR1 first-order autoregressive correlation structure CBD Convention on Biological Diversity CI confidence interval cm centimetres CPUE catch per unit effort CTDEP Connecticut Department of Environmental Protection DNREC Delaware Department of Natural Resources and Environmental Control, Division of Fish and Wildlife EEZ Exclusive Economic Zone exp exponent FAO Food and Agriculture Organization FMP Fishery Management Plan g grams GEE generalized estimating equation GLM generalized linear model GLMM generalized linear mixed model GSO University of Rhode Island, Graduate School of Oceanography IUCN World Conservation Union kg kilograms L liters log natural logarithm

xv n sample size MDNR Maryland Department of Natural Resources, Fisheries Service NCDMF North Carolina Division of Marine Fisheries NMFS National Marine Fisheries Service NOAA National Oceanic & Atmospheric Administration POP Pelagic Observer Program SEFSC Southeast Fisheries Science Center s.d. standard deviation s.e. standard error SC South Carolina Department of Natural Resources SEAMAP Southeast Area Monitoring and Assessment Program, UNC University of North Carolina at Chapel Hill - Institute of Marine Sciences VIMS Virginia Institute of Marine Science

xvi Acknowledgements

First and foremost, I would like first to thank my supervisor, Ransom Myers, for his mentorship throughout my graduate studies. Ram provided an excellent environment for me to grow as a scientist, always believing in and encouraging me, allowing me the freedom to pursue my own research interests, and inspiring me through his own research. Ram's enthusiasm for science and for life in general, his support and friendship, and his passion for the ocean are greatly missed and will be always.

Sincere thanks to my co-supervisor, Jeffrey Hutchings, for guidance and helpful ecology and conservation discussions throughout, and for stepping in during these last few months. Thanks also to my committee, Sandra Walde and Doug Swain, for their questions and insightful comments about this research.

This thesis is based primarily on the analysis of research survey and fishery-dependent data, and I would like to sincerely thank all of the individuals involved in the collection of these data. I am also indebted to the individuals who made the data accessible and who generously answered my enquiries about the data, fisheries, monitoring programs and surveys: Larry Beerkircher, Guillermo Diaz, Dennis Lee, Mark McDuff, Scott Nichols, and Gerry Scott of the U.S. National Marine Fisheries Service; Glenn Ulrich of the South Carolina Department of Natural Resources, Ken Frank of the Canadian Department of Fisheries and Oceans, and many other individuals from U.S. universities and State Departments, who are listed in the acknowledgements of Chapter 4.

xvn Thanks also to my collaborators on Chapter 4, Travis Shepherd, Sean Powers, and especially to Charles (Pete) H. Peterson for demanding succinct and grammatically correct writing, for always being direct, and for caring so much. Pete is a true gem.

I am also indebted to the Rampire's statistical guru, Wade Blanchard of the Department of Statistics, Dalhousie University for statistical advice, explanations, and discussions throughout my graduate research. Valuable statistical advice about generalized linear mixed models was also provided by Joanna Flemming of the Department of Statistics, Dalhousie University.

Thanks are also due to Isabelle Cote and John Reynolds for providing me a space to work in their lab during my time at Simon Fraser University, Burnaby, British Columbia in the winter and spring of 2006.

Thanks also to everyone in the Myers lab, especially Gretchen Fitzgerald for holding it all together and Ian Jonsen, office-mate extraordinaire. The 'Rampire' is a great lab, and I am grateful to have worked with each of you and for your friendship and support.

I am also grateful for the support of my family. To Maureen, for always being there to listen, advise, and encourage, and to Joe for instilling in me his incredibly hard work ethic, thank you both. Thank you to Trevor Davies whose willingness to always listen means he now knows more about sharks and trophic cascades than he ever wanted to. Trevor's love, unwavering support and encouragement have been invaluable to me.

Finally, I would like to also acknowledge the financial support provided by the Natural Sciences and Engineering Research Council of Canada (NSERC) through a Julie Payette Postgraduate B Scholarship and by the Killam Trusts through a Pre-Doctoral scholarship, as well as funding provided by Ransom Myers through grants from the PEW Charitable Trusts and Lenfest Ocean Program.

xvm CHAPTER 1

Introduction

1 "Sharks ... may be considered as one of the richest 'strikes' in the only inexhaustible mine - the sea" -Barrett, 1928, Scientific Monthly

Human impacts on natural ecosystems are overarching and accelerating (Vitousek et al. 1997, Balmford & Bond 2005, Mace et al. 2005). Habitat degradation and loss (Brook et al. 2003), overexploitation of wild organisms (Pauly et al. 1998, Brashares et al. 2004), species invasions (Mack et al. 2002), and climate change (Thomas et al. 2004) are exacting a heavy price on biodiversity and ecosystem integrity globally (Chapin et al. 2000, Millennium Ecosystem Assessment 2005). Concern about these threats is reflected in the commitment made in 2002 by the Conference of the Parties to the Convention on Biological Diversity (CBD) to 'achieve by 2010 a significant reduction of the current rate of biodiversity loss at the global, regional and national levels' (UNEP 2002, CBD 2007).

In the ocean, fishing is the most widespread exploitative human activity and is generally considered to be the primary perturbation impacting marine species (Pauly et al. 2002, Dulvy et al. 2003). Marine fisheries developed at an unprecedented rate over the past century (Smith 1994, Pauly et al. 2005) leading to substantial increases in global fisheries yields but also to increases in the proportion of fished populations considered to be overexploited and depleted (now 17% and 7% respectively, with an additional 52% fully exploited) (FAO 2007). Indeed, fisheries collapses have become increasingly common in the past few decades (Myers et al. 1996, Hutchings 2000, Mullon et al. 2005) and many marine fish populations are estimated to have declined to extraordinarily low levels (e.g. Myers & Worm 2003, but see Sibert et al. 2006; Christensen et al. 2003, Jennings & Blanchard 2004, Hutchings & Reynolds 2004).

In addition to reductions in targeted species, the unintended impacts of exploitation on marine communities, including incidental catches, habitat damage, and altered species interactions, also are now widely recognized (Botsford et al. 1997, Pikitch et al. 2004, Mangel & Levin 2005). Bycatch in commercial fisheries threatens several endangered

2 species of marine mammals, seabirds, and sea turtles (Lewison et al. 2004), as well as many fish species (Jennings & Kaiser 1998, Hall & Mainprize 2005). Because variability among species in life history characteristics renders some more vulnerable to overexploitation than others (Jennings et al. 1998, Denney et al. 2002), target species that can sustain high fishing mortality may continue to drive fisheries while less productive incidentally caught species become imperiled (Musick 1999, Schindler et al. 2002). Fisheries also may affect species indirectly through altered competitive or predator-prey interactions. The excessive removal of top predators, for example, may initiate trophic cascades, with potential ramifications for ecosystem structure and function (Pace et al. 1999, Duffy 2002). Assessing these broader consequences of exploitation remains a challenge given the often limited knowledge of bycatch species' biology and of species interactions in marine food webs, as well as the dearth of data collected over the relevant temporal scales of ecological processes and the vast spatial scales over which species and fishing fleets range in the ocean.

Sharks are among the most vulnerable fishes to overexploitation (Bonfil 1994, Barker & Schluessel 2005, Lack & Sant 2006). Along with skates and rays, sharks comprise the Subclass Elasmobranchii within the Class Chondrichthyes, distinguished from teleost (bony) fishes by their cartilaginous skeletons (Helfman et al. 1997). These species differ markedly from teleost fishes in their life history traits (Rogers et al. 1999, Hutchings et al. unpublished) and, despite variation in the life histories of the 437 extant shark species (Last & Stevens 1994, L. Lucifora, pers. comm.), they are in general characterized by slow growth, a late age of maturity, long reproductive cycles, and low fecundity (e.g. Table 1.1; Holden 1973, Musick et al. 2000) resulting in low population growth rates (Cortes 2002, Hutchings et al. unpublished). This vulnerability to overexploitation is evidenced both by the pattern of rapidly collapsed shark fisheries in the mid-20th century (Ripley 1946, Olsen 1959, Parker & Stott 1965, Holden 1968) and by recent retrospective analyses of fisheries data for sharks taken as bycatch, which show large declines and extirpations of some species over the last century (Aldebert 1997, Rogers & Ellis 2000, Graham et al. 2001, Jukic-Peladic et al. 2001, Ferretti et al. 2005, Garcia et al. 2007).

3 Despite their low productivity and history of population depletions, shark exploitation has intensified globally in recent decades. Bycatch of sharks increased from the 1960s as fisheries targeting other large pelagic fishes developed in the open ocean (Bonfil 1994). Targeted shark fisheries (typically in more coastal waters) underwent a resurgence in the 1980s and 1990s, stimulated both by the decline of traditional food fish species and by increased demand and value for shark products (Rose 1996, FAO 1998). In particular, rising demand driven by rapid economic growth in mainland China has made shark fins one of the most highly valued marine commodities (Rose 1996, Clarke 2004, Clarke et al. 2006). The practice of finning, removing the fins and returning the carcass to the water, is illegal in several countries (Australia, Brazil, Costa Rica, Oman, South Africa, and the European Union (Fowler et al. 2005)), including Canada (DFO 1997) and the United States (Anon. 2002a); the first international bans were implemented in 2004 in the Atlantic Ocean (ICCAT 2004) and in 2005 in the eastern Pacific Ocean. However, shark finning remains unregulated in many regions of the world, and finning bans are notoriously difficult to enforce (Clarke et al. 2006). Recent studies by Clarke et al. (2005, 2006) estimated that about 38 million sharks are traded annually in the global shark fin trade, three to four times more than that recorded in official FAO statistics, and that the trade in mainland China has been increasing at a rate of 5% per annum.

Understanding the direct and indirect impacts of shark exploitation is imperative given the high potential for these vulnerable species to be overexploited (Musick et al. 2000, Stevens et al. 2000). This research focused on a group of apex and near-apex predatory pelagic shark species that are intensely exploited around the world and whose fins are thought to comprise the majority of those in the global shark fin trade (Clarke et al. 2006). The thesis sought to estimate changes in the population abundances of these large pelagic sharks in the Northwest Atlantic Ocean and to explore the broader community-level effects of declining numbers of these and other oceanic top predators.

Monitoring data for marine populations from their initial period of exploitation are rarely available or utilized, such that analyses often capture population changes only in the recent most portion of the exploitation history. This problem is particularly acute for

4 incidentally captured species, like large pelagic sharks, which have not typically been recorded in fisheries data until recently (if at all). Sharks often went unreported in fisheries records in the past because of their low economic value (prior to development of the fin trade), the incidental nature of their catches, and inadequate fisheries monitoring systems (Stevens et al. 2000). The practice of finning exacerbated this problem because fishers could readily fin sharks and discard the carcasses without recording the catches. Chapter 2 therefore sought to compare the relative abundance of shark populations prior to and following intense commercial exploitation. The analysis focuses on sharks in the Gulf of Mexico where data were available from research surveys conducted in pelagic waters during the mid-1950s.

In contrast, Chapter 3 examines recent changes in the relative abundance of large pelagic sharks across a large geographic expanse of the Northwest Atlantic Ocean. The chapter models fishery-dependent data from observers monitoring the U.S. Atlantic pelagic longline fishery between 1992 and 2005, updating estimates from an earlier analysis of logbook data for 1986 to 2000 from the same fishery (Baum et al. 2003), and informally comparing the two data sources.

The aim of Chapter 4 was to investigate potential community-level consequences of overexploiting large pelagic sharks, centering on the hypothesis that weakened top-down control by these species would most likely affect their elasmobranch (shark, skate, ray) prey. Drawing upon a compilation of research survey data from the east coast of the United States for the period 1970 to 2005, the chapter first provides further estimates of trends in abundance for the elasmobranch-consuming large sharks and then examines trends for their elasmobranch mesopredatory prey. These analyses are combined with evidence from long-term field observations and controlled experiments to document an apparent trophic cascade through one mesopredator, cownose ray {Rhinoptera bonasus), down to bay scallops (Argopecten irradians).

In the final research chapter the thesis broadens beyond sharks to synthesize the evidence for top-down control in oceanic ecosystems. Reviewing the recent ecological

5 and fisheries literature, Chapter 5 highlights the methods being used to assess evidence for top-down effects at the ecosystem-scale, the inferred mechanisms triggering top-down effects like mesopredator-release and trophic cascades, and the management implications of these ecosystem changes.

Chapters 2 and 4 of this thesis have been published in the primary literature. The latter represents a collaborative piece of research conducted by myself, Ransom Myers and Travis Shepherd of Dalhousie University, Charles Peterson of the University of North Carolina, and Sean Powers of the University of Alabama. Copyright permission to reproduce these papers herein has been obtained and comprises Appendix 2.

Shark Species

The large pelagic shark species included in this research are from the families Alopiidae (thresher sharks), Carcharhinidae (requiem sharks), Lamnidae (mackerel sharks), and Sphyrnidae (hammerhead sharks). Throughout the thesis, the species are classified either as oceanic or large coastal sharks following the U.S. National Marine Fisheries Service's (NMFS) Atlantic Highly Migratory Species Fishery Management Plan (FMP) (NMFS 2006). These divisions are based primarily on where the species are typically caught, rather than on taxonomic or ecological divisions (Branstetter 1999).

The six oceanic shark species included in this research are highly mobile species whose populations (with the exception of common thresher) range across entire ocean basins (Compagno 2001). The blue shark {Prionace glauca) is one of the most common and widest-ranging sharks, distributed throughout tropical, subtropical and temperate waters, from the surface to depths of at least 150 m (Compagno 1984). Its relatively early maturity and high fecundity (Table 1.1) make it one of the more productive shark species. Within the mako sharks, a small group of large to gigantic, fast-swimming fishes, the shortfin mako (Isurus oxyrinchus) is an apex predator found in tropical and warm-temperate waters throughout all oceans from the surface to depths of at least 500m (Compagno 2001). In comparison, the longfin mako (/. paucus), which is found in

6 epipelagic tropical and warm-temperate waters, is much less common and little known (Compagno 2001). The thresher sharks are very large, wide-ranging coastal and oceanic sharks distinguished by the enormously long upper lobe of their caudal fin (Castro 1983). The bigeye thresher (Alopias superciliosus) is virtually circumglobal in tropical and temperate waters, where it is found inshore, on continental shelves and slopes, and in the epipelagic open ocean, from the surface to at least 500m depth (Compagno 2001). Distributed almost circumglobally in cold-temperate to tropical waters, the common thresher {A. vulpinus) most commonly occurs in temperate waters, and although it ranges from the surface to at least 350m, it is usually found nearer the surface than the bigeye thresher (Compagno 2001). Unlike the other oceanic species, transatlantic migration has not been demonstrated for common thresher, and there are likely separate populations in the Northwest and Northeast Atlantic (Compagno 2001). The oceanic whitetip shark (Carcharhinus longimanus) is an epipelagic species found throughout the world's oceans in tropical and warm temperate waters (Compagno 1984), and occupying similar habitat and depths as the blue shark (Last and Stevens 1994). Although usually located offshore in the open ocean, this species also occurs close to shore where the continental shelf is narrow (Compagno 1984).

Hammerheads are large coastal sharks with distinctive flat wide heads (Compagno 1984). Of the three hammerhead species included in this research, the large schooling scalloped hammerhead (Sphyrna lewini) is the most abundant in U.S. Atlantic waters (Compagno 1984; Castro etal. 1999). This circumtropical species occurs over continental shelves and adjacent deep waters from the surface to at least 275m depth (Last and Stevens 1994), and is considered to be highly migratory (Kohler et al. 1998) although little is known about its population structure within the Northwest Atlantic. Unlike the scalloped hammerhead shark, the smooth hammerhead (S. zygaena) is not usually found in tropical waters, but rather on continental shelves and offshore cool temperate areas (Branstetter 2002). The great hammerhead (S. mokarrari), one of the largest shark species (Table 1.1), is a solitary fish found in warm waters both coastally and in open oceans (Castro 1983), from the surface to depths of 80m (Compagno 1984).

7 Table 1.1. Shark species included in thesis analyses [/ at genus level indicates species within the genus were grouped for analysis) and their life history characteristics (age at maturity (for females), fecundity, size at maturity (for females) and maximum size (FL=fork length, TL=total length)).

Species Modelledli n Life history characteristics chaptei! " Family Common name Latin name 2 3 4 Age at Fecundity Size at maturity Maximum size Reference maturity (years) Oceanic species Carcharhinidae Blue Prionace glauca • 5 25-50 185cm FL 310cm FL, 1,2 380cm TL Lamnidae Mako sharks Isurus species • Shortfin mako /. oxyrinchus 18 10-18 275-285cm FL 400cm TL 3-5 Longfin mako 1. paucus ? 2-8 - 245cm TL 420cm TL 3 Alopiidae Thresher sharks Alopias species • Bigeye thresher A. superciliosus 12-13 2-4 295-355cm TL 460cm TL 3 Common thresher A. vulpinus 7 2-4 ~315-400cmTL -575-610cm TL 3 Oceanic whitetip shark Carcharh'mus longimanus • • 6-7 1-15 180-200cmTL ~350cm TL 6 Large coastal species Sphyrnidae Hammerhead sharks Sphyrna species • • Scalloped hammerhead S. lewini • 15 2-21 240cm TL 250cm FL; 6-9 370-420cm TL Great hammerhead S. mokarran • ? 13-42 ~250-300cm TL -550-610cm TL 6 Smooth hammerhead S. zygaena V ? 29-37 -210-240cmTL 370-400cm TL 6 Carcharhinidae Tiger shark Galeocerdo cuvier • s 7 3-57 330-345cm TL ~550cm TL 6,10,11 Carcharhinidae Coastal shark group Carcharh'mus species • • Dusky shark C. obscurus • • 21 3-14 260-300cm TL ~400cm TL 6,12 Silky shark C. falciformis • 8-12 2-14 210-230cmTL ~330cm TL 6,7 Blacktip shark C. limbatus • 6-7 4-7 ~120-190cmTL 255cm TL 6,13,14 Bull shark C. leucas • 10-18 1-13 180-230cmTL 340cm TL 6,15 Sandbar shark C. plumbeus • 15 5-12 145-180cm TL ~240-300cm TL 6,16 Night shark C. signatus 10 4-12 ~180cmTL 280cm TL 6,17,18 References: 1 Skomal & Natanson 2003; 2 Pratt 1979; 3 Compagno 2001; 4 Natanson etal. 2006; 5 Francis & Duffy 2005; 6 Compagno 1984; 7 Branstetter 1987a; 8 Piercy et al. 2007; 9 Hazin era/. 2001; 10 Natanson etal. 1999; 11 Whitney era/. 2007; 12 Natanson etal. 1995; 13 Killam & Parsons 1989; 14 Branstetter 1987b; 15 Cruz-Martinezef a/. 2005; 16 Sminkey & Musick 1995; 17 Hazin etal. 2000; 18 Santana & Lessa 2004.

oo The remaining large coastal shark species in this research belong to the Family Carcharhinidae (Table 1.1). The tiger shark (Galeocerdo cuvier), the largest shark in this family, occurs in warm waters around the world, where it inhabits both deep oceanic and shallow coastal regions (Castro 1983; Castro et al. 1999). Its relatively fast growth and high fecundity make it less susceptible to overexploitation than many other shark species (Table 1.1). Several species of the genus Carcharhinus, a group of warm-water species distributed coastally around the world (Castro 1983, Compagno 1984), are also included. Notably, the dusky shark (C. obscurus), which is found in warm-temperate and tropical waters of continental shelves and adjacent oceanic areas (Compagno 1984), is among the latest maturing vertebrate species, requiring over twenty years to reach sexual maturity (Table 1.1). This, coupled with a long reproductive cycle and low fecundity, renders it one of the most vulnerable shark species to overexploitation (Musick et al. 2000, Cortes 2002). Reflecting this vulnerability, this species has been included on the Candidate Species List of the U.S. Endangered Species Act since 1997 (NOAA 2007). Finally, the silky shark (C. falciformis) is ecologically an oceanic species, but is classified in the NMFS FMP as a large coastal shark because of difficulty distinguishing it from its congeners. It is in fact one of the three most common and widespread oceanic sharks (along with blue and oceanic whitetip), occurring in warm tropical and subtropical offshore epipelagic waters around the world (Compagno 1984, Castro et al. 1999).

Data Sources

Large, pelagic sharks are exploited globally; the choice of the Northwest Atlantic Ocean as a focal region for this research was motivated by the availability of fishery- dependent data and long-term research surveys for these species. Although shark population monitoring in the Northwest Atlantic is better than in most other parts of the world, it should be noted that data limitations, including poor species identification in fishery-dependent data, inadequate observer coverage of commercial fisheries, and a paucity of long-term fishery-dependent data, are still a problem in this region, particularly in international waters.

9 The analyses in Chapters 2 to 4 are based on empirical data collected by the U.S. National Marine Fisheries Service and its predecessor the U.S. Bureau of Commercial Fisheries, as well as by state departments on the U.S. east coast and the University of North Carolina. For large pelagic sharks, longlines are the most suitable sampling gear, and most of the fishery-dependent and research survey data examined for these species in this thesis used this gear. Pelagic longlines are a free-floating fishing gear used to catch high-valued tunas and billfishes in the epipelagic and mesopelagic waters of the ocean. Sharks comprise the main bycatch of pelagic longline fisheries, and conversely, these fisheries are the world's main source of shark bycatch (Bonfil 1994). The gear consists of a mainline, suspended from buoys floating at the sea surface, which can run 10s of kilometers in length thereby allowing a single vessel to distribute effort over a large area to harvest patchily distributed fishes. Baited hooks are connected to the mainline by branchlines (also called gangions) at regularly spaced intervals, and typically number in the hundreds.

10 CHAPTER 2

Shifting Baselines and the Decline of Pelagic Sharks in the Gulf of Mexico

Published as Baum, J.K. and Myers, R.A. (2004). Shifting baselines and the decline of pelagic sharks in the Gulf of Mexico. Ecology Letters, 7, 135-145.

11 Abstract

Historical abundances of many large marine vertebrates were tremendously greater than today. However, while pelagic sharks are thought to have declined rapidly in the Northwest Atlantic in recent years, there, as elsewhere, little is known about the abundances of these species prior to industrial exploitation. Here, we compare 1950s (initial) and late-1990s (recent) standardized catch rates of pelagic sharks in the Gulf of Mexico, the area where methods of exploitation between these two periods were most comparable. We estimate that oceanic whitetip and silky sharks, formerly the most commonly caught sharks, have declined by over 99% and 90%, respectively. The former prevalence of oceanic whitetip sharks in this ecosystem is unrecognized today, providing an example of shifting baselines. Our analysis provides the missing baseline for pelagic sharks in the Gulf of Mexico that is needed for the rational management and restoration of these species.

INTRODUCTION

Understanding the full extent and manner in which anthropogenic forces have affected natural ecosystems requires knowledge of their unimpacted state. Although human influences on terrestrial and coastal ecosystems are highly evident (MacPhee 1999, Jackson et al. 2001), the open ocean has been regarded as pristine until recently. Precipitous declines in many oceanic species and concomitant fisheries collapses are, however, clear demonstrations that these ecosystems also have been significantly impacted. In particular, large predators can structure aquatic ecosystems (Duffy 2002), but that role may have changed substantially due to large-scale declines. Estimates for whales, tunas, billfishes, and large demersal fishes suggest that as in terrestrial and coastal ecosystems, the former natural abundances of many large predators were enormous compared with recent observations (Myers & Worm 2003, Roman & Palumbi 2003). For some species, however, a historical perspective is obscured by a reliance on recent data in analyses. Without this knowledge our baseline of what was natural in the

12 open ocean will continue to shift, and we risk becoming complacent about the rarity of species (Pauly 1995).

In the open ocean, sharks are among the remaining pelagic apex predators for which baseline population abundances are unknown. Large pelagic sharks include species that range across entire ocean basins and species whose range is more neritic, here termed oceanic and coastal sharks respectively. Establishing a baseline for these shark populations is necessary to fully understand how industrial fisheries have impacted them, and thus essential to their informed management and recovery. Sharks are among the least resilient fishes to intense exploitation because of their life histories, which are characterized by a late age at maturity and low fecundity (Musick et al. 2000). Populations of different oceanic and coastal shark species in the Northwest Atlantic are estimated to have declined by between forty and eighty-nine percent since the mid-1980s (Baum et al. 2003). Despite the magnitude of these losses we hypothesize that they are underestimates because they do not account for changes that occurred during the first several decades of industrial exploitation of these species (i.e. 1957-1985). Moreover, because rates of change are unlikely to have been the same among species (due to differences both in fishing mortality and life history characteristics), we hypothesize that the composition of unexploited pelagic shark assemblages may have been considerably different than that recognized today.

Quantifying the former natural abundance of pelagic shark populations should be facilitated by the short history of anthropogenic impacts in offshore pelagic ecosystems relative to that in most other aquatic ecosystems. Beyond continental shelves these populations were largely protected by their distant location and vast ranges until the past half century when industrialized fisheries developed in offshore waters to target other large predatory fishes, namely tunas and billfishes. As with other incidentally caught species, however, shark catches were not systematically recorded in these fisheries until recently. Instead, baseline information for the Northwest Atlantic is obtained from exploratory surveys that were conducted in the 1950s along the east coast of the United States, in the Gulf of Mexico, and Caribbean, to acquire information for the development

13 of commercial tuna fisheries. Longlines used in these surveys, and in the commercial fisheries that subsequently developed, resemble a transect through the pelagic ecosystem, consisting of a mainline suspended horizontally in the water column by floatlines and buoys, with baited hooks on branchlines attached at set intervals. We focus on the Gulf, of Mexico, because unlike in the other regions where target species, fishing methods, and specific locations fished have changed substantially over time, the initial surveys in this region are similar to contemporary methods of exploitation. We compare shark catch rates in the Gulf of Mexico on pelagic longline gear set to target yellowfin tuna (Thunnus albacares) during the 1950s surveys and the late-1990s commercial fishery. Because no sampling of the pelagic fish community in the Gulf of Mexico has exactly replicated the methods of the 1950s exploratory surveys, this comparison not only required that we standardize catch rates to account for variation in the fishing operation, location, and timing among sets, but also that for characteristics of the fishery which have changed over time that we incorporate independent estimates of their effects on each species' catchability. Although there will inevitably be some imprecision in our estimates, we believe that establishing a benchmark against which to judge present conditions is critical. Here, we use this approach to estimate the former natural abundance of oceanic whitetip (Carcharhinus longimanus), silky (C. falciformis), and dusky (C. obscurus) sharks in the Gulf of Mexico.

METHODS

Data

Data for the initial exploitation of sharks in the Gulf of Mexico are from sixteen exploratory pelagic longline research cruises conducted on the research vessel Oregon between 1954 and 1957 (n=170 longline sets). These surveys aimed to determine the distribution of surface tuna in the region, but quickly focused on the most frequently caught species, yellowfin tuna. Exploratory cruises occurred in all seasons and covered large areas of the Gulf of Mexico (Figure 2.1). Data were obtained from the U.S. National Marine Fisheries Service and supplemented and cross-checked with cruise summaries by Bullis and Captiva (1955), Wathne (1959), and Iwamoto (1965).

14 The surveys were conducted using standard gear and fishing methods adapted from the Japanese longlining method (Table 2.1; described in Shapiro 1950, Murray 1953, Bullis & Captiva 1955, Wathne 1959). Gear was usually deployed before dawn (n=145; 85% of sets), and allowed to soak for only a few hours. Gear retrieval usually began mid-morning and lasted for three to seven hours. We retained the longline sets that began in the late morning or afternoon (n=25), because the time of these sets is covered by the longer soak times of sets in the 1990s, but removed the few night sets (n=6).

Table 2.1. Comparison of 1950s and 1990s pelagic longline fishing data from the Gulf of Mexico used in analysis. Mean values ± 1 standard deviation.

Descriptor 1950s 1990s Data source U.S. Bureau of Commercial Fisheries U.S. National Marine Fisheries Service researchers observers Purpose Exploratory tuna surveys Commercial fishery Years 1954-1957 1995-1999 Target species Yellowfin tuna Yellowfin tuna Fishing effort 170 sets, 82 972 hooks 275 sets, 219 461 hooks Hooks per set 488 + 251 798 ±181 Hooks between floatlines 10 ± 1 5±2 Estimated hook depths 72 ± 19m 110±28m Hook type 9/0 Japanese style tuna hook 7/0, 8/0,15/0,16/0 primarily J-hook, some circle hooks Mainline material 132-fhread type-E filament nylon (10001b Nylon monofilament (600-11001b test) test) Branchline material "Gulf-lay" nylon (11/64" diameter), 3/32" Nylon monofilament (300-450lb test), diameter 7x7 stainless steel wire leaders often with nylon monofilament leaders Light sticks No No Bait Mackerel, herring, squid, menhaden, Mackerel, herring, squid, sardine, scad mackerel scad, Atlantic croaker Set time 5:30am ± 2 hours 8:30am ± 3 hours Retrieval time 10:00am ±2 hours 7:00pm ± 4 hours Months fished January - December January - December

Recent data are from the scientifically trained observers monitoring fishing activity by U.S. commercial longline vessels between 1995 and 1999. The fleet in the Gulf of Mexico primarily targets yellowfin tuna, but also swordfish. The target species of the set is identified prior to gear retrieval, and is not based on the catch, but rather on characteristics including time of set, hook depth, and use of light sticks. For comparison with the 1950s data, we only included data from the most similar sets, those that targeted yellowfin tuna with soak times during the day (n=275 sets from 63 trips; Table 2.1).

15 These sets occurred in all seasons and were located throughout the northeast and northwest Gulf of Mexico (Figure 2.1).

Our comparison between the 1950s and 1990s is contingent upon the similarity of the fishing operations and on our ability to correct for the effects on shark catch rates of differences between them. Comparing survey and fisheries data may be tenuous because of the different sampling methods employed, but is possible here because the surveys operated similarly to a fishery, actively searching for and concentrating in areas with yellowfin tuna (Wathne 1959, Iwamoto 1965). In addition, because the datasets from the 1950s and 1990s represent groupings of four and five years of data respectively, the effect of any anomalous years, is minimized. The main differences then in comparing the 1950s and 1990s data are the improvements in searching for and targeting yellowfin tuna, changes in the leader material (at the end of the branchlines), hook type, and hook size, increased depth fished, and the intensified fishing effort, including increases in the number of hooks set and the soak time of the gear in the water (Table 2.1). Of these, we account for increases in the number of hooks set, hook depth, and soak time in our models (methods described below). Correlations of yellowfin tuna catch rates with each of the commonly caught shark species were all highly nonsignificant, and yellowfin tuna catch had little effect on our results when included as a model variable (see Supplementary Material), thus changes in catch rates of the target species should not bias relative abundance estimates for sharks. A paucity of data precluded modelling the effect of the other gear changes on shark catchability, but the limited available data (Berkeley & Campos 1988, Branstetter & Musick 1993) suggest that these may either have had little effect or actually increased shark catch rates, which would minimize (rather than exaggerate) any estimated declines between the 1950s and 1990s (see discussion in Supplementary Material).

Shark Species

Pelagic longline fisheries primarily catch oceanic shark species, but also catch coastal shark species when operating in relatively close proximity to land, as in parts of

16 the Gulf of Mexico (Table 2.2). Of the species caught, we grouped shortfin and longfin mako, bigeye and common thresher, and scalloped and great hammerhead sharks, because these species were sometimes only identified to genus. The high proportion of unidentified sharks in the 1990s data resulted from fishers releasing unmarketable large non-target fishes prior to positive identification by observers (L. Beerkircher pers. comm., National Marine Fisheries Service (NMFS), Southeast Fisheries Science Center, Miami, Florida). We account for unidentified sharks in our presentation of unstandardized catch rates and in our models, by distributing them across all shark species caught, except mako sharks, which have high value and are usually retained.

Table 2.2. Recorded oceanic and coastal shark species and sample sizes in 1950s and 1990s datasets, listed within each category in declining order of abundance in the 1950s.

Species No. caught Common name Latin name 1950s 1990s Oceanic species Oceanic whitetip Carcharhinus longimanus 397 5 Silky Carcharhinus falciformis 158 24 Mako species Isurus oxyrinchus, Isurus paucus 17 24 Coastal species Dusky Carcharhinus obscurus 61 30 Tiger Galeocerdo cuvieri 6 18 Blacktip Carcharhinus limbatus 6 0 Hammerhead species Sphyrna lewini, Sphyma mokarran 4 8 Sandbar Carcharhinus plumbeus 0 16 Spinner Carcharhinus brevipinna 0 4 Thresher species Alopias superciliosus, Alopias vulpinus 0 4 Atlantic sharpnose Rhizoprionodon terraenovae 0 1 Unidentified sharks - 62

We attempted to model each of the shark species that was caught in both the 1950s and 1990s and found that we did not have statistical power to detect changes in abundance for species with fewer than 25 observations total. We therefore report model results for oceanic whitetip, silky, mako, and dusky sharks.

Modelling Change in Abundance We used generalized linear models with a negative binomial error structure and a log link to standardize the catch rates for operational, spatial, and temporal variation among

17 longline sets (Venables & Ripley 1999). Thus, for each species, s, the observed catch,

Csj, on set i is assumed to follow a negative binomial distribution, and the model predicts the mean number of the species that would be caught by a standard longline set, at a standard location and time. We estimated the effect of the fishing period with an indicator variable /,, which was defined as 0 if the year was between 1954 and 1957, and 1 if the year was between 1995 and 1999. The exponent of its estimated parameter, e^, can then be interpreted as the change in relative abundance between the two fishing periods.

We begin with a basic model of the expected mean catch jusi,

2 logGu,,) = pis>i +q(ds.) + fimlNgJ + pn2N s. + PwlWs>i + PW2W\ + flolOtJ + P^O),

+ logtfT,,) where Is i is the variable for fishing period, q{ds,.) is the seasonal cycle described below, the variables for area (degrees latitude north, Nsi, degrees longitude west, Wsi) and ocean depth Osi are fit as quadratic terms, and Hs., the number of hooks set is a known value, treated as an "offset" (McCullagh & Nelder 1989). The seasonal cycle is determined by fitting sines and cosines with periods, j, of Yz and 1 year in the model as

2 q(dsi) = ^[$j cos(27gdsi /365.25) + cry sin(27gdsl /365.25)], where ds(is the sequential 7=1 day of the year that set i occurred on, and the estimated parameters are q .f and a•., Including the log hooks as an offset is equivalent to dividing the catch by effort to determine a catch rate, but is used to preserve the probabilistic sampling model for the catch data.

We then develop our model to include independent estimates of the effects of both soak time and hook depth on catch rates of the species being modelled. We use this approach because these variables, both of which are known to significantly influence catch rates of pelagic fishes (Suzuki 1977, Uozumi & Okamoto 1997, Ward et al. 2004, Ward & Myers 2005a), have changed considerably between the two time periods (Table

18 2.1). Estimation of the effect of soak time in particular, if included as a variable in the model, would be largely confounded with the estimates of relative abundance between the two periods, because there is little overlap in this variable between the 1950s and 1990s. Instead, our approach is to include independent estimates of the effect of these variables in the offset of our model. That is, we include more information in the "offset" than simply the number of hooks, since, for example, if we know that the effectiveness of the hooks in set / is Gt times the average, we can replace the offset, log(/f;), with log(tf,.G,.).

We used the effect of soak time on catch rates estimated by Ward et al. (2004), in which the catchability of species s is defined as a function of soak time, t, in hours, to be gs (t) = as exp(Sst). Ward et al. (2004) used generalized linear mixed models to estimate this effect for five pelagic longline fisheries in the Pacific in which the soak time and catch of each hook were recorded. For each species, we weighted the slope coefficients,

Ss, estimated for the different fisheries by their inverse variance and used the mean value (Table 2.3). Catch rates of the sharks modelled here increased with soak time, by between 1.3 and 2.2 times after ten hours, the average difference in soak time between the 1950s and 1990s (Table 2.3). In our datasets, because we do not have records of the catch and soak times of individual hooks, we use data on gear deployment and retrieval times to calculate the shortest and longest soak times of hooks on each set. To include the soak time effect in our model for each species, we assume that the longline is recovered at a constant rate, and integrate over the soak time from the beginning to the end of the haul. We then estimate a mean soak time effect for each species for each set as

STES. = — where bi and e;- are the times that the hooks with shortest and longest soak times, respectively, have been in the water.

For the effect of depth on catch rates, we used estimates from Ward and Myers (2005a), where the catchability of species s is defined as a cubic function of hook

19 2 3 depthZ), in kilometers, to be fs (D) = exp(as + yslD + ys2D + ys3D ). Ward and Myers (2005a) derived the estimates by combining data from three pelagic longline fisheries in the Pacific in which the catch and depth of individual hooks were recorded. Catch rates of each of the shark species modelled declined with depth: the difference in efficiency between hooks at the mean 1990s depth (110m) and those at the mean 1950s depth (72m) ranges from 0.65 to 0.85 among species (Table 2.3). To estimate the depth of hooks on each longline set, we assumed that hooks are spaced evenly along the mainline between floatlines and that the shape of the catenary curve formed by the longline, and thus the corresponding depth of hooks, does not vary systematically over the entire set. We then estimated the hook depth at each position using the floatline and branchline lengths, the length of the mainline between floats, and the equation and corrections for the catenary curve (Suzuki et al. 1977, Uozumi & Okamoto 1997, Mizuno et al. 1999). Our depth estimates are similar to those obtained from depth sounders (Wathne 1959). Estimates for each longline set i result in a vector, pt (D), which describes the proportion of hooks in the set at depth D. As with soak time, to include the depth effects in our model, for each species we calculate a mean depth effect for each set, using the proportion of hooks in a given set at depth D and the depth effect at each discrete depth

DEti=JdDfl{D)Pt{D).

Table 2.3. Modelled species, coefficients for soak time (6S) and depth (ysV ys2, ys3) used to calculate the offset, and efficiency of hooks in the 1990s relative to the 1950s for the mean change in soak time and depth respectively.

Soak effect Depth effect Species § Relative Relative Ys2 'S3 Efficiency Efficiency Oceanic whitetip 0.0458 1.56 -9.886 11.720 -2.188 0.74 Silky 0.0260 1.29 -3.562 -4.423 9.895 0.85 Dusky 0.0778 2.18 -19.048 47.475 -35.911 0.65 Mako spp. 0.0343 1.41 -9.105 26.325 -22.572 0.83

We standardized both the soak time, STE*si, and depth, DE*si, effects so that the mean value of each, and hence its mean effect on the efficiency of hooks, was 1. The standardized effects were then included in the offset such that the offset in our basic

20 model is replaced with log(HsjSTE*sjDE*si). We explore the effects of soak time and depth on our model estimates in the Supplementary Online Material.

Model performance was compared by backward-selection using the Akaike Information Criterion (AIC), which penalizes the deviance by twice the number of parameters^. We then adjusted the final model estimates to account for the unidentified sharks in the 1990s. We assumed that among the species caught in the 1990s there is no bias in which are unidentified by observers, such that the species composition of this group should reflect that of the identified sharks (except for mako sharks). Distributing the 62 unidentified sharks among these species according to their proportion of the total 1990s shark catch increases the total catch of each by 1.56 times. Thus, to account for the effect these sharks would have on our estimates of change in abundance for oceanic whitetip, silky, and dusky sharks we simply reduce our estimated decline by this amount.

Finally, we evaluated the robustness of our results by fitting additional models to test the precision of our soak time effect estimate, the effect of including yellowfin tuna catches as a variable, and the effect of including sets targeting any tuna species, and to examine a reduced area in the Gulf of Mexico, excluding the southern area where there were few sets in the 1990s (see Supplementary Material).

Estimating Change in Size

We estimated the mean size of the species modelled in each time period using data on weights from the 1950s, and lengths in the 1990s applied to length-weight conversions. For the 1990s estimates, we used fork length data from any pelagic longline tuna set in the Gulf of Mexico between 1992 and 1999 to increase the sample size for each shark species. We estimated the weight of oceanic whitetip sharks using fork length to total length conversions (Lessa et al. 1999, Lessa unpublished data) and total length to weight conversions (Strasburg 1957). Silky, dusky, and mako shark weights were estimated using fork length to weight conversions (Kohler et al. 1995). Reported size at maturity is the fork length.

21 RESULTS

Unstandardized Catch Rates Catch rates of sharks were substantially higher in the 1950s than in the 1990s, declining from a mean of 7.30 (± 7.94 s.d.) to 0.92 (± 2.51 s.d.) per 1000 hooks. Sharks comprised on average 17.2% of the total catch on the exploratory pelagic longline sets in the 1950s, but only 2.2% of the total catch on pelagic longline sets in the 1990s. The decline cannot be attributed to increased catches of the target species: yellowfin tuna declined from 57% to 35% of the total catch between the two time periods.

In the 1950s, oceanic whitetip and silky sharks were the second and fourth most commonly caught fishes respectively on the pelagic longline surveys. Both species were found throughout the Gulf of Mexico: the oceanic whitetip shark was caught on 64% of sets, the silky shark on 35% (Figure 2.1a-d). Oceanic whitetip sharks accounted for 61% of all sharks caught, while silky sharks accounted for 24%. Between the mid-1950s and the 1990s, catch rates of oceanic whitetip and silky sharks declined from an average of 4.62 (± 6.47 s.d.) and 1.71 (± 3.49 s.d.) per 1000 hooks to only 0.02 (±0.18 s.d.) and 0.10 (± 0.42 s.d.) per 1000 hooks respectively (Figure 2.2). These two species, which together comprised on average almost 15% of the total pelagic longline catch in the 1950s, accounted for only 0.3% of the total catch in the 1990s. Catch rates of the next most commonly caught shark in the 1950s, dusky shark, were also substantially lower in the 1990s (Figure 2.1e, f), declining from a mean of 0.61 (± 1.72 s.d.) in the 1950s to 0.16 (± 1.24 s.d.) in the 1990s (Figure 2.2). Despite the decrease, dusky sharks had the highest catch rate among sharks in the 1990s. Mean catch rates of oceanic whitetip (0.04), silky (0.17), and dusky (0.21) sharks still show substantial declines when their total catches in the 1990s are adjusted to include a proportion of the unidentified sharks.

Catch rates of all other shark species caught in either time period were low. Catch rates of mako sharks declined from 0.19 (± 0.77 s.d.) in the 1950s to 0.09 (± 0.37 s.d.) in the 1990s (Figures 2.1g,h, 2.2). The 1950s surveys caught only three other shark species:

22 Figure 2.1. Map of the Gulf of Mexico showing unstandardized mean catches per 10,000 hooks on yellowfin tuna targeted sets during the day in the 1950's (a, c, e, g) and the 1990's (b, c, f, h) for oceanic whitetip (a, b), silky (c,d), dusky (e, f) and mako (g, h) sharks. Empty hexagons are set locations where none of the specified shark species was caught. The 200m and 1000m coastal isobaths (dotted lines) are shown for reference.

23 a > i ^ (• * »* { ^S^* -~s • • . «,- * ^ • * • • • f r „ . - r«-V i * -\*. V --e \ » r f "V. ,. •' ! 8- , % { V ^ 4 t\. 1954-1957 i 1995-1999 c d 1

•/ . IT" P

--01I •* - \ f CC f *-> -1 I >""* 1 » I . ' \ u J *- V . • *' ' /.- • .- ft i *' 2 ^ ,.<

I 28 24 20 I" 12 «# • >8 M t

^V $ 1954-1957 •\ ' .^f* V* 1995-1999

on.»'-C& - h / vrj^>, k^ \ j- r iix { •Ol K( H I • » f s t \ f V M4i 1954-1957 - ^<* 4/. 1995-1999 -95 1950s 1990s

" Ell -• Silky shark

1 §11 * • Dusky shark

i—I —> Mako sharks

n-1 Tiger shark

HP Hammerhead sharks

Hi Blacktip shark

-• Sandbar shark

-• Spinner shark

H Thresher sharks i 1 1 1 1 1 1 1 12 10 8 6 4 2 0 2 Catch per 1000 hooks

Figure 2.2. Mean catch rates (± s.d.) of each shark species recorded in the 1950s data and/or 1990s data. Catch rates are unstandardized for changes in spatial, temporal, and operational variation. Species are listed in declining order of abundance in the 1950s. blacktip, tiger,an d hammerhead sharks (great hammerhead (S. mokarrari) and two sharks identified only to the genus Sphyrna), each of whose mean catch rate was less than 0.06 per 1000 hooks. The blacktip shark was not recorded by observers in the 1990s, while catch rates of tiger and hammerhead sharks were similar (Figure 2.2) or slightly higher when adjusted for unidentified sharks. Of the shark species unique to the 1990s data, the mean catch per 1000 hooks of sandbar sharks was 0.07 (0.11 when adjusted for unidentified sharks), while those of spinner and thresher sharks were less than 0.03, even when adjusted.

25 ocs FAL DUS MAK Species

Figure 2.3. The estimated change in abundance (± 95% CI) between the 1950s and 1990s of oceanic whitetip (OCS), silky (FAL), dusky (DUS), and mako (MAK) sharks. A value of one indicates no change in abundance, a value often indicates a tenfold, or 90%, decline.

Relative Abundance and Size Estimates Model estimates suggest that the abundance of oceanic and coastal sharks declined substantially between the 1950s and late-1990s. The most commonly caught shark species in the 1950s, the oceanic whitetip shark, is estimated to have declined by more than 150-fold, or 99.3% (95%CI: 98.3-99.8%), during this time (Figure 2.3). We estimate over a ten-fold decline, 91.2% (95% CI: 84.8-94.9), in silky shark abundance and almost a five-fold decline, 79.2% (95% CI: 58.8-89.5%), in dusky shark abundance (Figure 2.3). Mean estimated declines in our robustness analysis were very similar for oceanic whitetip and silky sharks (within 1 and 2% respectively), but ranged from 70.6- 87.2% for dusky sharks (see Supplementary Material). Fishing period (1950s, 1990s) was the most important explanatory variable for each of these species in all models.

26 We estimate that there has been less of a decline in mako shark abundance (45%, 95% CI: 0.02-70.6%; Figure 2.3). However, in the final model for mako sharks fishing period explained less variance than ocean depth, and in general estimates for this species group are much less precise than for the other modelled species because of the lower sample size. The change in abundance of each of the other shark species was poorly determined because of their rarity in the pelagic longline catches.

The mean weight of individuals of each of the modelled species is also estimated to be much lower than in the 1950s (Table 2.4). The average size of oceanic whitetip sharks is near the size at maturity. The mean size of other species is now below the size at maturity: dusky sharks mature at 230cm (Kohler et al. 1995) and catches now average 165cm, shortfin mako sharks mature at over 180cm (Kohler et al. 1995) and now average 171cm, and in the most extreme case, silky sharks, which mature at about 180cm (Kohler et al. 1995) now average only 97cm.

Table 2.4. Modelled species weights in kilograms ±1 standard deviation (sample size). Weight Species 1950s 1990s1 Oceanic whitetip 86.4 ± 25.9 (370) 56.1 ±52.2 (30) Silky 102.3 ±47.0 (157) 16.7 ± 31.8 (346) Dusky 154.2 ±36.9 (61) 60.9 ±43.5 (65) Mako spp. 158.2 ±103.7 (17) 73.6 + 69.5 (50) 1 Weights for the 1990s include data from all pelagic longline tuna targeted sets in the Gulf of Mexico between 1992-1999 because of the otherwise limited sample size.

DISCUSSION

Pelagic shark species are estimated to have declined precipitously in the Gulf of Mexico since the onset of industrialized pelagic fisheries. The shark species that were initially most commonly caught underwent the greatest declines. In particular, the oceanic whitetip shark, the most prevalent shark in the 1950s, is estimated to have declined by over 99%. Silky and dusky sharks are estimated to have declined by 91% and 79%) respectively. Our results for oceanic whitetip and silky sharks are very robust,

27 and although the magnitude of the declines fluctuated among models of dusky sharks, each model estimates that this species also has declined substantially.

Differences in the magnitude of declines among shark species have likely altered the composition of the Gulf of Mexico's pelagic shark assemblage. Whereas in the 1950s, catches of oceanic whitetip sharks accounted for over 60% of sharks captured, by the 1990s this species comprised only 2% of shark catches. The oceanic whitetip has not been replaced by other shark species, but rather the remaining catches are comprised of other depleted species, dusky, silky, and mako sharks, and by several coastal species that were either extremely rare or not caught at all in the 1950s. The rarity of these latter species is likely due to the location of the observed sets in the pelagic longline fishery, which were almost all beyond the continental shelf (Figure 2.1), at the margins of these species' more coastal distributions. Catches of new shark species in the 1990s may simply be an artifact of the increased sample size or increased hook depth of sets (e.g. sandbar sharks), although it is not inconceivable that their distribution may have expanded slightly offshore to occupy niches left by the pelagic sharks that have declined.

Mean sizes for each modelled species are now at or below the size at maturity, which if reflective of real declines, may accelerate declines and cause local extinction. Decreased sizes are often observed in heavily exploited species because fishing gear often selectively catches the largest individuals, but declines below sizes at maturity are of particular concern. Although a change in pelagic longline gear from wire to monofilament leaders was partially intended to reduce catches of large incidental species like sharks, studies comparing wire and monofilament have in fact found very similar mean shark catch rates and sizes, notably documenting mean lengths of silky sharks below the size at maturity on both gear types (Berkeley & Campos 1988, Branstetter & Musick 1993). Indeed, large fishes are still caught on the latter gear type: the mean size of the target species in the Gulf of Mexico was higher than that of oceanic whitetip, silky, and dusky sharks in the 1990s. Large sharks are however more likely to bite-off the monofilament lines, thus while the limited available evidence from other studies suggests

28 that the mean sizes of shark species in these data might reflect real declines, further research on this topic warranted.

The magnitude of the shark declines estimated here likely reflects both the life histories and levels of exploitation of these species. Along with other elasmobranchs, sharks are generally more vulnerable to overexploitation than teleost fishes, because their slow growth, late sexual maturity, and low fecundity result in low intrinsic rates of increase. Declines estimated here range from being less (dusky sharks) to an order of magnitude greater (oceanic whitetip sharks) than those estimated for pelagic teleost predators (Myers & Worm 2003). However, intrinsic rates of increase decline in order from oceanic whitetip, silky, to dusky sharks (Cortes 2002), opposite to the estimated magnitude of their declines. We attribute this to differences in the distribution of these species, which have translated into different exploitation rates. The oceanic whitetip is primarily an offshore species whose entire population has been vulnerable to intense pelagic longline fishing effort for over four decades (Compagno 1984). In contrast, silky sharks are more abundant offshore near land than in the open ocean and have nursery areas along the outer continental shelf edge, and dusky sharks only range into offshore waters adjacent to continental shelves (Compagno 1984; Branstetter 1987a). Thus, these two species have lower catch rates in pelagic longline fisheries and their populations have been partially protected from pelagic longline fishing effort. Directed shark fisheries that catch these sharks developed only in the 1980s.

The near disappearance of oceanic whitetip sharks from the Gulf of Mexico is a clear example of our shifting baseline in marine ecosystems (Pauly 1995). This species was initially described as the most common pelagic shark beyond the continental shelf in the Gulf of Mexico (Wathne 1959, Bullis 1961). Between two and three to as many as twenty-five oceanic whitetip sharks were usually observed following the vessel during longline retrieval on the exploratory surveys, and the abundance of these sharks was considered a serious problem because of the high proportion of tuna they damaged (Bullis & Captiva 1955, Backus et al. 1956, Wathne 1959). In contrast, recent papers on pelagic sharks in the Gulf of Mexico either have not mentioned the oceanic whitetip or

29 have dismissed it as a rare species, with no apparent recognition of its former prevalence in the ecosystem (e.g. Witzell 1985, Anderson 1990, Russell 1993, Gonzalez Ania et al. 1997, Brown 1999). The oceanic whitetip shark has been assessed by the World Conservation Union only as Lower risk/Near threatened (IUCN 2002). Similarly, the extent of declines in the once abundant silky shark has not been recognized (e.g. Branstetter 1987a, Russell 1993, Beerkircher et al. 2003). How did declines in these conspicuous predators go undetected? Population collapses of many coastal shark species were noted because of declining catches in the fisheries targeting them (e.g. Ripley 1946, Olsen 1959, Parker & Stott 1965, Holden 1968), but oceanic sharks have never been the primary target of commercial fisheries in the Gulf of Mexico and thus have received little research and management attention despite their high levels of exploitation. Consequently, dusky sharks, which are caught in the directed U.S. coastal shark fishery, are listed as a Candidate Species for the U.S. Endangered Species Act because of their high susceptibility to overexploitation, while oceanic whitetip and silky sharks have been overlooked.

Taken together, our model estimates suggest that the overall shark assemblage in the Gulf of Mexico's offshore waters declined considerably during this time period. Estimating the precise change in abundance of this assemblage is difficult however given that catchability on pelagic longlines may differ among species. The estimated total decline in the assemblage would be about 92% if species' catchabilities were equal, such that the prevalence of each species was indicative of their abundance in the sampled area. But even if the catchability of any of the four modelled species was an order of magnitude different than the other species (e.g. oceanic whitetip sharks lOx more catchable and hence less abundant) the total decline in the shark assemblage would still be about 82%. Although the ecosystem impacts of overexploitation in the open ocean remain largely unexplored, consumers like sharks may exert important controls on food web structure, diversity, and ecosystem functioning (Paine 2002, Worm et al. 2002). Altering entire assemblages of large predators to this extent might have had a considerable impact on the pelagic ecosystem.

30 Our results represent snapshots of the Gulf of Mexico's pelagic shark assemblage in the 1950s and in the 1990s, thus it is not clear when during this period the estimated declines occurred. Evidence that in other areas marine predators might have declined by an average of 80% within the first fifteen years of industrial exploitation (Myers & Worm 2003) suggests that shark populations (particularly oceanic species) could have been reduced rapidly following initial exploitation in the Gulf of Mexico. Although this may be the case, shark populations in the Gulf of Mexico have not stabilized, but rather recent analyses indicate that these species continued to decline in the 1990s (Baum et al. 2003). Using independent data from that examined here, Baum et al. (2003) estimated that in the Gulf of Mexico oceanic whitetip, dusky, and silky sharks declined by about 10% per annum between 1992 and 1999. Combined with our results, this suggests that each of these species is under a real risk of extirpation in this region.

The precipitous declines estimated here may be reflective of a general phenomenon for oceanic sharks. Early research surveys described the oceanic whitetip as the most common pelagic shark throughout the warm-temperate and tropical waters of the Atlantic and Pacific (Mather & Day 1954, Strasburg 1957), but as in the Gulf of Mexico current pelagic fisheries in these oceans apparently catch very few of this species (Williams 1999, Matsunaga & Nakano 2000, Matsushita & Matsunaga 2002). Considering that oceanic sharks are heavily exploited throughout the world's oceans, and that these fishes are more vulnerable to overexploitation than the teleost fishes recently estimated to have declined by a factor often (Myers & Worm 2003), it is quite possible that declines estimated for the Gulf of Mexico have occurred elsewhere. In most other regions it will not be possible to quantify the former natural abundance of these species because of changes in methods of exploitation over time, and because much of their exploitation occurs in international waters and is incidental to other target species such that there has been little monitoring of their catches. As with many marine organisms (Casey & Myers 1998, Dulvy et al. 2003), we may therefore fail to detect the risk of local extinctions of oceanic shark populations until after the fact.

31 This study contributes to the growing awareness that human impacts on natural ecosystems extend to our oceans, and that retrospective analyses are essential to understand the full magnitude and nature of these impacts (Jackson 2001, Jackson et al. 2001). The perception of what was natural in the open ocean has clearly changed over a very short period (less than half a century), and our results suggest that it may be particularly easy for baselines of incidentally harvested species to shift because they are usually poorly monitored. We provide the first estimates of baseline abundances for pelagic sharks in the open ocean, which suggest that these species were enormously more abundant than today. Our results imply that oceanic whitetip sharks are now ecologically extinct in the Gulf of Mexico, but also provide a benchmark needed to set clear goals for their restoration.

SUPPLEMENTARY MATERIAL

Effect of Soak Time and Depth on Model Estimates

In our main model, we used independent estimates of the effect of soak time and hook depth on catch rates of the species being modelled, because of changes in these variables between the time periods examined. Here, we demonstrate the impact of failing to account for these effects on our model estimates. Because shark catch rates increase with soak time and the mean soak time increased considerably, from about 4 to 14 hours, with little overlap between the 1950s and 1990s, the impact of ignoring this effect is to underestimate the declines in shark abundance (Table 2.5). The effect is small for oceanic whitetip and silky sharks, but considerable for dusky sharks because the relationship between soak time and catch rate is strongest for this species (Table 2.3). The depth of sets increased over time, but there is still considerable overlap in the range of hook depths in the 1950s and 1990s datasets (Table 2.1). In contrast to soak time, because catch rates of these shark species decline with depth, the effect of ignoring hook depth is to overestimate the declines (Table 2.5).

32 Model Robustness We evaluated the robustness of our results by fitting our main model (described in Methods) with changes to the variables, the offset, or the datasets used. We fit these models for oceanic whitetip, silky, and dusky sharks, the species whose change in abundance we can estimate most precisely with our model. (1) We first included yellowfm tuna catch as a variable because the data are from pelagic longline sets targeting this species. (2) To explore the effect of the soak time coefficient estimate, we calculated the soak time offset for each species using the soak time coefficient + or - one standard error: oceanic whitetip 0.0458 ± 0.0168, silky 0.0260 ± 0.0031, dusky 0.0778 ± 0.0329. (3) In the 1990s dataset, the target species for a few sets was noted only as "tuna species" (n=17). Characteristics of these sets were similar to those targeting yellowfin tuna, except that they occurred at greater depths, and we examined the effect of including these sets in the model. (4) Finally, because sets in the 1990s occurred in a more concentrated area in the northern Gulf of Mexico than sets in the 1950s, we examined the effect of including only sets from both datasets that were within this region (between 25 and 30° latitude north and 82 and 100° longitude west).

Table 2.5. Estimated declines1 (± 95% CI) between the 1950s and 1990s of oceanic whitetip, silky and dusky sharks in the Gulf of Mexico from the main model, from models examining the effect of the soak time offset and of the depth offset, and from robustness models.

Model Species Oceanic whitetip Silky Dusky Main model 99.4 (98.3-99.8) 91.2(84.8-94.9) 79.2 (58.8-89.5) Effect of offset Main model without soak time offset 99.0 (97.5-99.6) 88.3 (78.3-93.7) 55.7 (9.2-78.4) Main model without depth offset 99.5 (98.7-99.8) 92.4 (86.9-95.6) 86.8 (74.4-93.2) Model robustness Yellowfin tuna included as variable n.s. n.s. 70.6 (40.2-85.6) Soak time coefficient + 1s.e. 99.6 (98.6-99.8) 91.4(85.3-95.0) 85.2 (70.9-92.5) Soak time coefficient - 1s.e. 99.2 (98.0-99.7) 91.0(84.4-94.8) 71.1 (42.0-85.6) Dataset for 1990s expanded to include sets 99.1 (98.1-99.6) 89.7(81.8-94.2) 80.4(61.2-90.1) targeting any tuna species Datasets from reduced area 99.4 (98.3-99.8) 89.1 (79.7-94.1) 87.2(71.5-94.25) 1 All model estimates are adjusted to account for unidentified sharks in the 1990s, as described in Methods.

Model estimates for oceanic whitetip and silky sharks were very stable, while estimates for dusky sharks fluctuated somewhat (Table S2.1). In general, the estimated

33 change in abundance from our main model falls mid-range of estimates from the robustness models. Across the suite of models, mean estimated declines for oceanic whitetip sharks were all greater than 99%, and the mean estimated declines for silky sharks were between 89.1-91.4%. The variable yellowfin tuna catch was nonsignificant for oceanic whitetip and silky sharks, but decreased the mean estimated change in abundance of dusky sharks to a 70.6% decline (Table 2.5). Incorporating the error around the soak time coefficient estimate had the greatest effect on dusky shark estimates because the error was greatest for this species. Expanding the dataset to include sets targeting any tuna species had very little effect on estimates for dusky shark; examining data from a reduced area increased the magnitude of the estimated dusky shark decline to 87% (Table 2.5). Our results for oceanic whitetip and silky sharks are very robust, and although the magnitude of the declines fluctuated among models of dusky sharks, each model estimates that there this species has also declined substantially.

Effects of Other Operational Factors

In addition to the operational variables included in our models, several other gear changes, including leader material, and hook type and size, occurred between the 1950s and 1990s, but could not be included in our models because of a lack of overlap in their use between the two time periods.

The effect on shark catchability of different hook types and sizes is not well known. Any effect of hook type on catch rates should be lessened by the fact that three-quarters of sets in the 1990s data used a somewhat similar type of hook (J-hook) as those in the 1950s (tuna hook). We are not aware of any studies that have examined differences in shark catch rates between these two hook types. The impact on shark catch rates of changes in hook size and the introduction of circle hooks into the fishery has, however, been studied, and apparently has been either negligible, or to slightly increase shark catch rates (Bacheler & Buckel 2004, Watson et al. 2005, Yokota et al. 2006; also see Chapter 3), implying that estimated shark declines would be greater if we could include these factors in our analysis.

34 A change between the 1950s and 1990s from wire to monofilament leaders has substantially increased shark bite-offs (with unknown survival thereafter), but seems to have little effect on shark catch rates (those that are retained). The effect appears to be species-specific: in one of the two known experiments addressing this issue, of the species that were included in our analyses, catches of dusky, scalloped hammerhead, and tiger sharks were each higher on monofilament leaders, whereas catches of shortfin mako sharks were higher on steel leaders (Branstetter & Musick 1993). Sample sizes in both studies were very small (n<13 for any of the shark species in our analyses; Berkeley & Campos 1988, Branstetter & Musick 1993). We consider this an important area for future research: as improved estimates of the effects of these technical characteristics of the fishing operations on shark catch rates become available it will be possible to refine the estimates made in this study.

ACKNOWLEDGEMENTS

We thank S. Nichols and M. McDuff for the 1950s data, C. Brown for the 1990s data; R. Lessa for access to unpublished oceanic whitetip data; D. Swan for technical support; W. Blanchard for statistical advice; P. Ward for discussions about longline fisheries; B. Worm and D. Kehler for comments on the manuscript. This research is part of a larger study on pelagic longlining, initiated and sponsored by The Pew Charitable Trusts (R. A.M.), and was also supported by grants from the Natural Sciences and Engineering Research Council of Canada (J.K.B., R.A.M), the Pelagic Fisheries Research Program (R.A.M.), and the Future of Marine Animal Populations project of the Sloan Foundation Census of Marine Life (R.A.M.).

35 CHAPTER 3

Estimating Trends in Abundance from Fishery-Dependent Data: Recent Changes in Northwest Atlantic Pelagic Shark Populations

36 Abstract

Oceanic and large coastal sharks appear to have declined precipitously in the Northwest Atlantic Ocean over the past three decades, but continue to be heavily exploited both directly and incidentally. To assess their recent trends in abundance, I standardized shark catch rates for each of eight species groups from the U.S. Atlantic pelagic longline fishery's observer program, which has sampled the fleet since 1992. I used generalized linear mixed models and generalized estimating equations to account for correlations amongst fishing sets made on the same trip and by the same vessel, and compare trend estimates for each species across multiple subareas of the Northwest Atlantic, amongst three model types, and between the current analysis and a previous analysis of the pelagic longline logbooks (1986-2000). Shark catch rate data from this fishery show concordance between logbooks and observer data, suggesting that trends estimates from the logbook data are reliable. Trend estimates from the observer data suggest moderate declines (34 - 53%) in blue, mako, and oceanic whitetip sharks and precipitous declines (76%) in both hammerhead and large coastal sharks (a group comprised dusky, silky, and night sharks) between 1992 and 2005. Individual year estimates, however, show these populations stabilizing at reduced levels, while the tiger shark population appears to have increased in the past few years. These apparent recent changes must be set in the context of the significant declines in shark abundance that occurred to the observer program, and these depleted populations managed accordingly.

INTRODUCTION

Understanding the impacts of anthropogenic perturbations on wild populations is essential if they are to be effectively managed and conserved. The high levels of fishing mortality now experienced by many sharks and their inherent vulnerability to overexploitation have brought this group of fishes to the forefront of marine conservation in recent years (FAO 1998, 2000, Musick et al. 2000, ICC AT 2004, CITES 2006). Among shark species, large pelagic sharks are heavily exploited top predators. These sharks, which are circumglobal in distribution, include wide-ranging oceanic species,

37 such as blue (Prionace glauca), mako {Isurus species), oceanic whitetip {Carcharhinus longimanus), thresher (Alopias species), and silky (C. falciformis) sharks, and more coastal species, like hammerhead {Sphyrna species), tiger (Galeocerdo cuvier), and dusky (C. obscurus) sharks (Compagno 1984, 2001).

Exploitation of large pelagic sharks has increased greatly over the past four decades (Bonfil 1994, Rose 1996, Barker & Schluessel 2005). Oceanic shark species are primarily caught incidentally or as secondary targets, in pelagic longlines, drift gillnets, purse seines, and floating aggregating devices, with some also taken in directed commercial and recreational fisheries, while the large coastal species are caught in directed multi-species shark fisheries, recreational fisheries, and as bycatch in pelagic longline and other fisheries (Bonfil 1994, NMFS 2001, Megalofonou et al. 2005, Scott 2007). Only a few large pelagic shark species are valued for their meat (e.g. shortfin mako), but this group of species comprises the majority traded in Asia's rapidly escalating shark fin trade (Rose 1996, Clarke 2004, Clarke et al. 2006). In particular, fins of hammerhead sharks {Sphyrna lewini, S. mokarran, S. zygaena) are among the most valuable and preferred fin types in the shark fin trade (Clarke 2003, Abercrombie et al. 2005); fins from highly valuable species are one of the most expensive marine commodities globally (Parry-Jones 1996, Clarke 2004). Blue shark fins, despite not being of particularly high quality, are estimated to be the most commonly traded in Hong Kong's shark fin trade, likely because of the large numbers of blue sharks that are caught in pelagic fisheries globally (Clarke et al. 2006).

Determining reliable, unbiased trends in abundance for sharks is critical if we are to identify any threatened species, and mitigate the impacts of fisheries on their populations. Indices of abundance are not only key components of the complex population dynamics models used in stock assessments (Hilborn & Walters 1992, Maunder & Punt 2004), but for many shark species, for which there are inadequate catch records and biological information, will form the sole indicator of the direction and magnitude of changes in their abundance.

38 Quantifying trends in abundance for large pelagic sharks is complicated by several factors. Distributed in epipelagic and upper mesopelagic waters, these species are rarely caught in fishery-related research surveys, most of which are designed for groundfish and conducted with trawls. Surveys that have sampled sharks often are limited by low sample size to provide estimates only for the most frequently caught shark species. On the east coast of the U.S., however, two dedicated shark-targeted longline research surveys conducted annually in North Carolina and Virginia since 1972 and 1974 respectively, provide valuable multi-decadal records for many large coastal shark species (Musick et al. 1993, Ha 2006, Myers et al. 2007). These surveys do not, however, sample any of the oceanic shark species, and are limited by low sample size for some species. Conversely, fisheries sample intensely over large regions closer in size to the geographic ranges of shark populations, but because fisheries operations are much more variable than designed research surveys, standardizing these data can be a challenge (Maunder & Punt 2004, Bishop 2006). What is more, there is a dearth of long-term fishery-dependent data for sharks. In most commercial fisheries, shark catches were first recorded at the species level only in the 1990s, and reliable species identification remains a challenge. Among fisheries data types there also is a tradeoff between logbook data, which are self-reported by fishermen, and scientific observer data, which should be more accurate but are usually less common because observer programs tend to monitor only a small proportion of a commercial fleet. The situation is exacerbated for oceanic sharks because much of their exploitation occurs on the high seas, where their catches by the many fleets targeting tunas have been unrestricted and largely unreported.

In the Northwest Atlantic Ocean, many large pelagic shark species appear to have declined precipitously. Baum et al. (2003) estimated that between 1986 and 2000 changes in abundance for 17 such shark populations in this region ranged from 40% declines for two species of mako sharks up to 89% declines for three species of hammerhead sharks. Estimates were based on logbooks from the U.S. Atlantic pelagic longline fishery, and for each of the populations analyzed included both the largest sample size of any dataset and the most extensive sample of their range in the Northwest Atlantic, with coverage throughout the U.S. Atlantic EEZ, and in international waters as

39 far north as Newfoundland, Canada (50°N) and south to Brazil (0°S). Generalized linear models (GLM) were fitted to these data with the truncated negative binomial distribution. The aim of modelling only the non-zero catches for each species was to avoid the potential bias of any change in fishermen's tendency to record their shark catches over time (Baum et al. 2003). Six additional analyses using different statistical distributions and subsets of the data (based on the tendency of sharks to be recorded on different vessels) led to some quantitative differences in trends, but similar conclusions of significant declines in abundance (Baum et al. 2003 Supplementary Online Material). This research has, however, been criticized for inferring trends in abundance from a single data source, particularly since the data were from logbooks (Burgess et al. 2005, but see rebuttal in Baum et al. 2005).

To address these concerns and examine more recent changes, here I build upon this earlier research, estimating recent trends in relative abundance for pelagic shark populations in the Northwest Atlantic using observer data from the U.S. pelagic longline fleet. Sharks included in this analysis are classified either as oceanic or large coastal species, according to the U.S. Atlantic Highly Migratory Species Fishery Management Plan (FMP) (NMFS 2006). At the time of Baum et al. 's (2003) study, the observer data set comprised samples of only about 3-4% of the fleet's effort from nine years (1992- 2000). The limited amount of data, high variability of longlining operations by the fleet, and autocorrelation within trips, made the observer data a challenge to standardize and rendered it insufficient to estimate trends for some shark species (Baum et al. 2003, 2005). Now, using five more years of data (2001-2005) and recently implemented statistical methods (S AS 2005), I standardize catch rates from these data for temporal, spatial, and operational differences among sets, while also accounting for non- independence among sets made on the same trip and among trips made by the same vessel, to obtain estimates of trends in abundance for eight shark species groups. To examine the accuracy of the logbook data used in Baum et al. (2003), I compare shark catch rates by species between the logbook and observer data for the overlapping time period (1992-2000) using maps and summary statistics of the catch rates. This presentation of data is also useful for showing the spatial distribution of shark catches, as

40 well as comparing shark catch rates among species and areas. Finally I compare estimated trends in relative abundance of each shark species between data sets, and make suggestions for future improvements to models of the observer data.

METHODS

Data and Shark Species

The U.S. pelagic longline fishery is the major source of exploitation for large pelagic fishes off the eastern coast of North America. The fleet primarily targets swordfish (Xiphias gladius) and yellowfin tuna {Thunnus albacares), and to a lesser extent bigeye tuna (T. obesus), mahi mahi (Coryphaena hippurus), and pelagic sharks (e.g. shortfin mako sharks). Substantial numbers of oceanic and large coastal sharks are also caught incidentally in the fishery. Detailed descriptions of the fishery can be found in Berkeley & Campos (1988), Hoey & Moore (1999), and Beerkircher et al. (2002).

I obtained the dataset collected by the Pelagic Observer Program (POP) for the U.S. pelagic longline fishery operating in the Northwest Atlantic Ocean directly from the U.S. National Marine Fisheries Service (NMFS) Southeast Fisheries Science Center (SEFSC), and met with POP staff in Miami, in October 2005 to discuss the fishery, observer program and dataset. Detailed information on this observer program, as well as a version of this dataset (which does not contain data on the vessel that observations were made on), are available on the NMFS SEFSC website (http://www.sefsc.noaa.gov/pop.jsp). The logbook dataset from this fleet used here is identical to that analysed by Baum et al. (2003). Both the observer and logbook datasets consist of counts of the species caught per longline set, as well as detailed information on the timing and location of fishing activity, the gear deployed, and other operational characteristics. Logbook data have been compiled from fishers in the pelagic longline fleet since 1986 as part of the Pelagic Longline System (Cramer 1995) and form the largest available sample of these species in the Northwest Atlantic. Records up to 2000 comprised over 214,000 sets and, with an average of 550 hooks per set, totaled over 110 million hooks. Scientific sampling of the fleet was initiated in 1992 under the POP, and observers have monitored between 2.2 and

41 11.5% of the sets (mean=5.5%) in the fishery each year since (Beerkircher et al. 2004). The observer dataset was available for the years 1992 to 2005 and (excluding sets in the experimental fishery conducted to test measures for reducing sea turtle bycatch) totaled almost 7,000 sets and over 4.8 million hooks.

Both datasets underwent extensive checks for consistency in content and recording details. Data corrections and selection criteria applied to the logbook data are detailed in Baum (2002) and Baum et al. (2003); notable among these were the exclusion both of fishing sets that used bottom longline gear (to target large coastal sharks) because of differences in fishing techniques, and pelagic longline operations that directly targeted sharks (e.g. shortfin makos). No bottom longline sets were included in the POP data, and I excluded the shark-targeted sets (n=32) because these formed only a small proportion of the observer data (<0.5%), were not distributed evenly in the time series, and had higher shark catch rates than swordfish and tuna targeted set, and thus could have biased catch rate models and misled conclusions about shark trends in relative abundance. I then performed simple summary statistics, plots, and range checks on all variables of interest in the observer data, and corrected obvious errors. For example, implausible dates and locations (e.g. on land) could often be corrected using information from other sets on the same fishing trips. Any outstanding queries were discussed with POP staff and corrected wherever possible.

Over twenty-five shark species have been caught in the U.S. pelagic longline fishery (Table 3.1), but difficulties with species identification and the infrequency of catches for some species precludes estimation of trends in abundance for most individual shark species. Blue, tiger, and oceanic whitetip sharks are easily identified and caught in sufficient numbers to model their catch rates (Tables 3.1-3.3). Reliable identification of some shark species, however, is difficult even for trained observers particularly since much fishing effort occurs at night and sharks are generally not brought onboard. For example, members of the hammerhead (Sphyrna species), thresher (Alopias species), mako (Isurus species), and requiem shark (Carcharhinus species) genera are each fairly

42 Table 3.1. Total of each elasmobranch species recorded in the U.S. Atlantic pelagic longline observer program between 1992 and 2005. Oceanic and large coastal shark species are arranged as grouped in the analyses. Note that silky shark is ecologically an oceanic species, but is classified in the NMFS FMP as a large coastal. Species recorded fewer than 5 times not shown.

Species Number Common name Latin name caught Oceanic shark; species Blue Prionace glauca 28,317 Mako shark species Isurus species 3,433 Shortfin mako shark /. oxyrinchus 2,705 Longfin mako shark 1. paucus 217 Unidentified makos 1. species 511 Thresher shark species Alopias species 921 Bigeye thresher shark A superciliosus 627 Common thresher shark A. vulpinus 148 Unidentified thresher shark A. species 146 Oceanic whitetip shark Carcharhinus longimanus 506 Porbeagle shark1 Lamna nasus 192 Large coastal shark species Hammerhead shark species Sphyrna species 1,292 Scalloped hammerhead S. lewini 742 Great hammerhead S. mokarran 93 Smooth hammerhead S. zygaena 15 Unidentified hammerhead S. species 442 Tiger shark Galeocerdo cuvier 1,190 Coastal group 1 (dusky, night, silky) Carcharhinus species 7,212 Dusky shark C. obscurus 1,924 Night shark C. signatus 1,649 Silky shark C. falciformis 3,639 Coastal group 2 (bignose, blacktip, bull, dusky, night Carcharhinus species 9,799 sandbar, silky, spinner, all unidentifiedI sharks) Bignose shark C. altimus 47 Blacktip shark C. limbatus 125 Bull shark C. leucas 42 Sandbar shark C. plumbeus 550 Spinner shark C. brevipinna 31 Sand tiger shark1 Carcharias taurus 6 Other shark ispecie s Atlantic sharpnose shark1 Rhizoprionodon terraenovae 20 Collared dogfish1 - 6 Crocodile shark1 Pseudocarcharias kamoharai 162 Reef shark1 - 7 Smooth dogfish1 Mustelus canis 59 Spiny dogfish1 Squalus acanthias 95 Unidentified dogfish1 - 38 Unidentified sharks Unidentified requiem sharks Carcharhinus species 179 Unidentified sharks - 1,613 Total (all sharks) 46,052 Other elasmobranchs Pelagic ray2 Pteroplatytrygon violacea 1,646 Manta ray2 Mobulidae 92 Skates and rays2 - 8,072 1 Species not included in analysis because of small sample size. 2 Skates and rays not analysed.

43 Table 3.2. Total number of each oceanic and large coastal shark species (group) caught, and the total number of fishing sets monitored, each year between 1992 and 2005 as recorded in the U.S. Atlantic pelagic longline observer program (after data checks and removing shark targeted sets). Analyzed species (groups) and their numbers caught are in bold.

19921 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Oceanic shark species Blue shark 2006 5128 4446 3862 262 2726 1191 1126 2298 639 589 533 2953 558 Mako shark species 289 451 332 443 40 209 76 127 237 112 176 211 519 211 Shortfin mako 132 324 235 282 33 201 68 111 216 101 163 174 485 180 Longfin mako 8 17 16 15 7 6 8 13 15 9 11 33 32 27 Unidentified mako 149 110 81 146 0 2 0 3 6 2 2 4 2 4 Thresher shark species 49 142 46 77 34 27 21 42 2 67 88 50 91 115 Bigeye thresher 30 87 35 70 30 1 19 36 52 45 60 27 47 72 Common thresher 19 54 10 5 0 0 0 1 10 1 1 5 24 18 Unidentified thresher 0 1 1 2 4 10 2 5 10 21 27 18 20 25 Oceanic whitetip 22 56 45 61 38 39 37 46 14 16 32 20 37 43 Large coastal shark species Hammerhead shark species 191 307 108 132 28 52 69 34 79 71 37 54 62 68 Scalloped hammerhead 95 128 66 95 10 43 32 11 34 57 31 38 47 55 Great hammerhead 5 9 2 7 10 2 3 15 28 8 0 1 2 1 Smooth hammerhead 1 0 0 1 2 0 1 0 0 0 1 5 3 1 Unidentified hammerhead 90 170 40 29 6 7 33 8 17 6 5 10 10 11 Tiger shark 38 93 51 60 63 46 61 41 81 62 88 107 209 190 Coastal shark group 1 565 739 920 571 342 442 390 454 683 387 145 359 404 811 Dusky 415 283 471 135 120 38 142 26 110 35 8 7 64 70 Night 1 2 24 48 61 71 124 104 254 95 64 221 174 406 Silky 149 454 425 388 161 333 124 324 319 257 73 131 166 335 Coastal group 22 846 1046 1080 713 414 501 438 609 918 507 255 754 624 1094 Other Carcharhinus spp. 237 152 97 71 29 27 16 60 27 12 6 11 33 17 Unidentified shark species Unidentified requiem sharks3 0 0 0 0 0 0 0 0 0 0 0 0 63 116 Unidentified sharks 44 155 63 71 43 32 32 95 208 108 104 384 124 150 Total number of sets monitored 328 808 645 682 360 458 286 430 473 403 335 555 643 546 1 Monitoring for 1992 did not cover the entire year, as the program did not begin until May. 2 Coastal group 2 = Coastal group 1 + Other Carcharhinus species + both unidentified shark categories. 3 'Unidentified requiem shark' category was not created until 2004 Table 3.3. Total number of each oceanic and large coastal shark species (group) caught, by fishing area, as recorded in the U.S. Atlantic pelagic longline observer program from 1992 to 2005 (after data checks and removing shark targeted sets). For each analyzed species (in bold) , the percentage of monitored sets with a catch of that species is shown in brackets and the areas with more than 50 recorded (in bold) were included in models Species (groups) Fishing Area 1 2 3 4 5 6 7 8 91 Oceanic shark species Blue 48 (30.6%) 15 363 (25.9%) 413 (29.0%) 5987 (78.3%) 5084 (84.6%) 15611 (99.4%) 630 (56.9%) 166(69.7%) Mako spp. 8 291 (9.7%) 104(11.0%) 116(11.9%) 1095(43.3%) 439 (47.6%) 1293(61.4%) 74(14.4%) 13(12.1%) Shortfin mako 5 210 64 101 875 318 1081 47 4 Longfin mako 3 71 40 11 42 11 3 27 9 Unidentified mako 0 10 0 4 178 110 209 0 0 Thresher spp. 20(16.5%) 169(5.7%) 63 (6.5%) 122(10.7%) 449 (14.6%) 10(1.8%) 10(2.0%) 72 (12.8%) 6 (7.8%) Bigeye thresher 18 103 43 87 298 3 6 64 5 Common thresher 0 10 6 15 107 7 3 0 0 Unidentified thresher 2 56 14 20 44 0 1 8 1 Oceanic whitetip 58 (42.4%) 84 (3.0%) 216(17.9%) 78 (8.9%) 3 (0.3%) 1 (0.2%) 1 (0.2%) 28 (3.6%) 37 (5.6%) Large coastal shark species Hammerhead spp. 2(1.2%) 145 (2.6%) 178(11.2%) 349(18.0%) 599 (13.5%) 9(1.0%) 4 (0.8%) 4 (0.9%) 2 (3.0%) Scalloped hammerhead 0 104 125 182 327 1 1 2 0 Great hammerhead 2 17 28 33 12 1 0 0 0 Smooth hammerhead 0 2 0 3 9 0 0 0 1 Unidentified hammer- 0 22 25 131 251 7 3 2 1 Mhead Tiger 32(21.2%) 221 (7.8%) 335 (24.6%) 245 (23.6%) 175(8.9%) 123 (13.4%) 0 (0%) 56 (10.7%) 3 (4.5%) Coastal group 1 140 (64.7%) 1380(19.6%) 1245(43.4%) 3443 (74.8%) 862 (20.5%) 105 (8.9%) 1 (0.2%) 8(1.8%) 28 (28.8%) Dusky 5 175 232 903 535 72 1 0 1 Silky 134 1121 912 1157 253 32 0 8 22 Night 1 84 101 1383 74 1 0 0 5 Coastal group 22 159(68.2%) 2343(35.1%) 1584 (54.2%) 3973 (83.5%) 1503(32.6%) 167(14.8%) 10 28 32 Other Carcharhinus spp. 0 158 119 160 331 27 0 0 0 Unidentified sharks Unidentified requiem 0 82 11 64 16 4 0 2 0 sharks Unidentified sharks 19 723 209 306 294 31 9 18 4 1 Area 9 not included in analysis because sampling only occurred in this area in three years. 2 Coastal group 2 = Coastal group 1 + Other Carcharhinus species + both unidentified shark categories. distinctive from other sharks, but species within each of these genera may not be easily distinguished from one another. Thus, in the logbook data analysis, Baum et al. (2003) modelled trends in abundance for each of these genera rather than for the individual species. For the observer data analysis, I also modelled these species by genus because observers identified a substantial proportion of these sharks only to this taxonomic level (e.g. 34% of hammerhead sharks, 16% of thresher sharks, and 15% of mako sharks; Table 3.1). Scalloped hammerheads (S. lewini), bigeye threshers (Alopias superciliosus), and shortfin makos (Isurus oxyrinchus) are thought to be the most common species of each of these genera. A systematic bias occurred in observers' recording of Carcharhinus species whereby night shark (G signatus) were seldom correctly recorded (but were instead recorded as unidentified shark or misidentified as dusky (C. obscurus) or silky (C. falciformis) sharks) until species identification training improved in the late- 1990s (L. Beerkircher pers. comm, NMFS, SEFSC; Beerkircher et al. 2002). In recognition of misidentifications of Carcharhinus species a new species code 'unidentified requiem shark' was added to the observer data collection system in 2004. Because I was concerned that trends for individual Carcharhinus species might reflect changes in observers' recording tendencies rather than trends in relative abundance of the species, I modelled these species in two ways to encompass the two extreme possibilities, first by grouping only the three most commonly recorded species (dusky, silky, and night sharks), and second by grouping these three species with all other recorded Carcharhinus species (bignose, blacktip, bull, sandbar, and spinner), unidentified requiem sharks, and unidentified sharks (which I presume are most likely belong to the genus Carcharhinus) (Tables 3.1-3.3). Herein, I use the word "species" to refer to species groups as well as to individual species.

I divided the sampled region of the northwest Atlantic into the same nine areas as Baum et al. (2003) for my analyses, so that the two could be easily compared (l=Caribbean, 2=Gulf of Mexico, 3=Florida east coast, 4 =South Atlantic Bight, 5=Mid Atlantic Bight, 6=Northeast coastal, 7=Northeast distant, 8=Sargasso Sea & North central Atlantic, 9=Tuna north and south). These areas correspond closely with those used by NMFS (and I use the same names as NMFS), the primary difference being that four areas

46 with very little fishing effort were combined into two areas (areas 8 and 9 reassigned as Area 8, areas 10 and 11 reassigned as Area 9). In models of the observer data, I excluded records from Area 9 because monitoring occurred in this region only from 1996 to 1999, and only a few sets (n= 62) were observed.

Shark Catch Rates in U.S. Pelagic Longline Observer and Logbook Data

I compare unstandardized shark catch rates among areas and among species in the 1992-2005 observer data, and between the observer and logbook data for the common time period 1992-2000. Comparisons of shark catch rates between these two data sources on a set by set basis would have necessitated the use of identifiers for individual trips, which I could not obtain for privacy reasons. Instead, for each species I visually compared the spatial distribution of catches between these datasets by plotting maps of their mean catch rates for both datasets. I then made boxplots of the catch rates recorded in the two datasets for each species, in each of the nine areas and overall, including only the positive sets (i.e. sets in which recorded catch of the species was greater than zero), in order to evaluate the main assumption of Baum et al. 's (2003) logbook analysis, that fishermen recorded positive shark catches approximately correctly. Because the observer program sampled such a small proportion of the fleet, I used these comparisons only as a gross check on the magnitude of and patterns in the catch rates. Finally, for each species I plotted the overall raw catch rates from the observer data across years.

Observer Data Models

Indices of relative abundance for each species are derived from models that 'standardize' catch rates to remove the impact of factors that affect the efficiency of the fishing gear, which is otherwise assumed to be constant among operations (Maunder & Punt 2004, Ward & Myers 2005b). The model predicted catch rates are assumed to be proportional to abundance.

47 Generalized Linear Mixed Models and Generalized Estimating Equations

I modelled catch rates in the observer data for each shark species using both generalized linear mixed models (GLMM) and generalized estimating equations (GEE). These models are extensions of the generalized linear model (GLM), which is commonly used to model fishery catch rates (Maunder & Punt 2004, Venables & Dichmont 2004) because it allows for nonlinear relationships between explanatory variables and the response variable (via the link function) and for a non-normal error structure of the response. In the observer data, the response variable (the number of sharks (of a given species) caught on a longline set) was highly variable, with most sets catching zero (Table 3.3), a smaller proportion catching one, and even fewer catching several to many sharks. For all models, I specified a negative binomial distribution for the catches:

f O r + ~ f(y;k,M) = y^ ^/ _teLjfor y = 0> 1,2,... + y T r(j+i)r (1 + kju) \KJ with T the gamma function, n the mean, and k the negative binomial dispersion parameter, which is an appropriate probability distribution for discrete, overdispersed data like these that are more variable than expected in a Poisson distribution (McCullagh & Nelder 1989). Essentially the negative binomial distribution allows for extra variation in that its variance, y, + ky2, is greater than the mean rather than being equal to it as it is in the Poisson distribution, and as k—>0 the Poisson distribution is recovered. I used a log link to link the mean catch to a linear combination of explanatory variables, such that the expected mean catch, juSii, of species s on set i in a base GLM model is:

r log(//J,)=jC; /f + log(//,.)

where xt is the vector of covariates for observation i, T indicates the vector transpose, p is the vector of estimated parameters, and the number of hooks on set i, ht, is a known value that is treated as an offset, which is equivalent to dividing the catch by the effort to determine a catch rate, but maintains the discrete probabilistic sampling distribution of the data.

48 Exploratory analysis of the observer data showed that among the 993 trips and 225 vessels on which fishing was observed, for each species, high catch rates were aggregated on certain trips and certain vessels, thereby violating the assumption of independent observations. Essentially fishing sets made on the same trip (and those made by the same vessel) can be thought of as repeated measures in a longitudinal analysis, and the problem is similar to pseudoreplication because the data violate the assumption of independence among observations. To address this non-independence in the data, I used GLMM and GEE, both of which also allow for correlated response data (Diggle et al. 2002, Dobson 2002, Venables & Dichmont 2004).

Initially I modelled a single level of non-independence in the data, for sets made on the same trip. Based on examination of the data, I assumed that the correlation among sets made on the same trip would decline the farther apart in time sets occurred from one another. To do so, I specified a first-order autoregressive correlation structure (AR1) with trip as the clustering variable:

in both the GLMM and GEE models. I refer to these models as the GLMM-t and GEE models. An alternative correlation structure, 'exchangeable', in which all observations in a cluster are equally correlated did not represent patterns observed in the raw data, nor did models fitted with this structure converge easily.

Then, to account for both levels of correlation, I fitted GLMMs in which I assumed a common correlation among trips made by the same vessel as well as the AR1 correlation structure for sets made on the same trip. A motivation for implementing GLMM in addition to the simpler GEE was that only in the mixed model was it possible to model both levels of correlation (i.e. for vessel and trip). I present this last model, referred to as

GLMM-vt, as the main model in the results. The expected mean catch, /usj, of species s on set i in the GLMM-vt model is:

49 where xt is a vector of covariates for observation i, T indicates the vector transpose, /? is the vector of estimated parameters (the fixed effects described below), z, is a vector of covariates, y is the vector of predicted random effects, the offset term hi is the number of hooks on the set.

Fixed Effects

I considered a suite of temporal, spatial, and operational factors that could affect catch rates as candidate variables to include in models, and for each examined the available data. The variables included in the models are described below.

To model the rate of change in catch rates (which is interpreted as the trend in relative abundance), I treated 'year' as a continuous variable in the models. I also fitted models treating 'year' as a categorical variable to obtain individual year estimates. Fishing occurred throughout the year, and I modelled the seasonal cycle by fitting sines and cosines with periods, y, of Vi and 1 year as

2 q(dt) = Yj[gjcos(2nids. /365,25) + xj sin(2njdsi /365.25)],

where dt is the Julian day of the year that set i occurred on, and the estimated parameters are q} and r r

I accounted for spatial variability in the fishery by modelling the eight fishing areas. For each species, I included data only from areas in which a minimum of fifty sharks of that species had been recorded (Table 3.3) because the models often had difficulty converging when areas with fewer observations were included, and estimates based on few observations were imprecise and tended to be unreasonably high or low. Ocean depth, which can also be viewed as a proxy for distance from the coast, also may be an important spatial determinant of the prevalence of sharks, particularly for the coastal shark species. Using the recorded latitude and longitude of sets I obtained data on ocean depth from the International Research Institute for Climate and Society

50 (http;//ingrid.ldgo.columbia.edu/SOURCES/.WORLDBATH/.bath/) (Table 3.4) and included a quadratic term for it in our models. Since both seasonal effects on catch rates and trends over time in catch rates could differ among areas, I included interactions for area with season and area with year.

Table 3.4. Characteristics of observed sets in the U.S. pelagic longline fishery between 1992 and 2005. Mean values ± 1 SD. Variable Mean ± 1 SD Hooks per set 702 ±259; range: 10-1548 Ocean depth (m below sea level) -1924 ± 1441; range: -7996 - -10 Average hook depth (m) 45.8 ± 19.8; range: 6.4 -182.9 Soak duration (decimal hours) 13.9 ± 2.7; range: 1.7 - 43.5 Number of light sticks 305 ± 277; range: 0 (24%) -1488 Bait species Mackerel (25%), herring (5%), squid (56%), artificial (1 %), sardine (9%), scad (2%), other (<1%) Surface water temperature (°C) 24.7 ± 4.1; range: 6.8 - 31.8 Hook type J-hook (n=3957), circle hook (n=1194)

Table 3.5. Total number of monitored sets directed towards each target species between 1992 and 2005 in the US Atlantic pelagic longline observer program. Note that shark-targeted sets were excluded from analyses.

Fishing,Are a Target species 1 2 3 4 5 6 7 8 9 All Bigeye tuna, Thunnus obesus 0 2 18 0 47 35 0 0 0 102(1.5%) Dolphinfish, Coryphaena 0 0 0 20 0 0 0 0 0 20 (0.3%) Multiple targeted species 1 772 27 193 515 285 36 0 2 1831 (26.2%) Shark 0 0 0 15 17 0 0 0 0 32 (0.5%) Swordfish, Xiphias gladius 80 419 824 516 162 52 452 439 43 2987 (42.8%) Tuna, Thunnus species 3 463 14 45 357 125 14 0 0 1021 (14.6%) Yellowfin tuna, T. albacares 1 916 0 9 35 9 0 0 21 991 (14.2%)

I also included variables for operational characteristics that can affect shark catch rates (Ward et al. 2004, Ward & Myers 2005 a, 2007) including hook depth, soak duration (defined as the amount of time from the midpoint of the gear setting to the midpoint of the gear hauling), number of light sticks, target species, timing of the fishing operation (e.g. early morning, day, evening, night), and the type of bait (Table 3.4, 3.5). I excluded dolphinfish-targeted sets because of the low sample size (<0.3% of data), but added these data back in when models indicated that the variable 'target species' was non-significant.

51 Two additional variables, surface water temperature and hook type (circle or J-hook) (Table 3.4), were not always recorded, and including them in models involved a tradeoff between missing potentially important sources of variation or modelling only the subset of the dataset for which the variables were recorded (Maunder & Punt 2004). I included surface water temperature because data were missing for only -2% (n=150) of sets. Although hook type was less consistently recorded (n=5151; 74% of sets), its potential effect on shark catch rates (e.g. Bacheler & Buckel 2004, Watson et al. 2005, Kaplan et al. 2007) combined with a mandated change in 2004 to use only circle hooks, warranted investigation of its effect on catch rates. I therefore fitted models for each species on the subset of data with hook type recorded (n=5151 sets; Table 3.6), including it as a model covariate, as well as on the full data set without hook type as a covariate. The latter models are presented for those species for which the effect of hook type was non significant. For species for which hook type was significant, I also fitted models on the 'hook type' data subset without this variable, to determine if the difference in year estimates between the full and hook type subsetted data was due to the inclusion of the hook type covariate or simply the difference in number of observations between the two. I found that the difference in each case was indeed due to the inclusion of hook type in the model, and thus present results only for the hook type subsetted data for these species. Other operational variables with many missing values (e.g. leader material) were not included in the models.

Fishing Area Monitored fishing effort (no. of sets) 1 2 3 4 5 6 7 8 9 All Full data set 85 2572 883 783 1116 506 502 439 66 6952 Trip subsetted 85 2558 859 773 1103 500 502 439 66 6885 Trip & vessel subsetted 85 2300 826 769 1071 459 455 402 66 6433 Hook type (HT) subsetted 81 2170 779 632 541 236 211 439 62 5151 HT & trip subsetted 81 2161 757 624 530 230 211 439 62 5095 HT, trip & vessel subsetted 81 1949 725 604 475 209 191 379 62 4675

52 Model Implementation

All analyses were conducted in SAS v.9.1 (SAS Institute, 2004); the GLMM used the more recent GLIMMIX procedure (SAS 2005). GEE were fitted using the 'repeated' statement in PROC GENMOD, which estimates parameters via the method of moments (SAS 2004). The GLMM models were fitted using PROC GLIMMIX, which estimates parameters using pseudo-likelihood and assumes that the random effects are normally distributed (SAS 2005). The GLMM-t models were implemented using the 'random' statement with trip as a random variable (G-side effect), and in order to fit these models, trips with only one set (n=67, <1% of data) had to be omitted (Table 3.6). The GLMM-vt models were specified using the 'random' statement for vessel and the 'random _residual_' statement for trip such that vessel was a random effect (G-side) but trip was a residual effect (R-side) and thus essentially treated the same as in the GEE model. Note that although conceptually it is possible to have both vessel and trip as true random effects, I found that computationally this was infeasible (e.g. even simplified models with few covariates would not converge). As in the GLMM-t models, in order to fit these models, vessels with only one observed trip and trips with only one observed set had to be omitted (n=519, 7.5% of data; Table 3.6). In order for the GLMMs to converge, starting parameter values had to be provided for the vessel and trip variances, and either the AR(1) parameter (p) or the negative binomial parameter (k) had to be held fixed. The GEE models provided some guidance in this regard, and in each case, various combinations of starting values were tried to ensure that the models converged to the same final values.

For each species, for each of the three model types, I implemented the full model containing all explanatory variables using the 1992-2005 observer data. I also fitted the GLMM-vt models for each species to the 1992-2000 observer data in order to compare trend estimates between the observer and logbook data. Because neither GEE nor GLMM are fitted using maximum likelihood it is not possible to compare models using the Akaike Information Criteria (AIC) in either case. Instead I compared models using backward-selection by statistical significance testing of regression coefficients with p- values at the a = 0.1.

53 RESULTS

Shark Catch Rates in U.S. Pelagic Longline Observer and Logbook Data

-100 -90 -80 -70 -60 -50 -40

Fishing effort sampled by the pelagic longline observer program is concentrated between the 200m and 1000m isobaths along the U.S. Atlantic coast and beyond the 1000m isobath in the Gulf of Mexico (Figure 3.1). The fleet itself covers a larger offshore area of the Northwest Atlantic, particularly in Areas 1, 8 and 9 (shown in logbook data, left column Figures 3.2, 3.3). The broad geographic coverage of this fishery makes it particularly useful for sampling wide-ranging oceanic sharks, as it

54 includes significant proportions of their Northwest Atlantic ranges. These data are, however, also suitable for assessing coastal shark species because of the concentration of effort from just inshore of the 200m isobath out to the 1000m isobath.

Shark catch rates are highly variable both among species, and for individual species among areas (Figures 3.2-3.4). Given this high variability, it is not surprising then that single locations (i.e. hexagons) with high catch rates differ between the two data sets (Figures 3.2, 3.3). Overall, however, the broad catch rate pattern is similar between the logbook and observer data for each shark species, with comparable regions of high and low catch rates respectively. Below, I overview the catch rates by species and areas, and compare the two datasets, noting discrepancies where they exist.

Blue shark, a widespread oceanic species, had the highest shark catch rates (Figures 3.2a,b, 3.4, 3.5). Blue shark catches were concentrated in the northernmost region (Area 7), located off the Grand Banks, where it was recorded on almost every (> 99%) observed longline sets (Table 3.3). Blue shark catch rates recorded by observers in this area between 1992 and 2005 typically ranged from 14 to 48 per 1000 hooks (interquartile range; mean = 39/1000 hooks). Catches over 50 blue sharks per set in Area 7 were also common (occurring on 18% of observed sets; maximum catch = 303), and observers recorded over 15,000 blue sharks there in total (Table 3.3). Blue sharks were caught frequently in other areas as well (except the Gulf of Mexico), especially the northeastern coast of the United States (Areas 5, 6) where their catch rates averaged just under 10 per 1000 hooks (Table 3.3, Figure 3.2a,b). Catch rates recorded by fishermen in Area 7 were higher than those of observers (Figures 3.2a,b, 3.4), apparently partially reflecting their tendency to round up the catch on sets with many blue sharks. Overall, however, blue shark (positive) catch rates recorded in the logbook and observer data were quite similar (Figure 3.4).

Of the other oceanic shark species, mako sharks also were commonly caught in the three northernmost areas (5-7; Figure 3.2c,d), occurring on about half of the observed sets (Table 3.3) with mean catch rates of 1.5 per 1000 hooks in Areas 5 and 6, and 3 per

55 Figure 3.2. Maps of the Northwest Atlantic Ocean showing unstandardized catch rates of oceanic shark species: blue shark (a,b), mako sharks (c,d), thresher sharks (e,f), and oceanic whitetip shark (g,h) as recorded in the U.S. pelagic longline logbook (a,c,e,g) and observer (b,d,f,h) data from 1992 to 2000. The mean catch per 10,000 hooks is plotted in each hexagon, and the scale differs among species to allow the greatest resolution of catch rates. Hexagons where no sharks of the plotted species were caught are displayed as empty; only cells with greater than 5 sets are plotted for the logbook data. Areas (modified from the U.S. National Marine Fisheries Service's classification for longline fisheries): 1 Caribbean, 2 Gulf of Mexico, 3 Florida east coast, 4 South Atlantic Bight, 5 mid Atlantic Bight, 6 northeast coastal, 7 northeast distant, 8 Sargasso/north central Atlantic, 9 tuna north & south. The 200m (dashed line) and 1,000m coastal isobaths (dotted line) are shown for reference.

56 ;ia r>?\-^< #£*?^S^' _^~&%

TWS I «0 sea

4K)

300

20S

1CSJ I *§2* 90 I ^ 0 #.- 0. /.-.

* !

•

A mi' r *.

8 • f' 'Nil . V

>< 9 ^

>h J 20

. '0

. • 9 I* »

-SO -80 -70 -30 -SO -4B -33 -K) -SO -70 -fO -13 -49

57 1000 hooks in Area 7 (maximum catch = 43). Elsewhere makos were infrequently caught (11% of observed sets). Thresher sharks were caught infrequently (7.8% of observed sets; Table 3.3) and at low rates in all regions (overall mean catch rate = 0.24/1000 hooks), with rare, isolated high offshore catches (maximum catch = 28; Figure 3.2e,f). Catch rates of oceanic whitetip shark were very low along the U.S. east coast and in the Gulf of Mexico, but slightly higher (~1 - 3 /1000 hooks) between the Yucatan Peninsula and Cuba (edge of Areas 1 & 2), in the eastern Caribbean, and along the northeastern coast of South America (Areas 1, 9; Figure 3.2g,h). Overall, the oceanic whitetip shark was caught on just over 5% of observed sets (Table 3.3), with a mean catch rate of 0.15 per 1000 hooks (maximum catch = 9). Catch rates (positive sets) of mako sharks were very similar in the two datasets, and for thresher and oceanic whitetip sharks were only slightly higher in the logbooks compared to the observer data (Figure 3.4).

Hammerhead shark catches were concentrated primarily within or along the 200m isobath of the southeastern U.S. coast (up to about Delaware, Areas 3-5) and Gulf of Mexico (Area 2) (Figure 3.3a,b). There, hammerheads were recorded on 35% of observed sets (1992-2005), and had a mean catch rate of 3 per 1000 hooks (maximum catch = 46). Outside of these areas, however, hammerhead sharks were seldom caught (6% of observed sets, Table 3.3; overall mean = 0.41/1000 hooks). A second coastal species, tiger shark, was caught on almost a quarter of observed sets (Table 3.3) along the southeastern U.S. coast (Areas 3,4) at a rate of about 1-3 per 1000 hooks (Figure 3.3c,d). Overall, this species was caught on 12% of observed sets, at the much lower rate of 0.31 per 1000 hooks (maximum catch =32). Catch rates (positive sets) of these species were similar in the two data sets, in each area and overall, with only slightly higher rates recorded in the logbook data (Figure 3.4).

Large coastal sharks of the genus Carcharhinus were the second most commonly recorded shark group (Table 3.1). Catches of this group appear to be dominated by silky sharks, followed by dusky and night sharks, although the extent to which misidentification of these carcharhinid species biases these numbers is unknown. These

58 Figure 3.3. Maps of the Northwest Atlantic Ocean (with areas as in Figure 2) showing unstandardized catch rates of large coastal shark species: hammerhead sharks (a,b), tiger shark (c,d), and large coastal sharks of the genus Carcharhinus (e,f) as recorded in the U.S. pelagic longline logbook (a,c,e) and observer (b,d,f) data from 1992 to 2000. The mean catch per 10,000 hooks is plotted in each hexagon, and the scale differs among species to allow the greatest resolution of catch rates. Hexagons where no sharks of the plotted species were caught are displayed as empty; only cells with greater than 5 sets are plotted for the logbook data. The 200m (dashed line) and 1,000m coastal isobaths (dotted line) are given for reference.

59 .."%

F '

* * --; 1 .j1 «* f • ><&

^I^S; >J

.•A" #

*• < - *& ^. * 9 % 1 3 ^ •"'::* t , v .*. J" 1". /TV*

"V ^ -t «-*4

•-' ** I& • *•."J3C*** f '

1 - : V/ \

*> ,*'•„ />>-H. r <4 :i-._3\ i -M -M -7D -90 -90 -40 -30 -»0 -80 -70 -80 -SO -40

60 100 Blue shark 80 60 40 20 (D to 0 i* ~ui.8i8iTJi.wa T3 12 Thresher sharks 5 ai—s 10 8 j2 7 t r I 1 "5(/>) 6 I I I I o r 7 i_ r i 7 I d) 4 N BE i 2 91 C »0i° o 0 *»—*c ' Hammerhead sharks w r T ^o 7 o 15 r JZ i o t o 10 I • To— CD 5 n ' Q. u .c If a i- J J- $10-*- — 0 Ml -•-o ' CD 35 Dusky, silky & night sharks O 30 25

20 I T I I 15 10 5 ?• «# M «• 8* ei 0 L IL Bi 9 5 6 7 8 9 All Area

Figure 3.4. Boxplots of catch per 1,000 hooks on non-zero sets for modelled shark species in each area and in all areas combined, according to the U.S. pelagic longline logbook (clear bars, left) and observer (grey bars, right) data for 1992 to 2000. species were caught regularly in Areas 1 to 5, peaking in the south Atlantic Bight (Area 4) where catches occurred on three-quarters of all observed sets (Table 3.3). The highest catch rates of these coastal species occurred in Areas 2 to 4, from the coast out to the 1000m isobath (Figure 3.3e,f): there observers recorded on average 7.9 per 1000 hooks (maximum catch = 75). Overall, these sharks were recorded on 12% of observed sets, and had a mean catch rate of 2.4 per 1000 hooks. Recorded catch rates of carcharhinid species were quite similar between the logbook and observer data (Figure 3.3e,f); the

61 discrepancy in catch rates on positive sets in Area 7 (Figure 3.4) is the result of the very low frequency of large coastal shark catches in this area (0.24% and 0.27% of sets between 1992 and 2000 in the observer and logbook data respectively) and six sets recorded (probably unrealistically) in the logbook data with catches exceeding 25 of these sharks. Note that data from this area were not included in the logbook analysis for this species group because of the low sample size and thus did not influence the logbook model estimates (Baum et al. 2003).

Blue shark Make shafts Oceanic wiWettp 10 0.5 0.5 1.5 !» 8 \\ 0.4 0.4 \ 6 10 0.3 #' 1°| I 0.3 II

4 0.2- 0.5 HA A • * 2 k • 0,1 • m 0 0.0 0.0 3.5- Siifcy shark 3.5 f pusly shark •is- 1.5 3.0- 3.0 2.5' 2.5' 10 2.0 2.0 1 " 1,5 • 15 :, -f 0.5 ' 4tu i ° 1.0' —• 0.5' 0) 0.5 j 0.5 vt 0.5 0,5 h ft 2' H+y+ • • 0.0 0.0 ««©t|>«d 0.0 0' 1992 1996 2000 2004 1932 1996 2000 2004 1992 1996 2000 2004 1992 1996 2000 2004 Year

Figure 3.5. Catch per 1000 hooks (± 1 standard error) in all areas combined, by year for each species (group), as recorded in the observer data from 1992 to 2005. Top row: oceanic shark species, middle and bottom row: large coastal shark species. Coastal group 1 is comprised of silky, dusky, and night sharks, while coastal group 2 also includes Other Carcharhinus species and Unidentified sharks.

62 Annual mean unstandardized shark catches rates in the observer data, for all areas combined, showed a fairly high degree of interannual variation and a trend of declining catch rates for many of the species (Figure 3.5). The increasing trend in night shark catch rates reflects increasing identification of this Carcharhinus species as observer training improved in the laste-1990s.

Model Estimates of Recent Changes in Shark Abundance

Herein I present results of the main model (GLMM-vt) as well the GLMM-t and GEE models for each species, focusing solely on the model estimated changes in relative in abundance. I note that for generalized linear models (which do not account for the correlations among observations, but were run for comparison), the standard errors of the trends in abundance were between fifteen and fifty percent smaller than their GLMM-vt counterparts, implying a false precision in the year estimate. Apart from year, many other covariates affected shark catch rates, confirming the importance of modelling these data to account for the effects of these otherwise confounding variables. In particular, the area, season, and ocean depth fished were highly significant for all of the modelled sharks. Covariates included in each of the final models and their significance levels are shown in Table 3.7. In Appendix 1 these covariates are discussed and the final parameter estimates from the GLMM-vt models are provided. Examination of residuals indicated that the models tended to overfit some of the zero catches and underfit the rare, highest catches, but otherwise fit much of the data relatively well. The log-linear form of 'year' when fitted as a continuous variable, however, does not appear to match the pattern over time in the standardized catch rates for individual years, for blue, mako, thresher, and tiger sharks (see Figure 3.6), complicating interpretation of the estimated trends in abundance for these species.

Oceanic Shark Trends from Observer Data

Blue shark abundance is estimated to have decreased by 53% (95% Confidence Interval (CI): 38 - 64%) between 1992 and 2005, based on an instantaneous decline rate of-0.057 (Figure 3.6A). From the individual year estimates it also appears that there is a

63 decline, but the pattern is not well matched by the estimated trend in that some years are being overfit and others underfit (Figure 3.6A). Examining the trend across areas, a significantly greater decline (-0.135) in Area 7, where blue shark is most frequently caught, was tempered by more moderate declines in other areas (Figure 3.7A). Estimated trends from the observer data for 1992 to 2005 were similar among the three model types (-0.063 (GLMMt) and -0.052 (GEE)). Although models of the observer data up to the year 2000 did not detect a trend in blue shark relative abundance, the rate of decline for 1992-2005 was only slightly less than that estimated in the analysis of logbook data for 1986 to 2000 (-0.066; Figure 3.8A).

8 A p<0.0001 0.6 D p=0.00£ 0.5 «0 a. J* ° O 0.4 o , JZ A x 1 Ill 0.3 4 h n o < s I *"*-*,. o l( ir 0.2 i r-1 -•jHllllh l 2 fSvL ii H ' 0.1 n ii ?s ji H £' o<5. ltft x: 0 0.0 I 0-8 E p<0,0001 5 H p^O.OGOS o X5 4' m 0.6 .bi 3 ig. „ M 0,4 ll c J§

0.0 W+W^i 0 1W2 19M 2000 2004 1992 1986 2000 2004 1982 1996 2000 2004 1892 1996 2000 2004 Year

Figure 3.6. Estimated change in relative abundance (standardized catch per 1000 hooks) between 1992 and 2005 based on the observer data for oceanic shark species: (A) blue, (B) mako, (C) thresher, (D) oceanic whitetip, and large coastal shark species: (E) hammerhead, (F) tiger, (G) coastal shark group 1, (H) coastal shark group 2. Plotted for each species are the overall trend (solid line) and the individual year estimates (• ± 95% CI) as estimated from generalized linear mixed models with vessel as a random effect and trip as a residual effect (GLMM-vt).

Mako sharks are estimated to have declined slightly in abundance, although the trend (instantaneous rate = -0.032), equating to a 34% decline (95%CI: 1 - 56%) between 1992

64 and 2005, was only marginally significant and imprecisely estimated (Figure 3.6B). The estimated decline for shortfm mako, which accounted for 79% of all recorded mako sharks, was slightly greater (instantaneous rate = -0.040, 95% CI: -0.005 - -0.074, p = 0.026). As with blue shark, the estimated rate of decline in mako sharks was significant and largest (-0.115) in Area 7 where they are most frequently caught. For makos, however, trends in the two other common areas (5 & 6) were nonsignificant (Figure 3.7B). Trends estimated for 1992 to 2005 from GLMM-t and GEE models were of a similar magnitude as the GLMM-vt model, although nonsignificant (Figure 3.8B). Mako shark decline rates estimated from the observer data for 1992 to 2005 and the logbook data for 1986 to 2000 (-0.037) were very similar, while that estimated from the 1992 - 2000 observer data indicated a much greater decline (-0.079; Figure 3.8B).

The only nonsignificant trend estimated from the 1992-2005 data was for thresher sharks (Figure 3.6C), but the small rate of decline estimated for this species group (-0.024) masks differences in its trends among areas and over time. The problem arises because the change in relative abundance of thresher sharks does not appear to be monotonic over this time period. According to both the observer (1992-2000) and logbook (1986-2000) data, estimates from which were virtually identical (-0.118 vs. -0.120), thresher sharks declined sharply up to the year 2000 (Figure 3.8C). This decline, however, is not captured well by the models of 1992-2005 observer data, which seem to underfit the earliest years, in order to better fit the data from the past few years, in which the abundance of thresher sharks appears to have stabilized or increased slightly (Figure 3.6C). Trend estimates for 1992-2005 also varied significantly among areas: a large rate of decline (-0.068) in the Mid-Atlantic Bight (Area 5) where thresher sharks are most commonly caught contrasts with the increasing trend estimated in offshore Area 8 (Figure 3.7C), where few threshers are caught (Table 3.3). Finally, the estimated decline for bigeye thresher, which comprised 68% of all recorded thresher sharks, was slightly greater (instantaneous rate = -0.030, 95%CI: -0.068 - 0.007) than that of the thresher group, but still nonsignificant (p=0.113).

65 The estimated rate of change in the oceanic whitetip shark (-0.053) was similar to that of blue shark (although less precisely estimated), amounting to an estimated 50% decline (95%CI: 17 - 70%) between 1992 and 2005 (Figure 3.6D). Significant and large declines were estimated in the Gulf of Mexico (Area 2) and South Atlantic Bight (Area 4), although the differences in trends among areas were nonsignificant (Figure 3.7D). Trends estimated by the different models for 1992-2005 were similar to one another (- 0.062 GLMM-t,-0.051 GEE; Figure 3.8D). A model ofthe observer data between 1992 and 2000 failed to detect a significant trend in abundance for this species, which may be attributed to the low sample size (n=358), while the trend estimated from the logbook data for the same years based on over 8,500 recorded individuals was significant and large (-0.145; Figure 3.8D).

p-0.0011 p=0.0C0l N9 0.2 A B Na

CO 0<1

o m .A. 0.0 • • ...ft. m m * * 1 * * -0.1 \ *M* * « § -0.2 O ® JE NS NS G W.0007^ H ipaO.001 jg 0.4 o.eJf

66 Coastal Shark Trends from Observer Data

Hammerhead shark abundance is estimated to have declined sharply (instantaneous rate = -0.109), by 76% (95% CI: 56-87%) between just 1992 and 2005 (Figure 3.6E). The estimated rate of decline for scalloped hammerhead, the most frequently recorded hammerhead shark, was less (-0.051, 95%CI: -0.090 to -0.013, p=0.009) than that of the hammerhead group, but this rate may be misleading because of the large proportion of 'unidentified hammerheads' (Table 3.3). Differences in the estimated rate of change among areas for the hammerhead group, although nonsignificant, include a nonsignificant trend in the South Atlantic Bight (Area 4) and slightly larger, significant rates of decline in other areas (Area 3) (Figure 3.7E). Trend estimates from the GLMM-t model (-0.097) were quite similar to the GLMM-vt while the GEE model showed a greater decline rate (- 0.144) (Figure 3.8E). The observer data model for 1992-2000 indicates a much larger decline rate (-0.223, 95%CI: -0.295 to -0.150, pO.OOOl), which is driven by the sharp decline in hammerhead catch rates in the early 1990s. This decline rate equates to an 83% decline (95%CI: 70-91%) between 1992 and 2000, and exceeds that estimated in the logbook data analysis for 1986 to 2000 (Figure 3.8E).

In contrast, tiger shark abundance appears to have increased: the estimated rate of change (0.037), which was marginally significant, equates to a 62% increase (95%CI: 39 to 252%) between 1992 and 2005 (Figure 3.6F). This figure should, however, be interpreted with caution since the increase is driven by data from the most recent few years, particularly by catches in Area 8, and the model may be underfitting the 1992 data. Individual year estimates suggest instead that tiger shark abundance either declined slightly or was stable from 1992 to the late-1990s, and then began increasing (Figure 3.6F). Indeed, the observer data model for 1992-2000 failed to detect a trend (Figure 3.8F); models of observer data do not show a significant increasing trend until the addition of the 2004 data. A significant decline in tiger shark abundance between 1986 and 2000 was estimated in the logbook analysis (Figure 3.8F): individual year estimates from these data suggest that tiger shark abundance declined from 1986 to 1992 and thereafter, as in the observer data, were stable (Baum et al. 2003).

67 A : B • ' C : 0 CD C 0,0 . _ ii_: (Q : ik x: £ J : o ;f i t ; 1 ' CO 0.1 E : G : H : •«*>^ tz CO 0,0 .'-lit .a w ,£ t r. XJ -0,1 * i fj + + • If H 4>' CO 0 : £ -0.2 •*-< £i : <« \ j UJ -0.3 UJ *5 "5 "S. "J 2 15 1> 2 "S 15 -*~* w S "5 "5 UJ J '2 41 -J 2E -*2J 2 O © s e> 2 s? • us "3* , j _J 0 ~J 1 C3 0 o O 0 o e> o o Model type

Figure 3.8. Comparison of estimated instantaneous rates of change (± 95% CI) amongst data sets and model types for oceanic shark species: (A) blue, (B) mako, (C) thresher, (D) oceanic whitetip, and coastal shark species: (E) hammerhead, (F) tiger, (G) coastal shark group 1, (H) coastal shark group 2. Estimates to the left of the dashed line are from Baum et a/.'s (2003) logbook analysis (o) of 1986 - 2000 data for (A), (B), (C), (E), (F) and 1992-2000 data for (D), (G), (H) (differences due to start date in recording different species), and from GLMM-vt models of the 1992-2000 observer data (•). Estimates to the right of the dashed line are of the three model types (GLMM-vt (•), GLMM-t (A), and GEE (•)) fitted to the 1992-2005 observer data.

Like hammerhead sharks, the large coastal shark group (1, dusky, silky, night shark) is estimated to have declined by 76% (95%CI: 63-85%) between 1992 and 2005 (instantaneous rate = -0.110; Figure 3.6G). Although declines in this group were significant in the southerly areas (1-3) and non-significant in the northerly areas (4-6) (Figure 3.7G), the statistically significant difference in the trends between Areas 1 and 6 is actually likely just an artifact of the small sample number of catches of these species in both these areas (Table 3.3). Trend estimates for the 1992-2005 data were similar among model types, and although the 1992-2000 observer data model suggested a smaller non significant decline, the 1992-2005 trends were very similar to that estimated in the 1992-

68 Table 3.7. Final models (generalized estimating equations with trip marginal (GEE), generalized linear mixed models with trip random (GLMM-t) and with vessel random, trip residual (GLMM-vt)) for each species showing covariates included in the final model (shaded) and statistical significance level of each: * = <0.1; **=<0.01; ***=<0.001; ****=<0.0001, otherwise non-significant. Statistical significance was based on the score statistic in the GEE and the Wald test in the GLMM.

Blue shark Mako sharks Thresher sharks Oceanic whitetip Hammerhead Tiger shark Coastal group 1 Coastal group 2 sharks GLMM- t GLMM- t GLMM-v t GE E GE E GLMM- t GLMM-v t GE E GLMM-v t GE E GE E GLMM- t GLMM-v t GLMM- t GLMM-v t GE E GLMM- t GLMM-v t GE E GLMM- t GLMM-v t GE E GLMM- t GLMM-v t year W*T* **** **** x * * -ft* ** *** **** **x* 4 * * **** **** 444* *** **** *x* ocean depth * **** **** **** **** **** **** ** *** **** *** **x* XXV* *, **** **** - X*** **** **** *4** ocean depth2 * *+ ** ** **** **** * X* *** **** * * * *** * **** **** ft*** **** **** Untie* sine *x * *»** * ** ** * **** **** *** * 4*4 **** ***» **** sine2 **** * **** * ** * X * cos * ** ** #* * * cos2 * X* * X* ** * soak time *** **** **** * x**x ** * * * *** ** *** X** temperature **** *«** **** * * *** * X* ** ** * **** *** ** ** * *** **** * no. light sticks * * * 4* * hook depth +* ** *** * * 4* X * area X*X* **** **** **** **** **** *** X*XX **** **** **** **** * x*x* XX** ***4 44*4 **** **** **** 4*4* **** *K** ***x target species **** **** **** * *** *** ** 4* xft* **** * **** **** **** X** **X* *** period * x **** ** *** * * ft ** ** * ** X hook type * * * * * * **x* **** **** ** ** ** bait type * * * * T * * area*sine X*XX **** **** * * *** *x*x ft*** * * * * ***x **** X* *#* **** * area*sine2 ***X **** X*** * * area*cos *** * * * * * xx*# * ** *ft *x xx* X*** * area*cos2 ** x X* ** ft* X *** * * * X *x* X

<3\ 2000 logbook data analysis (Figure 3.8G). Models for coastal group 2 showed a less steep trend (-0.062), equating to a 55% (95%CI: 32-71%) decrease (Figure 3.6H), but results showed the same pattern among areas (Figure 3.7H) and models (Figure 3.8H) as coastal group 1.

DISCUSSION

Shark Catch Rates in U.S. Pelagic Longline Observer and Logbook Data

As in many of the world's pelagic longline fisheries, blue shark dominated the shark catches of the U.S. Atlantic fleet, accounting for 61% of all those recorded by observers (Table 3.1). Catches of this oceanic species were concentrated in the northernmost areas (5-7) fished by the fleet, most so in the area offshore from the Grand Banks where they averaged about 40 per 1000 hooks, but catch rates were still high in other areas (except the Gulf of Mexico) relative to most other shark species caught (Figures 3.2, 3.3; Table 3.3). Silky and shortfin mako were the next two most commonly caught sharks (notwithstanding some uncertainty about the level of misidentifications). Like the blue shark, catches of the shortfin mako were concentrated primarily in the north (Figure 3.2), while the silky shark, a warmer water oceanic species, was primarily caught in the Gulf of Mexico and along the southeastern states (Figure 3.3). In these respective regions, catches of these species occurred on roughly half of the sets and averaged around two to four per 1000 hooks. The remaining shark species (threshers, oceanic whitetip, hammerheads, tiger, and other coastal species) were less commonly caught, but catch rates varied substantially among areas and, especially for the coastal species, with distance from shore. In the best areas catch rates of these species often reached 1-3 per 1000 hooks, while their overall mean catch rates were usually less than 0.5 per 1000 hooks.

Comparisons of shark catch rates between the observer and logbook data show them to be broadly similar, suggesting that the logbooks have provided a reasonable record of shark catches in the fishery. For each species, the spatial distributions and concentrations

70 of high and low catch rates in all data (Figures 3.2, 3.3) were comparable between the two data sources. Median catch rates in the non-zero data also were similar for each species between the datasets both within individual areas and overall (Figure 3.4); where discrepancies exist the tendency was for slightly higher shark catch rates to be recorded in the logbook than the observer data.

There are very few data from years prior to the beginning of the pelagic longline logbook and observer programs with which these recent catch rates can be directly compared. Apart from the exploratory research surveys analyzed in Chapter 2, the only comparable information that I am aware of is the paper by Berkeley & Campos (1988). In this study, swordfish-targeted sets (n=l 11) were monitored on a single vessel in the pelagic longline fishery off the east coast of Florida between 1981 and 1983, and yielded an overall mean catch rate of 4.16 sharks per 100 hooks. Unfortunately, the original data from this study have been lost (S. Berkeley pers. comm., March 2007), precluding formal statistical comparison of these and the more recent data from the fleet. As a rough comparison, swordfish-targeted sets in the 1992-2005 observer data from the same area (part of Area 3) had an overall mean of 1.03 per 100 hooks, about one-quarter of that in the early-1980s. Although these figures are not directly comparable without standardizing for the catch rates for sampling differences between the two periods, they are certainly suggestive of a significant decline in the relative abundance of sharks in this region, and consistent with the downward trends estimated from models of the logbook and observer data. Notably, Berkeley & Campos (1988) reported shark catch rates per 100 hooks, whereas they are now reported per 1000 hooks.

Estimates of Recent Changes in Shark Abundance

Estimated rates of change from the 1992-2005 observer data were very similar amongst the three model types for each of the modelled species, suggesting that the simpler GEE models, although not common in the fisheries literature, may be as appropriate for these types of data as the more complex mixed models. Formal comparison of these models was not attempted in this chapter (e.g. final models of the

71 different model types included different covariates because a separate backwards selection was done for each model), but this may be a useful area of investigation.

Of the modelled species, trend estimates from the 1992 - 2005 observer data models suggest the greatest declines have occurred in hammerhead sharks and the first coastal group (dusky, silky, and night sharks). Both groups are estimated to have decreased in abundance by 76% during this fourteen-year period, and the trends match their respective individual year estimates well (Figure 3.6E,G). Extending the estimated trend for hammerheads back even just to 1986 (the year from which the logbook data also show precipitous declines for this species group) suggests an 88% decline in relative abundance of hammerhead sharks in only twenty years. The steep decline estimated for coastal group 1 is nearly the same as that estimated in the logbook data (1992-2000), but must be balanced against the smaller, although still rapid decline estimated in the observer data for coastal group 2 (55% in fourteen years). Although it is frustrating that trends cannot be estimated for individual Carcharhinus species, assessments that do so (e.g. Cortes et al. 2006), ignoring the propensity for fishermen and observers alike to misidentify these species, may produce biased results. There is at present a trade-off for this group of species between obtaining trend estimates for the genus across geographic area from fishery-dependent data or obtaining species-specific trends from research surveys conducted in single locations.

Both blue and oceanic whitetip sharks are estimated to have declined in abundance by about half between 1992 and 2005, according to trend estimates from the observer data. For blue shark, the trend was similar to that estimated between 1986 and 2000 in the logbook analysis, and together models of these two datasets suggest a 68% decline in this species from 1986 - 2005. It should be noted, however, that individual year estimates from observer data models for blue shark show high interannual variability between 1992 and 2000 with no trend (Figure 3.8A), and then lower catch rates in the last five years of the time series (Figure 3.6 A), suggesting that models of this species require further refinement. Given the significant differences among areas in the estimated trends from blue sharks, one approach will be to further investigate trends in Area 7, where

72 catches of this species are concentrated. For oceanic whitetip, the observer data model appears more reasonable. Its estimated decline, while still considerable, is significantly less than that of the logbook analysis suggesting that the latter was an overestimate.

Interpretation of the trend estimates from the 1992-2005 observer data is less clear for mako, thresher, and tiger sharks, which appear to have stabilized or increased in abundance in the past few years following declines (Figure 3.6B,C,F). Because the trend in relative abundance seems to change substantially over this time period for-these species, as judged by their individual year estimates and the difference in estimated trends between the 1992-2000 and 1992-2005 observer data, the models I fitted (which assume a constant rate of change over time) do not capture the patterns well. The estimated trend for mako sharks may be a slight overestimate of its decline, but the fit seems least problematic for this species group. Implications of these poor model fits appear more serious for thresher and tiger sharks. Whereas models of the 1992-2005 observer data did not detect a trend for thresher sharks, models of both the 1992-2000 observer and 1986-2000 logbook data estimated a large (and equivalent) rate of decline that equates to an 80% decrease from 1986 to 2000. Individual year estimates for thresher sharks suggest that this group may now have stabilized. For tiger shark, the rate of increase estimated from the 1992-2005 observer data appears to be misleadingly high when compared to its individual year estimates, which instead suggest that the population may be at the same level now as it was in 1992. Even from the individual year estimates, however, it seems that the tiger shark population has fared the best over this fourteen- year period, which is reasonable given that it is one of the most productive shark species, and among the sharks has the highest survival rate from being caught on the longlines (Beerkircher et al. 2002).

Improved modelling of the changes in relative abundance for these three shark species (groups) will require alternative functional forms for the 'year' covariate or different model types. One possibility is to apply piecewise regression, in which separate trends are fitted to the data, one before and one after a breakpoint mid-way through the time series. Management changes such as time-area closures have been implemented

73 since 2000 by NMFS to reduce the pelagic longline fishery's impact on bycatch and the juveniles of target species, providing some rationale for this approach (NMFS 2006). I found, however, that at present there are too few years in the data after 2000 for these models to converge. A second possibility is to include a quadratic term for 'year' in the models. Initial models show that this term is highly significant for mako, thresher, and tiger sharks, and suggest that there has been a significant decline in thresher sharks between 1992 and 2005. The trade-off with this model formulation is that interpretation of the change in relative abundance over time is less straightforward than with the trend estimates of the current models. A similar problem arises with the non-parametric approach of modelling the data using a generalized additive model, although these models are commonly applied to fisheries-dependent data (Maunder & Punt 2004, Venables & Dichmont 2004).

In addition to improving estimates of the change in relative abundance over time, there are several other modelling options that could be explored for these data (and other similar fishery-dependent data) in the future. Two aspects of the observer data, apart from the large number of variables to be included in models, present a challenge for modelling: the correlated structure of the data and the high proportion of zeroes in the shark catch data. In these analyses I focused on the first problem and modelled two levels of correlations (among sets on the same trip and among trips on the same vessel). In addition to these sources of variation, however, the high proportion of unidentified sharks in the observer data and known reporting problem with night sharks suggests that observers' reporting tendencies may also cause correlations within sets reported by the same observer. It may therefore be worthwhile to investigate models with 'observer' as a random effect. Secondly, with the obvious exception of blue shark in Area 7, the catch data contain many zeroes. In the modelled data (i.e. even after removing areas with fewer than 50 catches) for thresher, oceanic whitetip, hammerhead, and tiger sharks, for example, fewer than 15% of sets caught these species. To better address the high proportion of zeroes in the data for these species, it may be of use to implement models using the zero-inflated negative binomial (ZINB) distribution, which is a mixture of two distributions, one for the zeros and one that includes zeros and positive values (i.e. the

74 negative binomial distribution). The ZINB has been applied recently to models of silky shark bycatch in an eastern Pacific Ocean purse-seine fishery (Minami et al. 2007). Models that both account for correlations within the data (e.g. mixed models) and use the ZINB distribution are (to my knowledge) not yet available in statistical software packages, and thus it was not possible to address these two major statistical issues in the observer data together. Such models are currently being developed for the R program by Joanna Flemming (Department of Statistics, Dalhousie University), Eva Cantoni (Econometrics Department, University of Geneva), and Alan Welsh (Department of Statistics, Australian National University) and may prove useful in future studies.

Conclusions

For wide-ranging species like large pelagic sharks, fishery-dependent data are normally the only source of time series data that sample intensively across a broad spatial scale similar to the range of the populations. In this chapter, I used data from the U.S. Atlantic pelagic longline fishery's monitoring program to describe the spatial distribution and concentrations of large pelagic shark species in the Northwest Atlantic Ocean, to estimate recent (1992-2005) trends in their relative abundance, and to compare these data and estimates to those from the same fleet's logbook data. For each species group examined, the summarized catch rates from these two datasets were similar overall and within each area, suggesting that the logbook data has provided a reasonable record of shark catches in the fishery, and can augment the more restricted observer dataset. Although observer data are considered to be an improvement over logbook data, in this case, the high proportion of sharks identified only at the genus level, misidentified, or unidentified has unfortunately precluded species-specific analyses for most sharks. The high variability in fishing areas, season, and gear in this fleet also leads to a large number of factors that can affect catch rates, underscoring the need for observers to provide complete and accurate records of the fishing operation, whereas at present some variables in the observer data set have many missing values. Thus, in addition to increasing the percentage of the fleet monitored by observers, both improved species identification and data recording will improve the utility of the observer data for monitoring sharks and other large pelagic fishes.

75 Notwithstanding the problematic trend estimates for some of the shark species, the observer data suggests that there were substantial changes in the relative abundance of many of the modelled shark species between 1992 and 2005. From a conservation standpoint, the most important result of the analysis is the precipitous decline estimated' for hammerhead sharks (comprised primarily of scalloped hammerheads). The estimated rate of decline (-0.109) was less than that estimated in the logbook analysis (-0.158) for 1986-2000, but very similar to that estimated in the analysis of the 1972-2003 shark- targeted UNC research survey data for scalloped hammerheads (-0.127 (Chapter 4)). Together, these three datasets provide a consistent indication of marked declines in hammerheads, implying that improved management measures are required for these sharks. Large declines also were estimated for the coastal shark group, but the species- specific data available from other sources for many of the species in this group will be more informative. Of the remaining species, blue, mako, and oceanic whitetip sharks appear to have declined moderately between 1992 and 2005, while improved models are needed for thresher and tiger sharks. Individual year estimates for these species suggest that tiger shark may have increased in recent years, while thresher sharks may be stabilizing following a significant decline. Indications from individual year estimates that the relative abundance of many of these shark populations may be stabilizing is a positive sign, and might be attributed to recent management changes in the fishery (NMFS 2006). Setting these recent changes in the context of the shark population declines that are estimated to have occurred in the decades prior to the implementation of the longline observer program (Musick et al. 1993, Baum & Myers 2004, Ha 2006, Myers et al. 2007) implies that if these populations are now stabilizing, it is at greatly reduced levels of abundance.

76 CHAPTER 4

Cascading Effects of the Loss of Apex Predatory Sharks from a Coastal Ocean

Published as Myers, R.A., Baum, J.K., Shepherd, T.D., Powers, S.P. & Peterson, C.H. (2007). Cascading effects of the loss of apex predatory sharks from a coastal ocean. Science, 315, 1846-1850.

77 Abstract

Impacts of chronic overfishing are evident in population depletions worldwide, yet indirect ecosystem effects induced by predator removal from oceanic food webs remain unpredictable. As abundances of all 11 great sharks that consume other elasmobranchs (rays, skates, and small sharks) plummeted over the past 35 years, 12 of 14 of these prey species increased in coastal northwest Atlantic ecosystems. Effects of this community restructuring have cascaded downward from one mesopredator, the cownose ray: its order of magnitude increase enhanced predation on its bay scallop prey sufficiently to terminate a century-long scallop fishery. Analogous top-down effects may be a predictable consequence of eliminating entire functional groups of predators.

MAIN TEXT

The potential impacts on ecosystem structure and function of eliminating apex predators, including release of mesopredator prey from predatory control (Crooks & Soule 1999) and induction of subsequent trophic cascades (Paine 1980, Estes et al. 1998, Pace et al. 1999), can be far-reaching because of the strong influence predators may exert on their prey populations (Duffy 2002). In the oceans, the most pervasive human exploitative activity, fishing, has disproportionately reduced abundances of high trophic level predators (Pauly et al. 1998, Jackson et al. 2001), eliciting concern about not only their conservation but also indirect ecosystem effects that might ensue from their removal. Yet while top-down effects are relatively well understood in rocky intertidal and kelp forest ecosystems (Paine 1980, Estes et al. 1998, Jackson et al. 2001), their importance and prevalence in other marine ecosystems remain unclear. Evidence of oceanic trophic cascades is limited (Estes et al. 1998, Daskalov 2002, Frank et al. 2005), and some have argued that in complex marine food webs, with many interacting species and opportunities for functional redundancy, top-down effects may attenuate (Strong 1992, Jennings & Kaiser 1998).

78 Fundamental constraints on studying apex predators, sampling vast ocean ecosystems, and conducting controlled experiments in the sea limit our capacity to predict indirect effects of oceanic predator removal or (in many cases) even to detect those that have occurred. This gap in understanding presents a critical challenge to mitigating human impacts and restoring marine ecosystems. To meet this challenge, we employed the two most powerful empirical methods in ecology, meta-analysis of multiple independent data sets and controlled experimental hypothesis testing replicated over space and time, using a unique compilation of research surveys, fisheries and landings data, and predator exclusion experiments, to investigate the consequences of functionally eliminating apex predatory sharks from the coastal ocean.

Large (>2m) sharks are top predators of coastal and oceanic ecosystems, many of which are now of conservation concern. Exploitation of the great sharks has intensified worldwide in recent decades, driven by an upsurge in demand for shark fins and meat (Fowler et al. 2005) and bycatch in many fisheries. Data to assess direct impacts of exploitation on these species are limited, but where available, consistently indicate that they have been driven to low levels (e.g. Stevens et al. 2000, Baum et al. 2003, Fowler et al. 2005). Whether functional elimination of great sharks also induces indirect ecosystem effects, however, is an open question (Stevens et al. 2000).

We hypothesized that weakened top-down control by elasmobranch-consuming sharks could lead to increasing abundances of their elasmobranch prey (rays, skates, and small sharks), and that the enhanced predation by these mesopredators might cascade to lower trophic levels. Because mesopredatory elasmobranchs are of a substantial size, even as juveniles, and are thus consumed almost exclusively by great sharks (see Supplementary Material, Table 4.5), we inferred that these prey would be the most likely affected. Moreover, interannual variability in elasmobranch populations is minimal because of their low reproductive rates, such that changes effected by predator removal should be detectable in time series data. We tested these hypotheses on the eastern seaboard of the United States, between Cape Cod, Massachusetts (41.5°N) to Cape

79 Canaveral, Florida (28°N), focusing on 1970 - 2005 because of data availability and intensified exploitation of large sharks during this period.

79* JL m m" 59" m," T^ZZZ 7*7 •«" {tsa

4V ; I ' rC\¥'>/W>ffl-$t^ NMFS-Off$hor» ESi •41"

MD}JRf 'QNREC Wl

"NC

W |-31*

79* 7T «6>' «* »•

Figure 4.1. Map of the U.S. Atlantic coast showing the location of each research survey, with 200m, 500m, and 1000m isobaths (dotted lines) given for reference.

We first assembled all available species-specific elasmobranch data from scientific research surveys conducted along the U.S. east coast that began before 1990 and employed a standardized methodology that was consistent over time. Seventeen surveys, which together cover the eastern U.S. coast (Figure 4.1), met these criteria (Tables 4.1, 4.2). Survey start years ranged from 1959 to 1989 with a median start year of 1976 and mean survey time span of 28 years (Table 4.2). Two of the surveys used longlines and were carried out specifically to sample sharks, the UNC survey (detailed in Supplementary Material) and the SC survey (detailed in Low & Ulrich 1984, Ulrich 1996). Fifteen other surveys used either bottom trawls or seines, and were designed to

80 Table 4.1. Data sources.

Data type Data Acronym Source Reference or web access

Connecticut Department of Environmental Protection, Fisheries Division http://dep.state.ct.us/burnatr/fishinq/fdhome.htm DNREC Delaware Department of Natural Resources and Environmental Control, http://www.fw.delaware.qov/. Division of Fish and Wildlife GSO University of Rhode Island, Graduate School of Oceanography http://www.qso.uri.edu/ MDNR Maryland Department of Natural Resources, Fisheries Service http://www.dnr.state.md.us/fisheries/ NCDMF North Carolina Department of Environment and Natural Resources, Division of http://www.ncfisheries.net/ Marine Fisheries NMFS-Off& National Oceanic & Atmospheric Administration (NOAA), National Marine http://www.nefsc.noaa.gov/ NMFS-ln Fisheries Service (NMFS), Northeast Fishery Science Center SC South Carolina Department of Natural Resources Low & Ulrich 1984; Ulrich 1996 SEAMAP Southeast Area Monitoring and Assessment Program, South Atlantic http://www.dnr.sc.qov/marine/mrri/SEAMAP/seamap.html UNC University of North Carolina - Institute of Marine Sciences, Longline shark http://www.marine.unc.edu/Research monitoring survey VIMS Virginia Institute of Marine Science http://www.fisheries.vims.edu/trawlseine/sbmain.htm Fisheries Logbook NOAA, NMFS, Southeast Fishery Science Center http://www.sefsc.noaa.gov/fls.isp Observer NOAA, NMFS, Southeast Fishery Science Center http://www.sefsc.noaa.gov/pop.isp Landings Landings NOAA, NMFS, Office of Science & Technology http://www.st.nmfs.gov/st1/commercial

Landings UN Food and Agriculture Organization, Fisheries Department, Fishery http://www.fao.org/fi/statist/statist.asp Information, Data and Statistics Unit Table 4.2. Survey, fisheries and landings data set descriptions, including area, gear type, season and years sampled, and total sample size. Species sampled in each data set: great shark species (L), elasmobranch mesopredator species (M), bivalve species (S).

Data type Acronym Area Gear Season Years Samples Species Survey CTDEP Long Island Sound Trawl Fall/Spring 1984-2004 - M DNREC Delaware Bay Trawl Year round 1966-2004 1874 L,M GSO Narragansett Bay, Rhode Island Trawl Year round 1959-2002 - M MDNR Chesapeake Bay Seine Summer 1960-2005 8022 M NCDMF Pamlico Sound, North Carolina Trawl Summer/Fall 1987-2004 1889 M NMFS-Off Northeast U.S. Offshore Trawl Spring 1968-2005 10185 L,M Fall 1963-2005 8829 L,M Summer 1963-1995 1758 L,M NMFS-ln Northeast U.S. Inshore Trawl Spring 1976-2005 2084 L,M Fall 1974-2005 2228 L,M Summer 1977-1981 351 L,M SC Coastal South Carolina Bottom longline Year round 1983-84,1993-95 131 L,M SEAMAP Coastal Southeast U.S. Trawl Spring 1989-2005 1441 L,M Fall 1989-2005 1389 L,M Summer 1989-2005 1393 L,M UNC Coastal North Carolina Longline April - November 1972-2003 760 L,M VIMS Chesapeake Bay Seine Summer 1968-2003 3166 M Fisheries Logbook Northwest Atlantic Pelagic longline Year round 1986-2000 214234 L Observer Northwest Atlantic Pelagic longline Year round 1992-2005 6967 L Landings NMFS Landings Coastal Eastern U.S. Various Year round 1950-2003 - S FAO Landings Atlantic Canada Various Year round 1950-2003 - S sample a variety of fmfish and invertebrate species. In total, 12 surveys caught large sharks; all 17 caught elasmobranch mesopredator species (Table 4.2).

For each elasmobranch, we modelled temporal population trends in each of several data sets and also meta-analytically to yield synthetic estimates of rates of change (see detailed methods in Supplementary Material). Trends in relative abundance of each elasmobranch were estimated using generalized linear models (GLMs; see Supplementary Material, Table 4.6) from each survey in which it was caught in at least three years. Trends in the length of large sharks caught in the UNC research survey were analyzed using generalized linear models with a gamma error structure, log link, and month as a covariate after removing any biologically implausible length values (e.g, five lengths for dusky sharks were smaller than the minimum size of neonates). Because not all great sharks were sampled in surveys and the U.S. pelagic longline fishery covers a much greater proportion of the sharks' northwest Atlantic ranges, trends also were estimated from this fishery's observer data (Tables 4.1, 4.2) using generalized linear mixed models (as detailed in Chapter 3, Table 4.6) and its logbook data (Tables 4.1, 4.2) using GLMs (see detailed methods in Supplementary Material, Table 4.6).

The eastern seaboard's longest continuous shark-targeted survey (UNC), conducted annually since 1972 off North Carolina, demonstrates sufficiently large declines in great sharks to imply their likely functional elimination. Declines in seven species range from 87% for sandbar sharks (Carcharhinus plumbeus) and 93% for blacktip sharks (C limbatus) up to 97% for tiger sharks (Galeocerdo cuvier), 98% for scalloped hammerheads, and 99% or more for bull (C. leucas), dusky (C. obscurus), and smooth hammerhead (S. zygaena) sharks (Figure 4.2, Table 4.3). Because this survey is situated where it intercepts sharks on their seasonal migrations, these trends in abundance may be indicative of coastwide population changes. The UNC survey also showed the loss of the largest individuals, with declines in mean length of blacktip, bull, dusky, sandbar and tiger sharks ranging from 17 to 47% (Figure 4.3), suggesting that intense size-selective exploitation has left few mature individuals in these populations. The remaining four elasmobranch-consuming great sharks were caught too rarely to detect trends from this

83 Figure 4.2. Change over time in species at each trophic level as estimated from individual data sources: trends in relative abundance (overall trend (solid line) and individual yearly estimates (•)) of great sharks (top row, UNC survey) and elasmobranch mesopredators (middle row, survey acronyms as in Table 4.1) estimated from GLMs, and the trend in North Carolina bay scallops (bottom row, NMFS landings) shown with loess curve from a generalized additive model. 00 Table 4.3. Model results for each species of great shark from each of the research survey and fisheries data sources used in the meta-analysis shown in Figure 4.4, including the first and last year of capture in the data set, the number of years caught, the total number of the species caught, the model estimate (± 95% confidence intervals (CI)) of the instantaneous rate of change, for all years of data (All) and for only those years onwards from the baseline of 1970 (1970-). Statistical significance levels for model estimates are * = <0.05; **=<0.01; ***=<0.001; ****=<0.0001, otherwise non-significant.

Species Data First Last n n Years Instantaneous rate of change in source year year years caught model Estimate upper CI lower CI Blacktip UNC 1972 2003 32 905 All -0.084**** -0.065 -0.103 SEAMAP 1990 2005 12 29 All 0.040 0.133 -0.053 Bull UNC 1973 1995 10 23 All -0.181**** -0.093 -0.270 Dusky UNC 1972 2003 29 1036 All -0.149**** -0.120 -0.179 NMFS-Off 1967 1999 11 38 All -0.075* -0.005 -0.144 1972 1999 8 26 1970- -0.068 0.004 -0.140 NMFS-ln 1974 2005 17 100 All -0.092** -0.035 -0.149 SEAMAP 1990 2004 6 24 All -0.199* -0.032 -0.365 Great hammerhead UNC 1975 1997 4 5 All -0.080 0.037 -0.197 Sandbar DNREC 1966 2004 26 242 All -0.048**** -0.035 -0.062 1970 2004 22 159 1970- -0.041**** -0.023 -0.060 UNC 1976 2000 23 310 All -0.077**** -0.039 -0.114 NMFS-Off 1967 2002 22 73 All 0.014 0.057 -0.028 1973 2002 20 68 1970- 0.009 0.054 -0.036 NMFS-ln 1974 2005 27 107 All 0.019 0.048 -0.010 SC 1983 1995 5 196 All -0.281**** -0.225 -0.337 SEAMAP 1990 2005 13 71 All -0.029 0.070 -0.128 Scalloped UNC 1972 2003 29 495 All -0 197**** -0.104 -0.149 hammerhead NMFS- In 1980 1995 3 4 All -0.110 0.065 -0.285 SEAMAP 1989 2005 17 126 All 0.094** 0.155 0.033 Smooth hammerhead UNC 1973 1989 4 5 All -0.172* -0.010 -0.334 Hammerhead SC 1983 1994 3 11 All -0.110 0.089 -0.308 species Logbook 1986 2000 15 60,402 All -0.158**** -0.143 -0.172 Observer 1992 2005 14 1,292 All -0.110**** -0.062 -0:157 Large coastal Logbook 1986 2000 15 80,480 All -0.118**** -0.103 -0.133 species Observer 1992 2005 14 8,186 All -0.084**** -0.048 -0.121 Mako species Logbook 1986 2000 15 65,795 All -0.037**** -0.025 -0.050 Observer 1992 2005 14 3,433 All -0.032* -0.001 -0.063 Tiger UNC 1973 2002 18 39 All -0 117**** -0.064 -0.169 SC 1983 1995 5 142 All -0.027 0.029 -0.083 Logbook 1986 2000 15 16,030 All -0.076**** -0.061 -0.091 Observer 1992 2005 14 1,190 All 0.037* 0.071 0.002 Great white Logbook 1986 2000 15 6,087 All -0.117**** -0.074 -0.146

85 St

BiieMSp BuM 9 Dusky —*— Sandbar • Scatopsd hammariwad T|ger

-0,03 -0,02 -0.01 0,0 0.01 0.02 0.03 Instantaneous rate d change in fork tengtw

Smooth hs»M#rtw«S

1970 1880 IB* 2000 1870 1980 1S90 3000

Figure 4.3. Change in length of great sharks between 1972 and 2003 from the UNC shark-targeted longline research survey: a) instantaneous rates of change (± 95% CIs), b) overall trend (solid line) and individual year estimates (•). Species with length samples in more than three years were modelled in a) and b); only raw data are shown for great and smooth hammerheads. survey. Two of those, great white {Carcharodon carcharias) and sand tiger {Carcharias taurus) sharks, were each caught only once and early in the UNC survey (in 1974 and 1978 respectively). The only survey that has caught enough sand tiger sharks to estimate trends in abundance targets sharks with longlines in Chesapeake Bay, and suggests a decline of over 99% between 1974 and 2004 (see Supplementary Material, Ha 2006).

86 Consistent with the UNC survey, all but one of the other six significant survey trends indicate decreasing great shark abundances (Table 4.3). The only significant increase is for juvenile scalloped hammerheads, and comes from a single survey that began in 1989; thus it may reflect recently increased survival following losses of their only predators, larger apex predatory sharks. Accordingly, meta-analytic results portray a consistent pattern of declines in great sharks (Figure 4.4A).

Fisheries data from the past two decades help confirm losses of elasmobranch- consuming great sharks (Figure 4.4A). Logbook data analyses show declines between 1986 and 2000 ranging from 40% in mako sharks (predominantly shortfin mako (Isurus oxyrinchus)) to 89% in hammerhead sharks (predominantly scalloped hammerheads, but also great {Sphyrna mokarran) and smooth (S. zygaena) hammerheads) (Baum et al. 2003, Table 4.3). Trend estimates from observer data collected between 1992 and 2005 differ substantially from logbook data only for tiger sharks, which is probably because of the differing temporal coverage of the two data sets: following a decline, this species appears to have increased in the last four years. Rates of declines for the other three species groups analyzed in the observer data were concordant with the logbook analysis: mako sharks declined moderately (38%), while large coastals (genus Carcharhinus, including dusky, sandbar, blacktip, bull) and hammerheads declined by 67% and 76% respectively (Table 4.3).

Concurrent with reductions in great sharks, their mesopredatory elasmobranch prey increased along the eastern seaboard (Table 4.4). This group of 14 rays, skates, and small sharks is taxonomically diverse (seven families), and includes demersal and pelagic species from estuaries and the inshore coast to the continental shelf and slope. Individual surveys suggest that little skate (Leucoraja erinacea), Atlantic sharpnose shark (Rhizoprionodon terraenovae), chain catshark (Scyliorhinus retifer), and smooth butterfly ray {Gymnura altavela) populations may have each increased by approximately an order of magnitude (Figure 4.2). Overall, meta-analyses of research survey data reveal increases over the past 16 to 35 years for 12 of the species, with mean instantaneous rates

87 of increase ranging from 0.012 for bullnose eagle ray (Myliobatis freminvillii) to 0.228 for smooth butterfly ray (Figure 4.4B).

Blacktip Cwchattninm limtmtus Bull Carclmrtilnus feoeas Dusky Carctwhtotts obsnurus Great hammerhead Sphyma mokarmn Sandbar Carchwftinus pkmiimm Scalloped hammerhead Sphyma tewini Smooth hammerhead Spbyrm zygama Tiger G#feoe«re*o euvfer Great white Catdiemtkm camftarfas Hammerhead species Sphyrrwssp. Large coastal species Camttarftfnua sip. Mato species isums sp, T

Atlantic sharprtose Rfteprfonotfort team, B Blacknase shark Gmchattimmscronotut Bennetnead s Spttyrm tlinm BuMnose eafile ray Chain catshark Scynortimus nMf&r Ctearnose skate Cpwnose ray Flneteoth shark I Lesser devil ray Mobtm hyprntoma UMte skate .Rosette s i«uc

-0.3 -0.2 -0.1 0.0 0.1 0.3

Instantaneous rate of change in abundance

Figure 4.4. Instantaneous rates of change in relative abundance (±95% CIs) for (A) great sharks and (B) elasmobranch mesopredators, as estimated by random-effects meta-analyses of research survey (•) and fisheries (•) data.

88 Table 4.4. Model results for each elasmobranch mesopredator species from each of the research surveys used in the meta-analysis shown in Figure 4.4, including the first and last year of capture in the data set, the number of years caught, the total number of the species caught, the model estimate (+ 95% confidence intervals (CI)) of the instantaneous rate of change, for all years of data (All) and for only those years onwards from the baseline of 1970 (1970-). Statistical significance levels for model estimates are * = <0.05; **=<0.01; ***=<0.001; ****=<0.0001, otherwise non-significant.

Species Data First Last n n Years Instantaneous rate of change source year year years caught in model Estimate upper CI lower CI Atlantic UNC 1973 2003 31 2239 All 0.084**** 0.098 0.071 sharpnose NMFS-Off 1974 2003 15 39 All 0.084** 0.138 0.031 1974 2003 15 39 1970- 0.084** 0.138 0.030 NMFS-ln 1974 2005 26 331 All -0.025 0.002 -0.053 SEAMAP 1989 2005 17 13187 All 0.065**** 0.079 0.051 SC 1983 1995 5 135 All 0.103*** 0.159 0.047 Blacknose SEAMAP 1989 2005 17 156 All 0.043 0.091 -0.004 UNC 1972 2003 32 1304 All -0.090**** -0.073 -0.107 Bonnethead shark SEAMAP 1989 2005 17 4925 All 0.028** 0.045 0.010 Bullnose eagle DNREC 1966 2004 28 3701 All 0.008* 0.015 0.001 ray 1970 2004 24 3153 1970- 0.010* 0.019 0.001 NMFS-Off 1967 2005 23 297 All 0.056* 0.104 0.009 1973 2005 21 279 1970- 0.053* 0.102 0.003 NMFS-ln 1974 2005 32 2230 All -0.003 0.014 -0.020 SEAMAP 1989 2005 17 5300 All 0.041*** 0.065 0.018 Chain catshark NMFS-Off 1963 2005 43 778 All 0.052**** 0.065 0.038 1970 2005 36 715 1970- 0.070**** 0.087 0.053 Clearnose CTDEP 1984 2004 21 - ' All 0.199**** 0.281 0.118 skate DNREC 1966 2004 28 5778 All -0.049**** -0.042 -0.057 1970 2004 24 3359 1970- -0.008 0.001 -0.018 NCDMF 1988 2004 7 9 All 0.053 0.206 -0.099 NMFS-Off 1967 2005 39 1053 All 0.034**** 0.047 0.022 1970 2005 36 1029 1970- 0.029**** 0.042 0.016 NMFS-ln 1974 2005 32 2678 All 0.047**** 0.057 0.036 SEAMAP 1989 2005 . 17 6991 All 0.014 0.034 -0.007 Cownose ray DNREC 1979 2003 14 76 All 0.117**** 0.168 0.065 1979 2003 14 76 1970- 0.111**** 0.168 0.054 MDNR 1976 2003 12 26 All 0.063** 0.102 0.024 NCDMF 1987 2004 17 230 All 0.175**** 0.219 0.132 NMFS-Off 1972 1976 3 23 All -0.265 0.011 -0.541 1972 1976 3 23 1970- -0.432 0.063 -0.928 NMFS-ln 1974 2005 27 544 All 0.044* 0.081 0.006 SEAMAP 1989 2005 17 4817 All 0.059** 0.105 0.014 VIMS 1992 2003 7 11 All 0.104* 0.201 0.008 1992 2003 7 11 1970- 0.101* 0.200 0.002 Finetooth shark UNC 1977 1997 14 93 All 0.039 0.114 -0.037 SEAMAP 1990 2005 7 23 All 0.092 0.261 -0.078 Lesser devil ray SEAMAP 1990 2005 15 347 All 0.105** 0.173 0.037

Little skate CTDEP 1984 2004 21 7 All -0.008 0.010 -0.025 DNREC 1966 2004 25 2499 All 0.048**** 0.058 0.039 1970 2004 21 2378 1970- 0.082**** 0.096 0.068

89 Species Data First Last n n Years Instantaneous rate of change source year year years caught in model Estimate upper CI lower CI NMFS-Off 1963 2005 43 161330 All 0.018**** 0.022 0.015 1970 2005 36 151031 1970- 0.015**** 0.019 0.011 NMFS-ln 1974 2005 32 142760 All 0.076**** 0.084 0.067 GSO 1959 2002 44 - All 0.054**** 0.064 0.044 1970 2002 33 - 1970- 0.056**** 0.071 0.041 Rosette skate NMFS- Off 1963 2005 43 1014 All 0.039**** 0.052 0.025 1970 2005 36 939 1970- 0.037**** 0.053 0.022 Smooth DNREC 1967 1999 4 11 All -0.108* -0.025 -0.191 butterfly ray 1971 1999 2 3 1970- n.a. n.a. n.a. NCDMF 1989 2004 6 44 All 0.344**** 0.474 0.215 SEAMAP 1989 2005 17 5247 All 0.131**** 0.148 0.114 Spiny butterfly DNREC 1966 2002 22 55 All -0.029* -0.004 -0.053 ray 1970 2002 19 45 1970- -0.030 0.002 -0.062 NMFS-Off 1973 2005 19 52 All 0.041 0.085 -0.003 1973 2005 19 52 1970- 0.030 0.073 -0.014 NMFS- In 1974 2005 32 589 All 0.016 0.033 -0.001 SEAMAP 1989 2005 17 317 All 0.102**** 0.152 0.053 Spotted eagle ray SEAMAP 1990 2005 16 159 All -0.070** -0.017 -0.122

Most conspicuous among the increasing mesopredators is the cownose ray (Rhinoptera bonasus) (Fahrenthold 2004). Six of seven surveys covering the U.S. Atlantic population's range (southeast Florida - Raritan Bay, New Jersey, with recent expansion to Long Island, New York) show significant increases (Figure 4.2, Table 4.4). Together, these rates of change (meta-analytic mean = 0.087, 95%CI: 0.021-0.127) indicate an order-of-magnitude increase in cownose rays coast-wide since the mid-1970s and, when combined with earlier population estimates from aerial surveys in Chesapeake Bay (Blaylock 1993), suggest that there may now be over 40 million rays in the population. When considered with its known late maturity and low fecundity, this present high population growth rate would make the cownose ray anomalous among fishes in its combination of life-history traits (see Supplementary Material). Only if its natural mortality rate were substantially greater than at present, would the life history conform, implying that higher predation by great sharks prevailed in the past and possible reduction in bycatch is insufficient to explain the ascent of this ray.

Collectively, the hyperabundant cownose ray population consumes a large quantity of bivalves, implying a high potential for trophic cascades. Cownose rays migrate

90 southward in autumn from northerly estuaries to overwintering grounds on the Florida shelf (Grusha 2005), often entering bays and sounds en route. Their diet consists largely of bay scallops (Argopecten irradians), soft-shell clams (Mya arenaria), hard clams (Mercenaria mercenarid), oysters {Crassostrea virginica), and several smaller, non commercial bivalves (Blaylock 1993, Smith & Merriner 1985, Powers & Gaskill 2005). Annual bivalve demand within the Chesapeake Bay, based on our abundance estimate, individual daily consumption rates of -21 Og shell-free wet weight (see Supplementary Material) and 100-day occupancy each year, may approach 840,000 metric tons. In comparison, the 2003 commercial bivalve harvest in Virginia and Maryland totaled only 300 metric tons, substantially lower than historic landings (see Supplementary Material).

The second link in an apparent trophic cascade has emerged over the past two decades as the cownose ray population grew coast-wide (Figures 4.2, 4.5). Field sampling in 1983 and 1984 before and after ray presence during late-summer migration showed no impacts of ray predation on bay scallops (Figure 4.5 A; Peterson et al. 1989). Analogous recent sampling, confirmed by controlled ray-exclusion experiments using stockades, demonstrates that since 1996 migrating cownose rays have caused almost complete scallop mortality by early fall (Figure 4.5 A; Peterson et al. 2001) at every site with initial adult scallop densities above a threshold for intensive ray foraging (~2m" , Figure 4.5B,C). Bay scallop abundance declined much less inside cownose ray exclosures than on unprotected grounds (Figure 4.5A) and, in the absence of scallop emigration, numbers inside stockades would probably have remained nearly constant (Peterson et al. 2001). Unlike the fishery harvest, which occurs after, ray predation occurs before spawning of this annual species (Peterson et al. 1996). By 2004, ray predation had terminated North Carolina's century-old bay scallop fishery because too few scallops survived into fall to sustain fishing and a consequent Allee effect (apparently induced at adult densities below ~l-2m"2) limited reproductive success (Peterson et al. 1996). The fishery has remained closed through 2007 (Figure 4.2) because of low recruitment and continued ray predation on any high-density patch of scallops. Having depleted the more readily targeted epibiotic bay scallops, it is reasonable to expect future expansion of cownose ray foraging on infaunal bivalves, with

91 8384 02 03 04 Year

« Mane' 34 5 Oscar SAoo/ 3 Banks Say 100 100 100 r £ 5 50 !so so 0 0 0 4—k •»—f- -i—I—+- 83 64 0I2 03 04 83 84 92 93 94 96 98 990002 03 04 83 84 02 03 04 Year Year Year

Figure 4.5. (A) Map of the southeastern U.S. indicating the study location (inset) and North Carolina bay scallop monitoring sites. Total mortality (black bars) compares August (pre-cownose ray migration) to late September/October (post-migration) densities. Low scallop densities before ray migration are indicated by asterisks (<1- 2m-2) or Os (Onv2). Hatched bars represent mortality within experimental stockades that exclude rays (performed in a subset of years). Scallops were free to emigrate from stockades. Arrows denote direction of ray migration. (B) Mean scallop density measured in midsummer and mortality from early summer to early fall at Oscar Shoal for 10 years. (C) Scallop density trends at Oscar Shoal, based on 12 weekly surveys in 1998 and 8 in 2002 and 2003.

92 associated uprooting of seagrass and thus loss of nursery habitat (Orth 1975, Smith & Merriner 1985).

Increased predation by cownose rays also may now inhibit recovery of hard clams, soft-shell clams, and oysters (Fahrenthold 2004) compounding the effects of overexploitation, disease, habitat destruction, and pollution, which have already depressed these species (Jackson et al. 2001). Landings data for these bivalves and bay scallops from within the cownose ray's range show them falling without substantial recovery (Figure 4.6) as the rays increased despite active shellfish enhancement and habitat restoration. In contrast, areas beyond the ray's northernmost limit show some examples of stable or increasing bivalve landings (Figure 4.6).

Analogous elasmobranch community inversions and trophic cascades are probably occurring in other coastal oceans. Studies in the northeast Atlantic Ocean have shown increasing abundances of several mesopredatory elasmobranchs despite substantial exploitation (Dulvy et al. 2000, Rogers & Ellis 2000). Our ability to follow the cascading effects of increases in cownose rays to the decline of one of its prey, the bay scallop, is related to the fishery value of this benthic invertebrate. Similarly, in Japan's Ariake Sound in the northwest Pacific Ocean, where exploitation on apex predatory sharks is probably intense, wild stocks and cultured populations of multiple species of benthic bivalve shellfish (mostly manila clams and penshells, but also oysters and razor clams) are now decimated annually by expanding numbers of another elasmobranch mesopredator, the longheaded eagle ray (Aetobatus flagellum) (Yamauchi et al. 2005). Many other prey depletions may be going unrecognized because little monitoring and research exists for non-commercial marine species.

Our study provides evidence for an oceanic ecosystem transformation that is most parsimoniously explained by the functional elimination of apex predators, the great sharks, instead of assuming numerous coincidental increases in their mesopredatory prey. In addition to directly threatening the long-term persistence of these species, their

93 overexploitation appears to have led to a region-wide proliferation of mesopredatory elasmobranchs. Consequences of this community restructuring have cascaded down the food web through cownose rays to bay scallops, and possibly other bivalves. This cascade potentially could extend to seagrass habitat, exacerbating stresses on already highly degraded coastal benthic systems. Thus, like the classic killer whale - sea otter - urchin - kelp trophic cascade (Estes et al. 1998), eliminating great sharks carries risks of broader ecosystem degradation. Prevailing theory suggests that trophic cascades arise only in simple food webs lacking functional redundancy (Strong 1992, Pace et al. 1999), but we propose that top-down effects must be widely expected when entire functional groups of predators are depressed, as can occur with industrial fisheries. Illuminating the operation of indirect species interactions within marine and other environments brightens the future for development of what is now so widely sought, ecosystem-based management to achieve sustainability of natural living resources.

SUPPLEMENTARY MATERIAL

Materials and Methods Species

Great sharks

Large shark species in the northwest Atlantic were considered for inclusion in this category based on their size and the occurrence of elasmobranchs in their diet. Eleven species met these criteria (Table 4.5). These sharks are among the largest (notable exclusions being basking and whale sharks, which feed at much lower trophic levels), reaching maximum lengths ranging from ~2.0m in blacktip and sandbar sharks up to 5- 6m in great hammerhead and great white sharks (Compagno 1984, 2001, Cortes 2002). Bull, blacktip, sandbar, and scalloped hammerhead reach sexual maturity below or close to 2m, but all others mature at a greater length (Compagno 1984, 2001, Cortes 2002). These large fishes are all tertiary consumers (trophic level >4) with catholic diets. Five species (bull, great hammerhead, tiger, sand tiger, and great white sharks) are true apex predators, while the remaining six species feed at and near the top of the food web.

94 Smaller elasmobranchs form a key component of the diet of many large shark species (Clark & von Schmidt 1964, Compagno 1984, 2001, Cortes 1999), and conversely, sharks are the most common predators of other elasmobranchs (Springer 1967, Heithaus 2004). Among the large sharks, however, there is considerable variation in the proportion of elasmobranchs in their diet. Bull, great hammerhead, sand tiger, and great white sharks are each considered to be important predators of other elasmobranchs, with about 30-40% of their diet comprised of these fishes (Cortes 1999). For the other species, the proportion of elasmobranchs in their diet has ranged in different studies between approximately 1 and 15% (see Table 4.5 references; Cortes 1999). We compiled data on elasmobranch consumption for each of the large shark species, with particular consideration of the species included in the elasmobranch mesopredator category (see below). At the species level there is evidence that large sharks are predators of seven of the elasmobranch mesopredator species, little and clearnose skates, bullnose eagle ray, spotted eagle ray, cownose ray, bonnethead and Atlantic sharpnose sharks (Table 4.5). Notably, two species, blacktip and sandbar sharks, are known to eat the cownose ray, and four of the other great sharks species are known to consume species within the cownose ray genus (Rhinoptera). In general, there is a dearth of species-specific prey information for most sharks (most information in the literature is reported at higher taxonomic levels (usually family or genus)), and a lack of biological information in general for several of the little known mesopredator species. For example, there is no information in the literature (that we are aware of) on predators of five of the mesopredatory species (rosette skate, spiny and smooth butterfly ray, lesser devil ray, and chain catshark). However, there is evidence that large sharks consume species in ten out of the 11 mesopredator genera, and on the family (Scyliorhinidae) of the only other genus, that of the chain catshark (Table 4.5).

We assessed trends in relative abundance for ten of these 11 great shark species, based on the criterion that for each species there had to be at least one source of longline data (the most effective gear type for sampling these species) available for analysis. Evidence from a shark-targeted longline research survey in Chesapeake Bay (not available for analysis) suggests that the eleventh species, sand tiger shark (Carcharias

95 taurus), has experienced declines similar to those of the other great sharks (Ha 2006). The sand tiger shark has been considered for proposed listing under the U.S. Endangered Species Act, and is currently listed as a Species of Special Concern and a prohibited species (to land) by the U.S. National Marine Fisheries Service (NOAA 2007).

Table 4.5. Taxa of elasmobranchs (sharks, skates, rays) consumed by the apex (or nearly apex) shark species included in the large shark group. Prey are listed by species level and at the genus and/or family level because of the paucity of species-specific diet data available in the literature. Numbers correspond to references in notes below the table.

Large Sharks

Elasmobranch Mesopredators Family Genus Common name, Scientific name Bul l shar k Blackti p shar k Dusk y shar k Sandba r shar k Grea t whit e shar k Shortfi n mak o Grea t hammerhea d Smoot h hammerhea d San d tige r Tige r shar k Rajidae (Skates) •j l Scallope d hammerhea Little skate, Leucoraja erinacea

Rosette skate, L garmani n Cleamose skate, Raja eglanteria Gymnuridae (Butterfly rays) Gymnura species Smooth butterfly ray, Gymnura micrura Spiny butterfly ray, G. altavela 13 Myliobatidae (Mantas & eagle rays) : .'.•/•**

Aetobatus species (eagle rays) ,*•'?$*••' Spotted eagle ray, A. narinari • '» r: -'it Mobula species (devil rays) Lesser devil ray, M. hypostoma Myliobatis species *W1 Bullnose eagle ray, M. freminvillii 1 Rhinopteridae (Cownose rays) S||fc m Rhinoptera species (cownose rays) * HI 1° •i Cownose ray, R. bonasus •III1] Scyliorhinidae (catsharks) •HEi§§iiiiii§I Chain catshark, Scyliorhinus retifer

Sphyrnidae (hammerhead sharks) *• 3 1 4 <•• _ Sphyma species * 1 L Bonnethead shark, S. tiburo 1 L 0 12. Carcharhinidae (requiem sharks) 831 ^J ••I Carcharhinus species !>8 i J j ..'ilift t Biiiillif ISfli 96 Large Sharks

•o

CD erhea d hea d * e CD V -?= -5 e E Elasmobranch Mesopredators

Family har k shar k ar k psha r nma k hha m M S arsh a ham m Dedh a §> -Q i o Genus 2 >> T3 m fS rtfi IS ^ •o O CO Common name, Scientific name =J CO Z3 co S> E co co co Q co i- £CD C

Mesopredatory elasmobranchs

We initially considered all small shark, and all skate and ray species from within the geographic range of our study that are preyed upon by larger shark species, and for which there were sufficient data available to assess their trends in relative abundance. However, because both low intrinsic rates of population increase and heavy fishing pressure limit the potential responses of elasmobranch mesopredator populations following a loss of their predators, we restricted our analyses apriori to a subset of these species based on the following criteria: (i) of the mesopredators subject to fishing pressure (whether as direct targets or as bycatch), we included only those species with female age at maturity <4

97 years and thus relatively high potential rates of population increase. These criteria were relevant to all mesopredatory shark species, and excluded species like spiny dogfish (Squalus acanthias) and smooth dogfish {Mustelus canis), which mature late, and Atlantic angel shark {Squatina dumeril), which is presumed to mature late based on its size and the age at maturity of other species in its genus. For skates, this meant that three northern species (thorny (Amblyraja radiatd), winter (Leucoraja ocellata), and barndoor (Dipturus laevis) skate) were excluded because of both late age at maturity and high rates of exploitation; (ii) for those mesopredators not subject to high exploitation rates we included species with female age at maturity up to 7 years; (iii) finally, we excluded stingrays (Family Dasyatidae) from our analysis because they are subject to high rates of post-discard mortality (Stobutzki et al. 2002), presumably as a consequence of their thin body type relative to thicker bodied skates and rays and their mistreatment by fishermen fearing their venomous spines.

Fourteen elasmobranch mesopredator species met our criteria (Table 4.5). They comprise 11 different genera from 7 families, and range from fairly well known species (e.g. Atlantic sharpnose, blacknose, and finetooth sharks, cownose rays) to very poorly known species (bullnose eagle ray, lesser devil ray, chain catshark, smooth and spiny butterfly ray).

Bivalves

We examined all northwest Atlantic bivalve species that are components of the cownose ray diet (Blaylock 1993, Smith & Merriner 1987), for which sufficient data were available. These included only commercially fished species: the eastern oyster {Crassostrea virginica), hard clam (Mercenaria mercenaria), soft-shell clam (Mya arenaria), and bay scallop {Argopecten irradians). Based on reports from fishermen, we suspect that cownose ray predation also now influences surf clam (Spisula solidissima) populations of the New Jersey and Delmarva Peninsula coasts but were unable to locate sufficient information to include this species in our analysis.

98 Data Sources

Research survey data

The long-term UNC research survey of sharks has been conducted each year since 1972 by Dr. F.J. Schwartz of the University of North Carolina at Chapel Hill Institute of Marine Sciences in Onslow Bay off the central coast of North Carolina near Cape Lookout. The UNC data set that we analyzed comprised a total of 760 longline sets from 1972-2003. Survey methods (Schwartz 1984) have remained identical over this 32-year period. Unanchored longlines have been set biweekly from about April 15 to November 1 each year using a design employing the same gear at two fixed stations. Prior to setting out the longline, fresh fish were collected by trawling and used as whole fish to bait the hooks. Two successive sets of baited hooks constituted the sampling for every date (except less than one quarter of days when bad weather prevented establishment of the second set). Sampling was carried out during the day between the hours of 0800 and 1500hr. The East-West set was established first, near shore and approximately parallel to the beach of Shackleford Banks in 13 m depth, running up to 4.8 km eastward from 34° 38.029' N, 76° 37.835' W. Sets employed between 27 and 483 hooks (mean = 151), with one plastic foam international orange buoy of 1.3-m diameter attached for every 10 hooks and hooks spaced every 4.5 m. Case-hardened steel 9/0 Mustad tuna hooks were attached to 1.8-m drop lines of No. 2 (95 kg) porch swing chain, which were snapped onto the 7.6- cm braided nylon main line. Soak time after setting was 1 hr. During the -45 minutes required to pull in the line, the species, sex, and fork length of each hooked shark was recorded and all live sharks were tagged and returned to the sea. After 35-40 minutes travel time, the North-South set was established further offshore in Onslow Bay in 22 m depth, running southwards from 34° 33.071' N, 76° 37.422' W. The procedures followed were identical to those of the East-West set. Trawling for additional bait was occasionally required between sets.

Fisheries data

For the large sharks, we also examined logbook (1986-2000) and observer (1992- 2005) data from the U.S. pelagic longline fishery. Fisheries-dependent data are the only

99 type that covers a substantial proportion of the geographic range of these shark populations, and pelagic longline gear is particularly suitable for catching these species. The U.S. pelagic longline fleet fishes offshore of the Grand Banks (50°N), along the U.S. eastern coast and within the Gulf of Mexico, and as far south as the equator. The broad geographic coverage of these data therefore serves to complement the long temporal coverage of the research surveys. These data also include two species, shortfin mako and the great white shark, that consume elasmobranchs but were almost never caught in the research surveys (for great white n=l in the UNC survey and n=0 for all other surveys; n=0 for shortfin mako in all surveys). In the fisheries data analyses, species within the same genera that could not reliably be distinguished from one another were grouped. This includes a grouping of hammerhead sharks, genus Sphyrna (scalloped, smooth, and great hammerheads), mako sharks, genus Isurus (primarily shortfin mako), and large coastal sharks of the genus Carcharhinus (blacktip, bull, dusky, sandbar, bignose, night, silky, spinner; the first four of which consume elasmobranchs).

Commercial landings data

Data on U.S. landings were obtained from the NMFS commercial landings database, while those for Canada came from the United Nations Food and Agriculture Organization (FAO). Data on eastern U.S. landings were available by state from Maine to Texas and are an aggregate of both fishery and aquaculture production. For one species, the hard clam, aquaculture has made up a large portion of production (up to 82%) since the mid- 1980s (based on comparison of FAO aquaculture production data and NMFS commercial landings data for the U.S. east coast). Without a reasonable method of partitioning these two production sources, we were required to obtain fishery landings data for hard clams in U.S. states from other sources. Data were available only for Virginia and Rhode Island, from the Virginia public fishery hard clam production database (1973-1999) and the Rhode Island shellfish management plan (Anon. 2006) respectively.

100 Figure 4.6. Changes in landings (metric tons) by individual states of the U.S.A. plus east coast of Canada for a) oysters, b) bay scallops, c) hard clams and d) soft-shell clams. Regions enclosed by red lines are those in which the east coast population of cownose rays is expected to interact with bivalves. a)

a . M0CM*U3«ttO 0

1 •

o •

Connecticut New Void O O 8 . CD 0 0 i © o oo 0 O o o o o §• 8 _ o o o o R ' o - o WT* o -

Souti Cardm 00 8 CC S O

s oo O H VV* C3 '

a Gtcrgii Flirica East Ccas! Florida W«t Coast a Alabama 0 O °°o o 8 1 e s 0° 0 0 o R O IS i» ° o r 00 «*? 0 «. oo o 20 C * ° 0 * o o o <%»

1970 1930 1990 XX

Texas 1- O I- y $ ° * • m0 0 «oc R o 0

100 0 <0 * o •

1970 1SB0 1990 2O30 1970 1990 1990 2000 !970 1583 1990 2G00

102 b)

o Hi

* ,

o If '

1S70 1909 5590 2000 19?0 1980 1990 2300

North Csrrlm Roiij > Wfesr Ccosl 8 o o

8 *; • O o s ooo o ° ° ^

o < * °VoV

1 9/0 1930 1930 2000 197D 1980 1990 2000 c)

Phnra Island Virginia

a oo §•

o H ' o • iir..i—i..i.Mgi.i««l.ll.il.|l , 1970 1983 1990 2000 1970 1960 1990 20C0 1 370 1930 1990 2000 d) Massachusetts Phode Iskind • i O O g • s o

§• R • o s • R • o « *> o° fcr. *^ o rfoqp 8 • °% o

C3 * o •

1970 1980 19S0 2000

New Ycrk New Jersey Maryland s • O a . % o a o 0 0t> 0 o g • % o o > % o « o ° o o

n° ° 150 0 250 s o%°0 ° o 3 • JD ° O 0 Q R - V * ° o a • a • C3 <

•97C 19BG 1990 2100 1 970 I960 1990 200G 1 370 1983 1993 2000 Trends in Relative Abundance Models

Research survey data

Trends in relative abundance of each species, from each fishery-independent survey, were analyzed using generalized linear models (GLMs) with a negative binomial error structure and a log link. The negative binomial is an appropriate probability distribution for discrete, overdispersed data like the survey data, which contain a large number of zero (no catch) observations and are more variable than expected in a Poisson distribution. According to the negative binomial distribution, the probability of catching Q individuals of a given species in survey tow / has mean ju{,

rfa+f (^)C piCrXMi) = —^ 7TT , for Q = 0, 1, 2 'O m r(c,+i)r (1+A/oMi ) w where T is the gamma function and k is the negative binomial dispersion parameter. Using a log link in the GLMs means that the log of the mean catch is assumed to be a linear combination of predictor variables. The expected mean catch of a given species is then,

log(//,-)=*:/? + log(o^) where x\ is a vector of explanatory covariates for observation i, p is a vector of unknown coefficients for the explanatory variables and offset is the offset term. All analyses were conducted using SAS v9.1 (SAS 2004).

The breadth of ancillary data which could be used as explanatory covariates in estimating trends in relative abundance varied among surveys. For all surveys and species, we employed the general strategy of using the following covariates in the generalized linear models as the vector of explanatory variables (x't): year, the second order polynomial of depth, the second order polynomial of bottom temperature and q (the

104 seasonal cycle) (Table 4.6). The seasonal cycle, q, was characterized by a series of sine and cosine terms, with periods, j, of Vz and 1 year as,

27gdj *(<*,)=S gj cos + (Tj sin 7=1 365.25 365.25

where d{ is the sequential day of the year that observation i occurred in, and g( and cr. are estimated parameters. Modelling the seasonal cycle, q, also allowed us to generate common estimates for surveys conducted during multiple, distinct time-periods each year (NMFS offshore, NMFS inshore and SEAMAP surveys).

Table 4.6. Summary of generalized linear models used to estimate trends in abundance for large sharks and elasmobranch mesopredators. All data were modelled using generalized linear models, except for the observer data, which was modelled using generalized linear mixed models. All models included year as a covariate; q represents a seasonal term composed of a series of sine and cosine terms with periods of one year and one half year. Data source acronyms as in Table 4.1.

Data Covariates Error Link Offset CTDEP no covariates available Gamma Log None DNREC depth, depth2, station, q Negative Log Swept area binomial GSO no covariates available Gamma Log None NCDMF depth, depth2, temperature, temperature2, latitude, q, Negative Log Swept area latitude**? interaction binomial NMFS-Off depth, depth2, temperature, temperature2, latitude, q, Negative Log Swept area latitude**? interaction binomial NMFS-ln depth, depth2, temperature, temperature2, latitude, q, Negative Log Swept area latitude*^ interaction binomial SEAMAP depth, depth2, temperature, temperature2, latitude, q, Negative Log Swept area latitude**? interaction binomial SC depth, depth2, q, time of set, soak time, Negative Log Number of binomial hooks MDNR month, temperature, temperature2, salinity, salinity2 Negative Log None binomial UNC station, q Negative Log Number of binomial hooks VIMS river basin Negative Log None binomial Logbook area, season, temperature, use of light sticks, Truncated Log Number of area*season, arealight sticks interaction negative hooks binomial Observer area, q, depth, depth2, temperature, time of set, number Negative Log Number of of light sticks, hook depth, hook type, soaktime, target binomial hooks SDecies. bait. area*a interaction, fishina trio, vessel

105 The NMFS surveys and the SEAMAP surveys covered relatively large latitudinal ranges (Figure 4.1) and there was some interannual variation in the timing of these surveys. For species that do not undertake seasonal migrations out of each survey area, this was not a concern. However, changes in the timing of the survey could have significant effects on estimates for those species that do migrate out of the area surveyed. To account for this effect, we used the additional term of latitude when modelling the NMFS and SEAMAP survey data. Furthermore, for these surveys we allowed the seasonal cycle, q, to vary by latitude by including the interaction term between latitude and q.

There were exceptions to our general strategy of parameter selection and error structure used for the generalized linear models (Table 4.6). Data from the CTDEP and GSO trawl surveys were available only in the form of mean annual estimates so only year could be included in the model. For these two surveys, the negative binomial error structure was not appropriate since it is used for discrete data only, and instead we used a gamma error structure and a log link for the generalized linear models. The probability of a mean catch C, of a given species in year i was assumed to follow a gamma distribution with the mean JU(,

r^..v ( cP P(Ct;v,fti) = exp for 0 < C, < oo, r(v)c,i V Mi ) Mi) where T is the gamma function and v is the gamma distribution scale parameter. The expected mean catch is then, log(/ir) = x'fi where x\ is the year in which observation i occurred, and ft is the coefficient for year.

When surveys followed a fixed station design (DNREC trawl survey, UNC longline survey), we included a unique station identifier as a model factor. In some cases, covariates other than those in our standard list were available, including river basin for the VIMS seine survey, and the second order polynomial of salinity for the Maryland seine survey (MDNR). Fisheries data

Methods of the logbook data analysis are reported in Baum et al. (2003) and its Supplementary Material. Trends in relative abundance for large shark species were estimated from the observer data (as detailed in Chapter 3) using generalized linear mixed models with a negative binomial error structure and log link in the GLIMMIX procedure of SAS v.9.1 (SAS 2004, SAS 2005). In these models, to account for non- independence of longline fishing sets made by the same vessel and on the same trip, we specified vessel as a G-side random effect and fishing trip as an R-side random effect with an autoregressive one (AR1) correlation structure. Additional details of the logbook and observer data analyses are found in Table 4.6.

Calculating change in abundance

Changes in abundance reported in the main text for individual species were computed by applying the estimated rate of change over the time period from the species' first appearance in the data set until the end of the data set. Thus, for example, in the UNC data set the rate of change for most species was calculated for the entire time period (1972-2003), but for sandbar sharks was calculated only from 1976 to 2003.

Meta-Analyses of Trends in Relative Abundance

We summarized trend estimates from multiple surveys for each species using meta- analytic techniques. The instantaneous rate of change of species s from survey i is estimated as 6si. The units of the instantaneous rate of change will have the same units for all surveys (estimated using generalized linear models with a log link as described in the previous section) and can be thought of as slopes on a log scale. The estimate of 0si will be approximately normal, that is, we assume

We plotted the log-likelihood profile to check the normality assumption. As sample sizes were large for most surveys, this assumption was reasonable in most cases.

107 For the random effects meta-analysis of the instantaneous rate of change for a given species, we assumed that the true rates of change came from a normal distribution, i.e.

Maximum likelihood estimation of the random effects meta-analysis was carried out in SAS using Proc Mixed (SAS 2004). Testing for heterogeneity is equivalent to testing Ho: a2 = 0 against Hi: a2 > 0. The standard likelihood ratio must be modified in this case because the null hypothesis is on the boundary of the parameter space (i.e. the variance cannot be less than zero), which in this case means the p-value of the naive likelihood ratio test must be divided by two (van Houwelingen et al. 2002).

It is common in meta-analysis to use a fixed effect meta-analysis if there is not statistically significant heterogeneity among studies; however, the power of this test is low for small numbers of studies and a fixed effect meta-analysis will underestimate the standard errors of the estimate if heterogeneity is present. We thus used the random- effects meta-analysis in all cases. In the results section of the supplement, we examine cases where there may be differences among surveys and over different time periods.

Cownose Ray Absolute Abundance and Invertebrate Consumption Estimates

To estimate the number of cownose rays that currently pass through Chesapeake Bay during their fall migration, we combined our survey-based meta-analytic estimated rate of increase with Blaylock's population estimate of 9.3 million (s.e. = 1.8 million) for Chesapeake Bay, which was based on aerial surveys conducted between 1986 and 1989 (Blaylock 1993). Then, to estimate the total food demand for benthic bivalve mollusks by cownose rays in the Chesapeake Bay area annually, we combined our cownose ray abundance estimate with its annual occupancy time in Chesapeake Bay of 100 days (Blaylock 1993) and its individual consumption rate. Blaylock (1993) estimated an individual daily consumption rate of 210g for the cownose ray, while Schwartz (1990) estimated that individual cownose rays consume up to 1.5L of bivalve mollusks per day.

108 This equates to 250g/day based on the conversion of 1.5L to 1kg bivalve mollusk, with about 25% meat, which is comparable to Blaylock's estimate. To be conservative, we used the lower estimate of 210g, yielding a total estimate of 840,000 metric tons (wet flesh) per annum.

Quantifying Cownose Ray Impacts on Bay Scallops

Trends in bay scallop commercial landings data

To illustrate changes over time in the magnitude of North Carolina bay scallop landings (Figure 4.1), we fitted these data with a generalized additive model (GAM) in R v2.2.1 (The R Project 2007), using a gamma error structure and log link, with a loess curve of span 1, degree 2. The gamma is an appropriate probability distribution for these data, which are continuous, positive, and have non-constant variance.

Bay scallop density

To evaluate the population-level impacts of fall migrating rays across the full geographic range of traditional scalloping grounds in North Carolina, scallop densities were measured bi-weekly at six seagrass beds located within Core (Cedar Island, Yellow Shoal), Back (Oscar Shoal, Straights) and Bogue (Marker 34 and 40) Sounds from August through October in 2002, 2003, and 2004. Bay scallop density was measured within each seagrass bed in early August and again in mid-October because this period brackets the fall migration of the cownose rays. At each site, 5 replicate 1-m2 quadrats were haphazardly thrown near the edge and at the center of each seagrass bed (10 quadrats total per bed). All bay scallops within each quadrat were counted, measured and returned to their original location. Physical parameters (% cover of seagrass, salinity, temperature, sediment type) were also recorded during sampling. In the Back Sound portion of our study area, the North Carolina Division of Marine Fisheries (NCDMF) allowed a limited hand harvest of scallops coincident with the expected timing of fall immigration by cownose rays. Six harvest days were permitted between mid August and early September with a daily harvest rate of 10 bushels/fisherman. Few fishermen participated and fishing impacts were trivial compared to estimated losses from ray

109 predation. Nevertheless, to prevent our density estimates from being confounded by this additional treatment and to quantify the relative impact of this harvest, the NCDMF established, and we conducted our sampling in, two 25 m shellfish sanctuary areas within all seagrass beds. A substantially longer data base exists for one of the sites, Oscar Shoals. Although some small differences exist among years in methodologies, adult bay scallop density was measured in late July or early-mid August and again in September or October in 1992, 1993, 1994, 1996, 1998, 1999 and 2000 (detailed methods are reported in Peterson et al. 2001). For all years, bay scallop survival was calculated by dividing densities measured on the last sampling date by the density measured on the initial sampling date.

Experimental assessment ofcownose ray predation

To determine to what extent any decrease in scallop density is attributable to ray predation, we established four 2 m2 exclosures at the center and four at the edge of the 6 seagrass beds where NCDMF shellfish sanctuary areas were established. The exclosures, short (50 cm) PVC poles arranged as a stockade, exclude cownose rays while allowing other predators (crabs and whelks) into the matrix of poles (Woodin 1981, Peterson et al. 2001). The number of scallops surviving within the stockade is compared to areas of free ray access (controls). The experiment was performed during the fall of 2002, 2003, and 2004. The stockades were constructed in situ and bay scallops were allowed to move freely into and out of the exclosure. Exclosures were erected in mid August of each year and bay scallop density was measured within the exclosure and in the controls at that time and again in late September. A similar set of experiments had been performed at the Oscar Shoal site in 1996 and 1998 (Peterson et al. 2001). As in the later experiments, , naturally occurring bay scallops were allowed free access to the exclosure, but in addition ten marked and tethered bay scallops were placed within and outside the stockades. Mortality within the stockade should be substantially less than in the control areas if large mobile consumers are the chief predator on bay scallops during this time period. Bay scallop mortality within the stockades was calculated as 1 minus survival, computed by dividing densities measured on the last sampling date within the stockade by the density measured on the initial sampling date prior to construction of the enclosure. The

110 difference between scallop survival inside and survival outside the stockades greatly underestimates the proportion of natural mortality attributable to large mobile consumers (of which cownose rays were the only ones observed) because bay scallops initially inside stockades emigrate throughout this period of time and thereby become susceptible to consumption by rays. Earlier tethering experiments (Peterson et al. 2001) indicate that emigration explains a large majority of the apparent mortality of bay scallops inside the stockades.

Results and Discussion

Trends in Relative Abundance

Overall, the trend estimates from the 17 research surveys and the two fisheries data sets give broadly consistent estimates of population declines of great sharks and population increases in elasmobranch mesopredators (Tables 4.3, 4.4). Earlier trend estimates for great sharks from logbook reports (Baum et al. 2003) have been criticized for using fisheries-dependent data reported by fishers and for relying on only one data source (Burgess et al. 2005, but see Baum et al. 2005). Here, we have analyzed the complementary scientific observer data set from the same fishery, and shown similar results for each species (group) except tiger sharks (discussed below). We have also analyzed all available, long-term scientific research surveys (n=12) for great sharks. Importantly, the longline research surveys designed to catch sharks (UNC, SC) suggest declines for every great shark species, and show large, statistically significant declines for each great shark that was caught in sufficient numbers to estimate trends.

In a few cases, there are qualitative differences (i.e. increasing vs. decreasing trends) amongst trend estimates from different data sets for individual elasmobranch species (Tables 4.3, 4.4). Such differences could arise for several reasons, including differences in the years (early vs. recent) or areas (e.g. north vs. south, inshore vs. offshore) sampled (Figure 4.1, Table 4.2). Here we discuss the details of each of these cases.

Ill Only 2 of the 30 trend estimates for great sharks are statistically significant increases. The first is for juvenile scalloped hammerhead sharks caught in the recent (1989-2005) SEAMAP survey. We also note that juvenile blacktip sharks in the SEAMAP survey and juvenile sandbar shark in the NMFS surveys show nonsignificant changes (Table 4.3). These data suggest that declines in the juveniles may have ceased for these species, and that juvenile survival could have increased because of declines of larger sharks. The second increasing trend occurs in the observer data set for the tiger shark: its abundance has apparently begun to increase in the past couple of years. This fishery catches mainly juvenile tiger sharks, and of all the shark species caught, tiger shark has the highest survival rate (Beerkircher et al. 2002). Thus, this change from a declining to an increasing trend also may represent an increase in juvenile survival associated with a decline in predation by large sharks.

For elasmobranch mesopredators, most differences in trend estimates within species appear to be caused by small sample size and/or high sampling variability at the edge of the species range. For example, there were 7 surveys for cownose ray, and 6 of these have estimates of instantaneous rates of increase between 0.044 and 0.17 per year (Table 4.4). Cownose rays in the NMFS-Offshore survey data have a non-significant trend estimate of-0.26 (95%CI: -0.54 - 0.01). NMFS-Offshore surveys caught cownose rays in only 3 years because of the survey location at the edge of the geographic region consistently inhabited by this species, and we thus do not regard this estimate as being indicative of temporal change in the population. A similar problem probably exists for the trend estimate derived from the NMFS-Inshore data for Atlantic sharpnose shark. This is the only non-significant estimate for this species, and these surveys occur at the extreme northern limit of this species' range.

For some of the smaller skates, differences in trend estimates may represent real differences among populations. For example, little skate shows statistically significant increases in three surveys, but does not appear to be increasing in Long Island Sound (CTDEP survey, Table 4.4). We suspect that this may represent a real pattern because

112 there is an intense fishery for lobster bait in this region that catches little skate * incidentally (Packer et al. 2003).

The only two exceptions to the general pattern of increasing abundance amongst the mesopredators are the blacknose shark, which has been decreasing according to the UNC survey, and the spotted eagle ray, which has been decreasing according to the recent SEAMAP survey (Table 4.4). Like the Atlantic sharpnose shark, the blacknose shark is caught in several recreational and commercial fisheries (Carlson & Lee 1999), however, its age at maturity is slightly greater (3.8 years on average) than the Atlantic sharpnose shark (2.3 years) (Carlson et al. 1999, Driggers et al. 1999, Carlson & Baremore 2003, Loefer & Sedberr 2003). Thus, it may be more susceptible to fishing than the Atlantic sharpnose, which is clearly increasing. Insufficient information about the fishing pressure on spotted eagle ray (which matures between the age of 4 and 6 years (Last & Stevens 1994)) limits our interpretation of abundance trends for this species.

Inferences from Life History Theory about the Cownose Ray Rate of Increase

Females in the U.S. Atlantic cownose ray population reach sexual maturity between age 7 and 8 (Smith & Merriner 1987) and have one pup per year (Smith & Merriner 1986). Like most other elasmobranch species, there are no direct estimates of natural mortality for the cownose ray. However, using the meta-analytic mean increase, 0.087 (95% CI: 0.034 - 0.14), as the rate of population increase (r), we can solve the Euler- Lotka equation to estimate the mortality that the cownose ray population must be subject to. The mortality rate was calculated as 0.076 (95% CI: 0.021-0.127), which is much lower than a species of fish with this population growth rate would be expected to have (compared to similarly sized species) under natural conditions (Pauly 1980, Hoenig 1983), implying that the cownose ray population has experienced substantially reduced natural mortality. In addition, because the mortality rate of the cownose ray population also must include some bycatch mortality, the natural mortality must actually be somewhat less then the estimate of 0.076. We conclude that given the life-history of cownose rays and its estimated rate of increase that the population must now have an extraordinarily low natural mortality rate compared to what it would experience under

113 normal levels of predation. We infer that the loss of naturally more intense predation by the great sharks explains why the cownose ray now deviates so greatly in mortality rate from what is expected on the basis of life history relationships (Pauly 1980, Hoenig 1983).

Comparison of Shrimp Fishing Effort between the Southeast U.S. & Northern Gulf of Mexico

Whereas the cownose ray population on the east coast of the U.S. has increased substantially, in the Gulf of Mexico, where shrimp trawl fishing effort is enormously greater, incidental catches have apparently reduced that cownose ray population (Shepherd & Myers 2005), and oyster landings have increased (Figure 4.6a). We compared shrimp fishing effort between the southeast U.S. (North Carolina to eastern Florida) and Gulf of Mexico within equivalent time periods for which data were available (1991-1993). Along the southeast U.S., the annual average number of shrimp fishing trips during that period was 55,878 (Vaughan & Nance 1998). This includes ocean waters, sound waters and some areas possibly unsuitable for cownose rays, such as rivers. We were not able to exclude unsuitable areas due to the resolution of the data. For the same time period, the Gulf of Mexico shrimp trawl fleet fished an annual average of 306,910 24-hour shrimping days or 7,365,829 fishing hours (Anon. 2002b). Typically, a southeast U.S. shrimp fishing trip in the early 1990s was approximately 5 hours long (Vaughan & Nance 1998). Thus the southeast U.S. shrimp fishing effort equaled 279,390 fishing hours (55,878 trips x 5 hours per trip), or just 3.8% of the Gulf of Mexico effort.

As cownose rays migrate northward in the spring and southward in the late summer and fall along the coast of the southeast U.S., they will be exposed to local shrimp fisheries over restricted time periods. In North Carolina for example, the majority of the shrimp fishing occurs in July and August (Anon. 2005) while cownose ray abundance does not peak in the region until September (Peterson et al. 2001), missing the time of most intense effort. In contrast, the northern Gulf of Mexico shrimp fishery maintains a very high intensity from May into December (NMFS commercial landings data), which includes the time at which cownose rays inhabit the area (Shepherd & Myers 2005).

114 Thus, we conclude that the difference in shrimp fishing effort, and spatial and temporal overlap between cownose rays and fishing effort, could explain the different trends in abundance for these two cownose ray populations.

ACKNOWLEDGEMENTS

We thank J. Boylan, J. Collie, E. Durell, D. Gaskill, J. Hoey, W. Hogarth, D. Kahn, J. Kraeuter, R. Lipcius, M. McDuff, P. Rago, D. Ricard, R. Seitz, D. Simpson, F. Schwartz, J. Smith, G. Ulrich, K. West, and the North Carolina Division of Marine Fisheries (NCDMF) for sharing data. North Carolina biological data were provided at the authors' request by the NCDMF. Analyses of these data and conclusions drawn there from are those of the authors and do not necessarily represent the views of NCDMF. The University of North Carolina, Institute of Marine Sciences provided support for conducting the longline shark survey. We also thank D. Gaskill for assistance conducting the field research and the UNC Institute of Marine Sciences for providing support to conduct the longline shark survey. V. Garcia, J. Hoenig, H. Lotze, L. Lucifora, A. Sugden, J. Valentine, B. Worm, and the referees provided helpful comments on the manuscript. We acknowledge funding from Pew Institute for Ocean Science, Sloan- Census of Marine Life, Natural Sciences and Engineering Research Council of Canada, Killam Trusts, NC Fisheries Resource Grants Program, NC Sea Grant, and NSF.

115 CHAPTER 5

Top-Down Control in the Ocean: Methods, Evidence, Inferred Mechanisms, and Management Implications Abstract

Top-down control can be an important determinant of the structure and function of ecological communities, and is well-documented in freshwater, rocky intertidal, and some terrestrial ecosystems. Yet, the prevalence of top-down control in oceanic food webs remains unclear. Ecosystem effects ensuing from the depletion of oceanic apex predators might be massive in scale, but evidence for such effects has been sparse until recently, and it has been argued that complex marine food webs with many interacting species may generally diffuse top-down effects. Here I argue that fundamental constraints on studying widely dispersed predators, sampling vast marine ecosystems, and replicating controlled experiments in the ocean have hindered detection of cascading effects of marine predator removal. Recently a number of studies have partly overcome these problems by utilizing long-term time series data, large-scale field studies, and appropriate statistical techniques such as meta-analysis of species interactions to make ecosystem-scale spatial and temporal contrasts in predator abundance and associated community dynamics. These studies have uncovered evidence of top-down effects in the open ocean, coral reefs, and continental shelves, which have been triggered by single keystone or dominant predators, or the functional elimination of entire guilds of oceanic predators. Synthesizing the available evidence I show that increases in mesopredatory fishes and invertebrates often followed predator depletions and that while changes can cascade through multiple levels of predators, attenuation of top-down effects at the herbivore - plant link may be typical for oceanic food webs. Effective fisheries management and species conservation will depend on improved understanding and explicit accounting of predator-prey interactions, but since anthropogenic perturbations to oceanic food webs will outpace our ability to predict outcomes of deteriorating networks of species interactions, it would be precautionary to maintain marine predators above thresholds of ecological extinction.

117 1 INTRODUCTION

The successive removal of large, high trophic-level predators and other megafauna seems to be the most consistent initial impact that humans have on natural ecosystems (Jackson et al. 2001, Pandolfi et al. 2003, Lotze et al. 2006); current species losses also appear to be skewed towards the apex of food webs (Petchey et al. 1999, Duffy 2002, 2003, Cardillo et al. 2005). Predator removals like these may have far-reaching consequences for ecosystem structure, functioning, and resilience because of the strong influence they can exert on other species (Duffy 2002, Soule et al. 2005). Large predators in particular are often strong interactors and hence may determine the diversity, abundance, and distribution of their prey (Paine 1980, Strong 1992), as well as other species through indirect effects that propagate downwards through the food web (Wootton 1994). Such top-down control has been has been documented in a wide range of ecosystems (Pace et al. 1999), including lakes (Carpenter & Kitchell 1993, Brett & Goldman 1997), streams (Power 1990), rocky coastlines (Menge 1995), kelp forests (Estes et al. 1998), salt marshes (Silliman & Bertness 2002), grasslands and forests (Terborgh et al. 2001, Schmitz et al. 2000, Berger et al. 2001). Recent theoretical and empirical research suggests that top predators also may confer stability to food webs (Rooney et al. 2006) and buffer the effects of perturbations like climate change (Wilmers & Getz 2005). Predicting how ecosystems will respond to predator removals remains a challenge, however, because of the many different species interactions in complex ecological networks (Polis & Strong 1996, McCann 2007).

In the ocean, top predators are being reduced in abundance and diversity globally (Pauly et al. 1998, Myers & Worm 2003, Worm et al. 2005), implying that if top-down control were prevalent (Verity & Smetacek 1996, Banse 2007), these ecosystems might be significantly restructured as a-consequence. Marine organisms in almost all ecosystems are now intensely exploited following unparalleled increases in the efficiency, effort, and range of fisheries over the past century. Since harvesting in the sea preferentially targets high trophic-level species (Pauly et al. 1998), particularly those at

118 the apex of the food web that tend to be highly vulnerable to overexploitation (Jennings et al. 1998,1999, Denney et al. 2002), it is these predators that have been disproportionately affected (Jackson et al. 2001, Whitehead 2002). Although there are few documented global marine extinctions (Carlton 1999), local extirpations have occurred and depletions in excess of 90% are not uncommon (Dulvy et al. 2003, Hutchings & Reynolds 2004, Lotze et al. 2006). Because species interactions weaken and may become functionally extinct as predators are depleted (Dayton et al. 1998; Soule et al. 2005) the potential for unintended indirect effects ensuing from these losses is high (Botsford et al. 1997, Pikitch et al 2004). Yet, the evidence of top-down effects in the ocean has come almost exclusively from nearshore benthic communities, especially rocky shores and kelp forests (reviewed recently by Pinnegar et al. 2000, Steneck & Sala 2005). Evidence for such effects in other oceanic ecosystems (offshore continental shelves, slopes and the open ocean) has been sparse, leading many to conclude that such effects will generally attenuate in oceanic food webs because of high diversity, omnivory, connectance among species, and the environmental stochasticity of these ecosystems (Strong 1992, Jennings & Kaiser 1998, Link 2002). Demonstrations of strong bottom-up effects (by which I mean population control through resource limitation) in the ocean (e.g. Aebischer et al. 1990, Chavez et al. 2003, Ware & Thomson 2005, Frederiksen et al. 2006) have reinforced the commonly held view that these ecosystems may be little altered by the removal of top predators.

Here, I propose that top-down control and cascading effects of predator removal may be more common in oceanic ecosystems than previously thought, but that our capacity to detect such effects has until recently been limited by the logistic constraints of conducting controlled experiments in the sea, sampling vast marine ecosystems, and studying widely dispersed apex predators. Most previous studies of top-down control in marine food webs are experimental in nature and involve small organisms, and it is unclear whether their findings can be justifiably scaled-up to large mobile predators in oceanic ecosystems (Carpenter 1996). And while bottom-up effects, driven by climatic variation or nutrient loading, can strongly influence marine populations (Hunt & McKinnell 2006), this does not preclude predation from also affecting community

119 dynamics. In reality, both top-down and bottom-up factors influence natural ecosystems and it is the relative strength of these forces that is important (Hunter & Price 1992, Power 1992). The traditional viewpoint that bottom-up control predominates in oceanic ecosystems may, however, reflect the fact that studies examining bottom-up influences typically focus on primary production, consumption at the bottom of the food web, and overall ecosystem productivity, while ignoring interactions among species at higher trophic levels (e.g. Ware et al. 2005). Understanding the importance of predation in the ocean instead requires studying the upper trophic levels, where presumably changes in top-down control will be strongest, and explicitly considering how anthropogenic perturbations like fishing modify species interactions in these ecosystems.

In this review, I synthesize the evidence from recent ecosystem-scale studies identifying top-down control in the ocean, with a particular focus on apex predators (trophic level > 4) and other large mobile vertebrate predators. I first outline how ecological theory and findings from small-scale manipulative experiments inform our expectations about top-down control in the ocean and consider reasons why these studies might not scale up to large oceanic predators. Next, I highlight the methods researchers are now using to examine trophic interactions at the ecosystem scale by comparing oceanic food web dynamics across temporal and spatial gradients of predator abundance. I then examine the new body of evidence for oceanic top-down control, reviewing the main patterns observed and the mechanisms inferred to have driven these food web alterations. Finally, I consider the ramifications for ecosystem management and species conservation. Because ecosystem-scale studies are so data intensive and time consuming, the literature documenting oceanic top-down effects lacks the breadth of the ecological literature examining consumer and resource control through small-scale experiments. Notwithstanding the limited number of examples, there is emerging evidence in accordance with a top-down view of the ocean, suggesting that the loss of top predators can lead to a fundamental reorganization of marine food webs.

120 2 INSIGHTS FROM ECOLOGICAL THEORY AND EXPERIMENTS

Understanding the relative strength of abiotic and biotic forces influencing community dynamics has long been an important research focus in ecology (Hairston et al. 1960, McQueen et al. 1989, Hunter & Price 1992, Power 1992), and modern ecologists have studied consumer-resource interactions intensively, employing manipulative experiments to distinguish between these effects (e.g. Paine 1980, Power et al. 1985, Carpenter et al. 1988, Worm et al. 2002). Much of this research has concentrated on trophic cascades (Figure 5.1a,b), defined here as strong predatory effects resulting in inverse patterns in abundance or biomass across at least two trophic links in a food web (Carpenter and Kitchell 1993 Pace et al. 1999), and although the trophic cascade concept was first proposed to explain terrestrial food web dynamics (Hairston et al. 1960), the best known examples of cascades are from aquatic ecosystems with discrete trophic levels (Paine 1980, McQueen et al. 1989, Power 1990, Carpenter & Kitchell 1993, Paine 2002). Indeed, the predominance of trophic cascades in lakes, streams, and rocky intertidal shores led Strong (1992) to assert that these 'potent forces' were 'all wet', primarily restricted to low diversity aquatic food webs with algal bases, in which one or a few keystone species (Figure 5.1c) could dominate (Paine 1966, Power et al. 1996). In more diverse ecosystems, although predation might still be intense, ecologists posited that the 'runaway consumption' characteristic of trophic cascades would be dampened by complex networks of trophic interactions (Strong 1992, Polis & Strong 1996).

In the past decade, however, trophic cascades have been documented in a variety of ecosystems, including highly diverse systems like tropical forests, demonstrating that these interactions are not so easily categorized (Pace et al. 1999). Still, meta-analyses of 102 trophic cascade experiments suggest that these predatory effects are strongest in simple aquatic (lentic and marine benthic) and weakest in terrestrial and marine planktonic ecosystems (Shurin et al. 2002). While predators significantly suppressed herbivore abundance in all six ecosystem types examined, trophic cascades were atypical

121 (a) Classic trophic cascade (b) Oceanic trophic cascade

(e) Omnivory (f) Diffuse predation

Figure 5.1. Predatory effects observed among different levels of predators (blue circles), and on herbivores (white circles) and plants (black circle). While the classic trophic cascade involves first- level predator(s) - herbivore (s) - plant(s), an alternative in the ocean is a trophic cascade involving three levels of predators. Solid arrows indicate direct effects, hatched arrows indicate indirect effects. both in terrestrial and marine plankton (n=9) experiments because of negligible plant community responses to the predator effect (Shurin et al. 2002). Similarly, a meta analysis of 47 marine mesocosm experiments showed top-down control of herbivorous zooplankton by zooplanktivores but a dampening of this effect at the herbivore-plant link

122 (Micheli 1999). More recent experiments, however, show that trophic cascades can operate in three and four trophic-level marine pelagic food webs, but can be masked by counteracting effects on algal biomass (Stibor et al. 2004).

Although one might conclude that predator effects do not cascade in plankton-based marine ecosystems, since these food webs include secondary and tertiary predators, attenuation of effects at the lowest trophic levels still leaves considerable scope for trophic cascades to occur amongst higher levels (Figure 5.1b). Experimental evidence of predatory control of herbivores in marine plankton food webs (Micheli 1999, Shurin et al. 2000), and across ecosystems of stronger effects by vertebrate than invertebrate predators (Shurin et al. 2002, Borer et al. 2005) suggests that top-down control by large vertebrate predators might play an important role in structuring oceanic communities.

Likewise, experimental manipulations of predator diversity are showing that rules about complex species interactions diffusing top-down effects are not as clear-cut as once was thought (Strong 1992, Polis & Strong 1996), but rather a variety of mechanisms mediate how diversity influences predator effects (Sih et al. 1998). Predator diversity . can enhance prey consumption through complementarity of diets or interspecific facilitation (Sih et al. 1998, Duffy 2002, Snyder et al. 2006), or because diverse assemblages are more likely to include keystone predators (Ives et al. 2005). A recent kelp mesocosm experiment showed that predator diversity also can strengthen trophic cascades through different behavioural responses of herbivores, which translated into complementary reductions in herbivore grazing rates and increased density of kelp (Byrnes et al. 2005). Alternatively, predator diversity may reduce impacts on prey, if predators engage in cannibalism or intraguild predation (Figure 5.Id; Polis et al. 1989, Sih et al. 1998, Finke & Denno 2004, 2005). Omnivorous predators also may dampen trophic cascades by weakening or reversing the expected indirect positive effect of predators on species two trophic links below (Figure 5.1e; Petchey et al. 2004, Bruno & O'Connor 2005). In estuarine mesocosms, for example, a strong trophic cascade induced in the presence of three strictly carnivorous predators was short-circuited in experimental treatments that included pinfish (Lagodon rhomboids) because this species consumed

123 both herbivores and algae (Bruno & O'Connor 2005). Whether predator effects cascade to lower trophic levels is context dependent and influenced by predator and prey traits and which predator-prey interactions predominate. In oceanic food webs, where numerous counteracting predator-prey interactions exist amongst the myriad of species interactions, quantitatively or even qualitatively predicting the ecosystem consequences of predator losses thus becomes enormously challenging.

Finally, even if Strong's (1992) assertion holds - that top-down effects arise in diverse ecosystems only as 'aberration[s] imposed by man' - such perturbations are no longer rare, and the nature of modern hunting and fishing in the ocean seems to promote cascading effects of predator depletions. First, by targeting apex and near apex predators, exploitation removes species that are likely to be strong interactors (Bascompte et al. 2005), and in some cases may be keystone species (Power et al. 1996, Soule et al. 2003). Second, although fisheries may target single species, the relatively non-selective nature of industrial fishing gear renders most operations de facto multi-species fisheries, in which numerous similarly sized and behaved species are caught. Consequently, these fisheries typically reduce the abundance of whole predator guilds and hence the functional diversity of the ecosystem, effectively negating the buffering capacity that multiple predator species might otherwise have afforded. In such cases, top-down effects may propagate through diverse oceanic communities in which there are no keystone species, but instead predatory control exerted through diffuse predation by groups of predators (Figure 5.If) is lost. Third, in systems like coral reefs, sequential depletion of functional groups has further diminished connectance among species and functional diversity, rendering them even more susceptible to perturbations. In essence, exploitation's impact is to simplify diverse oceanic ecosystems, rendering them in some sense more like those aquatic food webs in which cascading effects are the norm. Explicit consideration of anthropogenic modifications to food webs may therefore inform predictions about top-down effects in the ocean.

124 3 OVERCOMING METHODOLOGICAL & DATA LIMITATIONS

Controlled, manipulative experiments are the hallmark of modern ecological research on consumer-resource dynamics, yet have limited utility in the ocean. While predator-prey interactions involving small, sessile prey can often be tested experimentally, for example through tethering experiments that examine predation rates (Witman & Sebens 1992, Guidetti 2006) or predator exclusion experiments (Peterson et al. 2001), constraints on conducting controlled, replicated experiments in the ocean increase as one moves up the food web and away from shore, and quickly become prohibitive. In general, it is not possible to experimentally manipulate large, mobile oceanic predators or to conduct experiments at sufficiently large spatial or temporal scales to encompass the population dynamics of these species, yet studies of predator- prey interactions at, or near, the top of oceanic food webs may be most illuminating about top-down control.

Evidence of top-down forcing in the ocean instead must be deduced from large-scale observational studies in which the causes and underlying mechanisms are inferred through comparisons and correlations. Such an approach represents a fundamental departure from the experimental method and involves a trade-off between scale and control. Absent direct predator manipulation, sufficient variation in the abundance of wild populations of predators must be observed to discern contrasting responses of prey populations. This variation may be temporal or spatial, and be driven by natural processes or anthropogenic perturbations (see examples below). Detecting such changes is however non-trivial, requiring significant amounts of data, collected over long time- spans (e.g. decades) or large spatial scales (103- 105km2), for species at multiple trophic levels. The development of hypotheses about top-down control also requires detailed knowledge of predator-prey relationships in the food web of interest, for example through diet (i.e. gut content) or stable-isotope studies. Documenting the topologies of complex food webs is, however, a huge task and even these food web reconstructions may provide little insight into the relative strength of different species interactions (McCann 2007).

125 Obtaining high quality data at the temporal and spatial scales relevant to examining hypotheses about top-down control in the ocean has proven difficult. Fisheries provide an obvious source of ecosystem-scale data for this purpose, the caveats being that these data generally are restricted to commercially valuable species and vary tremendously in their taxonomic resolution and quality. Landings (catch) data are the most common fishery-dependent data and almost always cover longer time periods and more species than other types of time series data. Studies of landings from continental shelf ecosystems often have identified a pattern suggestive of predation-release (or competitive release), in which rising catches of small fish or invertebrates followed declining catches of larger predators. In northwest Africa, for example, octopus {Octopus vulgaris) landings were low despite a lucrative market for the species, and rose sharply only after larger, predatory fishes declined in the late-1960s (Caddy and Rodhouse 1998). Similarly, declining catches of Atlantic cod (Gadus morhua) in the northwest Atlantic Ocean have been mirrored by increasing catches of northern shrimp (Pandalus borealis), snow crab (Chionoecetes opilio) and American lobster (Homarus americanus) (Worm & Myers 2003). The use of landings data as indices of abundance and for inference of top- down effects is tenuous, however, since these data may be confounded by changes in fishing effort, gear, and locations. Unless information about these operational factors is considered, findings based solely on landings data may best be viewed as hypotheses requiring more rigorous testing.

Research surveys can provide more reliable indices of abundance than landings statistics, as may catch-per-unit-effort (CPUE) data from logbooks or observer programs when modelled to account for differences among operations that affect the efficiency of the fishing gear. However, because these time series data tend to be short and highly variable, they often yield low statistical power making it impossible to distinguish among competing hypotheses that might explain the observed patterns (Myers & Mertz 1998). For example, one of the first studies cited as evidence of an oceanic trophic cascade used research survey and CPUE time series data of only ten years length (Table 5.1; Shiomoto et al. 1997). In the central Bering Sea, the abundance of pink salmon (Oncorhynchus gorbuscha), an important pelagic planktivore, fluctuates by an order of magnitude on a

126 two-year cycle. In summers when pink salmon were plentiful, biomass of its macrozooplankton prey (composed mainly of Neocalanus copepods) was low and phytoplankton biomass high. In alternate years the pattern reversed. However, these patterns were evident only in the latter half of the time series, and were statistically significant only for the inverse relationship between salmon and macrozooplankton (Shiomoto et al. 1997). Thus, while pink salmon may control plankton biomass in some years, there are evidently other important controlling factors in this subarctic ecosystem, and like landings data patterns gleaned from single, short time series must be interpreted cautiously.

Additionally, time series data are prone to high autocorrelation, a problem similar to pseudoreplication in which the assumption of independence among years is violated, thus reducing the true degrees of freedom and further compromising the data's power for statistical hypothesis tests (Pyper & Peterman 1998). Similarly, spatial correlations may limit comparisons of population data from different regions: recruitment in marine fishes, for example, is typically correlated up to about 500km (Myers et al. 1997a).

Improving Methods for Ecosystem-Scale Analyses

New approaches to the study of oceanic food webs have partly overcome previous limitations by drawing upon multiple data sources, borrowing principles from manipulative experiments, and employing advanced statistical and modelling techniques. Comparisons of predator and prey abundances from multiple areas may treat populations either as replicates (if predator abundances varied similarly) or treatment levels (if they differed) in large-scale 'experiments', and hence have increased power to detect general patterns (Myers & Mertz 1998). In this context, exploitation may be regarded as a predator-removal experiment, and conversely, no-take marine reserves as a predator recovery experiment. Time series data of species interactions from different regions may be formally combined by meta-analysis; where population data from multiple areas are unavailable, patterns in independent data sets from within a region may be examined for consistency. Detection of consistent patterns across data sets increases confidence in

127 Table 5.1. Studies identifying top-down control in oceanic ecosystems. Examples are drawn from the recent literature (1997-2007) and are representative of the ecosystems where top-down effects have been observed and the types of methods being used to identify them. Ecosystem Region Species (Predator- Prey) Inferred Mechanism Evidence Estimated Effect Refs Open ocean Central N Sperm whales, swordfish, blue shark Exploitation reduces biomass of Dynamic model (Ecosim), Dominance shift in apex predator 1 Pacific - large squid squid-consuming apex predators pre-exploitation food web guild to large squid —> prey release reconstruction Central Tunas, billfish, sharks - pelagic Exploitation reduces biomass of Comparison of 1950s Increase in pelagic 2 Pacific stingray pomfret, snake mackerel, large pelagic fishes —• survey and 1990s fishery mesopredators skipjack tuna mesopredator release catch rates Subarctic N Pink salmon - zooplankton - Biennial cycle in salmon Ten-year time series from Two-fold phytoplankton increase 3 Pacific phytoplankton abundance —• trophic cascade surveys, fisheries CPUE when salmon abundant Open to U.S. E coast, Great sharks - small elasmobranchs Exploitation functionally eliminates Meta-analysis of research Four- to ten-fold increases in 4 coastal ocean NW Atlantic (skates, rays, sharks) apex predators —• surveys, fisheries CPUE elasmobranch mesopredators mesopredator release from 1970 to 2005 Great sharks - cownose ray - bay Exploitation functionally eliminates As above; cownose ray Eight-fold increase in cownose 4,5 scallop apex predators —• trophic exclosure experiments ray -> bay scallop crash -• cascade fishery closure NE Pacific, Killer whale - sea otter - sea urchin - Sea otter recovery from hunting, Long-term observations Sea otter decline ->• urchin 6-8 Aleutian kelp subsequent killer whale predation (1970-1996), behavioral overgrazing —• up to ten times Islands reverses recovery-* trophic data, consumption rates, less kelp cascade spatial contrast Continental E Scotian Cod & other benthic fishes - snow Overexploitation of cod-dominated Research survey, landings Dominance shift from benthic 9,10 shelf Shelf, NW crab, shrimp, small pelagic fishes - benthic fish community -»• (1965-2002) piscivores to small pelagic Atlantic zooplankton - phytoplankton trophic cascade fishes, macroinvertebrates NW&NE Cod - shrimp Overexploitation of cod -»prey Meta-analysis of research General increase in shrimp 11 Atlantic release surveys (1970-2000) from biomass, weaker at southern 9 regions range limits NW Atlantic Cod- snow crab Overexploitation of cod —> prey Meta-analysis of research General increase in snow crab 12 release surveys from (-1970- biomass 2000) 10 regions Newfound Cod - capelin Overexploitation of cod -> Research survey, landings Increase in capelin biomass 13 land mesopredator release data Celtic Sea, Largest size classes of fish - smallest Size-selective exploitation of large Size structure analysis of Increases in smallest size • 14 NE Atlantic size classes of fish fishes —> mesopredator release survey data (1987-2003) classes of fish "North Sea Largest size classes of fish - smallest Size-selective exploitation of large Size structure analysis of Increases in smallest size 15 size classes offish fishes —> mesopredator release survey data (1977-2000) classes of fish N Atlantic, N Small pelagic fish - zooplankton - Fluctuations in fish abundance -> Meta-analysis of time Zooplankton decline when fish 16 Pacific phytoplankton prey release series from 20 regions abundant, little response in phytoplankton to 00 Ecosystem Region Species (Predator- Prey) Inferred Mechanism Evidence Estimated Effect Refs Continental Gulf of Great sharks - Atlantic angel shark, Overexploitation of apex Research trawl surveys Increase in deepwater 17 shelf to coast Mexico spreadfin skate, smooth dogfish Predators—> mesopredator (1972-2002) elasmobranch mesopredators release fishing inhibits response of coastal mesopredators Polar coast to W Ross Sea, Adelie penguins, minke & killer whales High seasonal abundance of top Long-term field Seasonal decrease in silverfish, 18 continental Antarctica - Antarctic silverfish, krill - diatoms predators -* trophic cascade observations (1996-2005) krill, phytoplankton ungrazed shelf 'natural experiments' South Leopard seal-fur seal Absence of apex predator —> Long-term census and field Fur seal population recovery 19 Georgia, S prey recovery observations (9 years), only where apex predator is Atlantic spatial comparison, absent demographic model Semi-enclosed Black Sea Bonito, mackerel, bluefish - anchovy, Overexploitation of predators, Research survey, landings Two-fold zooplankton decline, 20,21 sea sprat, horse mackerel, jellyfish - invasion of jellyfish -»trophic data (late 1950s--2001) doubling of phytoplankton zooplankton - phytoplankton cascade biomass Coral reef Great Barrier Grouper (coral trout) - small prey fish Overexploitation of top predator Marine reserve spatial Prey densities double in areas 22 Reef, —•mesopredator release comparison, underwater where predator fished Australia visual census Fiji, Pacific Largest size classes of fish - smallest Overexploitation of largest Spatial comparison across Increases in biomass and 23 Ocean size classes of fish predatory reef fishes by 13-island exploitation numbers of smallest size subsistence fishing —• gradient; underwater classes, but response weak mesopredator release visual census Fiji, Pacific Large predatory reef fish - starfish - Overexploitation of largest As above Increased crown-of-thorns 24 Ocean coral predatory reef fishes by starfish grazing depletes coral subsistence fishing —> trophic builders, phase shift to cascade macroalgae Rocky subtidal Adriatic Sea Sea breams - sea urchins - Overexploitation of predator —• Marine reserve spatial Transition from macroalgae 25 macroalgae trophic cascade comparison, underwater dominance to coralline visual census, prey barrens at high sea urchin tethering experiment levels References: 1 Essington 2007; 2 Ward & Myers 2005b; 3 Shiomoto ef al. 1997; 4 Myers ef al. 2007; 5 Peterson ef al. 2001; 6 Springer ef al. 2003; 7 Estes et al. 1998; 8 Estes & Palmisano 1974; 9 Frank et al. 2005; 10 Choi ef al. 2004; 11 Worm & Myers 2003; 12 Myers et al. unpublished; 13 Carscadden ef al. 2001; 14 Blanchard ef al. 2005; 15 Daan ef al. 2005; 16 Micheli 1999; 17 Shepherd & Myers 2005; 18 Ainley ef al. 2006; 19 Boveng ef al. 1998; 20 Daskalov 2002; 21 Daskalov ef al. 2007; 22 Graham ef al. 2003; 23 Dulvy ef al. 2004a; 24 Dulvy ef al. 2004b; 25 Guidetti 2006. results, but interpretation of patterns in complex data sets must still be underpinned by a 'sound understanding of the natural history of the system' (Steele et al. 2006) and independent evidence of predation links between species. Additional analyses may test the plausibility of competing hypotheses, such as potential environmental drivers of observed patterns. Bioenergetics and demographic models that evaluate the plausibility of hypothesized predator-prey interactions can complement these large-scale analyses (Doak et al. 2007). Although correlative evidence indicative of top-down control cannot definitely prove predator-prey interactions are the underlying cause, predator-prey 'replicates' and 'controls' for alternative factors are powerful tools for dissecting these processes at the ecosystem-scale. In the following sections I present examples of large- scale studies utilizing these tools to examine evidence for top-down effects driven either by temporal or spatial variation in predator abundance.

Interannual Variation in Predator Abundance

In the first broad quantitative evaluation of trophic control in the pelagic ocean, Micheli (1999) assembled time series data of 7 to 45 years for zooplanktivorous fish, herbivorous zooplankton, phytoplankton, and nutrients in each of 20 ecosystems. Time series data were derived primarily from coastal and continental shelf ecosystems of the northeast Atlantic Ocean and Mediterranean Sea, but also included offshore areas of the north and eastern Pacific Ocean. Spearman rank correlations between adjacent trophic levels provided evidence of top-down control, with small pelagic fishes suppressing zooplankton biomass (Micheli 1999). This effect was, however, driven by data from one study (Shiomoto et al. 1997), and when I re-analyzed the data with Pearson correlation coefficients after applying the Fisher-transformation and then corrected for auto correlation (as in Worm & Myers 2003), the combined effect offish on zooplankton was non-significant. Likewise, in the original study the effect of zooplankton on phytoplankton was non-significant, but there was a significant effect of nutrient supply on phytoplankton biomass. Both consumer and nutrient effects attenuated at the zooplankton - phytoplankton link (Table 5.1; Micheli 1999). This suggests a generally weak herbivore-plant coupling as has been observed in both terrestrial and freshwater ecosystems (Brett and Goldman 1997, Shurin et al. 2002), and in Micheli's (1999)

130 parallel meta-analysis of marine mesocosm experiments. Thus, the data from marine pelagic ecosystems provide no general evidence of top-down control of fish on plankton (but see Daskalov 2002 and Frank et al. 2005 for recent exceptions).

Another meta-analysis, evaluating temporal variation in the abundances of Atlantic cod and a benthic prey species, northern shrimp, across nine regions of the northeast and northwest Atlantic Ocean, did find evidence for top-down control (Table 5.1; Worm & Myers 2003). Time series estimates of cod and shrimp biomass between 1970 and 2000 (minimum 10 years) were obtained from research surveys, population dynamics models, and standardized CPUE data in each area, and the Pearson's correlation coefficient for each predator-prey pair was corrected for autocorrelation as well as measurement error, which can bias correlation coefficients toward zero (Hedges & Olkin 1985). Cod declines in 7 of 9 regions were accompanied by increases in shrimp abundance, and although corrected correlations were statistically significant in only three areas (northern Newfoundland, eastern Scotian Shelf, Barents Sea), the meta-analytic mean across all regions indicated a strong, negative relationship overall ( r - -0.64, P=0.007; Figure 5.2; Worm & Myers 2003). Weakly negative or positive cod-shrimp correlations were only seen in the Gulf of Maine and Skagerrak, close to the southern range limit of both species. It was shown that indeed the predator-prey relationship weakened with increasing temperature (Worm & Myers 2003). In contrast, shrimp populations themselves did not respond in a consistent way to climatic variability, as evidenced by nonsignificant correlations in 8 of 9 regions and overall between shrimp biomass and ocean temperature ( r = -0.24, P=0.094).

Spatial Variation in Predator Abundance

Spatial gradients in predator abundance provide another comparative method of evaluating potential top-down effects in marine ecosystems. Here, areas of contrasting predator abundance can be viewed as the 'treatment levels' in an experiment, and unless directly tested for, the implicit assumption is that environmental conditions are equivalent

131 weignts (%) Labrador 7 N. Newfoundland 4 Flemish Cap 11 N. Gulf of St. Lawrence 4 Eastern Scotian Shelf 4 Gulf of Maine 19 Iceland 16 Barents Sea 26 Skagerrak 8 FE Weighted mean

RE Weighted mean

-0.99 -0.9 -0.5 0 0.5 0.9 0.99 Correlation

Figure 5.2 Meta-analysis of cod-shrimp interactions for nine regions in the North Atlantic Ocean. Circles represent correlation coefficients with 95% CIs. The weighted mean correlations (r ) with 95% CLs are shown as diamonds and were calculated using a fixed-effects (FE) and a random- effects (RE) meta-analysis. Relative weights of individual data sets in the analysis are shown. Adapted from Worm & Myers 2003.

among areas, with the only significant difference being the variation in predator abundance. In a notable example, Estes & Palmisano (1974) capitalized on the 'geographically discordant recovery pattern' (Estes et al. 1998) of sea otters (Enhydra lutris) in western Alaska, and by contrasting inhabited with uninhabited areas were first able to show evidence for a trophic cascade from sea otters to urchins to kelp (Table 5.1).

The 'space-for-time' approach also is being widely applied in studies comparing predator-prey abundances and community structure inside and outside of no-take marine reserves. In temperate rocky reef communities of the Adriatic Sea, for example, comparisons based on underwater censuses, diet data, and prey tethering experiments of

132 two unprotected areas and two protected areas in which densities of sea bream (Diplodus sargus, D. vulgaris) were much higher, provided evidence that these predatory fishes control sea urchin (Paracentrotus lividus, Arbacia lixula) populations (Table 5.1; Guidetti 2006). In turn, sea urchins had the expected impacts on the benthic community: turf-forming and erect-branched macroalgae cover were significantly higher inside protected areas where urchin densities were kept low, while at unprotected sites, overconsumption of macroalgae by grazing sea urchins led to extensive barrens of encrusting coralline algae (Guidetti 2006). Similar spatial comparisons have revealed cascading effects of predator recoveries within marine reserves in several other regions including New Zealand (Shears & Babcock 2002,2003) and eastern African (McClanahan et al. 1999). Although such marine reserve 'experiments' are proving useful for testing hypotheses about top-down control by species with small home ranges, the approach is less suitable for examining effects of large, mobile predators that move across reserve boundaries. Moreover, given the current locations of closely monitored, well protected marine reserves this approach currently is limited for the most part to shallow benthic marine ecosystems.

Seasonal Variation in Predator Abundance

Not all top-down effects are revealed by depletions or recoveries of predators. In ecosystems where predator abundance is not markedly changing across years such effects still maybe observed if there is large seasonal variation in predator abundance, as commonly occurs in highly migratory species. In Antarctica's western Ross Sea, for example, a trophic cascade is apparently triggered each summer when top predators converge at Ross Island (Ainley et al. 2006; Table 5.1, Figure 5.3). Long-term field data from this location revealed that seasonal predation by Adelie penguins (Psgoscelis adeliae) and minke whales (Balaenoptera bonaerensis) locally depleted crystal krill {Euphausia crystallorophias), causing krill-eating Antarctic silverfish (Pleurogramma antarcticum) to become cannibalistic, and penguins and minke whales to prey-switch from krill to silverfish. Together with killer whales (Orcinus orca), seals, and Antarctic, toothfish (Dissostichus mawsoni), these predators depleted silverfish numbers. The subsequent scarcity of prey forced penguins to make longer and more distant foraging

133 tops and the whales to depart the area. Effects from the reduction in krill, the most important phytoplankton consumer in coastal Antarctic waters, cascaded down the food web leaving the diatom community minimally grazed compared to adjacent areas (Ainley et al. 2006). Two 'natural experiments' helped to distinguish between competing hypotheses and confirmed predation as the likely cause of these changes, providing support for top-down control as the driver: first, the predators' effect was accentuated when a short-term polynya (opening in the ice) concentrated them in a confined area; second, prey-switching from krill occurred even when grounded icebergs inhibited mid- season dispersal of krill's sea-ice habitat from the area (Ainley et al. 2006). Annual observations serve as quasi-replicates in ecosystems with seasonal predators; here, the pattern was observed annually for eight years.

4 PREDATOR DEPLETIONS REVEAL TOP-DOWN CONTROL

Cascading Effects of a Massive Fisheries Collapse

Unprecedented declines in northwest Atlantic cod populations in the early 1990s begat one of the most spectacular industrial fisheries collapses (Hutchings & Myers 1994, Myers et al. 1996). This later provided fertile ground for examining the ecosystem effects of predator depletion, and is highlighted as an example here. Cod were the dominant species in these temperate continental shelf ecosystems for centuries (Jackson et al. 2001) and are among the best studied of all fishes. Ecosystem consequences of their functional elimination appear to be extensive, extending far beyond the increases in shrimp populations documented by Worm & Myers (2003). Indeed, despite the ecosystem complexity, rampant omnivory, and high levels of connectance among species (Link 2002), a considerable body of evidence now supports the hypothesis that top-down control has played a role in shaping these ecosystems and that recent ecological changes can largely be attributed to reduced predation mortality (Frank et al. 2005, 2006). Less clear, however, is whether the underlying cause of the changes was the decline of the entire functional group of large benthic predators, or of cod alone. Figure 5.3. Simplified oceanic food webs showing only the species and the direct species interactions involved in the documented trophic cascades: significant predator (thick black arrow), minor predator (thin black arrow); for sharks diet data are often known only at taxonomic levels above species: predator on species within genus (medium grey arrow), predator on species within family (thin pale grey arrow), (a) Polar sea: seasonal trophic cascade occurs as migratory apex predators deplete their prey (Ainley ef a/. 2006); (b) Temperate continental shelf: trophic cascade due to overfishing of cod and other benthic fishes, phase shift from benthic to pelagic dominance, increasing grey seals may now partially inhibit cod recovery (Choi ef a/. 2004; Frank ef a/. 2005; Trzcinski ef al. 2007); c) Temperate coastal, continental shelf & pelagic waters: overfishing apex and large predatory sharks (top two rows) leads to mesopredator release (rays, skates, small sharks), and a trophic cascade through cownose ray down to shellfish (Myers ef al. 2007); d) Coral reef: trophic cascade, as overexploitation of predatory reef fish community leads to outbreaks of coral-eating starfish, phase shift from coral to macroalgae (Dulvy ef al. 2004b).

135 (3) Minke Adelie Killer whales Seals (b) Gray seal whales penguins (type-C) Cod Other large demersal fishes

Snowxrab Small pelagic Northern .^••-Antarctic fishes shrimp f silverfish

Crystal Large herbivorous o krill zooplankton

Crown of thorns starfish

Bonnethead J^flnetooth antic Non-reef Hard,./ builders shark Spiny & shark rosette, sharpnose corals smooth Cownose clearnose Uhain shark (turf algae) devil ray butterfly rays ray skates catshark Coralline algae Oyster 'Bay scallop Hard clam Soft-shell clam Sea grass A fundamental restructuring of the ecological community on the eastern Scotian Shelf, off Nova Scotia, Canada, has been ascribed to overexploitation of cod and other large predatory fishes in the late-1980s and early-1990s (Table 5.1, Figure 5.3; Choi et al. 2004, Frank et al. 2005). Fisheries in this region depleted not only the two dominant predators, cod and haddock (Melanogrammus aeglefinus), but most other large benthic fishes as well, including other Gadiformes (pollock (Pollachius virens), white hake {Urophycis tenuis), silver hake (Merluccius bilinearis)); flatfish (American plaice {Hippoglossoid.es platessoides), yellowtail flounder (Limanda ferrugined)); skates (barndoor {Raja laevis), thorny {R. radiata), winter {R. ocellata)); and redfish {Sebastes spp.) (Choi et al. 2004, Frank et al. 2005). Substantive increases in prey species occurred concomitantly with this massive reduction in overall predation pressure. Time series from standardized research surveys and landings data revealed benthic fish biomass to be inversely correlated with small pelagic fishes (r=-0.61, n=33 years, mainly herring, capelin {Mallotus villosus), and sand lance {Ammodytes dubius)), snow crab (r=-0.70, n=24), and northern shrimp (r=-0.76, n=24), but the data were not corrected for autocorrelation and significance levels of the correlations were not reported (Choi et al. 2004, Frank et al. 2005). Frank et al. (2005) suggest that these effects have cascaded further down the food web: benthic fish showed the expected positive correlation with large (>2mm) herbivorous zooplankton (r=0.45, n=23 years) and negative correlation with phytoplankton (r=-0.72, n=24), and the decline of large zooplankton is consistent with increased size-selective predation by pelagic fishes and early-life stages of crab and shrimp. However, the patchy nature of the plankton data (time series data were unavailable from the early-1970s to early-1990s) precluded comparison of small pelagic fish and macroinvertebrates with these lower trophic levels, and necessitated the use of landings data for the correlations between fish and plankton. Thus, evidence of a trophic cascade in this plankton based marine ecosystem is less convincing than that for prey release of macroinvertebrates and pelagic fish.

Increases in small pelagic fish and macroinvertebrate prey species have followed cod depletions in other areas on the east coast of Canada and the U.S. (Table 5.1). Worm & Myers' (2003) meta-analysis suggests that cod predation controls northern shrimp

137 populations in many of these regions; a related meta-analysis suggests that the same may be true of snow crab populations (Myers et al. unpublished). In the cod-dominated waters around Newfoundland, increased capelin abundance in the 1990s is thought to have resulted from reduced predation by cod rather than changes in food availability or environmental factors (Carscadden et al. 2001). A broader study examining long-term research surveys from nine intensely exploited areas between Georges Bank, New England and the Labrador Shelf found evidence of spatial variation in trophic control. Correlations between the dominant benthic predatory fishes and their prey (primarily small pelagic fishes) tended to be negative in northern areas and either weak or positive in southern areas (Frank et al. 2006). When I combined the correlation coefficients from that study meta-analytically the overall effect size is not statistically significant, but it can be seen that a north-south gradient underlies the data (Figure 5.4). Frank et al. (2006) have suggested that gradients in temperature and species diversity, both of which are inversely related to latitude across these regions, might account for this spatial pattern. First, because temperature-dependent physiology translates into later sexual maturation and lower maximum population growth rates of cod in northern areas (Myers et al. 1997b), those populations might be more vulnerable to overexploitation and thus experience greater declines than southern populations. Second, because cod-dominated northern areas have lower functional diversity, compensatory responses by other large predators following these declines would be limited. Indeed, in northern areas, the depletion of cod was extreme (>85%) and most other large demersal predators also declined significantly, leaving little remaining predation pressure on prey species. In southern areas, despite similar rates of exploitation, cod populations declined less and other large predators showed compensatory increases. These compensatory changes have been well documented on Georges Bank, for example, where declines in cod, haddock, and yellowtail flounder (Limanda ferruginea) to historically low levels were followed by increases in spiny dogfish (Squalus acanthias) and skates (Raja spp.), apparently through competitive release (Fogarty and Murawski 1998). Thus, variation in the magnitude of depletion in the entire benthic predator guild appears to explain latitudinal differences in the strength of top-down effects observed in these continental shelf ecosystems.

138 weights (%) Labrador & N. Grand Banks • i 2 N Gulf of St. Lawrence —• f 14 —•——|— S. Gulf of St. Lawrence 8 1_» St. Pierre Bank 9 Southern Grand Bank —•-—| 23 Eastern Scotian Shelf »-i 11 Gulf of Maine • 12 Western Scotian Shelf 1 • 3 Georges Bank -I—•— 18 FE Weighted mean

RE Weighted mean

-0.99 -0.9 -0.5 0 0.5 0.9 0.99

Correlation

Figure 5.4 Meta-analysis of large benthic fishes - small pelagic fish interactions for nine regions in the Northwest Atlantic Ocean (data from Frank et al. 2006). Circles represent correlation coefficients with 95% CIs. The weighted mean correlations (r ) with 95% CLs are shown as diamonds and were calculated using a fixed-effects (FE) and a random-effects (RE) meta-analysis. Relative weights of individual data sets in the analysis are shown. Meta-analysis conducted as in Worm & Myers 2003.

Restructuring Oceanic Ecosystems: Mechanisms and Effects

Overexploitation of oceanic top predators appears to also have induced top-down effects in many other regions and ecosystems (Table 5.1). Recent ecosystem-scale studies examining the indirect effects of these predator depletions most commonly reveal patterns indicative of top-down control across one trophic link, either mesopredator release, the increased abundance of smaller vertebrate carnivores following declines of their predators (sensu Crooks & Soule 1999), or invertebrate population increases. Some evidence of trophic cascades also is emerging, revealing that predator effects can propagate across multiple trophic links even in reticulate food webs. Here I present the evidence for these three main patterns of food web alterations, noting the inferred underlying mechanisms and methods employed, as well as how exploitation of multiple trophic levels, spatial refuges, and size-structured predation can dampen these effects.

Mesopredator Release

For temperate continental shelf ecosystems evidence continues to accumulate that the depletion of large piscivorous fishes leads to a proliferation of smaller fish species (Table 5.1). Two recent studies used a size-based approach to infer changes in the fish community from long-term scientific research surveys. In the Celtic Sea, where large- scale fisheries expanded fairly recently, research surveys of benthic fish populations have monitored much of the history of industrial exploitation. Between 1987 and 2003, a decrease in the relative biomass of large, predatory fish (100-169g) was associated with a significant increase in the relative biomass of smaller fishes (4-25g) (Blanchard et al. 2005). Although environmental forcing related to the North Atlantic Oscillation is thought to be particularly strong in this ecosystem, this analysis indicated that fishing has had a greater effect on community size structure than climatic variation (Blanchard et al. 2005). Likewise in the North Sea, long-term trawl surveys showed steady and significant increases in the abundance of small fishes (<40cm) over the past thirty years as the larger species declined (Daan et al. 2005). These changes were consistent spatially, occurring throughout much of the North Sea, and across many species groups (Daan et al. 2005). Examination of 87 fish populations from the North Atlantic and Northeast Pacific Oceans in a third study also revealed a pattern of declining demersal species and increasing small pelagic species (Figure 5.5; Hutchings & Baum 2005). Although these studies did not examine specific predator-prey interactions, given the size-based nature of predation among fishes, the observed patterns are in accordance with the hypothesis that reduced predation mortality has triggered compensatory responses in smaller prey species.

On coral reefs, most examples of top-down control involve invertebrates (either urchins (e.g. Hughes et al. 1994) or sea stars (McClanahan et al. 2002)), whereas studies have found no cascading effects of predatory fish removals (Jennings & Polunin 1997, Russ & Alcala 1998). Recently, however, evidence has emerged suggesting that

140 («) (h) 1.2 (•2.0 i.o- 35 • 1.8 30 1.0 0.8- • 1.6 25 0.8 • 1.4 0.6* 1-20 0.6 •1.2 0,4- 15 -1.0 10 4> 0.4 0,2- -0.8 •5 i 0.2 ••0.6 o- « (c) W) > r6 2.0 • (-2.0 1.0 •5 1.5 • • 1.5 0.8 •4 •3 1.0- 1.0 0.6 •2 0.5- \ 0.5 0.4 • 1 •0 o- 0 0.2 1978 1983 1988 1993 1998. 2003 1978 1983 1988 1993 1.998 2003 years

Figure 5.5. Temporal changes in abundance of demersal and pelagic marine fish from four regions in north-temperate oceans between 1978 and 2001, standardized to a value of one for 1978. The relative abundance estimates for demersal species (filled triangles) are scaled according to the left vertical axis of each panel, and for pelagic species (open triangle) according to the right vertical axis in each panel. Number of populations represented in each time-series is as follows: (a) northeast Atlantic (demersal: n=27; pelagic: n=14); (b) northwest Atlantic (demersal: n=23; pelagic: n=2); (c) north mid-Atlantic (demersal: n=13; pelagic: n=2); (d) northeast Pacific (demersal: n=1; pelagic: n=5). Reproduced from Hutchings & Baum 2005.

reductions in large predatory reef fishes can lead to increased numbers of prey fishes (Table 5.1). On the Great Barrier Reef, coral trout (Plectropomus leopardus), a heavily exploited piscivorous grouper, rose significantly in abundance inside no-take marine reserves after 14 years of protection (Graham et al. 2003). Comparisons between protected and fished areas, controlled for differences in habitat structure and complexity, showed a significant inverse correlation between predator and prey biomass with prey densities in fished areas double that of reserves (Graham et al. 2003). This effect was fairly consistent among prey species: eight of nine had higher densities in the fished areas, six of which were statistically significant (Graham et al. 2003). In Fiji, a comparison across a 13-island gradient of subsistence fishing pressure revealed that significant biomass declines in the largest fish size classes (>26cm), resulting from size-

141 selective exploitation, were accompanied by minor increases in biomass (9%) and numbers (31%) of the three smallest fish size classes (1 l-25cm) (Dulvy et al. 2004a). These increases are consistent with a weak compensatory response to reduced predation, . but might also partly reflect competitive release among territorial herbivorous fishes because the larger species like parrotfishes (Scaridae) and surgeonfishes (Acanthuridae) were also removed by fishing (Dulvy et al. 2004a).

Evidence of mesopredator release also has been detected in the open ocean, in response to reduced numbers of the largest predatory fishes, tunas, billfishes, and sharks. In this vast ecosystem where dedicated surveys are prohibitively expensive, standardized catch rates from pelagic longlines may serve as indices offish abundance. A comparison of catch rates in the tropical Pacific Ocean from exploratory research surveys conducted in the early-1950s at the onset of industrial fishing with those collected by scientific observers in the late-1990s, suggested that biomass of all twelve apex predatory fish species had plummeted, often by more than 90% (Table 5.1; Ward & Myers 2005b). Although the exact magnitude of the declines has been disputed (Sibert et al. 2006), it is clear from all studies that the abundance of apex predators has been significantly reduced. Over the same period, abundances of smaller species were either stable, despite exploitation, or showed increases (e.g. snake mackerels (Gempylidae), skipjack tuna (Katsuwonus pelamis)) (Ward & Myers 2005b). In particular, the data for pelagic stingray (Dasyatis violacea) and pomfrets (Bramidae) suggest large increases in their abundance or expansion into habitats previously dominated by their predators (Ward & Myers 2005b).

Compensatory increases in elasmobranch (shark, skate, ray) mesopredators following the removal of larger shark species also has been documented on the U.S. east coast (Table 5.1, Figure 5.3; Myers et al. 2007). Meta-analyses of time-series data from 17 independent research surveys and two fisheries-dependent data sets revealed the functional elimination of 11 great shark (>2m) species over the past 35 years, along with four- to ten-fold increases in 12 of 14 of their elasmobranch prey species (Myers et al. 2007). The elasmobranch mesopredators affected by this community restructuring

142 occupy pelagic and demersal marine habitats from inshore estuaries out to the continental shelf and slope, illustrating that large mobile predators can couple the dynamics of spatially segregated food webs (McCann et al. 2005). These apparent changes occurred despite the common occurrence of omnivory and intraguild predation by the great sharks (Figure 5.1d,e). Great sharks consume a wide range of species from cephalopods and bony fishes to marine mammals (Compagno 1984, 2001) and for many of these predators elasmobranchs comprise but a small proportion of their diets (Cortes 1999). Importantly though, these species are the major predators of the smaller elasmobranchs (Springer 1967, Heithaus 2004), and thus together could exert strong top-down control on them. Moreover, because shark diets change ontogenetically, with elasmobranch consumption increasing with size, the size-selective nature of fisheries may exacerbate these effects (Lucifora et al. in review). Finally, consumption of juvenile great sharks by the true apex predators within this group (Figure 5.3) did not dampen the pattern of mesopredator release because these species too were reduced by exploitation (Myers et al. 2007).

Although fisheries removals of predators can trigger changes in prey populations, exploitation exerted simultaneously on predator and prey species may obscure this response. For example, in the northern Gulf of Mexico shrimp trawling yields high levels of bycatch including small benthic elasmobranchs. Spatial variation in trawling effort appears to explain the marked differences in the response of these elasmobranchs to depletions of great sharks, their main predators. Research surveys conducted on the continental shelf between 1972 and 2002 showed that despite substantial great shark declines, increases occurred only in Atlantic angel shark (Squatina dumerit), smooth dogfish (Mustelus canis), and spreadfin skate (Dipturus olseni), the three species whose deep-water habitat provided a refuge from the shrimp fishery (Table 5.1; Shepherd & Myers 2005). In contrast, in shallow coastal waters where shrimp trawling is intense and the major source of fishing mortality for small elasmobranchs (Cortes 2002), these mesopredators, including cownose ray, smooth butterfly ray (Gymnura micrura), and Atlantic sharpnose shark, all declined (Shepherd & Myers 2005). The same species showed tremendous increases following great shark declines on the U.S. east coast, where shrimp trawling is less common (Myers et al. 2007). The divergent responses of

143 shallow and deep-water species demonstrate how fishing mortality can override responses to predation-release.

Invertebrate Explosion

Responses of invertebrates to predator depletions comprise some of the best known examples of top-down control in marine ecosystems. The literature is replete with examples from nearshore benthic ecosystems of sea urchin population explosions following depletions of their predators (see Pinnegar et al. 2000, Steneck & Sala 2005 for recent reviews). Likewise, increases in shrimp, crabs, and possibly lobster, following cod depletions suggest responses by invertebrate species also are important on offshore continental shelf ecosystems (Worm & Myers 2003, Frank et al. 2005, Myers et al. unpublished). In contrast, the underlying causes of pelagic invertebrate increases are difficult to examine rigorously due to inadequate data (Hay 2006). It has been speculated that the global rise in cephalopod landings between 1974 and 1994 might reflect increases in abundance due to reduced predation pressure (Caddy & Rodhouse 1998), while population 'explosions' of jellyfish have been variously attributed to climatic changes (Mills 2001, Attrill et al. 2007) or loss of predatory control (Daskalov 2002, Lynam et al. 2006).

A dynamic mass-balance ecosystem model (Ecosim, see Pauly et al. 2000) of the central North Pacific Ocean supports the notion that rising numbers of squid in the pelagic ocean may be the result of diminished predation mortality. According to the model, which 'reconstructed' this pelagic food web prior to commercial whaling and fishing, the most perceptible changes in the ecosystem have been marked declines in the abundance of apex predators and associated increases in large squid and epipelagic fishes (Table 5.1; Cox et al. 2002, Essington 2007). Increases in large squid (e.g. red flying squid {Ommastrephes bartramii), purpleback squid (Sthenoteuthis oualaniensis), common clubhook squid (Onchyteuthis banksiss)), which are themselves apex predators (i.e. trophic level > 4), were attributed to reduced predation primarily by sperm whales (Physeter catodon, trophic level=4.7), but also by swordfish (Xiphias gladius) and blue sharks (Prionace glauca). These cephalopods may now constitute three-quarters of the apex predator biomass in this ecosystem (Essington 2007). If this was true, the total

144 biomass of the apex predator guild actually would have declined only slightly, but its composition shifted from dominance by long-lived, moderately productive species to short-lived, highly productive ones (Essington 2007).

Trophic Cascades

In contrast to Micheli's (1999) meta-analysis, a more recent study indicates that trophic cascades can occur in planktonic-based marine food webs: ecosystem changes in the Black Sea can be partially explained by a trophic cascade that was initially triggered by overexploitation of dolphins and large pelagic fishes. Time-series data spanning three decades (-1960-1990) showed significant inverse trends in abundance across four trophic levels, from piscivores to phytoplankton (Daskalov 2002). Pronounced declines in abundance of the dominant piscivores (bonito (Sarda sardd), mackerel {Scomber scombrus), and bluefish (Pomatomus saltator)) by the early-1970s may have led to increased abundances of anchovy, sprat, and horse mackerel. Together with increasing native and invasive jellyfish populations, these planktivores reduced zooplankton biomass, leading to a doubling of phytoplankton biomass (Daskalov 2002). Simulations of this ecosystem over the same thirty year time span, using Ecosim mass-balance models, reproduced the observed cascading patterns best when increases in productivity, reflecting known anthropogenic eutrophication, were combined with intense fisheries exploitation and associated top-down effects (Daskalov 2002).

Altered predation by killer whales, the oceans' true apex predator, may have triggered a trophic cascade spanning the entire length of an algal-based food web. Notwithstanding controversy about the exact cause of shifting killer whale predation (DeMaster et at 2006; Wade et at 2007), multiple lines of evidence point to this increased predation as the cause of recent declines in west Alaskan sea otters, which have reversed the classic sea otter - urchin - kelp trophic cascade (Table 5.1; Estes et at 1998, Springer et at 2003). Bibenergetics modelling suggests that predation by even a few killer whales would have been sufficient to control the sea otter population (Estes et at 1998, Williams et at 2004).

145 In other cases, trophic cascades stemming from apex predators may involve multiple levels of consumers without necessarily affecting primary producers. On the U.S. eastern seaboard, elasmobranch mesopredator release associated with the loss of great sharks cascaded from one species, cownose ray (Rhinoptera bonasus), downwards to its bay scallop prey (Table 5.1, Figure 5.3; Myers et al. 2007). Whereas field sampling in North Carolina in the early-1980s showed no impact of the rays on bay scallops (Peterson et al. 1989), analogous sampling since the ray increase, confirmed by four years of controlled, replicated ray-exclusion experiments, revealed migrating cownose rays as the cause of almost complete scallop mortality each fall (Peterson et al. 2001, Myers et al. 2007). Overexploitation of great sharks was indirectly damaging to a second commercial fishery, as intense ray predation led to the closure of North Carolina's century-old scallop fishery by 2004 (Myers et al. 2007). This cascade may possibly extend further to negatively impact primary producers, not through predation, but rather through physical disturbance of seagrass habitat as cownose rays intensify their foraging efforts for infaunal bivalves (Orth 1975, Myers et al. 2007).

The ecosystem role of predators on coral reefs is poorly understood, but a recent large-scale field study in Fiji suggests that predatory reef fishes can inhibit starfish- mediated trophic cascades (Table 5.1, Figure 5.3). Across a 13-island fishing gradient, a 61% decline in predatory reef fish density was linked to a three-order of magnitude increase in the coral-eating crown-of-thorns starfish (Acanthaster planci) (Dulvy et al. 2004b). Starfish predation reduced reef-building corals and coralline algae along the gradient, facilitating their replacement by non-reef building taxa like filamentous algae (Dulvy et al. 2004b). Predatory control of starfish, which are a major management problem on Indo-Pacific reefs (McClanahan et al. 2002), implies that large carnivorous fishes may positively affect coral reef structure and function.

Not surprisingly, as in other ecosystems (Strong 1992, Pace et al. 1999), refuges from predation can suppress trophic cascades in the ocean. Rocky reef communities of the Mediterranean, for example, exhibit alternate states, coralline barrens dominated by urchins and erect macroalgal stands with low urchin and high predator densities (Sala et al. 1998). While increased sea bream predation can promote a trophic cascade by reducing urchin numbers and shifting the system back to its macroalgal state (Guidetti 2006), spatial heterogeneity modulates its effectiveness. In heterogeneous areas, where shelters are abundant, urchins coexist with their diurnal predators and still maintain coralline barrens by hiding in shelters during the day and foraging at night (Sala et al. 1998). Prey size refugia, in which large prey escape predation, also may limit predatory control of populations, as demonstrated by a recent study contrasting predator-prey dynamics inside and outside a long-term no-take marine reserve in the Caribbean (Mumby et al. 2006a). A concern with coral reef reserves is that while protection from fishing may initially benefit herbivores, as predator populations rebuild and lost predator- prey interactions are restored (Micheli et al. 2004), grazing, a critical mediator of competition between coral and macroalgae, could become impaired (Mumby et al. 2006a). These anticipated cascading effects of predator recovery were, however, not realized within the Caribbean reserve, where parrotfishes are the dominant grazers, despite a doubling of parrotfish predators because large-bodied parrotfish escaped gape- limited predation by the dominant piscivore, Nassau grouper (Epinephelus striatus) (Mumby et al. 2006a). Instead, the density of overexploited large parrotfishes actually increased leading to a net doubling of grazing within the reserve (Mumby et al. 2006a). Theoretical models reinforce these findings, suggesting that trophic cascades will be inhibited in reserves where fishing mortality on herbivores before protection exceeds predation mortality after protection (Baskett 2007). Predator recoveries thus may have unexpected outcomes, and because of these ecological complexities even detailed, comprehensive food web models may yield inaccurate predictions about strong interactions and the potential for trophic cascades (Bascompte et al. 2005). The recovery of apex predators like sharks, which can feed on both predator and prey species, further complicates predictions about these food web dynamics and their effect remains unresolved (Chapman et al. 2006, Mumby et al. 2006b).

147 5 MANAGEMENT & CONSERVATION IMPLICATIONS

Top-down control of oceanic ecosystem structure implies that key predator-prey interactions must be understood and explicitly accounted for if species conservation and fisheries management plans are to have their intended outcomes. Contrary to the perception that effects on community structure following predator depletions are weak (Jennings & Kaiser 1998), industrial fisheries collapses have led to major ecosystem changes in several areas, which do not seem easily reversible (e.g. Daskalov 2002, Frank et al. 2005). The perception of 'weak effects' may, however, reflect that fact that effects of overexploiting top predators commonly cascade only to single species or species guilds rather than to entire oceanic food webs, and when poorly monitored, non commercial species are the only ones impacted, effects most likely go unnoticed. Conversely, increases in commercial species are almost always tracked by fisheries (Steneck 1998). For example, subsequent to the early-1990s cod collapse in Atlantic Canada, an offshore herring fishery developed on the central Scotian Shelf (Choi et al. 2004) and snow crab and lobster landings increased dramatically (Worm & Myers 2003, Frank et al. 2005). Similarly, in the Black Sea industrial fisheries began targeting anchovy and sprat after these planktivorous fishes rose in numbers (Daskalov 2002). Increases in prey species thus may be reversed due to changes in fisheries targeting. Cascading effects of predator depletions can also negatively impact commercial fisheries as the U.S. east coast example shows, where depletion of great sharks indirectly led to the termination of a century-long bay scallop fishery (Myers et al. 2007). Predicting how overexploitation of top predators will impact other fisheries is challenging, especially since this disturbance does not occur in isolation, but rather combines with natural variability as well as other anthropogenic perturbations that alter trophic interactions including simultaneous exploitation of multiple trophic levels, climate change (Perry et al. 2005), and species invasions. What is clear, however, is that effects of overfishing oceanic predators can cascade to lower trophic levels, and thus single-species fisheries management, which ignores species interactions, appears increasingly unwise. How stable these new ecosystem configurations are ecologically is currently an active area of research. There is some evidence that a predation-mediated feedback is a factor limiting recovery of collapsed Atlantic cod populations. In a size-based predator- prey reversal, increased predation of cod eggs and larvae by now abundant herring and mackerel may critically limit cod recruitment in the southern Gulf of St. Lawrence. Swain and Sinclair (2000) showed that cod recruitment there is strongly inversely related to herring and mackerel biomass, and indeed cod recovered slowly in the 1990s when those small pelagics were abundant, but rapidly from similarly low levels of abundance in the mid-1970s, which coincided with the collapse of herring and mackerel fisheries. Similar interactions between cod and small pelagic fishes have been observed in the Baltic Sea (Koster and Mollmann 2000) and on Georges Bank (Garrison et al. 2000). Walters & Kitchell (2001) termed this dynamic 'cultivation/depensation' hypothesizing that as the dominant species, cod 'cultivate' favorable conditions for the survival of its juveniles by 'cropping down' small pelagic fishes that are important predators or competitors. Conversely, once cod are severely depleted, increases in abundance of its prey can cause depensatory decreases in juvenile survivorship.

Grey seals (Halichoerus grypus) also have been implicated in the cod recovery failure, and for the eastern Scotian Shelf where the seal population is increasing exponentially, population dynamics and bioenergetics models based on diet and distribution data, and thirty years of abundance data support this mechanism (Figure 5.3; Trzinski et al. 2007). Detrimental impacts on commercially important species, whether real or perceived, have frequently led to calls to cull apex predators, despite the lack of evidence that past culls have benefited fisheries (see Yodzis 2001). Direct negative effects of seal predation on cod may be outweighed by indirect positive effects resulting from seal predation on herring. This makes it difficult to predict whether a seal cull would adversely or beneficially influence cod recovery (Swain & Sinclair 2000).

Predation by apex predators also can have conservation implications, by impeding or reversing recoveries of threatened species. For example, since being hunted to near extinction in the 19th century, fur seal (Arctocephalus gazella) recovery in the South

149 Shetland Islands has lagged behind that of the South Georgia population (Boveng et al. 1998). A nine-year comparison of fur seal abundance and reproduction between areas with and without leopard seals (Hydrurga leptonyx) revealed incidents of leopard seal predation as well as much higher fur seal pup mortality in the presence of this apex predator, suggesting that leopard seals may inhibit recovery of the population (Table 5.1; Boveng et al. 1998). In other regions, sharks might exert control over pinniped population dynamics. Researchers have suggested that population growth of endangered Hawaiian monk seal (Monachus schauinslandi) in the Northwestern Hawaiian Islands (Bertilsson-Friedman 2006) and harbour seal (Phoca vitulina) on Sable Island, Nova Scotia (Lucas & Stobo 2000) might be suppressed by shark-inflicted injuries and mortalities. Finally, as already discussed, strong circumstantial evidence implicates killer whale predation in the reversal of west Alaskan sea otter recovery (Estes et al. 1998). This can lead to a management dilemma, where the conservation of one top predator needs to be weighed against that of another.

6 SYNTHESIS & CONCLUSIONS

New ecosystem-scale studies documenting top-down effects from polar seas to the tropical open oceans indicate that predation can be an important controlling force in oceanic food webs. There is evidence of top-down control exerted by a range of taxa, including marine mammals (whales, seals), elasmobranchs (sharks, rays), and large teleost fishes (swordfish, tunas, cod and other demersal species, grouper), in many geographic areas (Black, Mediterranean, and North Seas; northwest and northeast Atlantic, northeast Pacific, and Southern Oceans), and in almost all types of oceanic ecosystems (Table 5.1).

Two factors have facilitated detection and understanding of top-down control in the ocean. First, commercial hunting and fishing over the past century led to unprecedented depletions of marine predators, and these removals, and in some cases subsequent recoveries, provided sufficient variation in predation for top-down control to be detected. Second, monitoring of oceanic community dynamics over large temporal and spatial

150 scales allowed observation and statistical scrutiny of top-down effects. As Estes et al. (1998) stated, such 'events could not [be] chronicled or even detected in a short-term study, were unanticipated, and thus seem poorly suited for analysis by a priori hypothesis testing' emphasizing 'the need for large-scale approaches in ecological research'. Indeed, ecosystem-scale spatial and temporal contrasts of predator abundances and associated community dynamics with predator-prey 'replicates' and 'controls' for alternative factors are proving to be powerful tools for examining top-down control in vast oceanic ecosystems not amenable to manipulative experiments. It is critical that these robust correlative and comparative methods, with all their inherent uncertainty, gain acceptance in ecology given the rate at which oceanic predators are being depleted and the potential for subsequent restructuring of oceanic food webs.

Cascading effects of predator losses have mostly been unanticipated and surprising, but as evidence of them accumulates some generalities are emerging. Top-down effects may be triggered by keystone predators, like killer whales, or in low diversity systems by dominant species like cod. In more diverse ecosystems, functional diversity of predators seems to provide insurance against top-down effects, in accordance with ecological theory. Non-selective fisheries that deplete whole predator guilds are, however, weakening this functional redundancy, and emerging evidence suggests that elimination of entire functional groups of predators is a mechanism by which top-down effects are triggered in the ocean. Contemporary anthropogenic perturbations can thus render previously resilient oceanic ecosystems susceptible to destabilizing cascading effects. Common responses to predator depletions, indicative of the loss of top-down control, include increases in mesopredatory fishes and invertebrates. As a consequence oceanic communities are becoming increasingly dominated by smaller, shorter-lived species with higher turnover rates (Jennings & Blanchard 2004). The few examples of oceanic trophic cascades suggest that such effects typically propagate through several trophic levels of predators and that apex predators can couple pelagic and benthic food web dynamics. Whether cascades generally will impact primary producers or instead attenuate at the herbivore - plant link remains to be seen, but even with this attenuation oceanic trophic

151 cascades can still cause massive food web restructuring. It is unclear if these ecosystem changes are reversible, and there is some evidence for stabilizing feedbacks.

The limited number of documented oceanic trophic cascades may reflect data limitations rather than evidence of absence. In Myers et al. 's (2007) study, for example, a trophic cascade was traced through the cownose ray, but consequences of increased abundances of 11 other elasmobranch mesopredators remain unknown. As others have suggested (Pace et al. 1999, Pinnegar et al. 2000), variation in intensity of study amongst ecosystem types, top predator species, and regions, hints that there are more examples of top-down control in the ocean still to be detected.

The ultimate goal of studying oceanic top-down control is to gain predictive capacity in order to prevent or effectively mitigate cascading effects of predator losses. In this vein, the retrospective studies reviewed here provide glimpses into the operation of top- down control and the mechanisms triggering cascading effects in the ocean. Through these studies we may better be able to anticipate and avoid deleterious effects, or at least more effectively manage ecosystems that have been altered by such indirect effects. This is not to say that there will always be cascading effects of oceanic predator removals, and certainly in many ecosystems bottom-up control may predominate (Frank et al. 2006, 2007). Instead, this is a call for increased consideration of top-down control in the ocean. The challenge henceforth will be to disentangle the combined effects of top-down and bottom-up control (Hunter & Price 1992, Hunt & McKinnell 2006, Litzow & Lorenzo 2007) in oceanic ecosystems impacted by exploitation, nutrient loading, and climate change. For the foreseeable future, however, oceanic predator depletions will continue to outpace our ability to predict ecosystem responses to their losses and given the potential for cascading effects to ensue it would be precautionary to maintain or restore predator populations above thresholds for ecological effectiveness (Soule et al. 2005).

152 CHAPTER 6

Conclusions

153 Fisheries collapses and unprecedented declines in abundance of a number of marine fishes (Dulvy et al. 2003, Hutchings & Reynolds 2004) have raised concerns and sparked debate about whether exploitation might be jeopardizing the long-term persistence of some populations (e.g. Myers & Worm 2003, Safina et al. 2005, Murawski et al. 2007, Hilborn 2007a,b). The potential for cascading effects of predator depletions in oceanic food webs, including increased abundances of their prey species and trophic cascades, is also now widely recognized (Botsford et al. 1997, Jennings & Kaiser 1998, Duffy 2002, Pikitch et al. 2004), but since evidence of these effects has been sparse the importance of top-down control in the ocean has remained unclear. This thesis focused on a group of large pelagic sharks that are vulnerable to overexploitation yet intensely exploited globally (Musick et al. 2000, Fowler et al. 2005, Clarke et al. 2006), and sought to estimate changes in their population abundances in the Northwest Atlantic Ocean, and to explore the broader community-level effects of declining numbers of these and other oceanic top predators. Here I briefly reiterate the approach used and key findings of the four research chapters and propose some future avenues for this research.

In the first half of this thesis, shark catch rates from several different fishery- dependent and research survey datasets were modelled using generalized linear models (and extensions thereof) to standardize for temporal, spatial, and operational differences among observations and hence obtain estimates that may be regarded as indices of relative abundance. Species-specific estimates were obtained from the research surveys, but poor taxonomic resolution of the fishery-dependent data necessitated analysis at the genus level for many sharks.

Chapter 2 provided evidence of significant declines in the abundance of oceanic whitetip {Carcharhinus longimanus), silky (C. falciformis), and dusky (C. obscurus) sharks in the Gulf of Mexico since the mid-1950s. The motivation for this chapter was to try to understand how shark abundances have changed since the onset of pelagic longline fishing in the open ocean. Because sharks catches were not systematically recorded in fishery-dependent data until recently, there are, to the best of my knowledge, no data sources for these species from any region, which extend continuously back to the 1950s.

154 Instead, in this chapter I compared shark catch rates from exploratory longline research surveys conducted between 1954 and 1957 in the Gulf of Mexico with those recorded by observers monitoring the recent U.S. pelagic longline fleet in the same region. Rather than trends over time then, models estimates from these data essentially provide a snapshot of each time period. Although the primary purpose of modelling these data was to account for sampling differences between the two time periods, not all gear modifications (e.g. leader material and hook type) could be modelled. Refinement of the model estimates in this chapter will require further data quantifying how these differences in fishing gear affect shark catch rates. Fine-tuned models may affect estimates for silky, dusky, and mako sharks to some extent, but I expect that they will have minimal impact on oceanic whitetip shark estimates given its very low catch rate in the recent period. An obvious alternative to fine-tuning these models is to resample the area using the exact methods as the initial surveys, which would provide a robust contrast to the past community of large pelagic fishes in the Gulf of Mexico. At present, however, the models suggest that the oceanic whitetip shark has declined by more than 99%, far greater than the amount estimated from recent data alone. The depletion of this species appears to be an example of shifting baselines since there is little recognition today of its former prevalence in the Gulf of Mexico.

In contrast to Chapter 2, Chapter 3 examined recent trends in abundance for large pelagic sharks, and did so over a larger geographic area of the Northwest Atlantic using observer data from the U.S. pelagic longline fishery. This research complements my previous study, which estimated trends in abundance for sharks using the logbook data from 1986 to 2000 for the same fishery (Baum et al. 2003). Although logbook data are typically considered to be of questionable reliability because they are self-reported by fishermen, comparisons of shark catch rates between the logbook and observer data showed them to be fairly similar for each species (group) within the different subareas fished and overall. To address non-independence in the observer data, three different model types were fitted, including generalized linear mixed models that accounted for correlations among sets made on the same trip and those made by the same vessel. Trend estimates were similar to those from the logbook data for some species and substantially

155 different for others, but differences were not consistent in one direction or the other. Both hammerheads (Sphyrna spp.) and a group of large coastal sharks (dusky, silky, night) are estimated to have declined by three-quarters since 1992, while estimated declines in blue, mako, and oceanic whitetip sharks were less. Models suggest that the abundance of thresher sharks (Alopias spp.) was stable and that of tiger shark {Galeocerdo cuvier) increased between 1992 and 2005, but individual year estimates suggest that they first declined, indicating that these species may have reached minimum levels in the mid-1990s and that further modelling is needed to capture their non-constant rate of change in this period and to avoid fallacious conclusions.

Trend estimates in Chapter 4 were based primarily on research surveys, of which the UNC shark-targeted longline survey provided the most data for large sharks. Conducted annually since 1972 off the coast of North Carolina, this survey data showed precipitous declines in the abundance of seven shark species, ranging from 87% for sandbar shark (C. plumbeus) to 97% or more for tiger, scalloped hammerhead {Sphyrna lewini), bull (C leucas), and dusky sharks. Significant declines also occurred in the average length of the sharks caught. Of the other research surveys examined for large sharks, six significant trends were estimated, all but one of which was a decline.

Considered together, model estimates from this thesis (as well as those from previous studies, e.g. Musick et al. 1993, Baum etal. 2003) present consistent evidence that the abundance of large pelagic sharks in the Northwest Atlantic Ocean has decreased enormously over the past two to four decades. There is evidence from the observer data that some of these species may now be stabilizing or starting to increase, but these recent changes appear almost trivial when viewed in the context of the longer term declines. This highlights the danger of assessments based on recent data alone, which can obscure long-term changes such that we become complacent about species that are now rare. For the sharks examined herein, it seems clear that more stringent management measures are required to safeguard and rebuild their populations, especially considering that even in the absence of fishing pressure, most of these species would take decades to recover given their life history characteristics.

156 The latter half of the thesis examines community-level consequences of overexploiting sharks and other top predators. Chapter 4 centered on the hypothesis that declines of apex and near-apex predatory sharks would most likely impact their elasmobranch prey. To test this hypothesis on the U.S. east coast, we utilized extensive empirical evidence including 17 independent research surveys, fisheries and landings data, as well as field observations conducted over twenty years and four years of controlled, replicated predator exclusion experiments. Models of these data indicate that as all eleven elasmobranch-consuming great sharks plummeted over the past 35 years, the abundance of twelve of fourteen of their ray, skate, and shark prey species increased significantly. Consequences of the elasmobranch-mesopredator proliferation have cascaded down the food web from one such species, cownose ray {Rhinoptera bonasus), to bay scallops (Argopecten irradians), causing the collapse and closure of North Carolina's scallop fishery. This research provides evidence of an apparent trophic cascade initiated with the removal of sharks, and of regional-scale elasmobranch community restructuring. It also suggests that rather than attenuating in complex marine food webs (Strong 1992, Jennings & Kaiser 1998), top-down effects can still occur, especially when entire functional groups are removed as is often the case with industrial fisheries.

Moving beyond this case study, I synthesized the recent literature to explore the general evidence for top-down control and cascading effects of predator removals (and recoveries) in the ocean. Recent studies assessing large-scale spatial and temporal contrasts of predator abundances and their associated communities have provided evidence of top-down control exerted by a diverse group of predators in coastal, continental shelf, and open ocean ecosystems. Top-down effects may be triggered by the loss of single species, but industrial fisheries also facilitate these effects by exploiting functional groups of predators. Increases in mesopredatory fishes and invertebrates were common responses to predator depletions; whether the paucity of trophic cascades reflects true rarity or the difficulty of documenting interactions between oceanic species in 'tangled' oceanic food webs remains to be seen. And while oceanic cascades seem to propagate through multiple levels of predators unlike classic 'predator-herbivore-plant'

157 trophic cascades, it is unclear if attenuation before primary producers is typical or not. Certainly, there is still much to be understood about the nature and prevalence of top- down control in the ocean, including disentangling the relative importance of top-down and bottom-up effects (Frank et al. 2006, 2007), and I expect this will be an active research area in the years to come given its real-world significance.

Overall, this research provides quantitative evidence from multiple data types and sources of substantial declines in large pelagic sharks in the Northwest Atlantic Ocean, documents cascading effects of the function elimination of these species through the elasmobranch community, and synthesizes emerging evidence from the literature to show that top-down control may be more prevalent in the ocean than has been appreciated. The former underscores the need for improved management and conservation measures for shark populations, while the latter calls for improved understanding of predator-prey interactions in oceanic food webs and increased attention to the indirect effects of overexploiting the ocean's top predators.

158 References

Abercrombie, D.L., Clarke, S.C. & Shivji, M.S. (2005). Global-scale genetic identification of hammerhead sharks: Application to assessment of the international fin trade and law enforcement. Conservation Genetics, 6, 775-788.

Aebischer, N.J., Coulson, J.C. & Colebrook, J.M. (1990). Parallel long-term trends across four marine trophic levels and weather. Nature, 347, 753-755.

Ainley, D.G., Ballard, G. & Dugger, K.M. (2006). Competition among penguins and cetaceans reveals trophic cascades in the Western Ross Sea, Antarctica. Ecology, 87, 2080-2093.

Aldebert, Y. (1997). Demersal resources of the Gulf of Lions (NW Mediterranean): impact of exploitation on fish diversity. Vie et Milieu, 47, 275-284.

Anderson, E.D. (1990). Estimates of large shark catches in the Western Atlantic and Gulf of Mexico, 1960-1997. In: Elasmobranchs as living resources: Advances in the biology, ecology, systematics, and the status of the fisheries. Proceedings of the United States-Japan Workshop (2nd), (eds Pratt, H.L., Gruber, S.H. & Taniuchi, T.), pp. 443-454. National Marine Fisheries Service, Silver Spring, MD (USA).

Anon. (2002a). Final rule for shark finning prohibition act. www.nmfs.noaa.gOv/sfa/hms/hmsdocuments.html#shark, Highly Migratory Species NOAA Fisheries.

Anon. (2002b) Effects of Trawling and Dredging on Sea-floor Habitat. The National Academies Press, Ocean Studies Board, Washington, D.C.

159 Anon. (2005). North Carolina Fishery Management Plan, Shrimp. North Carolina Department of Environment & Natural Resources, Division of Marine Fisheries, Morehead City, NC.

Anon. (2006) Management plan for the shellfish fishery sector, State of Rhode Island & Providence plantations. Department of Environmental Management, Division of Fish and Wildlife, Marine Fisheries.

Attrill, M.J., Wright, J. & Edwards, M. (2007). Climate-related increases in jellyfish frequency suggest a more gelatinous future for the North Sea. Limnology and Oceanography, 52, 480-485.

Bacheler, N.M. & Buckel, J.A. (2004). Does hook type influence the catch rate, size, and injury of grouper in a North Carolina commercial fishery? Fisheries Research, 69, 303-311.

Backus, R.H., Springer, S., & Arnold Jr., E.L. (1956). A contribution to the natural history of the white-tip shark, Pterolamiops longimanus (Poey). Deep-Sea Research, 3, 176-188.

Balmford, A. & Bond, W. (2005). Trends in the state of nature and their implications for human well-being. Ecology Letters, 8, 1218-1234.

Banse, K. (2007). Do we live in a largely top-down regulated world? Journal of Biosciences, 32, 791-796.

Barker, M.J. & Schluessel, V. (2005). Managing global shark fisheries: suggestions for prioritizing management strategies. Aquatic Conservation of Marine and Freshwater Ecosystems, 15, 325-347.

Barrett, O. W. 1928. Shark Fishing in the West Indies. Scientific Monthly, 27:125-133.

Bascompte, J., Melian, C.J. & Sala, E. (2005). Interaction strength combinations and the overfishing of a marine food web. Proceedings of the National Academy of Sciences of the United States of America, 102, 5443-5447.

Baskett, M.L. (2007). Prey size refugia and trophic cascades in marine reserves. Marine Ecology Progress Series, 328, 285-293.

Baum, J.K. (2002). Determining the effects of exploitation on shark populations using fishery-dependent data. M.Sc. thesis, Department of Biology, Dalhousie University, Halifax, N.S.

160 Baum, J.K. & Myers, R.A. (2004). Shifting baselines and the decline of pelagic sharks in the Gulf of Mexico. Ecology Letters, 1,135-145.

Baum, J.K., Myers, R.A., Kehler, D.G., Worm, B., Harley, S.J. & Doherty, P.A. (2003). Collapse and conservation of shark populations in the Northwest Atlantic. Science, 299, 389-392.

Baum, J.K., Kehler, D. & Myers, R.A. (2005). Robust estimates of decline for pelagic shark populations in the northwest Atlantic and Gulf of Mexico. Fisheries, 30, 27- 30.

Beerkircher, L., Cortes, E. & Shivji, M. (2002). Characteristics of shark bycatch observed on pelagic longlines off the Southeastern United States, 1992-2000. Marine Fisheries Review, 64, 40-49.

Beerkircher L., Shivji, M. & Cortes, E. (2003). A Monte Carlo demographic analysis of the silky shark (Carcharhinus falciformis): implications of gear selectivity. Fishery Bulletin, 101,168-174.

Beerkircher, L.R., Brown, C.J., Abercrombie, D.L. & Lee, D.W. (2004). SEFSC Pelagic observer program data summary for 1992-2002., NOAA Technical Memorandum NMFS-SEFSC-522. Southeast Fisheries Science Center, Miami, 25pp.

Berger, J., Stacey, P.B., Bellis, L. & Johnson, M.P. (2001). A mammalian predatory-prey imbalance: grizzly bear and wolf extinction affect avian neotropical migrants. Ecological Applications, 11,947-960.

Berkeley, S.A. & Campos, W.L. (1988). Relative abundance and fishery potential of pelagic sharks along Florida's east coast. Marine Fisheries Review, 50, 9-16.

Bertilsson-Friedman, P. (2006). Distribution and frequencies of shark-inflicted injuries to the endangered Hawaiian monk seal (Monachus schauinslandi). Journal of Zoology, 268, 361-368.

Bishop, J. (2006). Standardizing fishery-dependent catch and effort data in complex fisheries with technology change. Reviews in Fish Biology and Fisheries, 16, 21- 38.

Blanchard, J.L., Dulvy, N.K., Jennings, S., Ellis, J.E., Pinnegar, J.K., Tidd, A. & Kell, L.T. (2005). Ecological and environmental factors influence size-based metrics of Celtic Sea fish communities. ICES Journal of Marine Science, 62, 405-411.

161 Blaylock, R.A. (1993). Distribution and abundance of the cownose ray, Rhinoptera bonasus, in Lower Chesapeake Bay. Estuaries, 16, 255-263.

Bonfil, R. 1994. Overview of world elasmobranch fisheries. FAO Fisheries Technical Paper 341. FAO, Rome, 119pp

Borer, E.T., Seabloom, E.W., Shurin, J.B., Anderson, K.E., Blanchette, C.A., Broitman, B., Cooper, S.D. & Halpern B.S. (2005). What determines the strength of a trophic cascade? Ecology, 86, 528-537.

Botsford, L.W., Castilla, J.C. & Peterson, C.H. (1997). The management of fisheries and marine ecosystems. Science, 277, 509-515.

Boveng, P.L., Hiruki, L.M., Schwartz, M.K. & Bengtson, J.L. (1998). Population growth of Antarctic fur seals: limitation by a top predator, the leopard seal? Ecology, 19, 2863-2877.

Branstetter, S. (1987a). Age, growth and reproductive biology of the silky shark, Carcharhinus falciformis, and the scalloped hammerhead, Sphyrna lewini, from the northwestern Gulf of Mexico. Environmental Biology of Fishes, 19, 161-173.

Branstetter, S. (1987b). Age and growth-estimates for blacktip, Carcharhinus limbatus, and spinner, C. brevipinna, sharks from the northwestern Gulf of Mexico. Copeia, 4, 964-974.

Branstetter, S. (1999) The management of the United States Atlantic shark fisheries. In: Case studies of the management of elasmobranch fisheries (ed. Shotton, R), pp. 109-145. FAO Fisheries Technical Paper, Rome.

Branstetter, S. (2002). Hammerhead sharks. Family Sphyrnidae. In: Bigelow and Schroeder's Fishes of the Gulf of Maine (eds. Collette, B.B. & MacPhee, G.K.), pp. 45-47. Smithsonian Institution Press, Washington.

Branstetter, S. & Musick, J. A. (1993). Comparisons of shark catch rates on longlines using rope/steel (Yankee) and monofilament gangions. Marine Fisheries Review, 55, 4-9.

Brashares, J.S., Arcese, P., Sam, M.K., Coppolillo, P.B., Sinclair, A.R.E. & Balmford, A. (2004). Bushmeat hunting, wildlife declines, and fish supply in West Africa. Science, 306,1180-1183.

Brett, M.T. & Goldman, C.R. (1997). Consumer versus resource control in freshwater pelagic food webs. Science, 275, 384-386. Brook, B.W., Sodhi, N.S. & Ng, P.K.L. (2003). Catastrophic extinctions follow deforestation in Singapore. Nature, 424, 420-423.

Brown, S.T. (1999). Trends in the commercial and recreational shark fisheries in Florida, 1980-1992, with implications for management. North American Journal of Fisheries Management, 19,28-41.

Bruno, J.F. & O'Connor, M.I. (2005). Cascading effects of predator diversity and omnivory in a marine food web. Ecology Letters, 8, 1048-1056.

Bullis, H.R.J. (1961). Observations on the feeding behavior of white-tip sharks on schooling fishes. Ecology, 42,194-195.

Bullis, H.RJ. & Captiva, F.J. (1955). Preliminary report on exploratory long-line fishing for tuna in the Gulf of Mexico and the Caribbean Sea. Commercial Fisheries Review, 17, 1-20.

Burgess, G.H., Beerkircher, L.R., Cailliet, G.M., Carlson, J.K., Cortes, E., Goldman, K.J., Grubbs, R.D., Musick, J.A., Musyl, M.K. & Simpfendorfer, C.A. (2005). Is the collapse of shark populations in the Northwest Atlantic Ocean and Gulf of Mexico real? Fisheries, 30,19-26.

Byrnes, J., Stachowicz, J.J., Hultgren, K.M., Hughes, A.R., Olyarnik, S.V. & Thornber, C.S. (2006). Predator diversity strengthens trophic cascades in kelp forests by modifying herbivore behaviour. Ecology Letters, 9, 61-71.

Caddy, J.F. & Rodhouse, P.G. (1998). Cephalopod and groundfish landings: evidence for ecological change in global fisheries? Reviews in Fish Biology and Fisheries, 8, 431-444.

Cardillo, M., Mace, G.M., Jones, K.E., Bielby, J., Bininda-Emonds, O.R.P., Sechrest, W. et al. (2005). Multiple causes of high extinction risk in large mammal species. Science, 309, 1239-1241.

Carlson, J.K. & Lee, D. (1999). The directed shark drift gillnet fishery: catch and bycatch 1998-1999. Sustainable Fisheries Division Contribution No. SFD-99/00-87, NOAA Fisheries Service, Panama City, Florida.

Carlson, J.K. & Baremore, I.E. (2003). Changes in biological parameters of Atlantic sharpnose shark Rhizoprionodon terraenovae in the Gulf of Mexico: evidence for density-dependent growth and maturity? Marine and Freshwater Research, 54, 227-234.

163 Carlson, J.K., Cortes, E. & Johnson, A.G. (1999). Age and growth of the blacknose shark, Carcharhinus acronotus, in the eastern Gulf of Mexico. Copeia, 1999, 684- 691.

Carlton, J.T., Geller, J.B., Reaka-Kudla, M.L. & Norse, E.A. (1999). Historical extinctions in the sea. Annual Review of Ecology, Evolution, and Systematics, 30, 515-538.

Carpenter, S.R. (1996). Microcosm experiments have limited relevance for community and ecosystem ecology. Ecology, 7, 677-680.

Carpenter, S.R. & Kitchell, J.F. (1988). Consumer Control of Lake Productivity. Bioscience, 38, 764-769.

Carpenter, S.R. & Kitchell, J.F. (1993) The Trophic Cascade in Lakes. Cambridge University Press, Cambridge, UK.

Carscadden, J.E., Frank, K.T. & Leggett, W.C. (2001). Ecosystem changes and the effects on capelin (Mallotus villosus), a major forage species. Canadian Journal of Fisheries and Aquatic Sciences, 58, 73-85.

Casey, J.M. & Myers, R.A. (1998). Near extinction of a large, widely distributed fish. Science, 281, 690-692.

Castro, J.I. (1983) Sharks of North American Waters. Texas A&M University Press.

Castro, J.I. (1996). Biology of the blacktip shark, Carcharhinus limbatus, off the southeastern United States. Bulletin of Marine Science, 59, 508-522.

Castro, J.I., Woodley, CM. & Brudek, R.L. (1999) A preliminary evaluation of the status of shark species. Technical Report FAO Fisheries Technical Paper 380, Food and Agriculture Organization of the United Nations, Rome.

CBD (2007). 2010 Biodiversity target. Convention on Biological Diversity (CBD) http://www.cbd.int/2010-target/default.shtml (Updated 2007-06-01).

Chapin, F.S., Zavaleta, E.S., Eviner, V.T., Naylor, R.L., Vitousek, P.M., Reynolds, H.L., Hooper, D.U., Lavorel, S., Sala, O.E., Hobbie, S.E., Mack, M.C. & Diaz, S. (2000). Consequences of changing biodiversity. Nature, 405, 234-242.

Chapman, D.D. & Gruber, S.H. (2002). A further observation of the prey-handling behavior of the great hammerhead shark, Sphyrna mokarran: Predation upon the spotted eagle ray, Aetobatus narinari. Bulletin of Marine Science, 70, 947-952.

164 Chapman, D.D.F., Pikitch, E.K. & Babcock, E.A. (2006). Marine parks need sharks? Science, 312, 526-527.

Chavez, F.P., Ryan, J., Lluch-Cota, S.E. & Niquen C, M. (2003). From anchovies to sardines and back: multidecadal change in the Pacific Ocean. Science, 299, 217- 221.

Choi, J.S., Frank, K.T., Leggett, W.C. & Drinkwater, K. (2004). Transition to an alternate state in a continental shelf ecosystem. Canadian Journal of Fisheries and Aquatic Sciences, 61, 505-510.

Christensen, V., Guennette, S., Heymans, J J., Walters, C.J., Watson, R., Zeller, D.C. & Pauly, D. (2003). Hundred-year decline of North Atlantic predatory fishes. Fish and Fisheries, 4, 1-24.

CITES (2006). Implementation of CITES shark listings. Twenty-second meeting of the Animals Committee Lima (Peru), 7-13 July 2006. Convention on International Trade in Endangered Species of Wild Fauna and Flora, AC22 Doc. 17.2.

Clark, E. & von Schmidt, K. (1965). Sharks of the central Gulf coast of Florida. Bulletin of Marine Science, 15, 13-83.

Clarke, S.C. 2003. Quantification of the trade in shark fins. PhD Thesis, Imperial College London, UK.

Clarke, S. (2004). Understanding pressures on fishery resources through trade statistics: a pilot study of four products in the Chinese dried seafood market. Fish and Fisheries, 2004, 53-74.

Clarke, S.C, McAllister, M.K. & Michielsens, C.G.J. (2005). Estimates of shark species composition and numbers associated with the shark fin trade based on Hong Kong auction data. Journal of the Northwest Atlantic Fishery Science, 35, 453-465.

Clarke, S.C, McAllister, M.K., Milner-Gulland, E.J., Kirkwood, G.P., Michielsens, C.G.J., Agnew, D.J., Pikitch, E.K., Nakano, H. & Shiyji, M.S. (2006). Global estimates of shark catches using trade records from commercial markets. Ecology Letters, 9, 1115-1126.

Cliff, G. (1995). Sharks caught in the protective gill nets off Kwazulu-Natal, South- Africa. 8. The great hammerhead shark Sphyrna mokarran (Ruppell). South African Journal of Marine Science, 15, 105-114.

165 Cliff, G. & Dudley, S. (1991). Sharks caught in the protective gill nets off Natal, South Africa. 4. The bull shark Carcharhinus leucas Valenciennes. South African Journal of Marine Science, 10, 253-270.

Cliff, G., Dudley, S.F.J. & Davis, B. (1989). Sharks caught in the protective gill nets off Natal, South Africa .2. The great white shark Carcharodon carcharias (Linnaeus). South African Journal of Marine Science, 8, 131-144.

Cliff, G., Dudley, S. & Davis, B. (1990). Sharks caught in the protective gill nets off Natal, South Africa. 3. The shortfin mako shark Isurus oxyrinchus (Rafinesque). South African Journal of Marine Science, 9, 115-126

Compagno, L.J.V. (1984). FAO Species Catalogue Vol. 4, Part 2 Sharks of the world: an annotated and illustrated catalogue of shark species known to date. United Nations Development Programme, Food and Agriculture Organization of the United Nations, Rome.

Compagno, L.J.V. (2001). Sharks of the world: an annotated and illustrated catalogue of shark species known to date. Volume 2. Bullhead, mackerel and carpet sharks (Heterodontiformes, Lamniformes and Orectolobiformes). Food and Agriculture Organization of the United Nations, Rome.

Cortes, E. (1999). Standardized diet compositions and trophic levels of sharks. ICES Journal of Marine Science, 56, 707-717'.

Cortes, E. (2002). Incorporating uncertainty into demographic modeling: application to shark populations and their conservation. Conservation Biology, 16, 1048-1062.

Cortes, E., Apostolaki, P. & Brown, C.A. (2006) Stock assessment of dusky shark in the U.S. Atlantic and Gulf of Mexico, p.155. National Marine Fisheries Service, Southeast Fisheries Science Center, Panama City, USA.

Cox, S.P., Martell, S.J.D., Walters, C.J., Essington, T.E., Kitchell, J.F., Boggs, C.H. & Kaplan, I. (2002). Reconstructing ecosystem dynamics in the central Pacific Ocean, 1952-1998.1. Estimating population biomass and recruitment of tunas and billfishes. Canadian Journal of Fisheries and Aquatic Sciences, 59, 1724-1735.

Cramer, J. 1995. Large pelagic logbook newsletter-1994. U.S. Department of Commerce, NOAA Tech. Memo. NMFS-SEFSC-387, 33pp.

Crooks, K.R. & Soule, M.E. (1999). Mesopredator release and avifaunal extinctions in a fragmented system. Nature, 400, 563-566. Cruz-Martinez, A., Chiappa-Carrara, X., & Arenas-Fuentes, V. (2005). Age and growth of the bull shark, Carcharhinus leucas. Journal of Northwest Atlantic Fishery Science, 35, 367-374.

Daan, N. Gislason, H. Pope, J.G. & Rice, J.C. (2005). Changes in the North Sea fish community: evidence of indirect effects of fishing? ICES Journal of Marine Science, 62, 177-188.

Daskalov, G.M. (2002). Overfishing drives a trophic cascade in the Black Sea. Marine Ecology Progress Series, 225, 53-63.

Daskalov, G.M., Grishin, A.N., Rodionov, S. & Mihneva, V. (2007). Trophic cascades triggered by overfishing reveal possible mechanisms of ecosystem regime shifts. Proceedings of the National Academy of Sciences of the United States of America, 104,10518-10523.

Dayton, P.K., Tegner, M.J., Edwards, P.B. & Riser, K.L. (1998). Sliding baselines, ghosts, and reduced expectations in kelp forest communities. Ecological Applications, 8, 309-322. de Bruyn, P., Dudley, S.F.J., Cliff, G. & Smale, MJ. (2005). Sharks caught in the protective gill nets off KwaZulu-Natal, South Africa. 11. The scalloped hammerhead shark Sphyrna lewini (Griffith and Smith). African Journal of Marine Science, 27, 517-528.

DeMaster D.P., Trites, A.W., Clapham, P., Mizroch, S., Wade, P., Small, R.J. & Hoef, J.V. (2006). The sequential megafaunal collapse hypothesis: testing with existing data. Progress in Oceanography, 68, 329-342.

Denney, N.H., Jennings, S. & Reynolds, J.D. (2002). Life-history correlates of maximum population growth rates in marine fishes. Proceedings of the Royal Society of London B, 269, 2229-2237.

DFO. (1997). Canadian Atlantic pelagic shark management plan 1997-1999. Technical report. Department of Fisheries and Oceans, Canada.

Diggle, P.J., Heagerty, P., Liang, K-Y., and Zeger, S. (2002). Analysis of longitudinal data, 2nd ed. Oxford University Press, New York.

Doak, D.F., Williams, T.M. & Estes, J.A. (2007). Great whales as prey: using demography and bioenergetics to infer interactions in marine mammal communities. In.- Whales, whaling, and ocean ecosystems (eds. Estes, J.A., DeMaster, D.P., Doak, D.F., Williams, T.M. & Brownell Jr., R.L.), pp. 231-244. University of California Press.

167 Dobson, A.J. (2002). An Introduction to Generalized Linear Models, 2" ed. Chapman and Hall, Boca Raton.

Driggers, W.B., Carlson, J.K., Cullum, B., Dean, J.M., Oakley, D. & Ulrich, G. (2004). Age and growth of the blacknose shark, Carcharhinus acronotus, in the western North Atlantic Ocean with comments on regional variation in growth rates. Environmental Biology of Fishes, 71, 171-178.

Duffy, J.E. (2002). Biodiversity and ecosystem function: the consumer connection. 0ifow,99,2Ol-219.

Duffy, J.E. (2003). Biodiversity loss, trophic skew and ecosystem functioning. Ecology Letters, 6,680-687.

Dudley, S.F.J. & Cliff, G. (1993). Sharks caught in the protective gill nets off Natal, South-Africa .7. The blacktip shark Carcharhinus limbatus (Valenciennes). South African Journal of Marine Science, 13, 237-254.

Dudley, S.F.J., Cliff, G., Zungu, M.P. & Smale, M.J. (2005). Sharks caught in the protective gill nets off KwaZulu-Natal, South Africa. 10. The dusky shark Carcharhinus obscurus (Lesueur 1818). African Journal of Marine Science, 27, 107-127.

Dulvy, N.K., Metcalfe, J.D., Glanville, J., Pawson, M.G. & Reynolds, J.D. (2000). Fishery stability, local extinctions, and shifts in community structure in skates. Conservation Biology, 14, 283-293.

Dulvy, N.K., Sadovy, Y. & Reynolds, J.D. (2003). Extinction vulnerability in marine populations. Fish and Fisheries, 4, 25-64.

Dulvy, N.K., Polunin, N.V.C., Mill, A.C. & Graham, N.A.J. (2004a). Size structural change in lightly exploited coral reef fish communities: evidence for weak indirect effects. Canadian Journal of Fisheries and Aquatic Sciences, 61, 466- 475.

Dulvy, N.K., Freckleton, R.P. & Polunin, N.V.C. (2004b). Coral reef cascades and the indirect effects of predator removal by exploitation. Ecology Letters, 7, 410-416.

Ellis, J.K. (2003). Diet of the sandbar shark, Carcharhinus plumbeus, in Chesapeake Bay and adjacent waters, p. 126. M.Sc. Thesis. The Faculty of the School of Marine Science, The College of William and Mary, Virginia. Essington, T.E. (2007). Pelagic ecosystem response to a century of commercial fishing and whaling. In: Whales, whaling, and ocean ecosystems (eds. Estes, J.A., DeMaster, D.P., Doak, D.F., Williams, T.M. & Brownell Jr., R.L.), pp. 38-49. University of California Press.

Estes, J.A. & Palmisano, J.F. (1974). Sea otters: their role in structuring nearshore communities. Science, 185, 1058-1060.

Estes, J. A., Tinker, M. T., Williams, T. M., & Doak, D. F. (1998). Killer whale predation on sea otters linking oceanic and nearshore ecosystems. Science, 282, 473-476.

Fahrenthold, D. A. Washington Post, http://www. washingtonpost.com/wp- dyn/articles/A29969-2004Aug24.html (25 August 2004).

FAO. (1998). International plan of action for the conservation and management of sharks. Technical Document FI: FI: CSS/98/3, Food and Agriculture Organization Fisheries Department, Rome.

FAO. (2000). Conservation and management of sharks. FAO Technical Guidelines for Responsible Fisheries. No. 4, Suppl. 1. Food and Agriculture Organization, Rome, 37pp.

FAO. (2007). The state of world fisheries and aquaculture 2006. Food and Agriculture Organization of the United Nations, Rome.

Ferretti, F., Myers, R.A., Sartor, P. & Serens, F. (2005). Long term dynamics of the Chondrichthyan fish community in the Upper Tyrrhenian Sea. ICES CM 2005/N:25.

Finke, D.L. & Denno, R.F. (2004). Predator diversity dampens trophic cascades. Nature, 429,407-410.

Finke, D.L. & Denno, R.F. (2005). Predator diversity and the functioning of ecosystems: the role of intraguild predation in dampening trophic cascades. Ecology Letters, 8, 1299-1306.

Fogarty, M.J. & Murawski, S.A. (1998). Large-scale disturbance and the structure of marine systems: fishery impacts on Georges Bank. Ecological Applications, 8, S6-S22.

169 Fowler, S. L., Cavanagh, R.D., Camhi, M., Burgess, G.H., Cailliet, G.M., Fordham, S.V., Simpfendorfer, C.A. & Musick, J.A. (2005). Sharks, rays and chimaeras: the status of the chondrichthyan fishes (IUCN/SSC Shark Specialist Group), The World Conservation Union, Cambridge, UK.

Francis, M.P. & Duffy, C. (2005). Length at maturity in three pelagic sharks (Lamna nasus, Isurus oxyrinchus, and Prionace glauca) from New Zealand. Fishery Bulletin, 103, 489-500.

Frank, K.T., Petrie, B., Choi, J.S. & Leggett, W.C. (2005). Trophic cascades in a formerly cod-dominated ecosystem. Science, 308, 1621-1623.

Frank, K.T., Petrie, B., Shackell, NX. & Choi, J.S. (2006) Reconciling differences in trophic control in mid-latitude marine ecosystems. Ecology Letters, 9, 1096-1105.

Frank, K.T., Petrie, B. & Shackell, N.L. (2007). The ups and downs of trophic control in continental shelf ecosystems. Trends in Ecology & Evolution, 22, 236-242.

Frederiksen, M. Edwards, M., Richardson, A.J., Halliday, N.C. & Wanless, S. (2006). From plankton to top predators: bottom-up control of a marine food web across four trophic levels. Journal of Animal Ecology, 75, 1259-1268.

Galvan-Magana, F. & Nienhuis, H.J. (1989). Seasonal abundance and feeding habits of sharks of the lower Gulf of California. California Fish and Game. 75, 74.

Garcia, C.B., Duarte, L.O., Altamar, J. & Manjarres, L.M. (2007). Demersal fish density in the upwelling ecosystem off Colombia, Caribbean Sea: historic outlook. Fisheries Research, 85, 68-73.

Garrison, L.P., Michaels, W., Link, J.S. & Fogarty, M.J. (2000). Predation risk on larval gadids by pelagic fish in the Georges Bank ecosystem. I. Spatial overlap associated with hydrographic features. Canadian Journal of Fisheries and Aquatic Sciences, 57, 2455-2469.

Gelsleichter, J., Musick, J.A. & Nichols, S. (1999). Food habits of the smooth dogfish, Mustelus canis, dusky shark, Carcharhinus obscurus, Atlantic sharpnose shark, Rhizoprionodon terraenovae, and the sand tiger, Carcharias taurus, from the northwest Atlantic Ocean. Environmental Biology of Fishes, 54, 205-217.

Gonzalez Ania L.V., Ulloa Ramirez P.A., Lee D.W., Brown C.J. & Brown C.A. (1997). Description of Gulf of Mexico longline fisheries based upon observer programs from Mexico and the United States. In: International Commission for the Conservation of Atlantic Tunas (ICCAT) Collective Volume of Scientific Papers, 48,308-316.

170 Graham, K.J., Andrew, N.L. & Hodgson, K.E. (2001). Changes in relative abundance of sharks and rays on Australian South East fishery trawl grounds after twenty years of fishing. Marine & Freshwater Research, 52, 549-561.

Graham, N.A.J., Evans, R.D. & Russ, G.R. (2003). The effects of marine reserve protection on the trophic relationships of reef fishes on the Great Barrier Reef. Environmental Conservation, 30, 200-208.

Grusha, D.S. (2005). Investigation of the life history of the cownose ray, Rhinoptera bonasus (Mitchill 1815). p. 135. M.Sc. Thesis. The Faculty of the School of Marine Science, The College of William and Mary, Virginia.

Guidetti, P. (2006). Marine reserves reestablish lost predatory interactions and cause community changes in rocky reefs. Ecological Applications, 16, 963-976.

Ha, D.H. (2006). Ecology and conservation of Virginia shark species: analysis of 30 years of Virginia longline shark census data, 1974-2004. p. 198. Ph.D. Thesis. The Faculty of the School of Marine Science, The College of William and Mary, Virginia.

Hairston, N.G., Smith, F.E. & Slobodkin, L.B. (1960). Community structure, population control, and competition. American Naturalist, 118, 240-261.

Hall, SJ. & Mainprize, B.M. (2005). Managing by-catch and discards: how much progress are we making and how can we do better? Fish and Fisheries, 6, 134- 155.

Hay, S. (2006). Marine ecology: gelatinous bells may ring change in marine ecosystems. Current Biology, 16, R679-R682.

Hazin, F.H.V., Lucena, F.M., Souza, T.S.A.L., Boeckman, C.E., Broadhurst, M.K. & Menni, R.C. (2000). Maturation of the night shark, Carcharhinus signatus, in the southwestern equatorial Atlantic Ocean. Bulletin of Marine Science, 66,173-185.

Hazin, F.; Fischer, A. & Broadhurst, M. (2001). Aspects of reproductive biology of the scalloped hammerhead shark, Sphyrna lewini. Environmental Biology of Fishes, 61,151-159.

Hedges, L.V. & Olkin, I. (1985). Statistical Methods for Meta-analysis. Academic Press. San Diego, California. Heithaus, M.R. (2004). Predator-prey interactions. In: Biology of Sharks and their Relatives (eds. Carrier, JC, Musick, JA & Heithaus, MR), pp. 487-521. CRC Press, Boca Raton.

Helfman, G.S., Collette, B.B. & Facey, D.E. (1997). The Diversity of Fishes. Blackwell Science, Maiden, Massachusetts.

Hilborn, R. & Walters, C. (1992). Quantitative Fisheries Stock Assessment: Choice, Dynamics & Uncertainty. Chapman and Hill, New York.

Hilborn, R. (2007a). Moving to sustainability by learning from successful fisheries. Ambio, 36, 296-303.

Hilborn, R. (2007b). Biodiversity loss in the ocean: how bad is it? Science, 316, 1281- 1282.

Hoenig, J.M. (1983). Empirical use of longevity data to estimate mortality-rates. Fishery Bulletin, 81, 898-903.

Hoey, J.J. & Moore, N. (1999). Captain's Report: Multi-species catch characteristics for the US Atlantic pelagic longline fishery, pp. 1-78. NMFS and the National Fisheries Institute, Inc., Silver Spring, MD and Arlington, VA.

Hoffinayer, E.R. & Parsons, G.R. (2003). Food habits of three shark species from the Mississippi Sound in the northern Gulf of Mexico. Southeastern Naturalist, 2, 271-280.

Holden, M.J. (1968). The rational exploitation of the Scottish-Norwegian stocks of spurdogs, Squalus acanthias. Fisheries Investigations Series II, Ministry of Agriculture, Fisheries and Food, London, 25, 28pp.

Holden, M.J. (1973). Are long-term sustainable fisheries for elasmobranchs possible? Rapports et Proces-Verbaux des Reunions Conseil International Pour V Exploration de la Mer, 164, 360-367.

Hueter, R.E. & Manire, C.A. (1994). Bycatch and catch-release mortality of small sharks in the Gulf coast nursery grounds of Tampa Bay and Charlotte Harbor. Mote Marine Technical Report No. 368, Final report to NOAA/NMFS, MARFIN Project NA17FF0378-01.

Hughes, T. (1994). Catastrophes, phase shifts, and large-scale degradation of a Caribbean coral reef. Science, 265, 1547-1551. Hunt, G.L. & McKinnell, S. (2006). Interplay between top-down, bottom-up, and wasp- waist control in marine ecosystems. Progress in Oceanography, 68, 115-124.

Hunter, M.D. & Price, P.W. (1992). Playing chutes and ladders: heterogeneity and the relative roles of bottom-up and top-down forces in natural communities. Ecology, 73, 724-732.

Hutchings, J.A. (2000). Collapse and recovery of marine fishes. Nature,406, 882-885.

Hutchings, J.A. & Myers, R.A. 1994. What can be learned from the collapse of a renewable resource? Atlantic cod, Gadus morhua, of Newfoundland and Labrador. Canadian Journal of Fisheries and Aquatic Sciences, 51, 2126-2146.

Hutchings, J.A. & Reynolds, J.D. (2004). Marine fish population collapses: Consequences for recovery and extinction risk. Bioscience, 54, 297-309.

Hutchings, J.A. & Baum, J.K. (2005). Measuring marine fish biodiversity: temporal changes in abundance, life history and demography. Philosophical Transactions of the Royal Society B, 360, 315-338.

Hutchings, J.A., Myers, R.A., Lucifora, L.O. & Garcia, V.B. Life history and extinction risk: a comparison among sharks, bony fishes, and mammals. Unpublished manuscript.

ICC AT. (2004). Recommendation by ICC AT concerning the conservation of sharks caught in association with fisheries managed by ICC AT. BYC-04-10. http://www.iccat.int/Documents%5CRecs%5Ccompendiopdf-e%5C2004-10- e.pdf

IUCN. (2002). 2002 IUCN Red List of Threatened Species, http://www.redlist.org.

Ives, A.R., Cardinale, B.J., & Snyder, W.E. (2005). A synthesis of subdisciplines: predator-prey interactions, and biodiversity and ecosystem functioning. Ecology Letters, 8, 102-116.

Iwamoto, T. (1965). Summary of tuna observations in the Gulf of Mexico on cruises of the exploratory fishing vessel Oregon, 1950-1963. Commercial Fisheries Review, 27(1), 7-14.

Jackson, J.B.C. (2001). What was natural in the coastal oceans? Proceedings of the National Academy of Sciences of the United States of America, 98, 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. & Warner, R.S. (2001). Historical overfishing and the recent collapse of coastal ecosystems. Science, 293, 629-638.

Jennings, S., & Polunin, N.V.C. (1997). Impacts of predator depletion by fishing on the biomass and diversity of non-target reef fish communities. Coral Reefs, 16, 71-82.

Jennings, S. & Kaiser, M.J. (1998). The effects of fishing on marine ecosystems. In: Advances in Marine Biology, pp. 201-352.

Jennings, S. & Blanchard, J.L. (2004). Fish abundance with no fishing: predictions based on macroecological theory. Journal of Animal Ecology, 73, 632-642.

Jennings, S., Reynolds, J.D. & Mills, S.C. (1998). Life history correlates of responses to fisheries exploitation. Proceedings of the Royal Society of London B, 265, 333- 339.

Jennings, S., Greenstreet, S.P.R. & Reynolds, J.D. (1999). Structural change in an exploited fish community: a consequence of differential fishing effects on species with contrasting life histories. Journal of Animal Ecology, 68, 617-627.

Jukic-Peladic, S., Vrgoc, N., Krstulovic-Sifner, S., Piccinetti, C, Piccinetti-Manfrin, G., Marano, G. & Ungaro, N. (2001). Long-term changes in demersal resources of the Adriatic Sea: comparison between trawl surveys carried out in 1948 and 1998. Fisheries Research, 53, 95-104.

Kaplan, I.C., Cox, S.P. & Kitchell, J.F. (2007). Circle hooks for Pacific longliners: not a panacea for marlin and shark bycatch, but part of the solution. Transactions of the American Fisheries Society, 136, 392-401.

Killam, K.A. & Parsons, G.R. (1989). Age and growth of the blacktip shark, Carcharhinus limbatus, near Tampa Bay, Florida. Fishery Bulletin, 87, 845-857.

Kohler, N.E., Casey, J.G. & Turner, P.A. (1995). Length-weight relationships for 13 species of sharks from the western North Atlantic. Fishery Bulletin, 93, 412-418.

Kohler, N.E., Casey, J.G. & Turner, P.A. (1998). NMFS Cooperative shark tagging program, 1962-93: an atlas of shark tag and recapture data. Marine Fisheries Review, 60,1-87.

174 KQster, F.W. & MoTlmann, C. (2000). Trophodynamic control by clupeid predators on recruitment success in Baltic cod? Ices Journal of Marine Science, 57, 310-323.

Lack, M. & Sant, G. (2006). Confronting shark conservation head on. TRAFFIC International, Cambridge, UK.

Last, P.R. & Stevens, J.D. (1994) Sharks and Rays of Australia. CSIRO, Australia.

Lessa, R., Paglerani, R. & Santana, F.M. (1999). Biology and morphometry of the oceanic whitetip shark, Carcharhinus longimanus (Carcharhinidae), off North- Eastern Brazil. Cybium, 23, 353-368.

Lewison, R.L., Crowder, L.B., Read, A.J. & Freeman, S.A. (2004). Understanding impacts of fisheries bycatch on marine megafauna. Trends in Ecology & Evolution, 19, 598-604.

Link, J.S. (2002). Does food web theory work for marine ecosystems? Marine Ecology Progress Series, 230, 1-9.

Litzow, M.A. & Ciannelli, L. (2007). Oscillating trophic control induces community reorganization in a marine ecosystem. In press: Ecology Letters, 10, xxx. doi: 10.1111/j. 1461-0248.2007.01111.x

Loefer, J.K. & Sedberry, G.R. (2003). Life history of the Atlantic sharpnose shark (Rhizoprionodon terraenovae) (Richardson, 1836) off the southeastern United States. Fishery Bulletin, 101, 75-88.

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. & Jackson, J.B.C. (2006). Depletion, degradation, and recovery potential of estuaries and coastal seas. Science, 312, 1806-1809.

Low, R.A. & Ulrich, G.F. (1984). Survey of the shark resource in shelf waters off South Carolina. In. South Carolina Marine Resources Center Tech. Rep. No. 61.

Lucas, Z. & Stobo, W.T. (2000). Shark-inflicted mortality on a population of harbour seals (Phoca vitulina) at Sable Island, Nova Scotia. Canadian Journal of Zoology, 252,405-414.

Lucifora, L.O., Garcia, V.B., Menni, R.C., Escalante, A.H. & Hozbor, N.M. Fisheries affect traits that determine a shark's ecological function. In review. Lynam, C.P., Gibbons, M.J., Axelsen, B.E., Sparks, C.A.J., Coetzee, J., Heywood, B.G. & Brierley, A.S. (2006). Jellyfish overtake fish in a heavily fished ecosystem. Current Biology, 16, R492-R493.

Mace, G., Masundire, H., Baillie, J., Ricketts, T., Brooks, T. Hoffmann, M. et al. (2005). Biodiversity. In: Millenium Ecosystem Assessment, 2005. Current state and trends: findings of the condition and trends working group. Ecosystems and human well-being, Vol. 1. Island Press, Washington, DC.

Mack, R.N., Simberloff, D., Lonsdale, W.M., Evans, H., Clout, M. & Bazzaz, F.A. (2002). Biotic invasions: causes, epidemiology, global consequences, and control. Ecological Applications, 10, 689-710.

MacPhee, R.D.E. (1999). Extinctions in near time: causes, contexts, and consequences. Kluwer Academic/Plenum Publishers, New York.

Mangel, M. & Levin, P.S. (2005). Regime, phase and paradigm shifts: making community ecology the basic science for fisheries. Philosophical Transactions of the Royal Society B, 360, 95-105.

Mather, F.J.I. & Day, C.G. (1954). Observations of pelagic fishes of the tropical Atlantic. Copeia, 1954, 179-188.

Matsunaga, H. & Nakano, H. (2000). Report on information of bycatch fishes obtained from observer program for Japanese tuna longline fisheries in the Atlantic Ocean. In: International Commission for the Conservation of Atlantic Tunas (ICCAT) Collective Volume of Scientific Papers, 51, 1805-1811.

Matsushita Y. & Matsunaga H. (2002). Species composition and CPUE of pelagic sharks observed by Japanese observers for tuna longline fisheries in the Atlantic Ocean. In: International Commission for the Conservation of Atlantic Tunas (ICCAT) Data Collective Volume of Scientific Papers, 53,1381-1385.

Maunder, M.N. & Punt, A.E. (2004). Standardizing catch and effort data: a review of recent approaches. Fisheries Research, 70, 141-159.

McCann, K. (2007). Protecting biostructure. Nature, 446, 29.

McCann, K.S., Rasmussen, J.B. & Umbanhowar, J. (2005). The dynamics of spatially coupled food webs. Ecology Letters, 8,513-523. McClanahan, T.R., Muthiga, N.A., Kamukuru, A.T., Machano, H. & Kiambo, R.W. (1999). The effects of marine parks and fishing on coral reefs of northern Tanzania. Biological Conservation, 89, 161-182.

McClanahan, T.R., Polunin, N.V.C. & Done, T J. (2002). Ecological states and the resilience of coral reefs. Conservation Ecology, 6, 18.

McCullagh, P. & Nelder, J. A. (1989) Generalized Linear Models. 2nd edn. Chapman & Hall, Boca Raton.

McQueen, D.J., Johannes, M.R.S., Post, J.R., Stewart, D J. & Lean, D.R.S. (1989). Bottom-up and top-down impacts on freshwater pelagic community structure. Ecological Monographs, 59, 289-309.

Megalofonou, P., Yannopoulos, C, Damalas, D., deMetrio, G., Deflorio, M. & de la Serna, J.M. (2005). Incidental catch and estimated discards of pelagic hsarks from the swordfish and tuna fisheries in the Mediterranean Sea. Fishery Bulletin, 103, 620-634.

Menge, B.A. (1995). Indirect effects in marine rocky intertidal interaction webs: patterns and importance. Ecological Monographs, 65, 21-74.

Micheli, F. (1999). Eutrophication, fisheries, and consumer-resource dynamics in marine pelagic ecosystems. Science, 285, 1396-1398.

Micheli, F., Halpern, B.S., Botsford, L.W. & Warner, R.R. (2004). Trajectories and correlates of community changes in no-take marine reserves. Ecological Applications, 14,1709-1723.

Millenium Ecosystem Assessment. (2005). Ecosystems and human well-being: biodiversity synthesis. A report of the Millenium Ecosystem Assessment (MEA). World Resources Institute, Washington, DC. 100pp.

Mills, C.E. (2001). Jellyfish blooms: are populations increasing globally in response to changing ocean conditions? Hydrobiologia, 451, 55-68.

Minami, M., Lennert-Cody, C.E., Gao, W. & Roman-Verdesoto, M. (2007). Modeling shark bycatch: the zero-inflated negative binomial regression model with smoothing. Fisheries Research, 84,221 -221.

Mizuno, K., Okazaki, M., Nakano, H. & Okamura, H. (1999). Estimating the underwater shape of tuna longlines with micro-bathythermographs. Special report: IATTC 10, Inter-American Tropical Tuna Commission, La Jolla, California.

177 Mullon, C, Freon, P. & Cury, P. (2005). The dynamics of collapse in world fisheries. Fish and Fisheries, 6, 111-120.

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. & Gill, A.B. (2006a). Fishing, trophic cascades, and the process of grazing on coral reefs. Science, 311, 98-101.

Mumby, P.J., Micheli, F., Dahlgren, C.P., Litvin, S.Y., Gill, A.B., Brumbaugh, D.R., Broad, K., Sanchirico, J.N., Kappel, C.V., Harborne, A.R. & Holmes, K.E. (2006b). Marine parks need sharks? Response. Science, 312, 527-528.

Murawski, S., Methot, R. & Tromble, G. (2007). Biodiversity loss in the ocean: how bad is it? Science, 316, 1281-1281.

Murray, J.J. (1953) Gulf of Maine bluefin tuna exploration -1952. Commercial Fisheries Review, 15, 1-17.

Musick, J.A. (1999). Ecology and conservation of long-lived marine animals. In: Life in the slow lane (ed. Musick, J. A.), pp. 1-10. American Fisheries Society Symposium 23.

Musick, J.A., Branstetter, S. & Colvocoresses, J.A. (1993) Trends in shark abundance from 1974 to 1991 for the Chesapeake Bight region of the U.S. Mid-Atlantic coast. In: Conservation biology of elasmobranchs (ed. Branstetter, S), pp. 1-18. Seattle, Washington: National Marine Fisheries Service, 1993.

Musick, J.A., Burgess, G., Cailliet, G., Camhi, M. & Fordham, S. (2000). Management of Sharks and Their Relatives (Elasmobranchii). Fisheries, 25, 9-13.

Myers, R.A. & Mertz, G. (1998). Reducing uncertainty in the biological basis of fisheries management by meta-analysis of data from many populations: a synthesis. Fisheries Research, 37, 51 -60...

Myers, R.A. & Worm, B. (2003). Rapid worldwide depletion of predatory fish communities. Nature, 423, 280-283.

Myers, R.A., Hutchings, J.A. & Barrowman, N.J. (1996). Hypotheses for the decline of cod in the north Atlantic. Marine Ecology Progress Series, 138, 293-308.

Myers, R.A., Mertz, G. & Bridson, J.M. (1997a). Spatial scales of interannual recruitment variations of marine, anadromous, and freshwater fish. Canadian Journal of Fisheries and Aquatic Sciences, 54, 1400-1407.

178 Myers, R.A., Mertz, G. & Fowlow, P.S. (1997b). Maximum population growth rates and recovery times for Atlantic cod, Gadus morhua. Fishery Bulletin, 95, 762-772.

Myers, R.A., Baum, J.K., Shepherd, T.D., Powers, S.P. & Peterson, C.H. (2007). Cascading effects of the loss of apex predatory sharks from a coastal ocean. Science, 315, 1846-1850.

Myers, R.A., Worm, B. & Blanchard, W. Top-down control of snow crab abundance linked to overfishing of demersal predators in the Northwest Atlantic. Unpublished manuscript.

Natanson, L.J., Casey, J.G. & Kohler, N.E. (1995). Age and growth estimates for the dusky shark, Carcharhinus obscurus, in the western North Atlantic Ocean. Fishery Bulletin, 93,116-126.

Natanson, L.J., Casey, J.G., Kohler, N.E. & Colket, T. (1999). Growth of the tiger shark, Galeocerdo cuvier, in the western North Atlantic based on tag returns and length frequencies; and a note on the effects of tagging. Fishery Bulletin, 97, 944-953.

Natanson, L.J., Kohler, N.E., Ardizzone, D., Cailliet, G.M., Wintner, S.P. & Mollet, H.F. (2006). Validated age and growth estimates for the shortfin mako, Isurus oxyrinchus, in the North Atlantic Ocean. Environmental Biology of Fishes, 11, 367-383.

NMFS. (2001). United States National Plan of Action for the conservation and management of sharks. Technical report, Department of Commerce, NOAA, NMFS, www.nmfs.noaa.gov/sfa/Final%20NPOA.February.2001.htm

NMFS. (2006). Final consolidated Atlantic Highly Migratory Species Fishery Management Plan. National Oceanic and Atmospheric Administration, National Marine Fisheries Service, Office of Sustainable Fisheries, Highly Migratory Species Management Division, Silver Spring, MD. pp. 1600.

NOAA. (2007). Office of Protected Resources, Species of concern and candidate species. NOAA Fisheries, www.nmfs.noaa. gov/pr/species/concern/#rist

Olsen, A.M. (1959). The status of the school shark fishery in south-eastern Australia waters. Australian Journal of Marine & Freshwater Research, 10, 150-176.

Orth, R.J. (1975). Destruction of eelgrass, Zostera marina, by the cownose ray, Rhinoptera bonasus in the Chesapeake Bay. Chesapeake Science, 16, 205-208. Pace, M.L., Cole, J.G., Carpenter, S.R. & Kitchell, J.F. (1999). Trophic cascades revealed in diverse ecosystems. Trends in Ecology & Evolution, 14,483-488.

Packer, D.B., Zetlin, C.A.& Vitaliano, JJ. (2003) Essential fish habitat source document: Little skate, Leucoraja erinacea. Life history and habitat characteristics. NOAA Tech. Mem. NMFS-NE-175. Northeast Fisheries Science Center, Woods Hole, Massachusetts.

Paine, R.T. (1966). Food web complexity and species diversity. American Naturalist, 100, 65-75.

Paine, R.T. (1980). Food Webs - Linkage, Interaction Strength and Community Infrastructure - the 3rd Tansley Lecture. Journal of Animal Ecology, 49, 667-685.

Paine, R.T. (2002). Trophic control of production in a rocky intertidal community. Science, 296, 736-739.

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. & Jackson, J.B.C. (2003). Global trajectories of the long-term decline of coral reef ecosystems. Science, 301, 955-958.

Parker, H.W. & Stott, F.C. (1965). Age, size and vertebral calcification in the basking shark, Cetorhinus maximus (Gunnerus). Zool. Meded. (Leiden), 40, 305-319.

Parry-Jones, R. (1996). TRAFFIC report on shark fisheries and trade in Hong Kong. In: The world trade in sharks: a compendium of TRAFFIC'S regional studies, pp. 87- 143 . TRAFFIC International, Cambridge, United Kingdom.

Pauly, D. (1980). On the interrelationships between natural mortality, growth parameters and mean environmental temperature in 175 fish stocks. Journal du Conseil international pour I'Exploration de la Mer, 39, 175-192.

Pauly, D. (1995). Anecdotes and the shifting baseline syndrome of fisheries. Trends in Ecology & Evolution, 10, 430.

Pauly, D., Christensen, V., Dalsgaard, J., Froese, R. & Torres, Jr. F. (1998). Fishing down marine food webs. Science, 279, 860-863.

Pauly, D., Christensen, V. & Walters, C. (2000). Ecopath, Ecosim, and Ecospace as tools for evaluating ecosystem impact of fisheries. ICES Journal of Marine Science, 57, 697-706.

180 Pauly, D., Christensen, V., Guennette, S., Pitcher, T.J., Sumaila, U.R., Walters, C.J., Watson, R. & Zeller, D. (2002). Towards sustainability in world fisheries. Nature, 418,689-695.

Pauly, D., Watson, R. & Alder, J. (2005). Global trends in world fisheries: impacts on marine ecosystems and food security. Philosophical Transactions of the Royal Society B-Biological Sciences, 360, 5-12.

Perry, A.L., Low, P.J., Ellis, J.R. & Reynolds, J.D. (2005). Climate change and distribution shifts in marine fishes. Science, 308, 1912-1915.

Petchey, O.L., McPherson, P.T., Casey, T.M. & Morin, P.J. (1999). Environmental warming alters food-web structure and ecosystem function. Nature, 402, 69-72.

Petchey? O.L., Downing, A.L., Mittelbach, G.G., Persson, L., Steiner, C.F., Warren, P.H. & Woodward, G. (2004). Species loss and the structure and functioning of multitrophic aquatic systems. Oikos, 104,467-478.

Peterson, C.H., Summerson, H.C., Fegley, S.R. & Prescott, R.C. (1989). Timing, intensity and sources of autumn mortality of adult bay scallops Argopecten irradians concentricus say. Journal of Experimental Marine Biology and Ecology, 127, 121-140.

Peterson, C.H., Summerson, H.C. & Luettich, R.A. (1996). Response of bay scallops to spawner transplants: A test of recruitment limitation. Marine Ecology Progress Series, 132, 93-107.

Peterson, C.H., Fodrie, F.J., Summerson, H.C. & Powers, S.P. (2001). Site-specific and density-dependent extinction of prey by schooling rays: generation of a population sink in top-quality habitat for bay scallops. Oecologia, 129, 349-356.

Piercy, A.N., Carlson, J.K., Sulikowski, J.A. & Burgess, G.H. (2007). Age and growth of the scalloped hammerhead shark, Sphyrna lewini, in the north-west Atlantic Ocean and Gulf of Mexico. Marine and Freshwater Research, 58, 34-40.

Pikitch, E.K., Santora, C, Babcock, E.A., Bakun, A., Bonfil, R., Conover, D.O., Dayton, P., Doukakis, P., Fluharty, D., Heneman, B., Houde, E.D., Link, J., Livingston, P.A., Mangel, M., McAllister, M.K., Pope, J. & Sainsbury, KJ. (2004). Ecosystem-based fishery management. Science, 305, 346-347.

Pinnegar, J.K., Polunin, N.V.C., Francour, P., Badalamenti, F., Chemello, R., Harmelin- Vivien, M.L., Hereu, B., Milazzo, M., Zabala, M., D'Anna, G. & Pipitone, C. (2000). Trophic cascades in benthic marine ecosystems: lessons for fisheries and protected-area management. Environmental Conservation, 27, 179-200.

181 Polls, G.A. & Strong, D.R. (1996). Food web complexity and community dynamics. American Naturalist, 147, 813-846.

Polis, G.A., Myers, C.A. & Holt, R.D. (1989). The ecology and evolution of intraguild predation - potential competitors that eat each other. Annual Review of Ecology and Systematics, 20, 297-330.

Power, M.E. (1990). Effects offish in river food webs. Science, 250, 811-814.

Power, M.E. (1992). Top-down and bottom-up forces in food webs: do plants have primacy? Ecology, 73, 733-746.

Power, M.E., Matthews, W.J., Stewart, A.J. (1985). Grazing minnows, piscivorous bass, and stream algae: dynamics of a strong interaction. Ecology, 66, 1448-1456.

Power, M.E., Tilman, D., Estes, J.A., Menge, B.A., Bond, WJ., Mills, L.S., Daily, G.C., Castilla, J.C., Lubchenco, J. & Paine, R.T. (1996). Challenges in the quest for keystones: identifying keystone species is difficult but essential to understanding how loss of species will affect ecosystems. BioScience, 46, 609-620.

Powers, S.P. & Gaskill, D. (2005). Bay scallop-cownose ray interactions, Final Report, North Carolina Fishery Resources Grant Program, North Carolina Sea Grant, Raleigh, NC.

Pratt, H.L. (1979). Reproduction in the blue shark, Prionace glauca. Fishery Bulletin, 77, 445-470.

Pyper, B.J. & Peterman, R.M. (1998). Comparison of methods to account for autocorrelation in correlation analysis of fish data. Canadian Journal of Fisheries and Aquatic Sciences, 55, 2127-2140.

Ripley, W.E. (1946). The soupfin shark and the fishery. Fishery Bulletin, 64, 7-37.

Rogers, S.I. & Ellis, J.R. (2000). Changes in the demersal fish assemblages of British coastal waters during the 20th century. ICES Journal of Marine Science, 57, 866- 881.

Roman, J. & Palumbi, S.R. (2003). Whales before whaling in the North Atlantic. Science, 301, 508-510.

Rooney, N., McCann, K., Gellner, G. & Moore, J.C. (2006). Structural asymmetry and the stability of diverse food webs. Nature, 442, 265-269. Rose, D.A. (1996). An overview of world trade in sharks and other cartilaginous fishes. TRAFFIC International, Cambridge, UK, 106pp.

Russ, G.R. & Alcala, A.C. (1998). Natural fishing experiments in marine reserves 1983- 1993: roles of life history and fishing intensity in family responses. Coral Reefs, 17,399-416.

Russell S.J. (1993). Shark bycatch in the northern Gulf of Mexico tuna longline fishery, 1988-91, with observations on the nearshore directed shark fishery. In: Conservation biology of elasmobranches, ed Branstetter, S. Seattle, Wash.: National Marine Fisheries Service, pp. 19-29.

Safina, C, Rosenberg, A.A., Myers, R.A., Quinn II, T.J. & Collie, J.S. (2005). U.S. ocean fish recovery: staying the course. Science, 309, 707-708.

Sala, E., Boudouresque, C.F. & Harmelin-Vivien, M. (1998). Fishing, trophic cascades, and the structure of algal assemblages: evaluation of an old but untested paradigm. Oikos, 82, 425-439.

Santana, F.M. & Lessa, R. (2004). Age determination and growth of the night shark (Carcharhinus signatus) off the northeastern Brazilian coast. Fishery Bulletin, 102,156-167.

SAS. (2004). SAS/STAT® 9.1 User's Guide. SAS Institute Inc., Cary,'NC

SAS (2005). The Glimmix Procedure, Nov. 2005. SAS. Available from http://support.sas/com/rnd/app/papers/glimmix.pdf [accessed December 2005].

Schindler, D., Essington, T., Kitchell, J., Boggs, C. & Hilborn, R. (2002). Sharks and tunas: fisheries impacts on predators with contrasting life histories. Ecological Applications, 12, 735-748.

Schmitz, O.J., Hamback, P.A. & Beckerman, A.P. (2000). Trophic cascades in terrestrial systems: a review of the effects of carnivore removals on plants. American Naturalist, 155, 141-153.

Schwartz, F.J. (1984). Occurrence, abundance, and biology of the blacknose shark, Carcharhinus acronotus in North Carolina. Northeast Gulf Science, 7, 29-46.

Schwartz, F.J. (1990). Mass migratory congregations and movements of several species of cownose rays, genus Rhinoptera: a world-wide review. Journal of the Elisha Mitchell Scientific Society, 106, 10-13.

183 Scott, M.D. (2007). IATTC research on reducing shark bycatch in the tuna purse-seine fishery in the Eastern Tropical Pacific Ocean. Scientific Committee Third Regular Session, Western and Central Pacific Fisheries Commission, WCPFC-SC3-EB SWG/IP-3, Honolulu, U..S.A.

Shapiro, S. (1950). The Japanese long-line fishery for tunas. Commercial Fisheries Review, 12, 1-26.

Shears, N.T. & Babcock, R.C. (2002). Marine reserves demonstrate top-down control of community structure on temperate reefs. Oecologia, 132,131-142.

Shears, N.T. & Babcock, R.C. (2003). Continuing trophic cascade effects after 25 years of no-take marine reserve protection. Marine Ecology Progress Series, 246, 1-16.

Shepherd, T.D. & Myers, R.A. (2005). Direct and indirect fishery effects on small coastal elasmobranchs in the northern Gulf of Mexico. Ecology Letters, 8, 1095-1104.

Shiomoto, A., Tadokoro, K., Nagasawa, K. & Ishida, Y. (1997). Trophic relations in the subArctic North Pacific ecosystem: Possible feeding effect from pink salmon. Marine Ecology Progress Series, 150, 75-85.

Shurin, J.B., Borer, E.T., Seabloom, E.W., Anderson, K., Blanchette, C.A., Broitman, B., Cooper, S.D. & Halpern, B.S. (2002). A cross-ecosystem comparison of the strength of trophic cascades. Ecology Letters, 5, 785-791.

Sibert, J., Hampton, J., Kleiber, P. & Maunder, M. (2006). Biomass, size, and trophic status of top predators in the Pacific Ocean. Science, 314, 1773-1776.

Sih, A., Englund, G. & Wooster, D. (1998). Emergent impacts of multiple predators on prey. Trends in Ecology & Evolution, 13, 350-355.

Silliman, B.R. & Bertness, M.D. (2002). A trophic cascade regulates marsh primary productivity. Proceedings of the National Academy of Sciences of the United States of America, 99,10500-10505.

Simpfendorfer, C.A., Goodreid, A.B. & McAuley, R.B. (2001). Size, sex and geographic variation in the diet of the tiger shark, Galeocerdo cuvier, from Western Australian waters. Environmental Biology of Fishes, 61, 37-46.

Skomal, G.B. & Natanson, L.J. (2003). Age and growth of the blue shark (Prionace glauca) in the North Atlantic Ocean. Fishery Bulletin, 101, 627-639.

184 Smale, M.J. (2005). The diet of the ragged-tooth shark Carcharias taurus Rafinesque 1810 in the Eastern Cape, South Africa. African Journal of Marine Science, 27, 331-335.

Sminkey, T.R. & Musick, J.A. (1995). Age and growth of the sandbar shark, Carcharhinus plumbeus, before and after population depletion. Copeia, 4, 871- 883.

Smith, J.W. & Merriner, J.V. (1985). Food habits and feeding behavior of the cownose ray, Rhinoptera bonasus, in Lower Chesapeake Bay. Estuaries, 8, 305-310.

Smith, J.W. & Merriner, J.V. (1986). Observations on the reproductive biology of the cownose ray, Rhinoptera bonasus, in Chesapeake Bay. Fishery Bulletin, 84, 871- 877.

Smith, J.W. & Merriner, J.V. (1987). Age and growth, movements and distribution of the cownose ray, Rhinoptera bonasus, in Chesapeake Bay. Estuaries, 10, 153-164.

Smith, T. (1994) Scaling fisheries: the science of measuring the effects of fishing, 1855- 1955. Cambridge University Press, Cambridge.

Snyder, W.E., Snyder, G.B., Finke, D.L. & Straub, C.S. (2006). Predator biodiversity strengthens herbivore suppression. Ecology Letters, 9, 789-796.

Soule, M.E., Estes, J.A., Berger, J. & Del Rio, CM. (2003). Ecological effectiveness: Conservation goals for interactive species. Conservation Biology, 17, 1238-1250.

Soule, M.E., Estes, J.A., Miller, B. & Honnold, D.L. (2005). Strongly interacting species: conservation policy, management, and ethics. Bioscience, 55, 168-176.

Springer, A.M., Estes, J.A., van Vliet, G.B., Williams, T.M., Doak, D.F., Danner, E.M., Forney, K.A. & Pfister, B. (2003). Sequential megafaunal collapse in the North Pacific Ocean: an ongoing legacy of industrial whaling? Proceedings of the National Academy of Sciences of the United States of America, 100, 12223-12228.

Springer, S. (1967) Social organization of shark populations. In: Sharks, skates and rays (eds. Gilbert, PW, Mathewson, RW & Rail, RP). John Hopkins Press, Baltimore, MD.

Steele, M.A., Schroeter, S.C., Carpenter, R.C. & Kushner, D.J. (2006). Top-down vs. bottom-up effects in kelp forests. Science, 313,1738-1738.

185 Steneck, R.S. (1998). Human influences on coastal ecosystems: does overfishing create trophic cascades? Trends in Ecology & Evolution, 13, 429-430.

Steneck, R.S. & Sala, E. (2005). Large marine carnivores: trophic cascades and top-down controls in coastal ecosystems past and present. Ch. 7 In: Large carnivores and the conservation of biodiversity, (eds. Ray, J.C., Redford, K.H., Steneck, R.S. & Berger, J.), pp. 110-137. Island Press: Washington DC.

Stevens, J.D., Bonfil, R., Dulvy, N.K. & Walker, P.A. (2000). The effects of fishing on sharks, rays, and chimaeras (chondrichthyans), and the implications for marine ecosystems. ICES Journal of Marine Science, 57, 476-494.

Stibor, H., Vadstein, O., Diehl, S., Gelzleichter, A., Hansen, T., Hantzsche, F., Katechakis, A., Lippert, B., Loseth, K., Peters, C, Roederer, W., Sandow, M., Sundt-Hansen, L. & Olsen, Y. (2004). Copepods act as a switch between alternative trophic cascades in marine pelagic food webs. Ecology Letters, 7, 321- 328.

Stillwell, C.E. & Kohler, N.E. (1982). Food, feeding-habits, and estimates of daily ration of the shortfin mako {Isurus oxyrinchus) in the Northwest Atlantic. Canadian Journal of Fisheries and Aquatic Sciences, 39, 407-414.

Stillwell, C.E. & Kohler, N.E. (1993). Food habits of the sandbar shark Carcharhinus plumbeus off the United States northeast coast, with estimates of daily ration. Fishery Bulletin, 91, 138-150.

Stobutzki, I.C., Miller, M.J., Heales, D.S. & Brewer, D.T. (2002). Sustainability of elasmobranchs caught as bycatch in a tropical prawn (shrimp) trawl fishery. Fishery Bulletin, 100, 800-821.

Strasburg, D.W. (1957). Distribution, abundance, and habits of pelagic sharks in the Central Pacific Ocean. U.S. Fish and Wildlife Service, Fishery Bulletin, 58, 335- 361.

Strong, D.R. (1992). Are trophic cascades all wet? The redundant differentiation in trophic architecture of high diversity ecosystems. Ecology, 73, 747-754.

Suzuki, Z., Warashina, Y. & Kishida, M. (1977). The comparison of catches by regular and deep tuna longline gears in the western and central equatorial Pacific. Bulletin of Far Seas Fisheries Research Laboratory, 15, 51-89.

Swain, D.P. & Sinclair, A.F. (2000). Pelagic fishes and the cod recruitment dilemma in the Northwest Atlantic. Canadian Journal of Fisheries and Aquatic Sciences, 57, 1321-1325.

186 Terborgh, J., L. Lopez, P. Nunez V., M. Rao, G. Shahabuddm, G. Onhuela, M. Riveros, R. Ascanio, G. H. Adler, T. D. Lambert & Balbas, L. (2001). Ecological meltdown in predator-free forest fragments. Science, 294, 1923-1925.

The R Project. (2007). The R Project for Statistical Computing, http://www.r-proiect.org

Thomas, CD., Cameron, A., Green, R.E., Bakkenes, M., Beaumont, L.J., Collingham, Y.C., Erasmus, B.F.N., de Siqueira, M.F., Grainger, A., Hannah, L., Hughes, L., Huntley, B., van Jaarsveld, A.S., Midgley, G.F., Miles, L., Ortega-Huerta, M.A., Peterson, A.T., Phillips, O.L. & Williams, S.E. (2004). Extinction risk from climate change. Nature, All, 145-148.

Trzcinski, M.K., Mohn, R. & Bowen, W.D. (2007). Continued decline of an Atlantic cod population: how important is gray seal predation? Ecological Applications, 16, 2276-2292.

Ulrich, G.F. (1996) Fishery independent monitoring of large coastal sharks in South Carolina (1993-1995), final report. U.S. NOAA and Interjurisdictional Fisheries Act NA47FI0347-01. NOAA Fisheries.

UNEP. (2002). Report on the Sixth Meeting of the Conference of the Parties to the Convention on Biological Diversity (UNEP/CBD/COP/6/20/Part 2) Strategic Plan Decision VI/26, United Nations Environment Program.

Uozumi, Y. & Okamoto, H. (1997). Research on the hook depth of longline gear in the 1995 research cruise of the R/V Shoyo Marun. Technical Report Working Paper 3, Secretariat of the Pacific Community, Noumea, New Caledonia. van der Elst, R.P. (1979). A proliferation of small sharks in the shore-based Natal sport fishery. Environmental Biology of Fishes, 4, 349-362. van Houwelingen, H.C., Arends, L.R. & Stijnen, T. (2002). Advanced methods in meta analysis: multivariate approach and meta-regression. Statistics in Medicine, 21, 589-624.

Vaughan, D.S. & Nance, J.M. (1998) Estimates of bycatch of mackerel and cobia in the U.S. South Atlantic shrimp trawls. Stock assessment report for the Gulf of Mexico Fishery Management Council and the South Atlantic Fishery Management Council. NOAA Fisheries Service, Galveston, Texas.

Venables, W.N. & Ripley, B.G. (1999). Modern Applied Statistics with S-Plus. 3rd edn. Statistics and computing. Springer, New York. Venables, W.N. & Dichmont, CM. (2004). GLMs, GAMs and GLMMs: an overview of theory for applications in fisheries research. Fisheries Research, 70, 319-337.

Verity, P.G. & Smetacek, V. (1996). Organism life cycles, predation, and the structure of marine pelagic ecosystems. Marine Ecology Progress Series, 130, 277-293.

Vitousek, P.M., Mooney, H.A., Lubchenco, J. & Melillo, J.M. (1997). Human domination of Earth's ecosystems. Science, 227, 494-499.

Wade, P.R. et al. (2007). Killer whales and marine mammal trends in the North Pacific - a re-examination of evidence for sequential megafauna collapse and the prey- switching hypothesis. Marine Mammal Science, published online December 2006, doi: 10.1111/j.l 748-7692.2006.00093.x.

Walters, C. & 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, 39-50.

Ward, P. & Myers, R.A. (2005a). Inferring the depth distribution of catchability for pelagic fishes and correcting for variations in the depth of longline gear. Canadian Journal of Fisheries and Aquatic Sciences, 62, 1130-1142.

Ward, P. & Myers, R.A. (2005b). Shifts in open-ocean fish communities coinciding with the commencement of commercial fishing. Ecology, 86, 835-847.

Ward, P. & Myers, R.A. (2007). Bait loss and its potential effects on fishing power in pelagic longline fisheries. Fisheries Research, 86, 69-76.

Ward, P., Myers, R.A. & Blanchard, W. (2004). Fish lost at sea: the effect of soak time, on pelagic longline catches. Fishery Bulletin, 102, 179-195.

Ware, D.M. & Thomson, R.E. (2005). Bottom-up ecosystem trophic dynamics determine fish production in the Northeast Pacific. Science, 308,1280-1284.

Wathne, F. (1959). Summary report of exploratory long-line fishing for tuna in Gulf of Mexico and Caribbean sea, 1954-1957. Commercial Fisheries Review, 21, 1-26.

Watson, J.W., Epperly, S.P., Shah, A.K. & Foster, D.G. (2005). Fishing methods to reduce sea turtle mortality associated with pelagic longlines. Canadian Journal of Fisheries and Aquatic Sciences, 62, 965-981.

Whitehead, F£. (2002). Estimates of the current global population size and historical trajectory for sperm whales. Marine Ecology Progress Series, 242, 295-304.

188 Whitney, N.M. & Crow, G.L. (2007). Reproductive biology of the tiger shark (Galeocerdo cuvier) in Hawaii. Marine Biology, 151, 63-70.

Williams, P.G. (1999). Shark and related species catch in tuna fisheries of the tropical Western and Central Pacific Ocean. In: Case studies of the management of elasmobranch fisheries, ed Shotton, R. FAO, Rome, pp. 860-879.

Williams, T.M., Estes, J.A., Doak, D.F. & Springer, A.M. (2004). Killer appetites: assessing the role of predators in ecological communities. Ecology, 85, 3373- 3384.

Wilmers, C.C. & Getz, W.M. (2005). Gray wolves as climate change buffers in Yellowstone. Plos Biology, 3, 571-576.

Witman, J.D. & Sebens, K.P. (1992). Regional variation in fish predation intensity - a historical perspective in the Gulf of Maine. Oecologia, 90, 305-315.

Witzell, W.N. (1985). The incidental capture of sharks in the Atlantic United States fishery conservation zone reported by the Japanese tuna longline fleet. NOAA Tech. Rep. NMFS. 31, 21-22.

Woodin, S.A. (1981). Disturbance and community structure in a shallow-water sand flat. Ecology, 62, 1052-1066.

Wootton, J.T. (1994). The nature and consequences of indirect effects in ecological communities. Annual Review of Ecology, Evolution, and Systematics, 25, 443- 466.

Worm, B. & Myers, R.A. (2003). Meta-analysis of cod-shrimp interactions reveals top- down control in oceanic food webs. Ecology, 84, 162-173.

Worm, B., Lotze, H.K., Hillebrand, H. & Sommer, U. (2002). Consumer versus resource control of species diversity and ecosystem functioning. Nature, 417, 848-851.

Worm, B., Sandow, M., Oschlies, A., Lotze, H.K. & Myers, R.A. (2005). Global patterns of predator diversity in the open oceans. Science, 309, 1365-1369.

Yamaguchi, A., Kawahara, I. & Ito, S. (2005). Occurrence, growth and food of longheaded eagle ray, Aetobatus flagellum, in Ariake Sound, Kyushu, Japan. Environmental Biology of Fishes, 74, 229-238.

Yodzis, P. (2001). Must top predators be culled for the sake of fisheries? Trends in Ecology & Evolution, 16, 78-84. Yokota, K., Kiyota, M. & Minami, H. (2006). Shark catch in a pelagic longline fishery: comparison of circle and tuna hooks. Fisheries Research, 81, 337-341.

190 Appendix 1

Models of 1992-2005 Observer Data

In addition to the effects of year discussed in Chapter 3, many other covariates affected shark catch rates (Tables 3.7, Al.1-1.8). In particular, the area, season, and ocean depth fished were highly significant for all of the modelled sharks (including significant interactions between area and season for all species except the coastal shark group). Temperature also significantly affected the catch rates of most species: as expected, catch rates of the temperate species (blue and mako sharks) decreased with temperature, while catch rates of the tropical and subtropical species (oceanic whitetip, hammerheads, and large coastal group) increased with rising temperature.

Among the variables characterizing the fishing operation itself, target species was highly significant for all species except blue and hammerhead sharks (Table 3.7). For these species, however, the number of light sticks used and time of day fished, which also characterize the species being targeted, were both significant. The soak time of the gear also significantly affected the catch rates of several shark species: catch rates of blue, mako, and tiger sharks all increased with soak time, whereas hammerhead and thresher shark catch rates declined. These results agree qualitatively with a previous study of soak time effects on catch rates for blue and mako sharks, but not for threshers (Ward et al. 2004). Hook depth was less important in the models overall (Table 3.7), but was highly significant for thresher sharks, with their catch rates increasing at greater hook depth, as expected (Ward & Myers 2005a). Catch rates of oceanic whitetip, silky, and dusky sharks (the species modelled in Chapter 2) have previously been shown to decline as hook depths increase (Ward & Myers 2005a) and to increase with soak time (Ward et al. 2004), whereas here, hook depth positively affected oceanic whitetip catch rates and had no effect on the coastal shark groups, while soak time had no effect on these species. In contrast to the previous studies, which modelled hook-specific data for soak time and hook depth, the observer data modelled here were only collected for individual sets, not

191 individual hooks, and thus had to rely on approximations for these two variables, which I suspect is the cause of these differences.

Hook type significantly affected tiger, mako, and the coastal shark group catch rates, with more of each of these species caught on circle than J-hooks, but did not have any detectable influence on catch rates of the other sharks. Because many fisheries have switched (or are switching) from J-hooks to circle hooks, researchers have been conducting gear experiments to investigate differences in catch rates of target species and bycatch species, like sharks and sea turtles, between these hooks types (e.g. Watson et al. 2005, Kaplan et al. 2007). Such studies may be limited by low sample size to report differences only for the most frequently caught species (e.g. blue shark, Kerstetter & Graves 2006, Yokota et al. 2006), and further investigation of observer data from pelagic longline fisheries (e.g. including analysis of the status of shark species caught (alive vs. dead)) may prove to be fruitful.

Final GLMM-vt models

Table A1.1. Parameter estimates (including standard error and p-value) for fixed effects in final GLMM-vt model of 1992-2005 observer data for blue shark.

Standard Effect Area Period Estimate Error DF t Value Pr>|t| Intercept -6.1691 0.5663 97 -10.89 <.0001 nyear -0.05736 0.01038 3830 -5.53 <.0001 oceandl -0.03764 0.04542 3830 -0.83 0.4073 oceand2 -0.08014 0.02958 3830 -2.71 0.0068 sin -0.2154 0.8649 3830 -0.25 0.8033 sin2 -0.2765 0.2576 3830 -1.07 0.2831 cos 0.2305 0.08271 3830 2.79 0.0053 cos2 -0.6953 0.5224 3830 -1.33 0.1832 soaktime 0.05614 0.01153 3830 4.87 <.0001 tempi -0.4257 0.04685 3830 -9.09 <.0001 nlightst -0.00033 0.000162 3830 -2.02 0.0432 area 3 -1.5781 0.5533 3830 -2.85 0.0044 area 4 -1.4319 0.5683 3830 -2.52 0.0118 area 5 0.6069 0.5462 3830 1.11 0.2665 area 6 0.9583 0.5619 3830 1.71 0.0882 area - 7 2.2459 0.6086 3830 3.69 0.0002 area 8 0 period 1 0.1146 0.06761 3830 1.69 0.0902

192 period 2 -0.1188 0.1751 3830 -0.68 0.4976 period 3 0.07738 0.2081 3830 0.37 0.7101 period 4 -0.08777 0.0494 3830 -1.78 0.0757 period 5 0 sin*area 3 1.339 0.8885 3830 1.51 0.1319 sirfarea 4 1.3839 0.9103 3830 1.52 0.1285 sin*area 5 0.256 0.8665 3830 0.3 0.7677 sin*area 6 0.1097 0.882 3830 0.12 0.901 sin*area 7 0.4743 0.9554 3830 0.5 0.6196 sin*area 8 0 sin2*area 3 -0.6167 0.2904 3830 -2.12 0.0338 sin2*area 4 0.2956 0.2883 3830 1.03 0.3053 sin2*area 5 0.1499 0.2713 3830 0.55 0.5805 sin2*area 6 -0.2817 0.2952 3830 -0.95 0.34 sin2*area 7 0.5303 0.3134 3830 1.69 0.0907 sin2*area 8 0 cos2*area 3 0.7932 0.5574 3830 1.42 0.1548 cos2*area 4 1.1152 0.5615 3830 1.99 0.0471 cos2*area 5 0.5165 0.5302 3830 0.97 0.33 cos2*area 6 0.3612 0.553 3830 0.65 0.5137 cos2*area 7 1.0454 0.5796 3830 1.8 0.0713 cos2*area 8 0

Table A1.2. Parameter estimates (including standard error and p-value) for fixed effects in final GLMM-vt model of 1992-2005 observer data for mako sharks. Hook Target Standard Effect type species Area Bait Estimate Error DF t Value Pr>| Intercept -17.4878 3.9215 127 -4.46 <.0001 year -0.03217 0.01596 4264 -2.02 0.0439 oceandl 0.5937 0.07396 4264 8.03 <.0001 oceand2 0.2204 0.04622 4264 4.77 <.0001 sin 9.0291 4.3278 4264 2.09 0.037 sin2 -4.1904 2.1105 4264 -1.99 0.0472 cos 7.0588 4.1404 4264 1.7 0.0883 cos2 0.4392 0.7995 4264 0.55 0.5828 soaktime 0.05344 0.01616 4264 3.31 0.001 tempi -0.127 0.07709 4264 -1.65 0.0997 area 2 7.6952 3.9104 4264 1.97 0.0491 area 3 6.987 3.9131 4264 1.79 0.0742 area 4 7.1785 3.9122 4264 1.83 0.0666 area 5 9.5069 3.9128 4264 2.43 0.0152 area 6 8.9432 3.9437 4264 2.27 0.0234 area 7 6.6919 4.4905 4264 1.49 0.1362 area 8 0 targetsp BET -0.2531 0.3715 4264 -0.68 0.4958 targetsp MIX 0.542 0.1894 4264 2.86 0.0042 targetsp SWO 0.9232 0.2047 4264 4.51 <.0001 targetsp TUN -0.1638 0.2177 4264 -0.75 0:4519

193 m m •<- •<- CM co co in N s en T- CO CO CM S CO S -st- •>* *-r -=t- co CM •* CD CM "=f- CM •>* co o o o i^- -<*• CD r- in o i- ••tf- •* CM -C*O CM o T— CO T— CD CM C7J ^ m ^ m co o O * CO CD O S S O) O CO N CM CD CO •* r-- X— CO •* CO OO 037 2 207 4 832 1 o o o o o Q Q CD CD CD •*— •*— o o •>- o o o *- *~ o o en co co I-— co co in ii. o en CD CD CD T— T— C5 o o o o d o d o CD CD CD CD CD o o o o o o o o o o o-o -5 A C3 CD CO oo co o ^ V o V CD CD CD CD ^ _l co CD T- CO CM CO •* CM f— (M CM CO CO t in CO CO T- CM T- CO in CD o CM CM O OS OT O O O CO CD co S CD CD N CO pj CM co CD CD co T- CM o en CJ3 CO CO •>* in T- O T— T— T— CM CM CM T- -r- 1— 1— 1— 1— t— 1— 1 r- CM CM CM CM o o CD T- CD o __CO_ CD co CO o> -3- CM C-M* CO 03 cd T— in CD CD CD CM

fi n • i i > CO 1 1 .E >

•*r- **• -* •* «* -* -d- -* -* •>* •**• "* "* •* •*!- •* •* •* •* •* •* co CD CD CD CD CD CD CD CD CO CO CO CD CO CO CO CO CD CO CO CO co co co co co co d> CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM ££z in CM CM CM CM CM CM •* •* •<3- U- •* >* •* **•<*• •<* -* "* •<*- -* "* •«*• -* •* -* -* •* -* QJ in *C*5 •C*D CD CD •C*D CD O «* -* ** "5R Ll_ m in in in IO m Sc o *^ *— co co in co co CD CO O) * CM CO CO O) Mo— O) IO CO CM O O CD (O * IO O CM CM CM CO CO CO CD CM O CO CO t- CD CO CD CD t-~ CM f- OO oo CM CD in OO

122 9 178 7 164 9 114 2 120 3 114 1 152 9 140 8 145 6 144 4 141 6 181 9 -=1- co n n ^ co s 007 7 co co co co co o 1— en co 1 P lue ) CO CD CO CD i-- in CM •«* •* •* CM CM CM CM •«*• CO T— T— -* CM LU o o o -a- in o o o o o -^ •o r-- CD r-~ r-~ "* •>* 2 CD c= o CD o o

Sta n o ai CD p- v C3 XJ co O CM o co o -* in co in CM co r~. r«. CD in •* CM O CO Tt U3 O) t ID o co o> co o> in r— CD co co •* co m CM C3) CM O) CD CM •* N m N co co • O) T- N CO C3) N CD in CM r— co "* o co co N CO IO CD CO •* N C~O~ " N -<- in ^ OJ CO O i- CO ^f CO r~-' co r~- co t-~ ^_ CM CM CO (D -+ CM S CM CM CO OV in CM CO CO CD T- in V •c— T— CO CO CO CO & CD CM CM

1349 6 in co r-- co 1 CO S "co co co CD co oo T— CM CO CO CO O) O •* ^J" "* CO CO CO N CO (D S S CM CD o o o o o o co co CD E CD CD fc CO CD CD CD CD "S5 1 • 1 o •g •s ii i • • CO CO XJ .£= a CO CO CO m CO T3 D> CO o CD 9> D. di n -i-^ CMcO'tiDcoNcoNco^mcoscocM co •* cocoscocMco^incosco ^

inc l ^ S2 CO "CD CD .» CO CO 5 •a S 'o CD CD 1CO CO CD P>'o E H- CO >-• $ £ Sj 2 est i CD •^ crT L_ JCDO CD 2£. 4—* O CD O CD in CO o CEO C3 s O -4> t2! CM < CO 1 Q_ CM CO O) CO CO CO CO CH CO CO CO CO CO CO CO CO CO «««gifflginSSepppSgSSS«gg££££S££ "S. CD CD CD CD CD CD CD CD SpjDjOjDIDpe:s:s:s:s;sis;cococDcocococp o < "o ID — CO CO CO CO CO CO CO cococococococo*. CD c .E CO CO CO CO CO CO CO £ CD "55 "o i_ CO CM ~3. CM CM CM CM CM CM CM e CD CO CO CO -a ^2 co £= i» -f -H -s -s -s O O co .*: -*: co CO CO .£.£.£ .c _E _c o o o o £ "£ CD I .C-C-Q-QXICOCOCOCOCOlOlOCOCOCOCOCOCOCOOOO LU o ' o O oCO Tab l E >> IT) s ^t to n oi •* •* a> co •<*• CM co co a> r— ••*• •<*• CM in •* co in t T- (O (O T- CD CD CD CO CO T- CD T- CM r- ** o •* CO T- CD CM co CM a> CD t m co oi ON t- O O O T- •<- o T- m m co co o m CM •* CO x- co o co T- o> CM •* o o CD o o o in O CM T- -<*• i-- CM CD CD CM N CM O OJ N Q CM A CO CO CO CO T-co-<--<-cOT-cocMr-- o o o CD CD d CD CD CD d CD CD CD CD CD CD CD CD O CD CD O S CDCDT-OCMCDCO dodo OOOOOOI-OO qoqqoOM-oo v" CD V V CD CD CD CD CD CM m ** m •* N in M CM CO CO T— in co in in co co •* •* prj CM •* r- CO i— CO <35 ,_' CO O in CO CM O T- T- CD CD t- CO T- CD T- CM CD CO CO CM T- O) O) CM CD d d d^ d d d d > CO K S !M T- co CM -=J- co ^ CD T- CM CM > co co a> •«- CM -* •* CO co CD CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM CM £ 3 -=i- ^ T -3- **• "* -* "* "* •* •* -* •* •**• <* •* <* o o o o o o CD CD O CD CD O CD CD O CD CD CD CD CD CD CD CD m io to in io m iCnD iCDn iCDn CiD n in in in in in in in in in in in in in in in in CD X inininioinininio cooocococococococo cococococococococo T— co co co co co co co co

T3 o> ^~ CM co co lO CD N S O) t- co CM in * co in co S * (M N CM T- in •«- co o> CM in in •* •<*• •* •* CM co CM co co s s co o •*• r— 1- co U) S N O) CO CO •<* OJ CM •* lO r- T- O O CJ> T- T~ CJ5 S CO CO N r-. co i-- m CM in co a> a> r-- CO CM CM CM ,-j CO CO T-; N CO (O N "S TD <=:C O r-~ t~- r— r-- •<* r- co co co co co CO co •* o co co •* c 8 o o o o O CD CD CD CD CD CD CD O CD CD CD CD CD CO CD CJ) CO CO CO in CO CO LU d d d d T— T— T— 0. 0 0. 0 CO CM in CO CO CT> I S o in CD T— T— CM CM CM CD CM co CO LLl CD CD CD CD CD CD CD CD CD TO a> CM CM CO CM i^. CO T— CM CD CO 03 CD - a> co ooioiinnoNcoraoiOT-soooio i_ & i^- CO V t— T— CD •* CM r~- co co o T— co p •*— co co •*- r^ co co oo in CD •**• CO CD O) O) i- CO •* O S CO T- Tf CO CM ~~CM co • n co co o) T— r— r-- •«— 1. 2 0. 2 CD CM CM "CD 0. 2 0. 3 CD CD CD CD CD CD O CD CD CD CD CD if ,216 9 .052 9 ,54 8 .541 4 ,830 4 307 1 189 7 261 1 328 4 1 CD CM co •* m a. is~~

~~-~ CO ' CD CD CD CO *- co P5 'o tig CD CD ™ Si §? si m 2 «i ^J "o aj io CM co •* m co CMcotincocMco^incocMco^inco ^ S CO CD ^ CM~~

CO CO CO CO CO "* ? CO CO CO CO CD CCDD CD CD CO CD T— CM CD O- O. Q_ CX D. co co co co co m at m m i— >= Q. CD T3 CO CO CO CO CO oSSSe.££S2§ COCOCDCOCO 5 o a> T3 T3 T3 aasaa-D-oTJ-oTi ro ro l_o_ cz c= J^ o _ - co co co co co S, 8, S, g, 8, -S .2 .2 .2 .2 Jo « Jo Jo fe * *<° * ««° *„ CM CM CM CM CM CD CD I— CO CO CM .CZ ex £ co P> P" po P5 -c: •?; -c; •=:•!= *_*_*_*_*_ CcOo C«Oo CcocoO CO CoD CcOo CcOo CcOo CcOo CcOo a> CO CD CD CO CO o> CO CO CO CO CO CD ter n LU 8 ° o o o o o >% O o CO o o CO -a-o-o"OD>o>o>. o>o> .. 0000000 CO CO CO CO o- o- cr 03 03 03 03 CD CD CD CD =L1 CD 00000000 m DI si 0!3j 03 03 03 03 < -S S Z3 CD CD CD CD 5. 3. S. ®5 ® m* » ~~CD CD CD CD 03 03 03 03 03 03 03 •Ii*P" 03 03 03 03 eo «"~~

~~1 CD S3 -(*. CO tvi -^~~

~~Ol -la- CO N3 •§•§" CD «> 3 c03o. -n cz g g m a. CD" -1 z o X -1 -o C0D3 c/3 CD 5T~~

~~=3- = 03 Q_~~

~~ist i CQ O • 1 O -»• CD O A i. CD 0 O 0 CD _*. N3 3~~

~~,222 3 ,543 8 .259 1 1010 4 ,508 5 ,103 5 c,307 6 b CD CO CO 395 3 240 4 in 183 6 CD ^-»- o bo o 1 -»• co 3 en —* ~j en CO CD 03 3-sr~~

N3 -ta. O OO CD CD N3 • CD CO CD CO CO —>• CD CD CO 03 1.540 1 1.510 6 1.45 8 1.123 2 rl-, <= ,843 4 ,744 6 ,0952 3 ,0199 3 479 2 464 7 023 4 678 2 464 6 a. 1 240 3 —». —». O- e— *—> CO 03 CO a 602 6 £98 9 o m of CD 8289 1 CD CD CD CD 0000 o CO CO O CD CO CD CO CD >- !->. Co j^. ^ co co co —»• co —*• co K3 -» -» M K3 M Ul en 00 o CD CD en —j co +>• co co o en -^ -»• co -U co 3 °- s ^ S CD O) co -f^- en -»• co co co a? ->• co i>o en -»• co CO CD CD • CD CD CD • CD CD CD • CD CD • CD 0 CD CD ' CD CO CD " 03 ,286 6 ,755 4 .271 1 ,302 7 ,721 6 ,267 6 ,306 1 ,749 1 ,324 9 ,346 4 ,433 7 .441 6 .439 2 .714 6 .213 8 .305 5 .389 6 CO (D M M . N3 OO -J Ol ~-J -» OO NJ -» OJ -1 *• CO CO —1 03

< 923 E 03 -t*. Ji. -tv Ji. CD O) O) Ol O) —I —I -~J —I CO eo eo • CO eo eo • eo eo co • CO co • eo eo eo eo • co co co en en en en CO CO o> 05 03 CO CO 03 CO CO 0 0 CT3 03 CO CO CO CO CO CO co co 00 00 00 00 00 CO 00 co 00 co 00 00 00 CO X en en en en en en en Ol en en en en en en en en en en CD CD- CD

CD *• CO -fc. IV3 ro 0. 3 , coco-»io^Lnp*M if N3 K3 • ->• CD CO r° 4*. en cocoj^bcncocobb ;-»• tV3 t\3 r\3 co -^ isi CD K3 •fe. en ^ en bo co co lp- ro<3>CDN>05CJl-|i.CJlJi. o bo en ££ CO Ko CD -4 K> en CD CO -t* -7 o CD CO -t*. CD l\3 (33 01 -t*. o> N3 ro -t^ en en 00 co w

Q3_ CO CD CO AOAOOOCDOOpA A A A P3 CD CD ^ OCOCOOOOO^-»-OOOOCOO CD CO CO CO CO OOO-»UOOC0O(0OOOO CD CD • CD CO co • CD CD CD • CD 0 eo ,073 4 ,010 4 ,050 7 ,179 9 CD CD CD CO " 059 5 091 2 024 1 003 3 M » N3 Ol oooocnoocncocnoooo 113 5 175 5 p A p cn CD -» co -»O)-»NS-^00vl(DM-»-'-i-' CD CD CD CD o o cn -» 03 CD CD Js. O 4^ 4*. cn eo en

89S 0 CO CO CO -J CO CO CD as In t co co co co co co tU 0) Q) 1U D) Q303a3Q303Q3Q3Cooo 0 O ••< oooococococn-a =>' 3' =J' 3" =J" =J"OBO.OfflOBlK-WSiajSjBjii j — — — — — — -000 0 O CD m o o o o 5" 5' 5" 5' CD ****** ^5: «-• ;=V ?».«-• «: S-" ^" ^" (Q CQ CO CQ CQ - -- (/> CD CD 03 CD o. 2 CO » M (O , » , » 2 Qj QJ Q) 03 03 03 "CJ "O CD CD CD CD CD ro CO cp_ ro -* -1 -* -1 -1 -3 =3 13 -> if ro to to ro Q] 03 03 03 O CD CD CD o) m Qi Qi m -*• -*• i3 "a XJ -a -a tO X w,o CD* CD* CD* CD* —^ —» —% —^ *—> •8 s 1 1 wy. a£ 0OJCflDiOJCDCDCDCD3 03 03 03 - - eo o>

~~eo 03 m co ui J^ u M co co 01 -^ co M en~~

~~§•$ oi*.cotocn*.coM ™CD C~~

~~CD 3^|sg oj. 1! CD" S3- co «- o 5f ° I O CD =*" X is ff o o ;-»• O p O ;-»• p 0103 03 CO A W o co o en ^- en 0> Ol -^ U IS3 ->• —4 —v 4* CO CO CO H OOlOICDOOlCO-iO 3.~~

• O 1 O c-> c-> O O O O 000^ —1 _^ OOO O O Q OO O O M M -» O 0 cb C3 0 0 —*• -» M CO M o ->• O ^ CO -»• 4=- iu co ho b cocneooeo-veD^-' CO CO CO -4 CO CO CO -fa. CO CO -»• tO O 4*. CD CD CO On 01 -»• en 4*. rocorococo^ioio) -t* IS1 CIO IV> en ro en S (D M CO O) -10(00 0)0 4^ —J -J en ~ " co co • O O O O —>• o> io en en en M P ro ^ COOOOOOOOOCO-^COCDCDCJIOO—»•• . -»• tO -»• O O —1 —»•—»• en —>• en -li CO to 4»- IO ~ CororotO-»-004i-a3CO-»-CO-40 -c -t>. 4^ 4=- •*» -P* co 00 * s s 0 co 01 05 en co en co en —»• co o -••^4kSMO*>J>IUW01 a> o> co 03 O) 03 Ol Ol N M -1^ 01 CO ^1 M ^ --I en co ro -P- MCOtOMCOCOCICSUCOC CD CO O 00 en co —4 —4 —4 —4-4—4 en en en en en en X o CD Ji * * * * .p.. .p» 4*. 4*. 4*. 4>.4*.4*-4*.4*-4^4>-4>-4*-4*-4*-4*. Q. CO CO 0 CO G) 03 CO CO CO CO O co CO CO CO CO COCOCOCOCOCOCOCOCOCOCOCOCOIO CD en . en en 01 en en en en en en en en . en en en en cncncncncncncncncncncncncn—4 oc f 4*. —x —». -»• Ji. ro co co -4 -P>. 00 00 co .

—>• ro ;-»• ro O p p p p p p p p p p pptoro^:»uo-' '->• ro co ro to co co o 4*- cobj^^cobois-i O -P-. co b ro ^ to bo co co ro ro —4 . co co -4 —4 COCOCONIIO-'NIÔO CO -»• N -vl M CO tO -» 4». en 9L A CD 0 CD A CD 000 1 097 2 CD 088 1 000 1 318 2 909 0 O p O p O p O 0 0 0 0 poooooooooooo A 031 8 co co bo bo bo lp» 091 7 017 1 285 7 013 6 ^jôoôoj^-rocoorooo co -4 4*- co -4 co A to ui to to -'Co-'-'Cocnocococo-'COCoo Ol O -> M CO 4^ ^1 -4 OO 4^ O SCOCOCOCOOlO-iSCO^COCOO co s to co -1 en CD o> -J en en . torocxiNioi-ii^to-'oiaicoco-» -J cos*area 2 -1.4406 0.9736 4165 -1.48 0.139 cos*area 3 -1.2284 0.9706 4165 -1.27 0.2057 cos*area 4 -1.1174 0.9764 4165 -1.14 0.2525 cos*area 5 -2.4467 0.9929 4165 -2.46 0.0138 cos*area 6 -2.7278 1.164 4165 -2.34 0.0192 cos*area 8 0 cos2*area 2 0.6145 0.9476 4165 0.65 0.5167 cos2*area 3 0.6976 0.9534 4165 0.73 0.4644 cos2*area 4 0.3104 0.9617 4165 0.32 0.7469 cos2*area 5 0.9128 0.9865 4165 0.93 0.3549 cos2*area 6 -0.6771 1.1095 4165 -0.61 0.5417 cos2*area 8 0

Table A1.7. Parameter estimates (including standard error and p-value) for fixed effects in final GLMM-vt model of 1992-2005 observer data for coastal group 1 (dusky, silky, night sharks). Target Standard Effect Area species Period Hook type Estimate Error DF t Value Pr>|t| Intercept -10.198 0.5875 122 -17.36 <0001 year -0.1103 0.01741 3805 -6.33 <.0001 oceandl 1.1471 0.09565 3805 11.99 <.0001 oceand2 0.6451 0.09714 3805 6.64 <.0001 sin 0.3239 0.09113 3805 3.55 0.0004 sin2 0.1316 0.07202 3805 1.83 0.0677 cos2 -0.2138 0.08394 3805 -2.55 0.0109 tempi 0.1735 0.08136 3805 2.13 0.033 area 1 3.06 0.5873 3805 5.21 <.0001 area 2 2.1492 0.4946 3805 4.35 <.0001 area 3 1.9213 0.4928 3805 3.9 <.0001 area 4 2.4889 0.4864 3805 5.12 <.0001 area 5 1.2891 0.4754 3805 2.71 0.0067 area 6 0 period 1 0.04481 0.1425 3805 0.31 0.7532 period 2 -0.8635 0.3955 3805 -2.18 0.0291 period 3 -0.76 0.2851 3805 -2.67 0.0077 period 4 -0.2993 0.1077 3805 -2.78 0.0055 period 5 0 targetsp BET 0.68 0.7323 3805 0.93 0.3532 targetsp MIX 1.4273 0.3344 3805 4.27 <.0001 targetsp SWO 1.77 0.3428 3805 5.16 <.0001 targetsp TUN 1.6641 0.3379 3805 4.92 <.0001 targetsp YFT 0 hktypel Circle Hook 0.8763 0.1595 3805 5.5 <.0001 hktypel J Hook 0

198 Table A1.8. Parameter estimates (including standard error and p-value) for fixed effects in final GLMM-vt model of 1992-2005 observer data for coastal group 2 (all Carcharhinus species and all unidentified sharks). Target Standard Pr> Effect Area species Period Hook type Estimate Error DF t Value N Intercept -8.5433 0.4252 122 -20.09 <.0001 year -0.06213 0.01657 3807 -3.75 0.0002 oceandl 1.0095 0.08491 3807 11.89. <.0001 oceand2 0.5604 0.08951 3807 6.26 <.0001 sin 0.3398 0.08326 3807 4.08 <.0001 tempi 0.1852 0.07345 3807 2.52 0.0117 area 1 2.2614 0.4833 3807 4.68 <.0001 area 2 1.658 0.3721 3807 4.46 <.0001 area 3 1.2481 0.3709 3807 3.37 0.0008 area 4 1.8361 0.3641 3807 5.04 <.0001 area 5 0.7653 0.3453 3807 2.22 0.0267 area 6 0 targetsp BET 0.9772 0.4613 3807 2.12 0.0342 targetsp MIX 0.4647 0.2249 3807 2.07 0.0389 targetsp SWO 0.8077 0.2403 3807 3.36 0.0008 targetsp TUN 0.8725 0.2371 3807 3.68 0.0002 targetsp YFT 0 period 1 0.08228 0.1345 3807 0.61 0.5407 period 2 -0.8019 0.318 3807 -2.52 0.0117 period 3 -0.5574 0.223 3807 -2.5 0.0125 period 4 -0.1901 0.0999 3807 -1.9 0.0571 period 5 0 hktypel Circle Hook 0.4602 0.144 3807 3.2 0.0014 hktypel J Hook 0

~~199~~