UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Complexity, cost and quality of ecosystem models and their impact on resilience : a comparative analysis,.. Morissette, Lyne 2007

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

Item Metadata

Download

Media
[if-you-see-this-DO-NOT-CLICK]
ubc_2007-318928.pdf [ 26.25MB ]
Metadata
JSON: 1.0074903.json
JSON-LD: 1.0074903+ld.json
RDF/XML (Pretty): 1.0074903.xml
RDF/JSON: 1.0074903+rdf.json
Turtle: 1.0074903+rdf-turtle.txt
N-Triples: 1.0074903+rdf-ntriples.txt
Original Record: 1.0074903 +original-record.json
Full Text
1.0074903.txt
Citation
1.0074903.ris

Full Text

COMPLEXITY, COST A N D QUALITY OF ECOSYSTEM M O D E L S A N D THEIR IMPACT O N RESILIENCE: A COMPARATIVE ANALYSIS, WITH EMPHASIS O N M A R I N E M A M M A L S A N D T H E G U L F OF ST. L A W R E N C E by LYNE MORIS SETTE B.Sc. Universite de Montreal, 1998 M.Sc. Universite du Quebec a Rimouski, 2001 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Zoology) THE UNIVERSITY OF BRITISH COLUMBIA October 2007 © Lyne Morissette, 2007 A b s t r a c t The development of an ecosystem approach is essential to improve understanding of the nature and dynamics of exploited marine ecosystems and the complexity of species interactions. A database of 393 Ecopath models was compiled to investigate questions pertaining to 1) the uncertainty in model parameters, 2) the relationship between ecosystem complexity and stability, 3) the effect of marine mammals on ecosystem productivity, and 4) the lack of recovery of commercially caught predatory fish species. Reviewing the strengths and weaknesses of Ecopath methods showed that few published models have addressed uncertainty even when the tools were provided with the software. Estimating uncertainty (with confidence intervals for parameters) was improved by combining the Ecopath approach with other modelling techniques. The second analyses, using 50 ecosystem models, showed that quality of input data affects conclusions concerning the relationship between complexity and stability in ecosystems. Findings further support the notion that stability also increases as community complexity (expressed as ascendancy and omnivory) increases. The third analyses used seven ecosystem models to explore whether the presence of marine mammals and their overlap with fisheries for food resources affects overall ecosystem productivity, and whether their extirpation could decrease productivity. Simulation results were equivocal about whether extirpation of marine mammals from ecosystems would benefit fisheries, while mixed trophic impact analyses suggested that beneficial predation does occur in these ecosystems. Finally, a case-study of four Ecopath models from the Gulf of St. Lawrence addressed the collapse in biomass of predatory fish and the potential causes of the lack of recovery of these species. Analyses revealed that the ecosystem structure shifted from one previously dominated by piscivorous groundfish during the mid-1980s to one dominated by small-bodied pelagic species during the mid-1990s. Part of the total mortality of groundfish was accounted for by seal predation when fishing mortality was reduced. These analyses collectively demonstrate the value of investigating ecological questions applicable to marine ecosystems using a multi-model comparative approach. ii T a b l e o f C o n t e n t s Abstract ii Table of Contents iii List of tables vi List of figures viii List of acronyms xii Acknowledgements Remerciements xiv Co-authorship statement xviii INTRODUCTORY CHAPTER: The role of ecosystem models, particularly Ecopath with Ecosim, as a transition to ecosystem based management 1 1.1 Introduction 1 1.1.1 Studying marine ecosystems '. 1 1.1.2 Ecosystem modelling 2 1.1.3 Census of Ecopath models : 6 1.2 Results: the models database 6 1.2.1 Geographical coverage 8 1.2.2 Structure and quality of Ecopath models 11 1.2.3 Rationale of Ecopath modelling projects 12 1.2.4 Publishing Ecopath models 13 1.3 Discussion & Conclusions 14 1.3.1 Research questions 17 1.4 Chapter summary 19 1.5 References 20 CHAPTER 2: Addressing uncertainty in marine ecosystem modelling 26 2.1 Introduction 26 2.2 Ecopath ecosystem modelling and its sources of uncertainty 26 2.3 Sensitivity analyses as a tool to address uncertainty 30 2.4 Sensitivity analysis in Ecopath 31 2.5 Pedigree and Ecoranger routines 33 2.6 Autobalance 40 2.7 A step further: combining Ecopath with other modelling approaches 43 2.8 Conclusions 45 2.9 Chapter summary 46 2.10 References 47 CHAPTER 3: Data quality and complexity in ecosystem models: does it affect resilience and model performance? 52 3.1 Introduction 52 3.2 Material and methods 54 3.2.1 Ecopath models and their pedigree 54 3.2.2 Costs of computing trophic interaction in Ecopath models 56 iii 3.2.3 Quantifying the complexity of models 58 3.2.4 Dynamic simulations of ecosystem's stability with Ecosim 60 3.2.5 Statistical methods 61 3.3 Results 61 3.3.1 Pedigree and stability (simple regression) 65 3.3.2 Predicting ecosystem's resilience (multiple regression) 67 3.3.3 Complexity versus stability 68 3.4 Discussion 69 3.4.1 About the modelling approach 69 3.4.2 Data quality and associated modelling outcomes 70 3.4.3 Cost-effectiveness of ecosystem models 71 3.4.4 On complexity-stability relationship 73 3.5 Chapter summary 74 3.6 References 75 CHAPTER 4: Modelling the indirect positive feedback of marine mammal predation: outcomes for competition with fishery and ecosystem structure .....85 4.1 Introduction 85 4.2 Materials and methods 87 4.2.1 Global ecosystem representation 87 4.2.2 Modelling approach 91 4.2.3 Static ecosystem indices (Ecopath outputs) 91 4.2.4 Dynamic simulations (Ecosim) 94 4.3 Results 96 4.3.1 Total annual food consumption of marine mammals and fisheries catches 96 4.3.2 Mean trophic level of marine mammals' consumption and fisheries catches 102 4.3.3 Primary production required to sustain marine mammal consumption and fisheries catch 103 4.3.4 Resource overlap between marine mammals and fisheries 104 4.3.5 Marine mammals and their impact on the trophic structure 104 4.3.5.1 Eastern Bering Sea 105 4.3.5.2 Northern Gulf of St. Lawrence 110 4.3.5.3 Benguela system 116 4.3.5.4 Eastern tropical Pacific Ocean 122 4.3.5.5 Gulf of Thailand 128 4.3.5.6 North Sea 135 4.3.5.7 Strait of Georgia 141 4.3.6 Mixed trophic impact versus catch or consumption 147 4.4 Discussion 148 4.4.1 Resource overlap and trophic levels 149 4.4.2 Primary production required to sustain marine mammals and fisheries 152 4.4.3 Comparing the mixed trophic impact of marine mammals and fisheries 152 4.4.4 What if marine mammals were not present in these ecosystems? 153 4.4.5 Marine mammals and beneficial predation 154 4.4.6 Strengths and weaknesses of the modelling approach 156 4.5 Conclusions 158 4.6 Chapter summary 158 4.7 References 159 iv CHAPTER 5: A case study in ecosystem modelling: Contrasting changes between the northern and southern Gulf of St. Lawrence ecosystems associated with the collapse of groundfish stocks 172 5.1 Introduction 172 5.2 Materials and Methods 173 5.2.1 Study areas 173 5.2.2 Structure of the models 174 5.2.3 Data used for the models 177 5.2.4 Modelling approaches 178 5.3 Results 180 5.3.1 Biomass 180 5.3.2 Consumption 181 5.3.3 Mortality 184 5.3.4 Trophic levels 189 5.3.5 Emergent properties of ecosystems 192 5.4 Discussion 193 5.4.1 Biomass structure of ecosystems 194 5.4.2 Changes in consumption 195 5.4.3 Predation by apex predators 196 5.4.4 Impact of fishing 197 5.4.5 Emergent properties of the models 198 5.5 Chapter summary 201 5.6 References 201 CHAPTER 6: Conclusion 207 6.1 Strengths and weaknesses of this thesis research 210 6.2 Future improvements and applications 211 6.3 Final conclusions and recommendations 211 6.4 References 211 APPENDICES 213 Appendix 1: Questionnaire sent to Ecopath users 214 Appendix 2: List of Ecopath models (models highlighted in grey are the one that were already available on the Ecopath website) and their respective reference, when published 215 v L i s t o f tables Table 1.1. Number of Ecopath models constructed for different habitat types 9 Table 1.2. The eleven most important regions in terms of Ecopath models constructed 9 Table 1.3. Number of Ecopath models for different size categories, in terms of trophic groups 11 Table 1.4. Number of Ecopath for different levels of quality (pedigree) 11 Table 1.5. Number of Ecopath models published by type of publication 13 Table 1.6. Most notable authors in terms of Ecopath models constructed, according to the survey 14 Table 2.1. Default options for the pedigree routine, for each input parameter used in Ecopath models. Defaults (percentage CI) are means based on actual estimates of CI in various studies. Modified from Christensen et al. (2000) 35 Table 2.2. Pedigree of biomass (E), production (P/B), consumption (Q/B), diet and catch inputs for the fish components of the Icelandic shelf ecosystem model for the 1950s constructed by Buchary (2001) with Ecopath. The overall pedigree of this model was 0.164 36 Table 2.3. Pedigree of biomass (B), production (P/B), consumption (Q/B), diet and catch inputs for the fish components of the northern Gulf of St. Lawrence ecosystem model constructed by Morissette et al. (2003) with Ecopath. The overall pedigree of this model was 0.651 36 Table 2.4. Ecoranger results for two models with different pedigrees 38 Table 2.5. Parameters causing failure for the first 50 runs of Ecoranger, with the related problematic value of each criterion 39 Table 3.1. List of Ecopath models used for the present study 55 Table 3.2. Default options for the pedigree of biomass used in Ecopath models. Defaults (Percentage confidence intervals [CI]) are means based on actual estimates of CI in various studies. Modified from Christensen et al. (2000) 56 Table 3.3. Relative value of publication by types as used in this study (from Rounsefell 1961) 57 Table 3.4. Types of publications describing the Ecopath models sampled for the present analysis 62 Table 3.5. Embodied costs and quality of Ecopath models depending on the type of publication they represent 62 Table 3.6. Effect of the significant regressors for predicting resilience in a multiple regression for all ecosystem models (n = 50) 67 Table 4.1. Ecopath models used for analyses of marine mammals consumption 90 Table 4.2. Eight prey categories used to apportion dietary information for marine mammals (adapted from Pauly et al. 1998c) 93 Table 4.3. Mean trophic level of marine mammals' consumption and fisheries catches. 103 Table 4.4. Primary production required to sustain marine mammals' consumption and fisheries catches. Numbers in parentheses are PPR in percentage of total primary production in the system 103 Table 4.5. Estimated resource overlap index between marine mammals and fisheries and connectance of the related ecosystem models 104 vi Table 4.6. Biomass (t-km"2-year') change from the start (S) to the end (E) of the 51-year Ecosim simulation in the Eastern Bering Sea ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species.... 110 Table 4.7. Biomass (t-km~2-year') change from the start (S) to the end (E) of the 43-year Ecosim simulation in the Gulf of St. Lawrence ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species.... 115 Table 4.8. Biomass (t-km"2-yeaf') change from the start (S) to the end (E) of the 25-year Ecosim simulation in the Benguela ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species 121 Table 4.9. Biomass (t-km'̂ year"1) change from the start (S) to the end (E) of the 89-year Ecosim simulation in the Eastern Tropical Pacific ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species 127 Table 4.10. Biomass (t-km -̂year1) change from the start (S) to the end (E) of the 24-year Ecosim simulation in the Gulf of Thailand ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species.... 134 Table 4.11. Biomass (t-km -̂year1) change from the start (S) to the end (E) of the 22-year Ecosim simulation in the North Sea ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species 140 Table 4.12. Biomass (tkm 2-year"') change from the start (S) to the end (E) of the 50-year Ecosim simulation in the Strait of Georgia ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species.... 146 Table 5.1. Trophic groups used in the models of the northern (NGSL) and southern (SGSL) Gulf of St. Lawrence ecosystems, for the 1985-1987 and the 1994-1996 periods 176 Table 5.2. Changes in predation mortality (M2), fishing mortality (F) and other mortality (MO) from the 1980s to the 1990s, for all trophic groups in the northern Gulf of St. Lawrence model 185 Table 5.3. Changes in predation mortality (M2), fishing mortality (F) and other mortality (MO) from the 1980s to the 1990s, for all trophic groups in the southern Gulf of St. Lawrence model 187 Table 5.4. Average trophic level of the 30 balanced scenarios for the Northern Gulf of St. Lawrence in the 1980s and the 1990s 190 Table 5.5. Average trophic level of the 30 balanced scenarios for the southern Gulf of St. Lawrence in the 1980s and the 1990s 191 Table 5.6. Emergent properties of the 1980s and 1990s models for the northern and southern Gulf of St. Lawrence 193 vii L i s t o f figures Figure 1.1. Occurrence of the word 'ecosystem modelling' or 'ecosystem modeling' in the title or abstract of publications, according to Google Scholar 7 Figure 1.2. Occurrence of the word "Ecopath" in the title or abstract of publications, according to Google Scholar 8 Figure 1.3. Ecosystem models constructed from 1984 to 2007 with Ecopath. A total of 393 models are shown on the map, 316 in marine habitats, 71 in rivers, lakes or reservoirs, and 6 terrestrial ecosystems 10 Figure 2.1. The Northwest Atlantic areas studied by the CDEENA program for ecosystem modelling 40 Figure 3.1. Bivariate fit of the embodied construction costs (in credits) of Ecopath models by the number of trophic groups they include 63 Figure 3.2. Pedigree of Ecopath models versus their embodied costs 64 Figure 3.3. Bivariate polynomial fit of the pedigree of Ecopath models versus their embodied costs 64 Figure 3.4. Additive and variance stabilizing (AVAS) transformations indicating how parameters (x-axis) may be transformed (y-axis indicates resilience, linear scale) to linearize the individual parameters while considering their joint effects to predict resilience. Results indicate that no transformation is required for connectance, while log transformations would be acceptable for pedigree and SOI, and exponential transformation for ascendancy. Note that the same transformations were obtained for predicting persistence 65 Figure 3.5. Resilience versus pedigree for 50 Ecopath models. Grey dots (with resilience = 0) were considered as outliers and not used in the regression analysis 66 Figure 3.6. Predicted versus observed values for the regression of resilience and pedigree. Grey dots (with resilience = 0) were considered as outliers and not used in the regression analysis 67 Figure 3.7. Predicted versus observed values for the model predicting models' resilience for the regression of resilience versus pedigree, ascendancy and SOI 68 Figure 4.1. Location of the ecosystem modeled with Ecopath and used for this analysis. 88 Figure 4.2. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km"2) for all ecosystems combined. Note the greater proportion of miscellaneous fish taken by the fisheries 97 Figure 4.3. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km"2) in the Eastern Bering Sea ecosystem. 98 Figure 4.4. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km"2) in the northern Gulf of St. Lawrence ecosystem 98 Figure 4.5. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km"2) in the Benguela ecosystem 99 Figure 4.6. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t km"2) in the Eastern Tropical Pacific ecosystem 100 Figure 4.7. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km"2) in the North Sea ecosystem 100 viii Figure 4.8. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km"2) in the Gulf of Thailand ecosystem. .101 Figure 4.9. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km"2) in the Strait of Georgia ecosystem.. 102 Figure 4.10. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the Eastern Bering Sea 105 Figure 4.11. Ecosim results for the Eastern Bering Sea model with original fishing effort. 107 Figure 4.12. Ecosim results for the Eastern Bering Sea model with extirpation of marine mammals 108 Figure 4.13. Biomass change after a 51-year simulation in the Eastern Bering Sea ecosystem, with (white) and without (grey) marine mammals 109 Figure 4.14. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the Gulf of St. Lawrence Ill Figure 4.15. Ecosim results for the Gulf of St. Lawrence model with original fishing effort 112 Figure 4.16. Ecosim results for the Gulf of St. Lawrence model after the extirpation of marine mammals 113 Figure 4.17. Biomass change after a 43-year simulation in the Gulf of St. Lawrence ecosystem, with (white) and without (grey) marine mammals (echinoderms, polychaetes and small zooplankton (not shown) have very high and nearly identical biomasses with and with out marine mammals in the system, i.e., 105, 58, and 51 t-km"2, respectively) 116 Figure 4.18. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the Benguela system 117 Figure 4.19. Ecosim results for the Benguela model with original fishing effort 119 Figure 4.20. Ecosim results for the Benguela model after the extirpation of marine mammals 120 Figure 4.21. Biomass change after a 25-year simulation in the Benguela ecosystem, with (white) and without (grey) marine mammals (phytoplankton and macrobenthos [not shown] have very high and nearly identical biomasses with and with out marine mammals in the system, i.e., 84 and 74 t-km"2, respectively) 122 Figure 4.22. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the Eastern tropical Pacific system 123 Figure 4.23. Ecosim results for the Eastern tropical Pacific model with original fishing effort 125 Figure 4.24. Ecosim results for the Eastern tropical Pacific model after the extirpation of marine mammals 126 Figure 4.25. Biomass change after a 89-year simulation in the Eastern tropical Pacific ecosystem, with and without marine mammals. Note that the miscellaneous epipelagic fish, miscellaneous mesopelagic fish, cephalopods, mesozooplankton, microzooplankton, large phytoplankton, small producers and detritus groups were excluded from the graph due to their large biomass. No significant difference was observed with or without marine mammals for these groups (see Table 4.9) 128 Figure 4.26. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the Gulf of Thailand system 129 Figure 4.27. Ecosim results for the Gulf of Thailand model with original fishing effort. 131 ix Figure 4.28. Ecosim results for the Gulf of Thailand model after the extirpation of marine mammals 132 Figure 4.29. Biomass change after a 24-year simulation in the Gulf of Thailand ecosystem, with (white) and without (grey) marine mammals 135 Figure 4.30. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the North Sea system 136 Figure 4.31. Ecosim results for the North Sea model with original fishing effort 138 Figure 4.32. Ecosim results for the North Sea model after the extirpation of seals 139 Figure 4.33. Biomass change after a 22-year simulation in the North Sea ecosystem, with (white) and without (grey) seals (polychaetes, other macrobenthos and phytoplankton [not snown] have very high and nearly identical biomasses with and with out marine mammals in the system, i.e., 32, 51, and 47 t-km"2, respectively) 141 Figure 4.34. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the Strait of Georgia system 142 Figure 4.35. Ecosim results for the Strait of Georgia model with original fishing effort. 144 Figure 4.36. Ecosim results for the Strait of Georgia model after the extirpation of marine mammals 145 Figure 4.37. Biomass change after a 50-year simulation in the Strait of Georgia ecosystem, with (white) and without (grey) marine mammals (shelfish, grazing invertebrates, and phytoplankton [not shown] have very high and nearly identical biomasses with and with out marine mammals in the system, i.e., 124, 229, and 82 t-km"2, with marine mammals, and 126, 224, and 81 t-km"2 without marine mammals, respectively).... 147 Figure 4.38. Total consumption by marine mammals (black dots) or total catch by fisheries (open dots) versus their respective overall mixed trophic impact for each studied ecosystem. Density ellipses represent the 90% confidence intervals 148 Figure 4.39. Trophic interactions (predation in black, competition in grey) between predators and their prey that can lead to beneficial predation. The numbers above arrows represent the different types of beneficial predation (see text) 156 Figure 5.1. The regions of the northern (North Atlantic Fisheries Organisation divisions 4R and 4S) and southern (NAFO division 4T) Gulf of St. Lawrence 174 Figure 5.2. Biomass change for marine mammals and seabirds, piscivorous fish, forage fish, and shrimp and large crustaceans in the northern Gulf of St. Lawrence (NGSL) from the 1985-1987 (grey histogram; B = 48+24 t-km2) to the 1994-1996 (white histogram; B = 23±23 t-km"2) periods. B = total biomass for the four 180 Figure 5.3. Biomass change for marine mammals and seabirds, piscivorous fish, forage fish, and shrimp and large crustaceans in the southern Gulf of St. Lawrence (SGSL) from the 1985-1987 (B = 32±3 t-km"2) to the 1994-1996 (B = 27+6 t-km"2) periods. B = total biomass for the four groups, or all commercial species. Error bars indicate the 95% confidence intervals of biomass for the 31 balanced scenarios 181 Figure 5.4. Total consumption (0 on all prey by marine mammals, piscivorous fish, forage fish, shrimp and large crustaceans in the northern Gulf of St. Lawrence (NGSL) in the 1980s (grey) and the 1990s (white). Error bars indicate the 95% confidence intervals of total consumption for the 31 balanced scenarios 182 Figure 5.5. Total consumption (Q) on all prey by marine mammals, piscivorous fish, forage fish, shrimp and large crustaceans in the southern Gulf of St. Lawrence (SGSL) in the 1980s (grey) and the 1990s (white). Error bars indicate the 95% confidence intervals of total consumption for the 31 balanced scenarios 183 x Figure 5.6. Seal consumption on various prey in the northern (NGSL) and southern (SGSL) Gulf of St. Lawrence in the 1980s and the 1990s. Error bars indicate the 95% confidence intervals of seals consumption for the 31 balanced scenarios. Q on capelin in NGSL 1990s = 0.47 ± 0.12 t-km"2-year"', Q on capelin in SGSL 1990s = 1.27 ± 0.13 t-km"2yeaf' 184 Figure 5.7. Breakdown of total mortality (Z) into fishing (F), predation (M2) and unexplained (MO) mortalities, in annual rate (year" ), for commercial species of the northern Gulf of St. Lawrence (NGSL). Plain bars represent the 1980s while patterned bars represent the 1990s results. For clarity, 95% CI were excluded 186 Figure 5.8. Breakdown of total mortality (Z) into fishing (F), predation (M2) and unexplained (MO) mortalities, in annual rate (year" ), for commercial species of the southern Gulf of St. Lawrence (SGSL). Plain bars represent the 1980s while patterned bars represent the 1990s results. For clarity, 95% CI were excluded 188 Figure 5.9. Mean trophic level of the catch change over a decade in the southern (SGSL, in grey) and northern (NGSL, in white) Gulf of St. Lawrence ecosystems. Error bars indicate the 95% confidence intervals of trophic level for the 31 balanced scenarios. ". 192 xi L i s t o f a c r o n y m s 4RS NAFO zones for the northern Gulf of St. Lawrence 4T NAFO zone for the southern Gulf of St. Lawrence A Ascandency, from network analysis AE Assimilation efficiency AVAS Additive and Variance Stabilizing transformation in the S-Plus 6 software B Biomass density (in t-km"2) parameter in EwE BorMiCon Boreal Migration and Consumption model BUM Best Unbalanced Model (in EwE) C Catch (in tonnes) Ca Capacity, from network analysis CDEENA Comparative Dynamics of Exploited Ecosystems in the Northwest Atlantic, research program from DFO Canada CSIRO Commonwealth Scientific and Industrial Research Organisation, in Australia DC Diet composition DFO Department of Fisheries and Oceans, Canada Dynumes Dynamic Numerical Marine Ecosystem model EE Ecotrophic efficiency parameter in EwE ERSEM European Regional Seas Ecosystem model EwE Ecopath with Ecosim F Fishing effort FAO Food and Agriculture Organization of the United Nations FISHCODE Special programme of global partnerships for responsible fisheries established by the FAO Fisheries department Gadget Globally Applicable Area-Disaggregated General Ecosystem Toolbox GE Gross Efficiency parameter in EwE IPN Instituto Politecnico Nacional, Mexico MO Unexplained mortality, in EwE M2 Predation mortality, in EwE MPA Marine Protected Area MSVPA Multispecies virtual population analysis MULTSPEC Area-structured modelling approach developed to simulate interaction between commercially important fish and marine mammals in the Barents Sea NATO North Atlantic Treaty Organisation NGSL Northern Gulf of St. Lawrence NMFS National Marine Fisheries Service in NOAA NOAA National Oceanic & Atmospheric Administration 0 Overhead, from network analysis 01 Omnivory index P/B Production per unit of biomass parameter (equivalent to Z) in EwE xii PPR Primary production required, in EwE Q/B Consumption per unit of biomass parameter in EwE SGSL Southern Gulf of St. Lawrence SOI Systems' Omnivory Index, in EwE TL Trophic Level TLc Trophic Level of the catch T L Q Trophic Level of the consumption of a particular trophic group Y Fisheries Catch Z Instantaneous total mortality rate (in year"1) xiii A c k n o w l e d g e m e n t s ^ R e m e r c i e m e n t s Working on my Ph.D. has been a wonderful and often overwhelming experience. I am indebted to many people for making the time working on this project an unforgettable experience. First of all, I would like to thank my 'Doktorvater', professor Dr. Daniel Pauly, director of the Fisheries Centre of University of British Columbia. I could not have imagined having a better advisor and mentor for my Ph.D. Without his innovative view, knowledge, and guidance, I would never have finished. I also show gratitude to my co-supervisor from the Department of Fisheries and Oceans of Canada, Dr. Martin Castonguay. He guided me through all the steps of my project with great patience, and his sharp, deep way of thinking pushed my own insights steadily further. Thanks also to the members of my committee: Dr. Villy Christensen for introducing me to ecosystem modelling, taking time to help me with my models and to answer my questions patiently; Dr. Tony Pitcher for many enlightening and fruitful discussions and his careful, critical comments on my manuscript; and Dr. Andrew Trites for sharing with me his vast knowledge of marine mammals. When I arrived at the Fisheries Centre in 2001, I received a warm welcome there; also, I found true friends; notably Robyn Forrest, Sheila Heymans, Vasiliki Karpouzi, Yajie Liu, Deng Palomares, Chiara Piroddi, and Pablo Trujillo. Thank you for putting up with me for time I was in Vancouver. I would also like to thank other academic and support staff at UBC for the inspiring working atmosphere, particularly Janice Doyle and Suzanne Mondoux from the Fisheries Centre, and Allison Barnes from the Department of Zoology who provided me with so much information, help and for being my eyes and ears all over UBC, helping me in the complicated process of submitting a thesis from so far away. Furthermore, many thanks to Gerard O'doherty for enabling me 'network' from home in Quebec like I was on the UBC campus. I appreciatively acknowledge the input of the many scientific experts in ecosystem xiv modelling who contributed to the implementation, quantification and evaluation processes of the Ecopath database. The authors of the 393 Ecopath models who responded to the questionnaire on the use of Ecopath for ecosystem modelling are particularly thanked for their valuable collaboration, comments and insights, which made this project possible. A special thanks for Claude Nozeres who was the first to convince me to seek for a PhD project at UBC's Fisheries Centre. He was a great inspiration to me and broadened my mind with remarkable discussions, both related and not to the topic of this thesis. I am very grateful that I had the chance to work with the wonderful team of scientists of the CDEENA program. From them, I learned a lot about marine ecology of the northwest Atlantic, particularly through the workshops devoted figuring out how ecosystem models could be used to explain the trophic changes that followed the collapse of groundfish species. Special thanks go to Dr. Claude Savenkoff, for his important collaboration with the inverse modelling method, and to Drs Alida Bundy and Sheila Heymans for most fruitful discussions on ecosystem modelling. The financial support of the Canadian CDEENA program (supported by the Science Strategic Fund of Department of Fisheries and Oceans Canada), and the Ocean Management Research Network is also gratefully acknowledged. On a different note, I would like to thank: the coffee producers of Peru for keeping me thinking, the Fairtrade Foundation for making me feel okay about drinking so much coffee; and the music of U2 for keeping me sane. Finalement, des remerciements speciaux sont dedies a ma famille et mes amis que j'ai quelque peu negliges au cours des dernieres annees. Merci a mes parents, Nicole et Pierre-Paul, pour votre education et votre amour. Merci aussi de m'avoir appris qu'a force de courage et de determination on peut aller jusqu'au bout de nos reves. Merci a mon oncle le Dr. Jean-Claude Dionne, qui a ete ma premiere source d'interet pour les sciences marines, sur les battures de St-Fabien-sur-Mer. Merci a mes amies Lorraine Banville, Marie-Eve Cordeau, Parmelie Demers, Martine Deschenes, Caroline Guimont, Stephanie Houde, Genevieve Landry, Danielle Loranger, et Sonia Lucas. J'espere que vous savez a xv quel point vous etes importantes pour moi. Merci a Andre Jacques, mon epoux et ami de toujours, pour avoir constamment cru en moi, et pour m'avoir pleinement supporte (emotionnellement et financierement) durant cet important projet. Ta technique d'encouragement particuliere mais efficace (c'est-a- dire, commencer la majorite de tes phrases par « quand Lyne aura fini son doc ») aura finalement porte fruit. Merci de tout mon cceur. Et vous Eliane et Annabelle, mes deux p 'tites filles a la vanille, vous etes les plus belles choses qui me soient arrivees durant ce projet de doctorat. Souvenez-vous toujours que maman vous aime plus gros que le soleil... Lyne 26 March 2007 xvi Accuse not Nature, she hath don her part Do thou but thine, and be not diffident Of Wisdom, she deserts thee not, if thou Dismiss not her, when most thou needst her nigh, By attributing overmuch to things Less excellent, as thou thy self perceav'st. John Milton Paradise Lost, 1667 xvii C o - a u t h o r s h i p s ta tement A version of chapter 5 has been submitted for publication to Deep Sea Research II. The paper was co-authored with Department of Fisheries and Oceans scientists Martin Castonguay, Claude Savenkoff, Doug Swain, Denis Chabot, Hugo Bourdages, Mike O. Hammill, and Mark Hanson. This project was realized under the DFO program Comparative Dynamics of Exploited Ecosystems in the Northwest Atlantic. Overall guidance and directions for the research was provided by Dr. Castonguay, my co- supervisor. The modelling and data analysis conducted to explore the structure of the two ecosystems were primarily conducted by myself. Preparation of the manuscript, mainly for discussion, involved significant inputs from all co-authors. Initial writing and development work was carried out by myself. xviii I n t r o d u c t o r y c h a p t e r T h e r o l e o f e c o s y s t e m m o d e l s , p a r t i c u l a r l y Ecopath with Ecosim, as a t r a n s i t i o n t o e c o s y s t e m b a s e d m a n a g e m e n t . Introduction 1.1.1 Studying marine ecosystems The world's ocean covers 70% of our planet, and marine ecosystems play a critical role in maintaining many species (including humans) health and well being. Relative to terrestrial systems, our knowledge of marine ecosystems is often restricted by the difficulty of direct observation. Indeed, despite their obvious richness, marine ecosystems have largely been left out of most discussions about biodiversity. The knowledge on abundance of organisms or the interactions they have within food webs is also fragmentary (Mooney 1998). What is established, however, is that, globally, many fish populations have been declining in abundance over the past 50 years (Christensen et al. 2003; Myers and Worm, 2003), with many large-scale fisheries around the world collapsing (Pauly et al. 2002; Worm et al 2006). What was perceived for a long time as an inexhaustible resource all of a sudden seems quite limited (Rosenberg et al. 1993). The most important pressures being exerted on the ocean ecosystem are overfishing, destruction of coastal ecosystems, pollution through oil spills and illicit disposal, land- based contamination, and climate change (Costanza et al. 1998). Throughout all oceans of the world, intensive exploitation has led to dramatic changes in the structure and productivity of marine ecosystems (Fogarty and Murawski 1998), directly (fisheries catches) or indirectly (changes in the foodweb structure, habitat disturbance). Overfishing has become more important, and simultaneously, more difficult to manage (Ludwig et al. 1993). Overfishing diminishes the catches, and the genetic diversity and ecological resilience of exploited populations (Botsford et al. 1997; Pauly et al. 1998). As a result, the long-term sustainability of many fish stocks and the stability of 1 large marine ecosystems appear threatened. It is essential to focus on that issue not only to preserve the biodiversity of our planet, but also because more than one billion people now rely on fish as their main source of animal protein, making it the fifth largest food commodity in the world (FAO 2002). For many years, marine systems have been studied and managed from a single- species point of view. However, there is awareness in scientific and general communication that this traditional way of managing fisheries is not sufficient (Hofmann and Powell 1998). With the necessity to understand in detail the nature and dynamics of exploited marine ecosystems, and more precisely the complexity of species interactions, the development of the ecosystem approach for management of marine systems is becoming more and more important (Sissenwine and Daan 1991; Kroger and Law 2005, Choi et al. 2005). The science of marine management evolved rapidly from compartmentalized programs around distinct species to a more global view encompassed by the phrase "ecosystem approach" (Kroger and Law 2005). To address the world's ever-growing environmental problems, a comprehensive understanding of the structure, function and regulation of major ecosystems is essential (Mann 1988; Pahl-Wostl 1993; McCann 2000). Constructing models to examine the behavior of ecosystems is therefore the focus of much contemporary research. 1.1.2 Ecosystem modelling Marine ecosystems are often very complex systems that are heavily influenced by humans. In most cases, it is impossible to directly experiment with the ecosystems to understand how they work. Computer models and simulations can however improve our understanding of these systems. Ecosystem models allow for a better understanding of how species are influenced by each other and by human activities. They can be used both to quantitatively describe the functioning of marine ecosystems, and to provide indications of how these are likely to change in response to different ecological perturbations. Thus, ecosystem modelling is a useful tool for investigating a wide range of scientific questions in the world's oceans. Moreover, when many ecosystem models of the same type are 2 available, it becomes possible to perform global scale analyses and quantification of important marine ecological processes. Different types of models are available, that focus on different levels of interactions in a studied ecosystem (see Whipple et al. [2000] for an exhaustive review). The model of Andersen and Ursin (1977) used a mass-balance approach to describe the North Sea ecosystem. It included a complete suite of ecosystem compartments (from nutrients to predatory fish), but also required a lot of data, making it costly and difficult to use (Whipple et al. 2000; Pauly and Christensen 2002). Its major merits were to demonstrate that modeling an entire sea was feasible, and to inspire Multispecies Virtual Population Analysis (MSVPA; Sparre 1991). Models using Gadget (Globally Applicable Area-Disaggregated General Ecosystem Toolbox), an environment for statistical multispecies modeling (Begley and Howell 2004), have been developed as a forward simulation model using statistical estimation through weighted combinations of several log-likelihood criteria (Stefanson and Palsson 1998). The BorMiCon model (Boreal Migration and Consumption; Stefansson and Palsson 1997) uses the Gadget approach to represent temporal and spatial differences in the interaction of organisms around Iceland. They have the advantage of being multispecies-multiarea-multifleet simulation models. MULTSPEC (Bogstad et al. 1997) is also an area-structured modelling approach developed to simulate interaction between commercially important fish and marine mammals in the Barents Sea. These models share the characteristics of being system-specific, modelling only a small component of the ecosystem for a specific purpose. Also, primary and secondary productions are poorly represented. The ERSEM model (European Regional Seas Ecosystem Model) was developed to simulate ecosystem dynamics of the North Sea (Baretta et al. 1995). This model simulates the annual cycles of carbon, nitrogen, phosphorus and silicate in the pelagic and benthic food webs. This interlinked set of biological data is useful to describe the biological and chemical interactions between trophic groups of the ecosystem, but the higher trophic groups are not well represented. On the other hand, ecosystem models such as Dynumes (Dynamic Numerical Marine Ecosystem model; Laevastu and Larkin 1981) include a wide range of trophic 3 groups, from phytoplankton to marine mammals. They also have a spatial component which represents fish movements among habitat types. However, this modelling approach has large data requirements for biological and fisheries parameters, and applications are thus time-consuming and costly to construct (Whipple et al. 2000). The common problem with the models presented above is that they are usually constructed for a specific area or ecological issue, and because very few different applications are available, neither comparison nor validation is feasible. Size-spectra models (Sheldon et al. 1977; Moloney and Field 1985; Shin and Cury 2004) constitute an alternative to species or guild-based models, and place species or guilds in size classes for analyzing ecological relationships within ecosystems. This is a potentially useful approach for coarse predictions of fish production and potential catch in data-sparse situations, though the method still needs refining. However, this is a steady- state model that cannot make multiyear projections (Whipple et al. 2000). Another method to estimate the parameters of a food web is known as inverse modelling. This approach aims to compute the solution that minimizes the imbalances between inputs and outputs. The inverse approach provides a global criterion for an optimal (balanced) solution. When the system of equations is strongly underdetermined, additional constraints must be added to limit the range of possible solutions and thus to obtain a meaningful solution (Vezina and Piatt 1988; Savenkoff et al. 2001). Models built with inverse approach are very robust in terms of uncertainty. However, this model is also steady-state and cannot make multiyear projections. In addition, few models are available, so global inferences based on inverse models are not presently possible. Agent-based modelling (ABM) is mainly used in evolutionary ecology, archaeology, economics and social sciences. However, this modelling approach has great potential for increasing the breadth of resource management questions that may be considered (see Fulton et al. 2007 for a review). Early ABMs were often tightly focused on only a small portion of the ecosystem, but newer hybrid models incorporate more trophic groups and use differential equations, with decision based agents. InVitro (Gray et al. 2006) is an example of such a hybrid model. It has been built for management strategy evaluation. It has the potential to simulate the effects of a wide range of human activities 4 on a regional ecosystem, including fisheries, tourism, oil and gas or salt production, and to forecast system response to different management regimes. Atlantis (Fulton et al. 2004; Fulton et al. 2007) is an biogeochemical, spatially explicit, 3D, ecosystem models intended for management strategy evaluation that has been applied in 13 marine systems around the world, primarily in Australia and the US. It contains sub-models to represent each step in the management strategy and adaptive management cycles. Finally, Ecopath models are very inclusive, based on the full range of organisms in an ecosystem, and can be used for long-range simulations. These models have been applied to many ecosystems, making comparisons possible. As a result, Ecopath models represent the best option for the present analysis. The Ecopath with Ecosim (EwE) modelling approach indeed allows the construction of ecosystem models (Christensen and Pauly 1992). In addition to making ecosystem analyses possible, this approach requires to collect and use the available biological information for species of a given ecosystem and to verify that it is compatible with information on other species. The principal objection often raised against multispecies models is that they generally require excessive amounts of data. The ecosystem analysis used here is relatively easy to use, allows the estimation of a number of missing parameters, and has already been used on many ecosystems. Ecopath with Ecosim is an ecological software suite that was developed at the University of British Columbia's Fisheries Centre, with applications widespread throughout the world. The Ecopath approach was developed in the early 1980s by J.J. Polovina and co-workers at the NMFS Laboratory in Honolulu. It was first applied to a coral reef system, as a tool to allow quantitative analyses of biological interactions within aquatic systems (Polovina 1984). The method was expanded to integrate a wide range of analytic routines (Christensen and Pauly 1992). This was further developed into Ecopath with Ecosim, which provides a dynamic modeling capability for exploring past and future impacts of fishing and environmental disturbances as well as for exploring optimal fishing policies (Walters et al. 1997; 1999). In 2007, the EwE modelling approach has been recognized as one of the top 10 breakthroughs in marine science by the NOAA, in a special Web site celebrating 200 5 years of science, service, and stewardship (http://www.celebrating200years.noaa.gov). NOAA recognized Ecopath modeling as the first to apply a type of statistics called "path analysis" to the field of marine ecology. The model's simplicity and its ability to accurately identify ecological relationships, it is stated, "have revolutionized scientists' ability worldwide to understand complex marine ecosystems" (NOAA 2007). 1.1.3 Census of Ecopath models In order to compile many comparable marine ecosystem models in a database, this thesis focused on Ecopath models. This modeling approach was chosen because its software (Ecopath with Ecosim) was easily available (and free to download from the website http://www.ecopath.org), and has been used by a wide range of research groups all over the world, thus providing coverage of a good number of the marine ecosystems of the planet. In summary: while good modelling techniques exist, only Ecopath could provide enough models for the kind of comparative analyses envisaged here. The database of Ecopath models used here was started by Dr. Villy Christensen assisted by Mr. Sherman Lai (see http://www.ecopath.org). I completed this database through an exhaustive questionnaire sent to all the Ecopath users who downloaded the software. Doing so, I made sure that, in the present study, as many existing Ecopath models were taken into account. Each Ecopath user who received the questionnaire (by email) was invited to give a series of information on their models (see the complete questionnaire in Appendix 1). Results: the models database The use of ecosystem modelling for research in fisheries and aquatic sciences has increased rapidly in the last decade. According to Google scholar, a search engine covering peer-reviewed papers, theses, books, abstracts, and other scholarly literature from all broad areas of research, there was an total of 201 publications in the 1980s containing the term "ecosystem modelling" or "ecosystem modeling" in either the title or the text of the publications, while during the 1990s this number increased to 1966 publications (Figure 1.1). In the years 2000s, this number reached 3593 publications. The 6 same thing is observed for the use of the term "Ecopath", which increased from 17 publications in the 1980s period, to 370 in the 1990s, and 968 in the years 2000s (Figure 1.2). 700 -| • 3 - m t-~ oo O N o — i ( N c < " i ^ l - K ' > 1 o t ^ o o a N O — ( N c < " > ^ t n ^ o O O O O O O O O O O O O O N . O ' N O ' N O O N O N O N O ' N O N O ' N O O O O O O O 0 < ^ O S O N O N O N O \ < 7 S O N O N O N O N O N O O N O N O O O O O O O Year Figure 1.1. Occurrence of the word 'ecosystem modelling' or 'ecosystem modeling' in the title or abstract of publications, according to Google Scholar. 7 200 180 160 140 120 100 80 60 40 20 0 181 felt" llfwllffg'l in vo r-» oo 0 0 O O O O 0 0 oo ON ON ON ON ON O N O — ' ( N c o ^ t > n ^ O t - ~ O O O N O ^ ( N r O T l - i n N O O O O N O N O N O N O N O N O N O N O N O N O O O O O O O O N O N O N O N O N O N O N O N O N O N O N O O O O O O O Year Figure 1.2. Occurrence of the word "Ecopath" in the title or abstract of publications, according to Google Scholar. 1.2.1 Geographical coverage The Ecopath with Ecosim software has more than 2800 registered users from approximately 120 countries. In May 2006, of all the users contacted, 347 had replied to my questionnaire. With additional information taken from an in-depth literature search, a total of 393 EwE models had been constructed and were available for different ecosystems of the world (Figure 1.3; Appendix 2). The great majority (80%) of EwE models described were constructed for marine habitats (Table 1.1). Another 18% of models describe lakes, rivers or other freshwater habitats. Finally, six models are constructed so far for terrestrial ecosystems. Interestingly, there is also an online game called the 'Meru Park wildlife game', simulating the management of a wildlife park in the Serengetti, constructed by Agence Franqaise de Developpement and the Fonds Franqais pour I 'Environnement Mondiale, in collaboration with France 5 TV, built mainly from an Ecopath model constructed by M. Villanueva (INP/ENSAT, Castanet Tolosan, France, pers. comm.). 8 Table 1.1. Number of Ecopath models constructed for different habitat types. Environment modeled Number of EwE models % Marine 316 80.4 Freshwater 71 18.1 Terrestrial 6 1.5 TOTAL 393 100.0 Out of the 393 models covering over 80 countries, half were constructed in only eleven countries, for which eight or more EwE models are available. The most important country in terms of EwE models available are the USA, followed by Mexico, and Canada (Table 1.2). Table 1.2. The eleven most important regions in terms of Ecopath models constructed. Country Number of EwE models % of all models USA 59* 15.0 Mexico 32 8.1 Canada 26* 6.9 Australia 11 2.8 France 11 2.8 Philippines 11 2.8 Brazil 10 2.5 Chile 9 2.3 Italy 9 2.3 China 8 2.0 South Africa 8 2.0 TOTAL 194 49.5 USA and Canada share an additional 9 models representing the Great Lakes. 9  1.2.2 Structure and quality of Ecopath models For a sub-sample of 50 models, more details were collected about the size and quality (see Chapter 3 of this thesis for more details). In terms of size, most Ecopath models (42%) have less than 20 trophic groups. Models with twenty to forty trophic groups represent 30%, while 28% of Ecopath models have more than 40 trophic groups (Table 1.3). Occurrence of larger models tend to continue to increase, as we with newer models (not part of in this sub sample) such as English Channel (Stanford and Pitcher 2004; 50 groups), Northern British Columbia coast (Ainsworth et al. 2002; 53 groups), Norwegian Sea (Skaret and Pitcher (in press.); 58 groups), and Raja Ampat (Ainsworth et al. 2007; 96 groups). Table 1.3. Number of Ecopath models for different size categories, in terms of trophic groups. Number of trophic groups Number of EwE models % Less than 20 21 42 20 to 40 15 30 More than 40 14 28 TOTAL 50 100 The overall quality of EwE models was calculated for a sub-sample of 50 models, and expressed as a fraction ranging from zero to one, i.e. their "pedigree" (see Chapter 3 of this thesis for complete methodology and analysis). My results show that few models show a very low pedigree (2% have a pedigree lower than 0.2), or a very high pedigree (10% have a pedigree higher than 0.6). Forty percent of EwE models have an intermediate pedigree from 0.200 to 0.399, and 48% of the models fall into the higher range with pedigree from 0.400 to 0.599 (Table 1.4). Table 1.4. Number of Ecopath for different levels of quality (pedigree). Overall pedigree Number of EwE models % < 0.200 1 2.0 0.200 - 0.399 20 40.0 0.400-0.599 24 48.0 > 0.600 5 10.0 TOTAL 50 100.0 11 1.2.3 Rationale of Ecopath modelling projects Few ecosystem models appear to have been developed in response to specific questions. Most models (42%) were constructed to describe the structure of ecosystems, from the simplest foodweb (e.g., the model developed by Palomares et al. [1993] to describe a littoral lagoon in the French Mediterranean with its 10 functional groups) to the most elaborate (e.g., the model developed by Ainsworth et al. [2007] to analyze the ecology and economics of Bird's Head seascape, Papua, in Indonesia, which has 96 functional groups). Another 30% of the models were used to deal with fisheries management issues such as in the Ebrie Lagoon of Ivory Coast (McNamara 2004). Eleven percent addressed explicit theoretical ecology questions (e.g., trophic networks and carbon flows in south eastern Baltic coastal ecosystems by Tomczak et al. 2005), while 9% concentrate on policy matters, as was the case for the Batabano ecosystem in Cuba (Wolff et al., in prep.). Finally, 6% of the models were constructed to assist in the creation or management of marine protected areas (e.g., the model developed by Libralato et al. [2006] for the Miramare MPA in Italy). Of the Ecopath models constructed to date, 32% were for Ph.D. projects (e.g., chapter 5 of this thesis), 9% for M. Sc. projects (e.g., Nsiku's [1999] model of Lake Malawi), and 7% for post-doctoral projects (e.g., Bundy [2001] model of the Newfoundland shelf ecosystem). Research scientists constructed 19% of Ecopath models, followed with 17% by faculty members, 5% by research assistants, and 11% by graduate students (without specification of their degree). In 55% of cases, models were constructed under university research projects (as is the case, for example, for the models developed at the Fisheries Centre, University of British Columbia by Ainsworth et al. [2002] for Northern British Columbia ecosystems). In 37% of the cases, models were constructed as part of governmental projects (for example, the Gulf of St. Lawrence model presented in this thesis was constructed under the Comparative Dynamics of Exploited Ecosystems of the Northwest Atlantic - CDEENA program from Department of Fisheries and Oceans of Canada). Finally, models were constructed by independent research institutions (neither university nor government) in 8% of the cases. 12 1.2.4 Publishing Ecopath models In terms of publications, of the 314 models published in some way, close to 60% are published in peer-reviewed journals; 15% in reports; 11% in book chapters; 9% are available as M.Sc. or Ph.D. theses, and about 5% were published in conference proceedings (Table 1.5). Table 1.5. Number of Ecopath models published by type of publication. Type of publication Number of EwE models % Book chapter 34 10.8 Report 47 15.0 Conference proceedings 18 5.7 Thesis 28 8.9 Peer-reviewed journal 187 59.6 TOTAL 314 100.0 While most authors stated that they constructed only one Ecopath model, some others are very notable in terms of the number of models they constructed and published (Table 1.6). These authors are mainly from one of the top-ten countries listed in Table 1.2, but also from Spain and Germany. 13 Table 1.6. Most notable authors in terms of Ecopath models constructed, according to the survey*. Number of EwE Author Institution models Jacques Moreau Ecole Nationale Superieure Agronomique de Toulouse, France 17 Sheila Heymans Fisheries Centre, University of British Columbia, Canada 14 Thomas A. Okey Commonwealth Scientific and Industrial Research Organisation, Australia 7 Francisco Arreguin- Sanchez Centro Interdisciplinario de Ciencias Marinas del IPN, Mexico 6 Marta Coll Institut de Ciencies del Mar, Spain. 6 Lyne Morissette Fisheries Centre, University of British Columbia, Canada 6 Matthias Wolff Center fur Tropical Marine Ecology, Germany 6 Beth Fulton Commonwealth Scientific and Industrial Research Organisation, Australia 5 Cathy Bulman Commonwealth Scientific and Industrial Research Organisation, Australia 4 Alida Bundy Bedford Institute of Oceanography, Canada 4 James Hagy University of Maryland, USA 4 Simone Libralato Universita Ca' Foscari - Dipartimento di Scienze Ambientali, Italy 4 TOTAL * 83 These numbers do not account for the inputs of Dr Villy Christensen, who advised on the construction of many, if not most of these models, of Dr. Tony Pitcher, who constructed many models with his research team, and of Dr Daniel Pauly, who inspired and assisted in the construction and documentation of many EwE models. Discussion & Conclusions Ecosystem modelling is definitely an active research field. It is clear from the many responses that I received that this represents a popular way to address management, conservation, and ecological issues for marine ecosystems. The use of Ecopath with Ecosim as an approach to address such issues is extensive, and widespread around the globe. 14 The Ecopath representation of a naturally dynamic system helps to identify major shortcomings in the data and provides important reference points for more specific research on ecosystems. Such an ecosystem point of view allows an understanding of the inter-connections between the species groups of an ecosystem. From a management point of view, knowing that impacting one group will cause changes in another can provide tools for long-term and multispecies management. The strengths of Ecopath modelling are also to identify knowledge gaps, and to allow a universal format for comparing different ecosystem models. Over the past years, we have seen a transition from single species approach to an ecosystem-based approach. Because the Ecopath with Ecosim software package is free and relatively easy to use, its increasing application in many countries certainly accelerated this transition to ecosystem-based approach. At first, aquatic ecosystem models were constructed with focus on a particular species, for which we were describing the ecosystem. Most of the time, these models had a small number of trophic groups, principally describing species that were commercially important for the fishery. However, it was then considered that species (not only the ones of interest for a given study) are interacting with each other within a foodweb and that the effects of predation or competition are generating multiple indirect effects that need to be analyzed. Ecopath with Ecosim offered a fairly simple way to address this issue and became one of the important modelling tools to conduct ecosystem-based analyses. My results show that more models are constructed for marine ecosystems than for freshwater ones. The fishing industry and the species caught are much more important in marine habitats, where commercial fishing plays an important role, than in freshwater ecosystems, where there is less commercial and more sports fishing. The top-ten countries generating the largest numbers of EwE model all tend to have large aquatic areas (coasts, rivers, or lakes), and their EwE models are mainly constructed for marine systems. All these countries depend on an important fishery industry, and most of them have governmental agencies starting to show interest towards ecosystem-based management. This is the case, for example, for CEFAS and SAMS in the UK, CSIRO in Australia, DFO in Canada, and NOAA in USA. 15 Because Ecopath with Ecosim is a software project that is being developed at UBC Fisheries Centre, it is not surprising to see that the most productive authors in terms of EwE models are current and past members or collaborators of this research institution. UBC Fisheries Centre is indeed a place where ecosystem modelers can share knowledge and new ideas about this modelling technique, and apply it to their ecosystem of interest. Moreover, these models can be used for more purposes than they were built for, as they are now part of an essential and global database of our ecological knowledge of marine ecosystems on an international scale. As is true at the ecosystem scale, where models representing whole ecosystems make the detection of emergent properties possible, a global approach, linking ecosystems together for an even more extensive analysis, is also fundamental for generalization applying to marine ecosystems. Great efforts were made to assemble as much information about Ecopath models as possible during this Ph.D. thesis. This ended up being a very collaborative process, where many Ecopath users helped by providing information about their models, and sharing their interest for understanding marine ecosystem. This is particularly valuable now that these ecosystems are threatened by factors such as habitat degradation, climate change, invasive species, overexploitation, or pollution. The use of an international ecosystem models database, which is freely available and easily accessible through the web (at http://www.ecopath.org), allows the comparison of similar issues encountered in different ecosystems of the world, and the exploration of diverse approaches to address them. My analysis allowed updating the original Ecopath web database, expanding it from originally 98 models to 393. Consequently, Ecopath models have become more accessible to the scientific community. This attempt to make available as many models as possible is meant to support studies of global ecosystems. Since the software along with the models is available without charge, this makes it beneficial for every scientist or manager aware of understanding of marine ecology, even in developing countries. 16 1.3.1 Research questions Now that these many models are accessible, we can study a selection for analyses addressing various concepts in marine ecology. Here, I use a variety of ecosystem models (with the purpose of covering a global scale and having a better representation of the possibilities) in order to focus on three specific, but different hypotheses. First, this thesis presents an analysis of the uncertainty of the ecosystem modeling approach, and the importance to consider it before to address any problem with an ecosystem model (Chapter 2). As ecosystem modelling has become a very important way to study marine ecosystems processes, an essential question to address is the capacity to compare results from the model with those from observations, based on an analysis of the remaining sources of error. However, few of the currently developed Ecopath models have been examined in terms of their uncertainty. Thus, when addressing this problem, it is necessary to clearly define the type of uncertainty that may be encountered in ecosystem modelling, and the means by which it may handled. When doing so, we gain more confidence in ecosystem modelling as a tool to address or manage marine ecosystem matters. In Chapter 3, an assemblage of models is used to test the hypothesis that the quality of ecosystem model inputs plays a significant role for the relationship between complexity and stability. My secondary hypothesis was that there is, in ecosystems, a direct relationship between complexity and stability. This analysis is applied to a total of 50 Ecopath models, all of them being 'pedigreed' for their quality in terms of their input data, in collaboration with the authors of the models. Simulations were done with Ecosim to test if high quality models perform better than poor ones to predict changes in biomass in the ecosystem. The construction costs of models is discussed, the actual dollar amount spent on models of different quality estimated. Finally I explore the possibility that information quality may be related to the complexity and the resilience of marine ecosystem models. In Chapter 4, a set of EwE models that all share the particularity of having important marine mammals groups (and covering the major ecosystems where these animals are important) is used to test the hypothesis that the presence of marine mammals in ecosystems and their overlap with fisheries for food resources does not have a 17 significant effect on the overall productivity of the ecosystem, and the secondary hypothesis that marine mammals extirpation can decrease this productivity because of the beneficial predation effects they can have on their prey. Marine mammals are a very important component of marine ecosystems, and as top predators, they may compete with fishery for the same target species. Marine mammals are also exploited by whaling, mainly by Norway and Japan. The latter justify their whaling activity by stressing that whales in the world consume annually some three to five times more marine fish and invertebrates than are fished for human consumption or for reduction into fish meal and oil (Anonymous 2001; Kaschner and Pauly 2005). This situation, it is alleged, is not compatible with the world's increasing need for a stable food supply. However, the impact of top predators in marine ecosystems is not always a negative one. In many cases, there seems to be a positive predation effect from top-predators to their prey. This analysis will address that issue by using Ecosim to simulate a decrease in marine mammals biomass and quantifying its positive and negative trophic impacts on all other species of the foodweb. Normally, one would expect the impact of marine mammals on foodweb to be negative for their prey. However, in some cases I can see that this overall effect of marine mammals predation is a positive one; which would mean that a loss of these species may also represent a loss in usable production in the ecosystem. Based on four Ecopath models that were constructed as part of this Ph.D. project, the Chapter 5 addresses the problem of collapsing fish stocks in the Northwest Atlantic. By comparing two ecosystems (Northern and Southern Gulf of St. Lawrence) and two time periods (pre- and post-collapse), I test the hypothesis that the biomass of important predatory fish in a particular ecosystem (here the Gulf of St. Lawrence) changed over the last 20 years, and that the changes are linked to environmental and anthropogenic (overfishing) factors. My analyses revealed that the ecosystem structure shifted dramatically from one previously dominated by piscivorous groundfish during the mid- 1980s to one now dominated by small-bodied pelagic species during the mid-1990s in both the southern and northern Gulf of St. Lawrence. The species structure was different between the northern Gulf and the southern Gulf, and this may explain why these two ecosystems did not collapse to the same degree in the early 1990s. 18 After almost five years at the Fisheries Centre, surrounded by leaders in the marine ecology field, I realized that constructing an ecosystem model is not enough to understand a marine ecosystem. Not only do we have to deeply understand the marine ecological processes, but we also have to apply our knowledge of specific ecosystems to the global scale, draw comparisons, and trace global trends for the whole planet. Now that it becomes clearer that ecosystems are under enormous stress, it is critical that we clearly understand the biological processes of food webs, and that we figure out what is needed to keep resources as intact as possible. Chapter summary I developed a global database of ecosystem models based on the Ecopath approach. In total, 393 models were collected, which represents an addition of 295 new models to the Ecopath with Ecosim database that is available from the website (http://www.ecopath.org). A questionnaire was sent to more than 2800 users of that freely- downloadable software, and close to 350 replies were analyzed. This was completed by a literature review of other published Ecopath models. The majority of Ecopath models represent marine habitats. The most important countries in terms of models available are the USA, Mexico and Canada, but overall, more than 120 countries are represented. Most models have between 20 and 40 trophic groups, and have a pedigree between 0.400 and 0.599. The rationale of Ecopath modeling projects was also examined. Describing the structure of the ecosystem seems to be the major objective when constructing a model. Other uses are fisheries management, policy, and creation of marine protected areas. On the 314 models published in some way, most were in peer-reviewed journals, followed by reports, book chapters and theses. The creation of this model database allowed me to compare ecosystems on a global scale regarding various issues in marine ecology. Having a common methodology behind each ecosystem model was definitely an asset to explore the research questions presented in this thesis. 19 References Ainsworth, C , Heymans, J J . , Pitcher, T.J., and M. Vasconcellos (eds.). 2002. Ecosystem models of Northern British Columbia for the time periods 2000, 1950, 1900 and 1750. Fisheries Centre Research Reports 10(4). Ainsworth, C. Varkey D., and T. Pitcher. 2007. Ecosystem simulation models for the Bird's Head seascape, Papua, fitted to data, pp 6-174 In Pitcher, T.J., Ainsworth, C.H., and M. Bailey (eds.) Ecological and economic analyses of marine ecosystems in the Birtd's Head seascape, Papua, Indonesia: I. Fisheries Centre Research Reports 15(5). Andersen, K.P. and E. Ursin. 1977. A multispecies extension to the Beverton and Holt theory of fishing, with accounts of phosphorus circulation and primary production. Meddeleser Fra Danmarks Fiskeri og Havundersogelser N.S. 7: 319-435. Anonymous. 2001. Increasing competition between fisheries and whales. Japan's whale research in the Western North Pacific (JARPN II). Fisheries Agency. Baretta, J.W., Ebenhoh, W., and P. Ruardij. 1995. An overview over the European Regional Sea Ecosystem Model, a complex marine ecosystem model. Netherlands Journal of Sea Research 33: 233-246. Begley, J. and D. Howell. 2004. An overview of Gadget, the Globally applicable Area- Disaggregated General Ecosystem Toolbox. ICES CM 2004/FF13. Bogstad, B., Hiis Hauge, K., and O. Ulltang. 1997. MULTSPEC A multi-species model for fishand marine mammals in the Barents Sea. Journal of the Northwest Atlantic Fisheries Sciences 22: 317-341. Botsford, L.W., Castilla, J.C., and C.H. Peterson. 1997. The management of fisheries and marine Ecosystems. Science 277: 509-515. Bundy, A. 2001. Fishing on ecosystems: the interplay of fishing and predation in Newfoundland-Labrador. Canadian Journal of Fisheries and Aquatic Sciences 58: 1153-1167. Choi, J.S., Frank, K.T., Petrie, B.D., and W.C. Leggett. 2005. Integrated assessment of a large marine ecosystem: A case study of the devolution of the Eastern Scotian Shelf, Canada. Oceanography and marine biology: an annual review 43: 47-67. 20 Christensen, V , and D. Pauly. 1992. ECOPATH Il-a software for balancing steady-state ecosystem models and calculating network characteristics. Ecological Modelling 61: 169-185. Christensen, V , Guenette, S, Heymans, J.J, Walters, C.J, Watson, R , Zeller D , and D. Pauly. 2003. Hundred year decline of North Atlantic predatory fishes. Fish and Fisheries 4: 1-24. Costanza, R , Andrade, F , Antunes, P, van den Belt, M , Boersma, D , Boesch, D.F, Catarino, F , Hanna, S, Limburg, K , Low, B , Molitor, M , Pereira, J.G, Rayner, S, Santos, R , Wilson, J , and M. Young. 1998. Principles for sustainable governance of the oceans. Science 281: 198-199. Dalsgaard, J.P.T. and R.T. Oficial. 1997. A quantitative approach for assessing the productive performance and ecological contributions of smallholder farms. Agricultural Systems 55: 503-533. FAO. 2002. The state of world fisheries and aquaculture. FAO Fisheries Department, Rome. Fogarty, M.P. and S.A. Murawski. 1998. Large-scale disturbance and the structure of marine systems: fishery impacts on Georges Bank. Ecological Applications 8: S6- S22. Fulton, E , Morato, T , and T.J. Pitcher, (in press). Modelling seamount ecosystems and their fisheries. Chapter 15 In Pitcher, T.J, Morato, T , Hart, P.J.B, Clark, M.R, Haggan, N. and R.S. Santos (eds.) Seamounts: Ecology, conservation and management. Fish and Aquatic Resources Series, Blackwell, Oxford, UK. Gray, R , Fulton, E .A , Little, L.R, and R. Scott. 2006. Operating model specification within an agent based framework. CSIRO, North West Shelf Joint Environmental Management Study Technical Report, 13(16). Hofmann, E E . and T.M. Powell. 1998. Environmental variability effects on marine fisheries: four case histories. Ecological Applications 8: S23-S32. Kaschner, K. and D. Pauly. 2005. Competition between marine mammals and fisheries: Food for thought, pp. 95-117 In Salem, D. J. and A. N. Rowan (eds.). The state of animals III: 2005. Humane Society of the United States Press, Washington, D.C. Kroger, S. and R.J. Law. 2005. Sensing the sea. Trends in biotechnology 23: 250-256. 21 Laevastu, T. and H.A. Larkins. 1981. Marine fisheries ecosystems: its quantitative evaluation and management. Fishing News Books Ltd. Farnham, UK. Libralato S, Tempesta, M , Solidoro, C , and M. Spoto. 2006. An ecosystem model applied to Miramare natural Marine Reserve: limits, advantages and perspectives. Biologia Marina Mediterranea 13: 386-395. Ludwig, D., Hilborn, R , and C. Walters. 1993. Uncertainty, resource exploitation, and conservation: Lessons from history. Science 260: 17-36. Luong, Vu Cam. 2004. Management strategies for cove aquaculture in Tri An Reservoir of Vietnam. Ph.D. Thesis, Asian Institute of Technology, Bangkok, Thailand. Mann, K.H. 1988. Towards predictive models for coastal marine ecosystems, pp. 291-316. In Pomeroy, L.R. and J.J. Alberts (eds.). Concepts of Ecosystem Ecology, Ecological Studies 67, Springer, New York. McCann, K. 2000. The diversity-stability debate. Nature 405:228-233. McNamara, J. 2004. An investigation into the ecosystem responses of tropical inland fisheries: The implications for biodiversity, fisheries and management. MSc Thesis, Imperial College, London, UK. Moloney, C.L. and J.G. Field. 1985. Use of particle-size data to predict potential pelagic- fish yields of some southern African areas. South African Journal of Marine Science 3: 119-128. Mooney, H.A. (ed). 1998. Ecosystem management for sustainable marine fisheries. Ecological Applications 8: SI-SI 74. Myers, R.A. and B. Worm. 2003. Rapid worldwide depletion of predatory fish communities. Nature 423,280-283. NOAA. 2007. ECOPATH modeling: precursor to an ecosystem approach to fisheries management. Retrieved March 1st, 2007, from http://celebrating200years.noaa.gov/breakthroughs/ecopath/welcome.html Nsiku, E. 1999. Changes in the fisheries of Lake Malawi, 1976-1996: Ecosystem-based analysis. M.Sc. thesis. University of British Columbia, Vancouver, Canada. Pahl-Wostl, C. 1993. Food webs and ecological networks across temporal and spatial scales. Oikos 66: 415-432. 22 Palomares, M.L., Reyes-Marchant, P. Lair, N., Zainure, M., Barnabe, G., and G. Lasserre. 1993. A Trophic model of a Mediterranean lagoon, Etang de Thau, France, pp. 224- 229 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Pauly, D. and V. Christensen. 2002. Ecosystem models. Chapter 10, pp. 211-227 In: Hart, P. and J. Reynolds (eds.). Handbook of Fish and Fisheries. Blackwell Publishing, Oxford, Volume 2. Pauly, D., Christensen, V., Dalsgaard, J., Froese, R., and F. Torres Jr. 1998. Fishing down marine food webs. Science 279: 860-863. Pauly, D., Christensen, V., Guenette S., Pitcher, T.J., Sumaila, U.R., Walters, C.J., Watson R., and D. Zeller. 2002. Toward sustainability in world fisheries. Nature 418: 689- 695. Polovina, J.J. 1984. Model of a coral reef ecosystem. I. The Ecopath model and its application to French Frigate Shoals. Coral Reefs 3:1-11. Rosenberg, A.A., Fogarty, M.J., Sissenwine, M.P., Beddington, J.R., and J.G. Shepherd. 1993. Achieving sustainable use of renewable resources. Science 262: 828 - 829. Savenkoff, C , Vezina, A.F., and A. Bundy. 2001 Inverse analysis of the structure and dynamics of the whole ecosystem in the Newfoundland-Labrador shelf. Canadian Technical Report of Fisheries and Aquatic Sciences 2354. Savenkoff, C , Castonguay, M., Vezina, A.F., Despatie, S.-P., Chabot, D., Morissette, L., and M.O. Hammill. 2004. Inverse modelling of trophic flows through an entire ecosystem: the northern Gulf of St. Lawrence in the mid-1980s. Canadian Journal of Fisheries and Aquatic Sciences 61: 2194-2214. Sheldon, R.W., Sutcliffe, W.H. Jr, and M.A. Paranjape. 1977. Structure of pelagic food chain and relationship between plankton and fish production. Journal of the Fisheries Research Board of Canada 34: 2344-2353. Shin, Y-J. and P. Cury. 2004. Using an individual-based model offish assemblages to study the response of size spectra to changes in fishing. Canadian Journal of Fisheries and Aquatic Sciences 61: 414-431. Sissenwine, M.P. and N. Daan. 1991. An overview of multispecies models relevant to management of living resources. ICES Marine Science Symposia 193:6-11. 23 Skaret, G. and T.J. Pitcher, (in press). An ecosystem model of the Norwegian Sea and Barents Sea fitted to data. Fisheries Centre Research Reports. Sparre, P. 1991. Introduction to multispecies virtual population analysis. ICES Marine Science Symposia 193: 12-21. Stanford, R. and T.J. Pitcher (eds.). 2004. Ecosystem simulations of the English Channel: climate and trade-offs. Fisheries Centre Research Reports 12(3). Stefansson, G. and O.K. Palsson (eds.). 1997. Bormicon: A boreal migration and consumption model. Reykjavik Marine Research Institute Technical Report. Stefansson, G. and O.K. Palsson. 1998. A Framework for multispecies modelling of boreal systems. Reviews in Fish Biology and Fisheries 8: 101-104. Tomczak, M.T, Jarv, L , Kotta, J , Martin, G , Minde, A , Muller-Karulis, B , Pollumae, A , Razinkovas, A , and S. Strake. 2005. Trophic networks and carbon flows in South Eastern Baltic coastal ecosystems. ICES. CM 2005 / M:01. Vezina, A .F , and T. Piatt. 1988. Food web dynamics in the ocean. I. Best-estimates of flow networks using inverse methods. Marine Ecology Progress Series 42: 269- 287. Walters, C , Christensen, V , and D. Pauly. 1997. Structuring dynamic models of exploited ecosystems from trophic mass-balance assessments. Reviews in Fish Biology and Fisheries 7: 139-172. Walters, C , Pauly, D , and V. Christensen. 1999. Ecospace: prediction of mesoscale spatial patterns in trophic relationships of exploited ecosystems, with emphasis on the impacts on marine protected areas. Ecosystems 2: 539-554. Whipple, S.J, Link, J.S, Garrison, L.P, and M.J. Fogarty. 2000. Models of predation and fishing mortality in aquatic ecosystems. Fish and Fisheries 1: 22-40. Wolff, M , Alpermann, T.J , Gonzalez-Yanez, A.A., Areces, A , Valle, S. Baisre, J.A, and R. Puga. (in preparation). The Batabano ecosystem (Cuba) and its fishing targets: multispecies assessment and simulation of use scenarios. 24 Worm, B., Barbier, E.B., Beaumont, N., Duffy, J.E., Folke, C , Halpern, B.S., Jackson, J.B.C., Lotze, H.K., Micheli, F., Palumbi, S.R., Sala, E., Selkoe, K., Stachowicz, J.J., and R. Watson. 2006. Impacts of biodiversity loss on ocean ecosystem services. Science 314:787-790. 25 C h a p t e r 2 A d d r e s s i n g u n c e r t a i n t y i n m a r i n e e c o s y s t e m m o d e l l i n g . 1 2.1 Introduction When we want to address issues at the ecosystem level, one of the options is to create models that will represent these ecosystems and simulate different scenarios to see how they will react to diverse situations. Ecosystem modelling has become very popular in applied ecology, and perhaps even more so for marine ecosystems, which are harder to investigate directly. Even as this approach increases in popularity, only a minority of studies address the uncertainty related to the model results. For the marine environment, a very popular tool to create ecosystem models is called Ecopath. With this approach, many options exist to address uncertainty. However, many sources of uncertainty exist, and they are virtually infinite in ecosystems modelling. It is thus important to identify the various sources of uncertainty and to recognize how to deal with each of them. This paper will review the different options available for uncertainty analyses in marine ecosystem modelling, and propose some alternatives for a better investigation of uncertainty. 2.2 Ecopath ecosystem modelling and its sources of uncertainty Ecopath is a modelling approach that creates a simple static model to describe the average interactions of the populations within an ecosystem during a certain period. The model assumes mass-balance, i.e., that we account for all flows in a foodweb. Hence, its parameters can change. Such an approach is much simpler than others that attempt to 'Based on an earlier version published as Morissette, L. 2005. Addressing uncertainty in ecosystem modelling, pages 127-142 In E. Levner, I. Linkov and J.-M. Proth, eds. Strategic Management of Marine Ecosystems. NATO Science Series: IV: Earth and Environmental Sciences, Vol. 50., with kind perission of Springer Sciene and Business Media. 26 model multispecies interactions such as MSVPA (Sparre 1991) for which an enormous quantity of catch-at-age data and stomach contents analyses is required (Morissette 2001). The principal advantage with Ecopath is that the input values (mainly total mortality, consumption and diet composition) are often already available for several species or groups in the ecosystem, and that they can easily be placed in an ecological model (Christensen and Pauly 1992). Ecopath is thus an approach allowing the construction and the rapid evaluation of balanced ecosystem models (Christensen and Pauly 1992). The main advantage of this model is that it makes it possible to use all available data for a given food web and put it in a more global context (Christensen 1991). However, because the information at the ecosystem level is never complete, there is no unique solution for a specific region or period of time. During the last decades, Ecopath models were constructed for more than 314 ecosystems, and close to 80 others are presently under development (see Chapter 1 of this thesis). Models were published for ecosystems as diverse as the Peruvian upwelling system (Jarre et al. 1991), coral reefs in the Philippines (Alino et al. 1993), the Southern Gulf of Mexico (Arreguin-Sanchez et al. 1993), Antarctica (Schalk et al. 1993), and Lake Victoria (Moreau et al. 1993). This type of modelling was also applied to various uses (comparison of the structure of estuaries (Monaco and Ulanowicz 1997), estimate of the trophic levels (Pauly et al. 1995) or the modelling of inundated rice fields in the Philippines (Lightfoot et al. 1993). Most Ecopath models constructed so far have been based on a single set of input parameters representing the mean of the model period, typically a given year (Christensen et al. 2000). The way we reach a balanced solution from the input datasets consists mainly in modifying manually the parameters so as to obtain mass balance and the outcome represents one of the many possible representations of how the trophic structure of the ecosystem may have been during the period covered. Most of the time, obtaining a balanced network with the Ecopath approach is left to trial and error, either in the form of user intervention or Monte-Carlo simulations. However, newer models use features such as autobalance to ease the process. In its simplest form, the master equation of Ecopath defines the mass-balance between consumption, production, and net system exports over a given time period for each functional group (i) in an ecosystem (Christensen and Pauly 1992): 27 B, - EE = Y + YBi DCiS Eq. 2.1 V-o J j where B{ and Bj are biomasses (the latter pertaining to j, the consumers of i); P/B/ is the ratio of production to biomass, equivalent to total mortality under most circumstances (Allen 1971); EEt is the ecotrophic efficiency which is the fraction of production (i.e., P = B(PIB)) that is consumed within, or caught from the system (by definition between 0 and 1); Yi is equal to the fisheries catch (i.e., Y = FB); Q/Bj is the food consumption per unit of biomass of j; and DC),- is the contribution of (i) to the diet of (/'), and the sum is over all predators (j). Biomass accumulation and migration can also be added to the right hand side of the equation. Each group can have an unknown parameter {B or EE; P/B or Q/B) that can be estimated by the model. Most often, when the datasets are relatively complete, the EE is left unknown and is then used as a verification parameter to see which compartment of the model does not meet mass balance constraints. The problem when we build such a model is that the estimates of biomass, production (P/B), consumption (Q/B) and diet composition (DC) do not necessarily result in an ecotrophic efficiency (EE) between 0 and 1, as required by mass-balance constraints. Having an EE higher than 1 for a trophic group means that the sum of predation and the catch for this group is exceeding its biological production. Traditionally, to reach a balanced solution when building an Ecopath model, its designer generally had to modify the diet composition of the major predators of species for which we had an excess of EE. The decision process was mainly based on ecological knowledge of the modeler, but presented the risk of modifying high-quality (= reliable) estimates to balance some that were of lower-quality (Kavanagh et al. 2004). When using a balanced model, it is crucially important to compare model results with observations and to analyze remaining errors. However, it is difficult to distinguish between errors which are related with the model structure and those which are due to the improper choice of parameter values (Schartau et al. 2001). Sources of uncertainty are virtually infinite in ecosystem modelling. However, the more we learn about species that are parts of marine ecosystems, the better we can address uncertainty in these systems. 28 The degree of predictability of ecosystem models is itself uncertain (Hilborn 1987). However, there are many other types of uncertainty that we need to take into account in ecosystem modelling. One of them can be called "predictable uncertainty" (Silvert 2004), which arises from the known stochastic nature of the environment (e.g., climate fluctuation that follows a historical pattern). These fluctuations in the environment, such as El Nino, can be incorporated in the model as a known fluctuation that would affect some species groups (e.g., the primary production). A more fundamental source of uncertainty (and one much more difficult to take into account) is called "structural uncertainty" (Silvert 2004). Our lack of knowledge on marine ecosystems and fisheries is a good example of that (Gomes 1993). In Canada, for example, there is no consensus on the causes of the collapse of cod stocks. Some authors argue that it was due to intensive exploitation combined with a period of reduced productivity of cod stocks (poor condition and growth, and increased natural mortality) (Dutil et al. 1999; Dutil and Lambert 2000). Others believe that the collapse of cod stocks can be attributed solely to overexploitation (Hutchings and Myers 1994; Hutchings 1996). However, a large part of cod mortality remains unexplained in the Gulf of St. Lawrence's Ecopath models. Savenkoff et al. (2004a) concluded that much of the unexplained or other mortality in the Northern Gulf of St. Lawrence ecosystem in the 1980s resulted from the under-reporting and discarding of catches. As a result, fishing mortality was substantially underestimated in the mid 1980s, just before the demise of a cod stock that historically was the second largest in the Northwest Atlantic (Morissette et al, Chapter 5 of this thesis). Sources of uncertainty in fisheries models are presented in a long list by Seijo et al. (1998) for the FAO, in an attempt to incorporate risk and uncertainty in bioeconomic modelling and to address some alternative ways of contending with it in a precautionary fishery management context. According to these authors, uncertainty can come from as many sources as abundance estimates, model structure, model parameters, future environment conditions, behaviour of resource users, future management objectives, and economic, political and social future conditions. Conceptually, it is easy to agree that ecosystem models are designed to fulfill a more complex task than simpler models. However, in practice they may or may not 29 perform better, because we have too many functional relationships whose true functional forms are poorly known (or not known at all) and too many ecosystem components that we cannot parameterize. On a case-by-case basis, improvements have to be demonstrated, not just asserted as occurring. Development and evaluation of new ecosystem models have to seek for optimal parameters which remain constant in time, as it is usually assumed (but not always true), in order to fully address uncertainty. Sensitivity analyses as a tool to address uncertainty In ecosystem modelling, there are different approaches to reach a balanced scenario. As a result, it is very important to examine how sensitive are the results (or outputs) of the model to changes in the way it was constructed and balanced. Each model constructed must carry on such sensitivity analyses in order to observe if the decisions taken when trying to get a balanced solution for an ecosystem model might conceivably have affected the results. Unfortunately, in Ecopath modelling, not all authors are using this approach. Indeed, my observation is that only 20% of published models explicitly present a sensitivity analysis, performed using input parameters. Inspired by Alderson et al. (2004) and their study on review methods for healthcare research, we can describe the types of decisions and assumptions that might be examined in sensitivity analyses, including: changing the input parameters to reach the mass-balance constraints of the model; compare the efficiency (predictability) of the model with other studies of ecosystem models; reanalyzing the data using a reasonable range of possible values as new inputs; - reanalyzing the data imputing a reasonable range of values for missing data; reanalyzing the data using different statistical approaches (e.g., using a random effects model instead of a fixed effect model, or vice versa). Sensitivity analyses are the key to address the consequences of uncertainty, and this is particularly true for ecosystem models. 30 According to Silvert (2004), there is a general tendency to put too much confidence in ecosystem models, a priori. This is why the use of sensitivity analyses has become such an important part of the modelling approach. The main idea when we use sensitivity analysis in an ecosystem modelling approach is to test if the results are robust or if they are very sensible to small changes in the way the models are constructed, or changes in the value of input parameters, in a way that a trivial change could radically affect the results. The approach that allows testing the performance of the models is called "sensitivity analysis". In other words, if the sensitivity analyses do not significantly change the results, this strengthens the confidence that can be placed in the model. If the results change in a way that might lead to different conclusions, this indicates a need for greater caution in interpreting the results and drawing conclusions. After perturbations of the input data within their range of uncertainty, the derived probability distributions are likely to be narrower than the original distributions indicating that we have gained information in the process of checking for mass balance constraints, and eliminating parameter combinations that violate thermodynamic constraints (Christensen and Walters 2004). Sensitivity can also be a biological phenomenon. Indeed, if we perturb an ecosystem, we cannot be entirely certain of how it will respond. There are some common ecological principles however, which allow us to do general predictions on what could happen (Silvert 2004). For example, if a fish population decrease in abundance, there is a high probability that the ecological niche left by this population will be re-occupied by other species. This happened in the Gulf of St. Lawrence, where the quasi-disappearance of Atlantic cod (Gadus morhud) after an intense period of fishing in the 1980s left an empty ecological niche that was quickly filled by forage species such as capelin (Mallotus villosus), herring (Clupea harengus), sandlance (Ammodytes dubius) and Arctic cod (Boreogadus saida) (Morissette et al, Chapter 5 of this thesis). Sensitivity analysis in Ecopath A simple sensitivity routine is included in Ecopath, to allow users to explore the effects of uncertainty on the model results. The method is quite simple, and consists in 31 plotting relative output changes against relative changes in the inputs. However, this does not allow testing of the "structural uncertainty" discussed above. The routine varies all basic input parameters (biomass [B], production / biomass ratio [P/B], consumption / biomass ratio [Q/B], ecotrophic efficiency [EE]) in steps from - 50% to +50% for each species or trophic group of the model, and then checks what effect each of these steps has for each of the input parameters on all of the "missing" basic parameters for each group in the system (Christensen et al. 2000). The output is then given as the proportion of the difference between the estimated and original parameter to the original parameter, and converted to a percentage (Christensen et al. 2000). Unfortunately, this method only re-estimate the parameters for which no data was available, and that were left to be estimated by the model, using the mass-balance constraints. If a model uses good data on biomass, consumption and production and has no missing parameter, it is the EE that will be left to be estimated by the model. As a consequence, this becomes the only parameter to be taken into account in terms of impacts of the variation of input parameters on outputs values. The other outputs provided by Ecopath (diet composition, catch, mortalities, system's emergent properties, etc.) can not be evaluated directly by this method, unless a specific routine is made to study this. The Ecosim part of the software (used for temporal simulations) also includes a Monte Carlo routine to find input parameters that minimize the sum of squares with time series. This Monte Carlo simulation interface can be used with initial Ecopath biomass estimates chosen from normal distributions centered on the initial input estimates (Christensen and Walters 2004). Each simulation starts with a random biomass combination and re-balancing the Ecopath model, keeping only values that result in improved model fit. Such a search or fitting procedure is known as a "Matyas search" (Christensen and Walters 2004). The resulting balanced Ecopath model can then be used as the baseline for an Ecosim run. This approach is now beginning to be used in many analyses, such as risk evaluation (Pitcher et al. 2005a) to evaluate risks of extirpation, depletion, and biodiversity reduction in studied ecosystems. In addition, Morissette and colleagues developed a complementary methodology with the CDEENA program (work in progress, described briefly in Pitcher et al. 2005b) to address uncertainty through temporal simulations. The idea was to construct a model for a 32 given period of time (in this case, the Gulf of St. Lawrence in the mid-1980s) and another one for the exact same ecosystem, but for a later period (ten years later, in the mid-1990s). The "real" 1990s model would then compared with a 1990s model obtained after a 10- year simulation of the 1980s model, including functions such as change if the fishing effort, or environmental factors (changes in water temperature, ice coverage, El Nino effect, etc.). Moreover, it is a common misconception to believe that because we have complete databases for some ecosystem models, we can necessarily construct models that are reliable. Indeed, all the input parameters used in models can have very wide ranges of variation. Even when the data are well known, there can still be a lot of uncertainty related to the inputs, and thus, the outputs of the models. Pedigree and Ecoranger routines When we compare the different research fields in marine ecology, we quickly realize that charismatic species such as seals or whales are much more studied than invertebrates or parasites. However, when we analyze the ecosystem as a whole, the latter species can become very important in the global response of the system to its environment. The interactions between the different species and the fishery in a marine ecosystem are complex, generally not well understood, and can become a major source of uncertainty. The interactions that are generally better understood are predator-prey relationships, primarily based on stomach contents analyses. On the other hand, information on production, consumption or mortality sources for each species of the system is usually less understood and thus represents a considerable part of the variability of input values in the marine ecosystems models that we construct. The pedigree of a model is a summary of the coded statements quantifying the uncertainty related to each input value in Ecopath models. For each input that we use in a given model, a choice can be made to describe the kind of data used, and thus the confidence we can have in these data. The routine uses percent ranges of uncertainty based on a set of qualitative choices relative to the origin of biomass, P/B, Q/B, catch and diet input or model estimates (model estimates have a high range of uncertainty) (Table 2.1). 33 When these choices are made for each single input values, an overall pedigree of the model is calculated as the average of the individual pedigree values (Pauly et al. 2000). This overall pedigree is then very useful for comparison with other models (Christensen and Walters 2004), allowing to compare models with a different amount of trophic compartments (and thus with different amount of input values and individual pedigrees). The overall pedigree is calculated as: Eq. 2.2 where \ p is the pedigree index value for group i and input parameter p for each of the n living groups in the ecosystem. Parameters (p) can be B, P/B, Q/B, DC or catch data (Christensen and Walters 2004). The confidence intervals associated to each parameter attributed in the pedigree table can be defined by the constructor of the model or else left to default values (Table 2.1). Specifying the pedigree of data used to generate Ecopath input make users aware of the danger of constructing an Ecopath model mainly from input taken from other models, but also provides defaults for the Ecoranger routine of Ecopath (see below) (Christensen et al. 2000). When the pedigree table is complete, models are then implemented with this "quality footprint" that will be unique and make comparisons between models possible, based on single parameters pedigree, or overall pedigree indices. 34 Table 2.1. Default options for the pedigree routine, for each input parameter used in Ecopath models. Defaults (percentage CI) are means based on actual estimates of CI in various studies. Modified from Christensen et al. (2000). Parameter Pedigree Default CI index (±%) Biomass Sampling based, high precision 1.0 10 Sampling based, low precision 0.7 40 Approximate or indirect method 0.4 50-80 Guesstimate 0.0 80 From other model 0.0 80 Estimated by Ecopath 0.0 n.a. P/B and Q/B ratios Same group/species, same system 1.0 10 Same group/species, similar system 0.8 20 Similar group/species, same system 0.7 30 Similar group/species, similar system 0.6 40 Empirical relationship 0.5 50 From other model 0.2 80 Guesstimate 0.1 90 Estimated by Ecopath 0.0 n.a. Diet compositions Quantitative, detailed, diet composition study 1.0 30 Quantitative but limited diet composition study 0.7 40 Qualitative diet composition study 0.5 50 General knowledge for same group/species 0.2 80 From other model 0.0 80 General knowledge of related group/species 0.0 80 Catches Local study, high precision/complete 1.0 10 Local study, low precision/incomplete 0.7 30 National statistics 0.5 50 FAO statistics 0.2 80 From other model 0.0 >80 Guesstimates 0.0 >80 For the purpose of this analysis, two ecosystems with very different pedigree indexes are compared. In my database, the model with the lowest pedigree (0.164) corresponds to the Icelandic shelf, for the 1950s (Buchary 2001). Because it tries to reproduce an ecosystem from the past, for which less information is available, most data in this model is guessed, estimated by Ecopath, or taken from another model (Table 2.2). Opposed to this is a model for the northern Gulf of St. Lawrence in the 1980s (Morissette et al. 2003). This model is mostly based on sampling-based estimates for the exact same species than what is used in the model (Table 2.3). The colour code used in the pedigree routine allows to have a rapid idea of the quality of a model and its inherent data. Shades of blue indicate low quality information, and shades of red refer to high quality data. 35 Table 2.2. Pedigree of biomass (B), production (P/B), consumption (Q/B), diet and catch inputs for the fish components of the Icelandic shelf ecosystem model for the 1950s constructed by Buchary (2001) with Ecopath. The overall pedigree of this model was 0.164. Pedigree Species or group B P/B Q/B Diet Catch Adult Cod 0.0 0.0 0.2 0.2 0.5 Juvenile Cod 0.0 0.0 0.2 0.0 - Haddock 0.0 0.0 0.2 0.2 0.5 Saithe 0.0 0.0 0.2 0.2 0.5 Redfish 0.0 0.0 0.2 0.2 0.5 Greenland Halibut 0.0 0.0 0.2 0.2 0.5 Other Flatfish 0.0 0.0 0.2 0.2 0.5 Other Dem. Fish ().() 0.0 0.2 0.2 0.5 Herring ().() 0.1 0.2 0.2 0.5 Capelin 0.0 0.0 0.2 0.2 0.5 Other Pelagics 0.0 0.2 0.0 0.2 0.5 Other fish 0.0 0.0 0.2 0.2 - Table 2.3. Pedigree of biomass (B), production (P/B), consumption (Q/B), diet and catch inputs for the fish components of the northern Gulf of St. Lawrence ecosystem model constructed by Morissette et al. (2003) with Ecopath. The overall pedigree of this model was 0.651. Species or group Large cod Small cod L. Greenland halibut S. Greenland halibut American plaice Flounders Skates Redfish L. demersals S. demersals Capelin Sand lance Arctic cod L. pelagics S. pisciv. pelagics S. plankt. pelagics Pedigree Q/B Diet Catch However, even when we have a parameter for which a high pedigree score is assigned, this does not necessarily mean that the range of uncertainty associated with this parameter is small. In the Gulf of St. Lawrence, for example, the biomass estimates come from very detailed sampling survey (such as DFO groundfish survey database that was used for many species of the Gulf of St. Lawrence models, see Morissette et al. [2003] for more information) and still, these have quite large confidence intervals. This is also true 36 for diet composition studies. For example, in the Gulf of St. Lawrence, some inshore and offshore diets had to be combined, or we had to assume that the diet for the 1980s was the same as during the 1990s, or else that the diet for a key species was representative of the functional group (Morissette et al. 2003; Savenkoff et al. 20046, c). When aware of such cases, users should change/overwrite the default values in Table 2.1. Ecoranger is a resampling routine based on input probability distributions for B, P/B, Q/B, EE, DC, and catches, and it uses a Monte Carlo approach. The distribution ranges of each parameter can be entered explicitly for each input or Ecoranger can pick up the confidence intervals from the pedigree tables and use these as prior probability distribution for all input data. After perturbing the input data, the routine attempt to reach a balanced solution that solves the physiological and mass balance constraints (Christensen and Walters 2004). Most of the time, a balanced solution is not found, and Ecoranger proposes a "best unbalanced model" (or BUM) that can then be worked on manually. However, when such a situation occurs, it does not necessarily means that there is no balanced solution available for the input ranges of possible values. After running the same inputs through other perturbation methods (see inverse approach, in a section below), many models ended up having more than one possible balanced solution, that were just not found by the Ecoranger processes (L. Morissette, unpublished data for the northern Gulf of St. Lawrence model). Because uncertainty changes with models' quality, I got very different results for the Ecoranger runs with the low-pedigree and the high-pedigree models (Table 2.4). For each model, the routine was set to reach a total of 200 successful runs, and to have a maximal number of 10,000 runs in total. The routines included a randomization process (random seed), and the selection criteria for best model was to minimize the sum of squared residuals. For the Icelandic shelf model, because of the high ranges of uncertainty surrounding data, the model was able to reach 200 balanced solutions after 430 attempts. On the other hand, the Northern Gulf of St. Lawrence model only got to 18 balanced solutions after reaching the maximum 10,000 runs. This happened because all data with a high pedigree had very narrow ranges of possible variation. It was thus difficult for Ecoranger to find different balanced solutions. The more confident we are in the data, the 37 highest the pedigree is, and the most constant the model will be. This is also true for the minimum residuals (Table 2.4), which are very high for the low-pedigree model showing a highest uncertainty, and relatively low for the high-pedigree model that reflects lowest uncertainty. Table 2.4. Ecoranger results for two models with different pedigrees. Model Pedigree Total Number of %of Minimum index runs balanced solutions success residuals Low pedigree (Icelandic shelf in 0.164 430 200 46.51 7571 the 1950s; Buchary2001) High pedigree (Northern Gulf of 0.651 10,000 18 0.18 111 St. Lawrence in the 1980s; Morissette et al. 2003) The Ecoranger routine also produces a summary of what went wrong in the rejected runs. For each rejected runs, the potential failure criteria are: 1. EE>1 2. Production/consumption is too high 3. EE < EEmin or EE > E E m a x - EE may have been recalculated and thus can have changed 4. Output parameter is outside the range 5. Electivity outside range 6. Detritus EE > 1 For the Icelandic shelf and the Gulf of St. Lawrence models, the first 50 failed runs were all caused by the fact that P/Q was too high (Table 2.5). According to Christensen and Pauly (1992), the gross growth efficiency (GE, or the ratio of production to consumption) should have values between 10 and 30% for most trophic groups. Exceptions are top predators, e.g., marine mammals and seabirds, which can have lower GE (between 0.1 and 1%), and small, fast growing fish larvae, nauplii or bacteria, which can have higher GE (between 25 and 50%). 38 Table 2.5. Parameters causing failure for the first 50 runs of Ecoranger, with the related problematic value of each criterion. Criterion Icelandic shelf Group Value Northern Gulf of St. Lawrence Criterion Group Value 2 Other pelagics 0.770 2 Large demersals 0.770 2 Other pelagics 1.034 2 Large demersals 1.034 2 Other fish 0.634 2 Large pelagics 0.634 2 Juvenile cod 0.638 2 Seabirds 0.638 2 Redfish 0.669 2 Large Greenland halibut 0.669 2 Northern shrimp 0.781 2 Capelin 0.781 2 Northern shrimp 0.743 2 Capelin 0.743 2 Other pelagics 0.714 2 Large demersals 0.714 2 Northern shrimp 0.913 2 Capelin 0.913 2 Other pelagics 0.635 2 Large demersals 0.635 2 Zooplankton 0.629 2 Small piscivorous pelagics 0.629 2 Juvenile cod 0.781 2 Seabirds 0.781 2 Molluscs 0.616 2 Sand lance 0.616 2 Other fish 0.658 2 Large pelagics 0.658 2 Other pelagics 1.070 2 Large demersals 1.070 2 Nephrops 0.748 2 Small demersals 0.748 2 Juvenile cod 0.779 2 Seabirds 0.779 2 Other pelagics 0.688 2 Large demersals 0.688 2 Redfish 0.607 2 Large Greenland halibut 0.607 2 Northern shrimp 0.632 2 Capelin 0.632 2 Other flatfish 0.698 2 American plaice 0.698 2 Juvenile cod 0.871 2 Seabirds 0.871 2 Other pelagics 0.721 2 Large demersals 0.721 2 Northern shrimp 0.853 2 Capelin 0.853 2 Other pelagics 1.113 2 Large demersals 1.113 2 Other pelagics 0.638 2 Large demersals 0.638 2 Juvenile cod 0.708 2 Seabirds 0.708 2 Nephrops 0.739 2 Small demersals 0.739 2 Other pelagics 0.654 2 Large demersals 0.654 2 Other flatfish 0.787 2 American plaice 0.787 2 Northern shrimp 0.667 2 Capelin 0.667 2 Northern shrimp 0.913 2 Capelin 0.913 2 Juvenile cod 1.143 2 Seabirds 1.143 2 Northern shrimp 0.611 2 Capelin 0.611 2 Other flatfish 0.686 2 American plaice 0.686 2 Juvenile cod 0.802 2 Seabirds 0.802 2 Other flatfish 0.622 2 American plaice 0.622 2 Other pelagics 0.604 2 Large demersals 0.604 2 Zooplankton 0.865 2 Small piscivorous pelagics 0.865 2 Greenland halibut 0.691 2 Small Greenland halibut 0.691 2 Zooplankton 0.647 2 Small piscivorous pelagics 0.647 2 Zooplankton 0.971 2 Small piscivorous pelagics 0.971 2 Juvenile cod 0.751 2 Seabirds 0.751 2 Zooplankton 0.651 2 Small piscivorous pelagics 0.651 2 Northern shrimp 0.607 2 Capelin 0.607 2 Juvenile cod 0.965 2 Seabirds 0.965 2 Other fish 0.700 2 Large pelagics 0.700 2 Nephrops 0.739 2 Small demersals 0.739 2 Zooplankton 0.810 2 Small piscivorous pelagics 0.810 2 Redfish 0.805 2 Large Greenland halibut 0.805 39 The main advantage of the Ecoranger routine is that it will change all parameters at once within the confidence interval limits defined by the constructor of the model. However, this method is not used a lot, and many users have encountered troubles using this routine (A. Bundy, Department of Fisheries and Oceans, Bedford Institute of Oceanography; and J.J. Heymans, University of British Columbia, Fisheries Centre; pers. comm.). Some authors mention that, indeed, the results of this procedure can be confusing and reverted to standard manual methods for most of their analyses (Silvert 2004). Autobalance An important step was done addressing uncertainty in Ecopath models with the Comparative Dynamics of Exploited Ecosystems in the Northwest Atlantic (CDEENA) program in 2003. Within this program, a total of 10 Ecopath models were constructed for 2 periods (the 1980s, prior to the collapse of groundfish species; and the 1990s, after that decline) and 5 different ecosystems: the Grand Banks of Newfoundland (NAFO zones 2J3KLNO), the eastern and western Scotian shelves (NAFO zones 4VsW and 4X), and the northern and southern Gulf of St. Lawrence (NAFO zones 4RS and 4T) (Figure 2.1) Figure 2.1. The Northwest Atlantic areas studied by the CDEENA program for ecosystem modelling. 40 The "autobalance" routine of Ecopath is a new parameter optimization for ecosystem models. This tool is used mainly to obtain reproducible mass-balanced models from an unbalanced state (Kavanagh et al. 2004), but can also be used (as it is the case here) for perturbation analyses in order to assess uncertainty associated with certain models. The CDEENA program represents the first case study where such an approach was used in an attempt to fully address uncertainty in Ecopath modelling. Autobalance is an algorithm that allows reaching mass-balance constraints of Ecopath in a quantitative way. This routine should be used jointly with the traditional manual balance process which is way more informative (and can take ecological knowledge of the modeler into account). Autobalance is a structured way (as opposed to the traditional manual approach) to use input data and get to a balanced solution for an ecosystem model, based on clear assumptions. The routine uses the uncertainty definitions associated to data provided in Ecopath, which are described by the pedigree (see section above on pedigree), and then apply perturbations (within these uncertainty ranges) to the data based on the degree of confidence we have in them. Perturbations on parameters with better pedigree (that we know more about) should be less important than the ones on parameters with lower pedigree. This assumes that if we know less about these parameters, there is a better chance they can have a different value than the one provided in the model. The main objective of the autobalance process is to obtain a model for which the EE's are all below one, without exceeding the confidence intervals on diet compositions or biomass. The method also allows choosing the magnitude of perturbations according to the amount of reduction needed to reach EE < 1 (Kavanagh et al. 2004). Unfortunately, only DC and B parameters are presently modified with this routine. Kavanagh et al. (2004) explain that they allow only DC and B to be varied, as these parameters are generally the most uncertain. However, there are models where the quality of information (i.e., the pedigree) is much better (and thus, values have narrower confidence intervals and more certain) for B or DC than it is for parameters such as P/B, Q/B or catch (L. Morissette, unpublished data). This is in fact the case for many models, such as a Caribbean coral reef (Opitz 1996), a lake of Sri Lanka (Moreau et al. 2001), the West 41 Florida shelf (Okey et al. 2004), or the coast of Guinea (Guenette and Diallo 2003). In these cases, when B or DC is well known, the magnitude of perturbations is reduced, and the sensitivity analysis can be less robust. When used as an uncertainty analysis, the autobalance routine presents some limitations. Firstly, EE < 1 is the only hard constraint to achieve mass-balance in the autobalance process. Moreover, the approach only deals with parameters affecting species that are unbalanced. This means that not all parameters are perturbed in such an uncertainty analysis and not all species as well. For the CDEENA models, 30 alternate models were created with the autobalance by perturbing the input B and DC within their pedigree confidence intervals, and then re- balanced by the automated process. Ranges of possible values used for pedigree information are fully described in Morissette et al. (2003), Heymans (2003), Bundy (2004), and Savenkoff et al. (20046, c). The re-balance process was not checked for ecological logic, as a complete model construction and balance would be. Thus, the 30 alternate solutions represent balanced scenarios, but may be incorrect. Since the new solution is within the confidence intervals of all parameters entered, the solution can be as logical as a manually fixed model. Each run started with a different set of conditions, and the routine searched for the combination that will produce a balanced model. The autobalance routine was programmed to run for 10,000 runs in order to reach this target (Bundy 2004). The thirty solutions were used to define 95 % confidence intervals for the model estimates, giving an idea of the uncertainty associated with the model output. One other weakness of this method is that only biomass and diet composition were perturbed to create a range of possible alternative scenarios for these ecosystems. The consumption and the production were not changed, but we know that these parameters can be uncertain for each species in a given ecosystem. Finally, the uncertainty analysis consisted in comparing the two models (for the mid-1980s versus the mid-1990s); with confidence intervals generated from the 30 autobalance runs. We then determined whether differences between models were plausible, or artifacts created by the uncertainty of the input parameters. 42 Even if this method has some weaknesses, it is the first time an attempt has been made to address uncertainty in a structured way when comparing Ecopath models. Thus, incorporating the autobalance into an ecosystem comparison process is already a huge step forward to understand changes between models. Part of the weaknesses of the approach presented in this section were however addressed in the new method used by Morissette et al. (see Chapter 5 of this thesis) for the Gulf of St. Lawrence, part of the same CDEENA program. A step further: combining Ecopath with other modelling approaches There is a real advantage to using different approaches on the same data to ascertain the robustness of inferred differences between periods and among ecosystems, as done by Morissette and colleagues (Chapter 5 of this thesis) for the CDEENA program on ecosystem modelling. To obtain a balanced solution as well as to test the sensibility of the models, Ecopath was coupled with the inverse approach for the analysis of four Gulf of St. Lawrence models (one model for the northern Gulf, one model for the southern Gulf, both for the mid-1980s and the mid-1990s). The use of inverse modelling to find balanced food-web solutions provided a substantial improvement in objectivity and quantitative rigor compared to previous ecosystem modelling approaches using only Ecopath, with or without its autobalance routine. This method solved the flows of the different mass-balance equation by minimizing the imbalances between inputs and outputs. This inverse approach provided a global criterion for an optimal (balanced) solution (Parker 1977; Enting 1985; Vezina and Piatt 1988; Vezina et al. 2000; Savenkoff et al. 2001). In addition to the basic steady-state constraints of ecosystem models, additional constraints had to be added to obtain a meaningful solution. Each flow was taken to be non-negative and the flows and ratios of flows (metabolic efficiencies) were assumed to fall within certain ranges to satisfy basic metabolic requirements. Gross growth efficiency (GE) is the ratio of production to consumption and for most groups should have values between 0.1 and 0.3 (Christensen and Pauly 1992). Exceptions were top predators, e.g., marine mammals and seabirds, which can have lower GE, and small, fast growing fish 43 larvae or nauplii or bacteria, which can have higher GE (Christensen and Pauly 1992). Following Winberg (1956), 80% of the consumption was assumed to be physiologically useful for carnivorous fish groups while the non-assimilated food (20% consisting of urine and feces) was directed to the detritus. For herbivores, the proportion not assimilated could be considerably higher, e.g., up to 40% in zooplankton (Christensen and Pauly 1992). Assimilation efficiency (AE) was also constrained to fall between 70 and 90% for all the groups except for large and small zooplankton (between 50 and 90%) (Savenkoff et al, in press a). Certain flows have a minimal and maximal value imposed (export for detritus, production, consumption, diet composition, etc.). The production and consumption values that were not estimated from local field studies were used as constraints. To avoid a model that was too severely constrained (constraints on production, consumption, and growth efficiency), we constrained growth efficiency and either production or consumption depending on data availability (e.g., confidence level and local sampling). Diets with reasonable estimates of uncertainty (SD greater than 0.6%) were also specified as constraints. To facilitate comparisons with other Ecopath models, constraints were also added on the EE (Savenkoff et al, in press a). When the system of equations was strongly underdetermined, additional constraints (inequality relations) were added to constrain the range of possible solutions and thus to obtain a meaningful solution. The mass-balance equations and the additional constraints reduced the potential range of flux values, and trophic flows were estimated using an objective least-squares criterion for an optimal (balanced) solution (sum of flows in the system is as small as possible). The solution process thus generated the simplest flow network that satisfied both the mass conservation and constraints. The best solution was the model that produced the smallest sums of squared residuals for the compartmental mass balances. The solution minimized the imbalances between inputs and outputs. The mass balance was closed by residuals (inputs-outputs) instead of ecotrophic efficiencies as in the Ecopath approach (Savenkoff et al. 2004c; Savenkoff et al. in press b). These operations were done with the Optimization Toolbox in the Mathlab® software, ad the equations solution process is fully described in the work of C. Savenkoff (Savenkoff et al. 2004c; Savenkoff et al. in press a, b). 44 To assess the solution's robustness to variations in the data, random perturbations were applied to both input data and right-hand sides of the mass balance equations. A total of 31 balanced solutions corresponding to 31 random perturbations (including a response without perturbation) on each model input to a maximum of its standard deviation was used for each model. A more complete description of these balanced scenarios is given in Savenkoff et al. (2004a) for the northern Gulf of St. Lawrence and in Savenkoff et al. (in press b) for the southern Gulf of St. Lawrence models. Each of the 31 balanced inverse solutions were then transposed into Ecopath to obtain the fishing mortality, predatory mortality, and other mortality, as well as the basic emergent estimates and network analysis indices of the two time periods, combined with their associated uncertainties. The confidence intervals generated by the 31 balanced solutions allowed a comparison of many indices between the two ecosystems (northern and southern Gulf) and/or time periods (mid-1980s or mid-1990s) (Morissette et al, Chapter 5 of this thesis). Using conjointly inverse and Ecopath modelling approaches represent a useful strategy in ecosystem modelling, each tool supplying the other with optimized solutions. The inverse model is very useful to obtain a first balanced solution and to supply Ecopath with diet compositions and efficiencies (metabolic and ecotrophic) using an objective least-squares criterion. This approach also generates complete perturbations on all input parameters for sensibility analyses. The Ecopath model is then used for its strengths to estimate biomass of each groups, generate global ecosystem indices such as network analyses or emergent properties of ecological systems. An approach combining the two modelling methods thus gains robustness and represent an important step further in the comparison of marine ecosystem models through time or space. Conclusions Uncertainty and variability are inherent with ecosystem modelling. Therefore we need to use appropriate tools to define, represent and analyze this uncertainty. Ecopath is only one of the many approaches that are used worldwide for ecosystem modelling. Within this sole approach, many tools has been created and used -with more or less success- to address uncertainty. Despite the fact that there has been no comprehensive study published so far on Ecopath or Ecosim parameter sensitivity, my analysis proposes 45 that the combination with other modelling approaches, in order to use the strengths of each approach and thus gain more robustness, is the best way to address uncertainty and parameter sensitivity for these ecosystem models. / Chapter summary Ecosystem modelling has become a very important way to study marine ecosystems processes, but at the same time few studies have addressed the uncertainty related to model results. This lack of information about how confident we can be in our modeling outcomes poses serious problems, as many sources of uncertainty exist. A valuable tool for model development is the use of the software package Ecopath with Ecosim, which enables the construction of foodwebs and their simulation over time and space according to different scenarios. An important part of the process of ecosystem modelling is to compare results from the model with those from observations, followed by an analysis of the remaining sources of error. However, few of the currently developed Ecopath models have gone so far as to examine the uncertainty in analyses. Thus, it is useful to address this problem, to clearly define the type of uncertainty that may be encountered in ecosystem modelling, and the means by which it may handled. Sensitivity analyses represent one solution by which one might address uncertainty in Ecopath with Ecosim. This approach functions by examining the sensitive elements as revealed in model results with differing scenarios of model-building and construction. In addition, other tools can also be used to perform uncertainty analysis routines. Examples are the Pedigree, Ecoranger, Autobalance, and Monte Carlo tools, all of which are included the Ecopath with Ecosim software package. The strengths and weaknesses of these different ways to address uncertainty are discussed. Finally, I propose a new approach, combining Ecopath with other modelling techniques in order to get an even stronger analysis of uncertainty. 46 References Alderson, P., Green, S., and J.P.T. Higgins (eds.). 2004. Cochrane handbook for systematic reviews of interventions 4.2.1. The Cochrane collaboration, Chichester, UK. Alino, P.M., McManus, L.T., McManus, J.W., Nanola, C.L. Jr., Fortes, M.D. Trono, G.C. Jr. and G.S. Jacinto. 1993. Initial parameter estimations of a coral reef flat ecosystem in Bolinao, Pangasinan, northwestern Philippines, pp. 252-258 In Christensen, V. and D. Pauly (eds.). Trophic Models in Aquatic Ecosystems. ICLARM, Manila. Allen, K.R. 1971. Relation between production and biomass. Journal of Fisheries Research Board of Canada 28: 1573-1581. Arreguin-Sanchez, F., Valero-Pacheco, E., and E.A. Chavez. 1993. A trophic box model of the coastal fish communities of the southwestern Gulf of Mexico, pp. 197-205 In Christensen, V. and D. Pauly (eds.). Trophic Models in Aquatic Ecosystems. ICLARM, Manila. Buchary, E.A. 2001. Preliminary reconstruction of the Icelandic marine ecosystem in 1950 and some predictions with time series data. pp. 296-306 In Guenette, S., Christensen, V., and D. Pauly (eds.). Fisheries impacts on the North Atlantic ecosystems: Models and analyses. Fisheries Centre Research Reports 9(4). Bundy, A. 2004. Mass balance models of the eastern Scotian Shelf before and after the cod collapse and other ecosystem changes. Canadian Technical Report of Fisheries and Aquatic Sciences 2520. Christensen, V. 1991. On ECOPATH, Fishbyte and fisheries management. Fishbyte 9(2): 62-66. Christensen, V., and D. Pauly. 1992. ECOPATH Il-a software for balancing steady-state ecosystem models and calculating network characteristics. Ecological Modelling 61: 169-185. Christensen, V., and C.J. Walters. 2004. Ecopath with Ecosim: methods, capabilities and limitations. Ecological Modelling 172: 109-139. Christensen, V., Walters, C.J., and D. Pauly. 2000. Ecopath with Ecosim: a user's guide. October 2000 Edition, Fisheries Centre, The University of British Columbia, Vancouver, B.C. and ICLARM, Penang, Malaysia. 47 Dutil, J.-D. and Y. Lambert. 2000. Natural mortality from poor condition in Atlantic cod (Gadus morhua). Canadian Journal of Fisheries and Aquatic Sciences 57: 826-836. Dutil, J.-D., Castonguay, M., Gilbert, D., and D. Gascon. 1999. Growth, condition, and environmental relationships in Atlantic cod (Gadus morhua) in the northern Gulf of St. Lawrence and implications for management strategies in the Northwest Atlantic. Canadian Journal of Fisheries and Aquatic Sciences 56: 1818-1831. Enting, I.G. 1985. A classification of some inverse problems in geochemical modelling. Tellus37B: 216-229. Gomes, M.C. 1993. Predictions under uncertainty: Fish assemblages and food webs on the Grand Banks of Newfoundland. ISER, Memorial University, St. John's, Newfoundland. Guenette, S. and I. Diallo. (in press) Exploration d'un modele preliminaire de l'ecosysteme marine de la Guinee. pp. 329-346 In Chavance, P., Ba, M., Gascuel, D., Vakily, J.M. and D. Pauly (eds.). Pecheries, ecosystemes et societes en Afrique de l'Ouest: un demi-siecle de changement. Institut de Recherche pour le Developpement, Paris, France and European Comission, Burssels, Belgium. Heymans, J.J. and T.J. Pitcher. 2002. A model of the marine ecosystem of Newfoundland and Southern Labrador (2J3KLNO) in the time periods 1985-1987 and 1995-1997. pp 5-43 In Pitcher, T.J., Heymans, J.J. and M.' Vasconcellos (eds.). Ecosystem models of Newfoundland for the time periods 1995, 1985, 1900 and 1450. Fisheries Centre Research Reports 10(5). Hilborn, R. 1987. Living with uncertainty in resource management. North American Journal of Fisheries Management 7: 1-5. Hutchings, J.A. 1996. Spatial and temporal variation in the density of northern cod and a review of hypotheses for the stock's collapse. Canadian Journal of Fisheries and Aquatic Sciences 53: 943-962. Hutchings, J.A., and R.A. Myers. 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. 48 Jarre, A., Muck, P., and D. Pauly. 1991. Two approaches for modelling fish stock interactions in the Peruvian upwelling ecosystem. ICES Marine Science Symposia. 193:171-184. Kavanagh, P., Newlands, N., Christensen, V., and D. Pauly. 2004. Automated parameter optimization for Ecopath ecosystem models. Ecological Modelling 172: 141-149. Lightfoot, C , Roger, P.A., Cagauan, A.G., and C.R. Dela Cruz. 1993. Preliminary steady- state nitrogen models of a wetland ricefield ecosystem with and without fish. pp. 56- 64 In Christensen, V. and D. Pauly (eds.). Trophic Models in Aquatic Ecosystems. ICLARM, Manila. Monaco, M.E. and R.E. Ulanowicz. 1997. Comparative ecosystem trophic structure of three U.S. Mid-Atlantic estuaries. Marine Ecology Progress Series 161: 239-254. Moreau, J., Villanueva, M.C., Amarasinghe, U.S., and F. Schiemer. 2001. Trophic relationships and possible evolution of the production under various fisheries management strategies in a Sri Lankan reservoir, pp. 201-214 In S.S. de Silva (ed.) Lakes and Reservoir fisheries management in South East Asia ACIAR publication 97. Moreau, J., Ligtvoet, W., and M.L.D. Palomares. 1993. Trophic relationship in the fish community of Lake Victoria, Kenya, with emphasis on the impact of Nile perch (Lates niloticus). pp. 144-152 In Christensen, V. and D. Pauly (eds.). Trophic Models in Aquatic Ecosystems. ICLARM, Manila. Morissette, L. 2001. Modelisation ecosystemique du nord du Golfe du Saint-Laurent. M. Sc. thesis, Universite du Quebec, Rimouski, QC, Canada. Morissette, L., Despatie, S.-P., Savenkoff, C , Hammill, M.O., Bourdages, H., and D. Chabot. 2003. Data gathering and input parameters to construct ecosystem models for the northern Gulf of St. Lawrence (mid-1980s). Canadian Technical Report of Fisheries and Aquatic Sciences 2497. Okey, T.A., Vargo, G.A., Mackinson, S., Vasconcellos, M., Mahmoudi, B., and C A . Meyer. 2004. Simulating community effects of sea floor shading by plankton blooms over the West Florida shelf. Ecological Modelling 172: 345-365. Opitz, S. 1996. Trophic interactions in Caribbean coral reefs. ICLARM Technical Report 43. 49 Parker, R.L. 1977. Understanding inverse theory. Annual Review of Earth and Planetary Sciences 5: 35-64. Pauly, D , Trites, A , Capuli E , and V. Christensen. 1995. Diet composition and trophic levels of marine mammals. ICES Council Meeting Papers 1995/N:13. Pauly, D , Christensen, V , and C. Walters. 2000. Ecopath, Ecosim and Ecospace as tools for evaluating ecosystem impact of fisheries. ICES Journal of Marine Science 57: 697-706. Pitcher, T.J, Ainsworth, C.H, Lozano, H , Cheung, W.W.L, and G. Skaret. 2005a. Evaluating the role of climate, fisheries and parameter uncertainty using ecosystem- based viability analysis. ICES CM 2005/M:24. Pitcher, T.J, Ainsworth, C.H, Buchary, E , Cheung, W.L, Forrest, R, Haggan, N , Lozano, H , Morato, T. and L. Morissette. 20056. Strategic management of marine ecosystems using whole-ecosystem simulation modelling: the "back to the future' policy approach. Pp. 199-258 In Levner, E, Linkov, I, and Proth, J.M. (eds.). Strategic management of marine ecosystems. NATO Science Series: IV: Earth and environmental sciences, Volume 50. Springer-Verlag, Heidelberg, Germany. Savenkoff, C , Vezina, A. F , and A. Bundy. 2001. Inverse analysis of the structure and dynamics of the whole Newfoundland-Labrador Shelf ecosystem. Canadian Technical Report of Fisheries and Aquatic Sciences 2354. Savenkoff, C , Castonguay, M. Despatie, S.-P, Chabot, D. and L. Morissette. 2004a. Inverse modelling of trophic flows through an entire ecosystem: the northern Gulf of St. Lawrence in the mid-1980s. Canadian Journal of Fisheries and Aquatic Sciences 61:2194-2214. Savenkoff, C , Bourdages, H , Castonguay, M , Morissette, L , Chabot, D , and M.O. Hammill. 20046. Input data and parameter estimates for ecosystem models of the northern Gulf of St. Lawrence (mid-1990s). Canadian Technical Report of Fisheries and Aquatic Sciences 2531. Savenkoff, C , Bourdages, H , Swain, D.P, Despatie, S.-P, Hanson, J .M, Methot, R, Morissette, L , and M.O. Hammill. 2004c. Input data and parameter estimates for ecosystem models of the southern Gulf of St. Lawrence (mid-1980s and mid-1990s). Canadian Technical Report of Fisheries and Aquatic Sciences 2529. 50 Savenkoff, C , Castonguay, M., Chabot, D , Bourdages, H , Morissette, L , and M.O. Hammill. (in press a). Effects of fishing and predation in a heavily exploited ecosystem: comparing pre- and post-groundfish collapse periods in the northern Gulf of St. Lawrence (Canada). Estuarine, Coastal and Shelf Science. Savenkoff, C , Swain, D.P, Hanson, J .M, Castonguay, M , Hammill, M.O, Bourdages, H , Morissette, L , and D. Chabot. (in press b). Effects of fishing and predation in a heavily exploited ecosystem: Comparing pre- and post-groundfish collapse periods in the southern Gulf of St. Lawrence (Canada). Ecological Modelling. Schalk, P.H, Brey, T , Bathman, U , Arntz, W, Gerdes, D , Dieckmann, G , Ekau, W , Gradinger, R , Plotz, J , Nothig, E , Schnack-Schiel, S.B, Siegel, V , Smetacek V.S, and J.A. VanFraneker. 1993. Towards a conceptual model of a Weddell Sea ecosystem, Antarctica, pp. 323-337 In Christensen, V. and D. Pauly (eds.). Trophic Models in Aquatic Ecosystems. ICLARM, Manila. Schartau, M , Oschlies, A , and J. Willebrand. 2001. Parameter estimates of a zero- dimensional ecosystem model applying the adjoint method. Deep Sea Research II 48: 1773-1802. Seijo, J.C, Defeo, O , and S. Salas. 1998. Fisheries bioeconomics. Theory, modelling and management. FAO Fisheries Technical Paper. No.368. Rome, FAO. Silvert, W. 2004. Managing uncertainty in ecosystem model dynamics and the implications and feasibility of spceific management scenarios. EFEP Work Package 5, Final Report (Deliverable 5) 65 p. Sparre, P. 1991. Introduction to multispecies virtual population analysis. ICES Marine Science Symposia 193: 12-21. Vezina, A.F. and T. Piatt. 1988. Food web dynamics in the oceans. I. Best-estimates of flow networks using inverse methods. Marine Ecology Progress Series 42: 269-287. Vezina, A .F , Savenkoff, C , Roy, S, Klein, B , Rivkin, R, Therriault J.-C, and L. Legendre. 2000. Export of biogenic carbon and structure and dynamics of the pelagic food web in the Gulf of St. Lawrence. II. Inverse analysis. Deep-Sea Research 47: 609-635. Winberg, G.G. 1956. Rate of metabolism and food requirements offish. Fisheries Research Board of Canada Translation Series 253. 51 C h a p t e r 3 D a t a q u a l i t y a n d c o m p l e x i t y i n e c o s y s t e m m o d e l s : d o e s i t a f fec t r e s i l i e n c e a n d m o d e l p e r f o r m a n c e ? Introduction In the past decades, biodiversity has experienced a major decline all over the planet, and increasing numbers of animal and plant species are at risk of extinction (McCann 2000). In the world's oceans, overfishing has strongly altered marine food webs, in some cases resulting in the large fish that were commercially exploited being depleted and replaced by smaller, less desirable species (Pauly et al. 1998). Combined with a massive degradation of the marine habitat (Chabanet et al. 2005), this situation threatens marine ecosystems and undermines the stable extraction of vital economic resources from the ocean (Nixon 1997). This problem is an important one, and forces us to examine in details how ecosystems work. There is, in fact, a need to understand how the ecosystems would react to decreases in biodiversity, as well as to other perturbations. For many years, there has been a debate focusing on the question whether simple ecosystems are less stable than complex ones. To date, studies addressing the "diversity- stability debate" (McCann 2000) have yielded ambiguous results. Theoretical studies have generally supported the notion of a negative relationship between stability and complexity (May 1972; Gilpin 1975; Pimm and Lawton 1978; Hogg et al. 1989). Most of the time, these theories were based on mathematical models constructed using parameters drawn randomly from a statistical universe (McCann 2000). A negative complexity-stability relationship was also proposed by Kohdoh (2003), based on adaptive foraging behaviour. Conversely, in nature, studies have shown that trophic links can be connected in a way that enhances foodweb stability (Lawlor 1978, Yodzis 1981; Neutel et al. 2002). It seems that empirical models assembled from real food-webs relationships, with plausible interaction strengths, are generally more stable than randomly constructed food webs (Yodzis 1981; Walters et al. 1997). Pauly et al. (2000), who presented preliminary data suggesting such a link, suggested that if the existence of such a link between model 52 quality and persistence could be confirmed, this would have profound implications for our understanding of ecosystem functioning and the research required for ecosystem-based management. This issue has probably been one of the most controversial in ecology, but most studies devoted to this topic focus on terrestrial systems (Steele 2006). However, considering that the oceans occupy 71% of the surface of biosphere and probably much more in terms of volume (Groombridge and Jenkins 2000), that they are responsible for almost half of the Earth's net primary production (Schlesinger 1997), that more than half the world's population lives within 60 km of the sea, and that marine fisheries can consume up to 35% of the net primary production on shelves (Pauly and Christensen 1995), it is important to examine the relationship between complexity and stability in a marine ecosystem context as well (Naeem 2006). This project represents the first study describing complexity/stability relationships using many marine ecosystems models based on realistic parameter values, and examining the impact of model quality on these relationships. I intend to provide insights on the relationship between diversity and stability by testing whether the quality of data used in ecosystems models, and/or the details of the models used, are related to their stability. Thus, the main hypothesis to be tested in this study addresses an important theoretical ecology issue: is there, in ecosystems, a direct relationship between complexity and stability. My secondary hypothesis is that the quality of ecosystem model inputs also plays a significant role in predicting the relationship between complexity and stability. To test these hypotheses, I chose to use an ecosystem modelling approach. Ecosystem modelling becomes more and more important in fisheries science, particularly for studying the properties of food webs. The Ecopath with Ecosim (EwE) approach allows the construction and validation of food webs (Christensen and Pauly 1992). In addition to making food web and other ecosystem analyses possible, the EwE approach allows researchers to collect and treat a variety of biological information that is available for an ecosystem and put it in a more global perspective. Since J.J. Polovina (1984) proposed the Ecopath approach and software in the early 1980s, many models have been created all over the world (see chapter 1), for different ecosystem and purposes. Having a plausible description of an ecosystem is very important, especially to test hypotheses at an 53 ecosystem scale, and for comparisons among ecosystems. Here, it is proposed to standardize the description of EwE models published so far and to describe the quality of input data used for their construction. A criticism of ecosystem comparisons is that we try to "compare apples and oranges", i.e., that ecosystems are different entities that are fundamentally incommensurable. However, it is easy to compare apples and oranges, which may differ in price, vitamin or water content, color spectrum, etc. (Pauly 2000). Similarly, it is equally possible to compare ecosystems for variables that they share (Downing 1991). In fact, ecosystem comparisons are an important, if underutilized, tool for ecologists. Indeed, many of the scientist recognized as the discoverers of major concepts in ecology, such as Charles Darwin (evolution through natural selection), Alexander von Humboldt (botanical geography and biogeography) and Ernst Haeckel (ecology, as the study of the relationship of organisms with their environment), compared and contrasted the components of ecosystems which lead to important hypotheses, and ultimately, theories (Downing 1991). Material and methods 3.2.1 Ecopath models and their pedigree Comparisons were carried out with 50 Ecopath models (Table 3.1). Each selected model was linked with a publication (thesis, peer-reviewed article, book chapter or report) where data sources were described in details. 54 Table 3.1. List of Ecopath models used for the present study. Location Reference Antarctica, Weddell Sea, 1980s Australia, Darwin Harbour, 1908s Australia, South East Shelf Fishery, 1994/2043 Burundi, Lake Taganyika, 1974-1976 Burundi, Lake Taganyika, 1980-1983 Canada, Northern Gulf of St. Lawrence 1985- 1987 Canada, Southern Gulf of St. Lawrence, 1985- 1987 Canada, Newfoundland, Grand Banks, 1900 Canada, Newfoundland, Grand Banks, 1985- 1987 Canada, Newfoundland, Grand Banks, 1985- 1987 Canada, Newfoundland, Grand Banks, 1995- 1997 Canada, Eastern Scotian Shelf, 1980-1985 Canada, Eastern Scotian Shelf, 1995-2000 Canada, Hecate Strait, 1990s Canada, Lake Ontario, 1983-1993 Caribbean, Puerto Rico-Virgin Islands shelf, 1990s Denmark, Faroe Islands, 1961 Denmark, Faroe Islands, 1997 France, Lac d'Ayda, 1988-1989t France, Bay of Biscay, 1970s France, Bay of Biscay, 1990s French Polynesia, Moorea Island, fringing reef, 1971-1985 French Polynesia, Moorea Island, barrier reef, 1971-1985 Galapagos, Floreana rocky reef, 1997-2000 Greenland, West coast, 1997 Guinea marine ecosystem, 1998 Jarre-Teichmann et al. 1997 Martin 2005 Bulman et al. 2006 Moreau era/. 1993a Moreau et al. 1993a Morissette et al. 2003 Savenkoffe/a/. 2004; Savenkoff et al. (in press) Heymans and Pitcher 2002a Bundy et al. 2000 Heymans and Pitcher 2002b Heymans and Pitcher 2002b Bundy 2004 Bundy 2004 Beattie 2001 Halfon and Schito 1993 Opitz 1996 Zeller and Reinert 2004 Zeller and Freire 2001 Reyes-Marchant 1993 Ainsworth et al. 2001 Ainsworth et al. 2001 Arias-Gonzalez et al. 1997 Arias- Gonzalez et al. 1997 Okey et al. 2004 Pedersen and Zeller 2001 Guenette and Diallo 2004 Table 3.1 (contd.) Iceland Shelf, 1950s Iceland Shelf, 1997 Israel, Lake Kinneret, 1980-1990 Kenya, Lake Victoria, 1971-1972 Kenya, Lake Nakuru, 1972 Mauritania EEZ, 1998 Mexico, Celestun Lagoon, 1990s Mexico, Tamiahua Lagoon, early 1990s Mexico, Northwest Mexico, 1990s Morocco, Atlantic coast, mid-1980s New Caledonia, Western tropical Pacific, 2001 Philippines, Laguna de Bay, 1968 Philippines, Laguna de Bay, 1980 Philippines, San Miguel Bay, 1979-1982 Philippines, Wetland Ricefield, 1990s Portugal, Azores archipelago, 1997 South Africa, Northern Benguela, 1990s Sri Lanka, Prakrama samudra, 1980=1990 Taiwan, Chiku Lagoon, 1990s Uganda, Lake George, 1970s United Kingdom, Scotland's west coast, 2000- 2005 USA, Alaska, Prince william Sound, pre-spill period, 1980-1989 USA, Alaska's Prince William Sound, 1994- 1996 USA, Atlantic Continental Shelf, 1995-1998 USA, Mid-Atlantic Bight, 1995-2005 USA, Southwestern Alaska, Bercharof Lake, 1982-1998 USA, West Florida Shelf, 1997-1998 Zimbabwe, Lake Kariba, 1975-1976 Buchary, 2001 Mendy and Buchary 2001 Wallineefa/. 1993 Moreau et al. 1993b Moreau et al. 2001a Sidi and Guenette 2004 Chavezes a/. 1993 Abarca-Arenas and Pacheco 1993 del Monte-Luna 2004 Stanford et al. 2001 Godinot and Allain 2003 Delos Reyes and Martens 1993 Delos Reyes and Martens 1993 Bundy 1997 Lightfoot et al. 1993 Guenette and Morato 2001 Heymans and Baird 2000 Moreau et al. 2001b Unetal. 1999 Moreau et al. 1993c Morissette and Pitcher 2005 Dalsgaard and Pauly 1997 Okey and Pauly 1999 Okey and Pugliese 2001 Okey 2001 Mathisen and Sands 1999 Okey and Mahmoudi 2002 Machenaefa/. 1993 First, the pedigree was calculated for each model, describing its overall quality. The pedigree of an Ecopath input is a coded statement categorizing the origin and quality of a given input (Christensen et al. 2000; see Table 3.2 for an example of the available choices for the quality of biomass input). Table 3.2. Default options for the pedigree of biomass used in Ecopath models. Defaults (Percentage confidence intervals [CI]) are means based on actual estimates of CI in various studies. Modified from Christensen et al. (2000). Parameter Pedigree Default CI index ( ± % ) Sampling based, high precision 1.0 10 Sampling based, low precision 0.7 40 Approximate or indirect method 0.4 50-80 Guesstimate 0.0 80 From other model 0.0 80 Estimated by Ecopath 0.0 n.a. After documenting and coding every input for each trophic group in a given model, an overall pedigree index for this model is obtained. The routine computes an overall index of model 'quality'; a model is of high quality when it is constructed mainly using precise estimates of various parameters, based on field data obtained from the system to be represented by the model. The pedigree index values can range from zero for a model that is not rooted in local data up to one for a model that is fully rooted in local data (Christensen et al. 2000). After the pedigrees were calculated for each of my 50 selected models, I contacted the authors in order to corroborate the choices taken to describe the pedigree of each input data, and therefore validate the final pedigree of their models. 3.2.2 Costs of computing trophic interaction in Ecopath models Another measure of quality can be identified for ecosystem models: the costs of their construction. Costs can be direct, such as the time and salary it would take for a person and/or a group to construct and validate a model; but they can also be indirect. By indirect costs, I mean the data requirements, i.e., the research that was done before and that is used as input information in the model, as embodied in publication of various types. Estimating these costs is not an easy task (Pauly 1995), but objective bibliometric 56 indicators (e.g., the number of citations it generates) can be used to assess the value of scientific research (Nederhof and Noyons 1992). Fisheries science has been subject to few bibliometric studies. To assess the diverse nature of the different publications used to construct a model, they were classified into categories, based on Rounsefell (1961) index of relative publication. This index was used previously by Dizon and Sadorra (1995) for quantifying scientific productivity, while accounting for different types of publications, i.e., peer-reviewed journal articles (category A in Table 3.3), reports (B and D), newsletter articles (E and F), etc. These categories proved helpful for understanding the reasons for difference in embodied modelling costs among the different types of sources used for the construction of models. As an example, Pauly (1995) reviewed the costs of scientific productivity from various sources in fisheries science, and estimated the costs of a scientific paper to be about US$ 1,000 per page, with higher values for international refereed journals and lower values for technical reports. Given this estimation of US$ 1,000 per page in 1995 for fisheries sciences, one credit, in that specific case, would represent US$ 100. Table 3.3. Relative value of publication by types as used in this study (from Rounsefell 1961). Categories Type of publications included Credit per page Category A Original research requiring extensive analysis or original 10 thought published in a scientific journal, or as a textbook or reference of college level. May include articles describing original techniques and apparatus. Category B Research employing known techniques such as stream 5 surveys, censuses, routine sampling when analysis is not extensive. Includes doctoral dissertations, revisions of textbooks or references, and taxonomic descriptions, keys, or reviews. Category C Narrative and informal reports of scientific merit not requiring 3 analysis. Includes general books or chapters on conservation, encyclopedia articles, and technically correct information bulletins. Category D Reports designed to make raw data available, which may 0.5 include short descriptions of methods and procedures but no analysis. Includes bibliographies, faunal lists, and glossaries. Category E Semi-popular and popular articles, short descriptions of 2 research programs, summaries of research accomplishments. Category F Short faunal notes on species range, etc., critical book 2 reviews, articles published only as abstracts, technical research notes in annual laboratory or commission journals but not in newspapers. 57 An important reason for classifying the different bibliographic sources used to construct ecosystem models into these six categories is to evaluate the relative value of different sources of information. Thus, I believe that using an embodied cost in terms of credits represents a way of adding up all publications and pages used to construct an ecosystem model on a comparable value criterion, whatever the publication date or the currency is. 3.2.3 Quantifying the complexity of models The Ecopath methods allow the calculation of many indices that can describe the complexity of the community structure (Christensen 1995). The attributes were chosen according to a previous study by Vasconcellos et al. (1997), who identified ascendancy, connectance index and system omnivory index as representative of Odum's (1971) attributes of ecosystem complexity. The ascendancy (A) of a food web is a measure developed from Ulanowicz as a measure of the network's potential for competitive advantage over other network configurations (Ulanowicz, 1986). It is computed by Ecopath as the product of 'total system throughput' and the mutual information inherent in the flow network (Field et al. 1989). When divided by the system's capacity (Ca), the relative ascendancy (A/Ca), a dimensionless ratio that excludes the influence of total system throughput and is a suitable index for comparing different ecosystems (Heymans and Baird 2000). It should increase as it matures through a series of development stages (Ulanowicz 1986). The connectance index (Q represents the number of actual trophic links in the food web relative to the number of possible links. Consider a contingency table showing all possible trophic interactions in an ecosystem, with r rows (e.g., prey species) and c columns (e.g., predators). Connectance is calculated as: C = — Eq.3.1 r * c where / is the total number of non-zero elements in the matrix. The connectance index does not use the strength of the flows in its calculation, but is only based on the presence or absence of trophic links. Pimm (1982) defined an omnivore as a 'species which feeds on more than one trophic level'. In Ecopath, an index 58 called system omnivory index (SOT) is computed based on the average omnivory index (01) of all consumers weighted by the logarithm of each consumer's food uptake (Christensen and Pauly 1992). This was inspired by perceived drawbacks of the connectance index. Indeed, the connectance index is strongly dependent on how the groups of the system are defined. As this is quite arbitrary in aquatic systems, where interactions of nearly all groups are possible at some development stage, connectance would be close to one in most systems that are well described. Moreover, a prey would have the same 'score' in the connectance index whether it contributes 1, 10 or 100% of its predators' diet. Both of these drawbacks are overcome by the system omnivory index, which is computed as the variance of the trophic levels of a consumer's food groups. According to Christensen et al. (2000), the use of this index is recommended to characterize the extent to which a system displays web-like features. The omnivory index can be calculated as: OI = fJ{TLj-TLprJ*DCij Eq.3.2 where i is the predator, j the n t h prey, TLj is the trophic level of a predator, TLprey is the average trophic level of n prey species of / and DC,y is the diet composition, expressing the fractions of each / in the diet of /. Finally, the Finn's cycling index (FCI; Finn, 1976) is the fraction of an ecosystem's throughput that is recycled. This index utilizes the Leontief matrix to assess the amount of material cycling within an ecosystem, and is calculated as: T FCI = -2L Eq. 3.3 where Tcyc = X"-iz " = t n e P o r t i ° n ° f t o t a l system throughput (TST) which represents recycled flow, and z ii = the total flow from i which returns to i (without recycling through i en route) over all pathways of all lengths. The FCI varies from 0 (no cycling) to 1, and is also an indicator of system's maturity (Odum 1969; Christensen 1995). 59 3.2.4 Dynamic simulations of ecosystem's stability with Ecosim Pimm (1991) defined a stable ecosystem as persistent, resistant and resilient. I used Ecosim to simulate the stability of the ecosystems after a major perturbation of the system. The resilience refers to the 'elasticity' of the ecosystem's structure, but also to the speed and power of recuperation when an ecosystem is displaced from its equilibrium. This was here calculated as the total time of the simulation (100 years), minus the duration (in years) within which the system returned to its previous equilibrium after a perturbation, if it did. Biomass of a key species was used to calculate resilience. The persistence of an ecosystem refers to the time a variable (here a species' or group's biomass) lasts before it is changed to a new value. In my analysis, this value was calculated in years after the simulated perturbation. Finally, the term 'resistance' expresses the notion of ability to withstand perturbation, and is calculated here as a quantitative index of the ecological consequences in the ecosystem resulting from a permanent change in a variable (ranging from zero = no change, to four = severe losses of species). Inspired by a similar study by Vasconcellos et al. (1997), I generated a perturbation based on a high fishing effort during a period of 10 years. The fishing pattern was a 5-fold increase in fishing mortality, F, on the most species with the largest fishery in the ecosystem. For ecosystems where no fishery was described, I created a fishing mortality of 2.5 year"1 on the fish species with the highest biomass in the ecosystem. This perturbation generated a situation where biomasses are severely depleted or displaced from their original condition in a relatively short period of time (Vasconcellos et al. 1997). After the period of 10 years with high F, this strong fishing mortality was returned to the baseline, and I let the model run for 90 years (for a total simulation of 100 years). Persistence, resilience, and resistance were calculated for each of the 50 models. An additional variable was used to describe the type of fishery occurring in the ecosystem. Indeed, ecosystems where the 'real' fishery was increased to create a perturbation were grouped as 'fishery', while ecosystems where the fishing mortality was created on the most important fish of the ecosystem were grouped as 'pseudo-fishery'. This allowed me to test if the 'shock' I gave to the unfished models was the same as the fishery. 60 3.2.5 Statistical methods Prior to perform multiple regressions, variables were examined for their linearity using the additive and variance stabilizing transformation (AVAS; Tibshirani 1988) implemented in the S-Plus 6 software. AVAS is a generalized additive method in which the predictors are non-parametrically transformed to optimize R 2 and the response is non- parametrically transformed to stabilize variances of residuals. These transformations can be used to suggest appropriate functional forms for standard linear models. I used a multiple regression technique on the optimally transformed values to predict the resilience and persistence (response variables) from a linear combination of several other variables (our ecosystems indicators) with a least squares fitting principle. We used a forward stepwise approach to fit a model with different combinations of regressors. The process selects the best combination of predicting variables, adding regressors until no more variables are significant. Predicting variables were then identified in terms of their importance in forecasting the model's response (resilience or persistence). Results The 50 models analyzed covered the period from 1993 to 2005, and were mainly published as reports, but also theses, book sections and articles (Table 3.4). Many models of my analysis were published in 1993 and in 2001, when the book Trophic models of aquatic ecosystems (Christensen and Pauly 1993), and an extensive Fisheries Centre research report on Fisheries impacts on the North Atlantic ecosystems: models and analyses edited by Guenette et al. in 2001) presented modelling results for many ecosystems. Apart from these two very productive years in terms of models published, the effort to construct and publish ecosystem models changed from models published in primary literature to more detailed models published in theses or reports. 61 Table 3.4. Types of publications describing the Ecopath models sampled for the present analysis. Type of publication Year of publication Article Book chapter Report Theses TOTAL 1993 2 9 - 3 14 1996 1 - 1 - 2 1997 1 - 3 1 5 1999 2 - 1 - ' 3 2000 - - 1 - 1 2001 1 1 10 1 13 2003 - - 7 - 7 2004 1 - 2 - 3 2005 - - 1 1 2 TOTAL 8 10 26 6 50 Constructing an Ecopath model implies using data whose acquisition cost an average of 1,716 credits (Table 3.5). Models published as book chapters and journal articles usually had lower construction costs, while models published as reports and theses appeared more expensive to create. These models are also more detailed in terms of trophic groups. When published as theses or peer-reviewed articles, Ecopath models tend to have a higher pedigree and thus be better in terms of overall quality than when published in reports (Table 3.5). Table 3.5. Embodied costs and quality of Ecopath models depending on the type of publication they represent. Publication Type Average embodied costs (in credits) Average pedigree Average number of groups Journal article 1,373 0.473 19 Book chapter 858 0.429 12 Report 2,035 0.416 36 Thesis 2,326 0.509 29 Global average 1,716 0.440 27 Over two decades, the average number of trophic groups in Ecopath models increased by close to 50%, from an average 24 groups in the 1990s to an average of 33 trophic groups since 2000. As it take more pages to describe more trophic groups (while maintaining a minimum of quality), this increase in the number of trophic groups is 62 significantly related to increased costs to construct these models (F = 10.95, df = 49, p = 0.002; Figure 3.1). u o 7.0 6.5 WD o 6.0 © •8 5.5 H £ 5.0 s w 0.0 r = 0.44 10 20 30 40 Number of groups 50 60 Figure 3.1. Bivariate fit of the embodied construction costs (in credits) of Ecopath models by the number of trophic groups they include. The quality of a model also increases when construction costs are higher. This trend (significant with F = 13.51, df = 49, p = 0.001) is based on the log of the embodied costs, meaning that for high quality models, reaching a higher pedigree level costs more than for low quality models (Figure 3.2). 63 0.8 0.7 0.6 -\ | °'5 if 0.4 "8 0.3 - 0.2 - 0.1 0.0 \ - / / - 0.0 5.5 6.0 6.5 Embodied costs (log credits) r = 0.42 7.0 Figure 3.2. Pedigree of Ecopath models versus their embodied costs. My results also show that the pedigree increase related to embodied costs eventually reaches a plateau, where even if more money is invested in constructing a model, the quality does not increase much (Figure 3.3). 0.8 i 0.7 0.6 I 0-5 If °- 4 0.3 0.2 - 0.1 - 0.0 0. 7^- 0 r = 0.50 5.5 6.0 6.5 Embodied costs (log credits) 7.0 Figure 3.3. Bivariate polynomial fit of the pedigree of Ecopath models versus their embodied costs. 64 For both dependent variables (resilience or persistence), the plot of the AVAS transformed values against the original values (Figure 3.4) showed the same transformations required to obtain linearity in the models. Based on this, it seems that no transformation was required for connectance, while log transformations would be adequate for pedigree and SOI, and exponential transformation for ascendancy. 0 20 40 60 80 Resilience 0.0 -0.1 o o 0.2 0.3 0.4 0.5 0.6 Pedigree 0.05 0.10 0.15 0.20 0.25 SOI 0.2 0.3 0.4 0.5 0.6 Connectance 0 20 40 60 80 100 Ascendency Figure 3.4. Additive and variance stabilizing (AVAS) transformations indicating how parameters (x-axis) may be transformed (y-axis indicates resilience, linear scale) to linearize the individual parameters while considering their joint effects to predict resilience. Results indicate that no transformation is required for connectance, while log transformations would be acceptable for pedigree and SOI, and exponential transformation for ascendancy. Note that the same transformations were obtained for predicting persistence. 3.3.1 Pedigree and stability (simple regression) The following results address stability in terms of resilience, as it is the factor that produced the most significant results for the regression analyses. When the resilience of ecosystems is tested against the pedigree for all models, there is a significant positive relationship (F = 6.5, df = 46, p < 0.015. r = 0.37), showing that resilience increase with higher pedigree (Figure 3.5). 65 Figure 3.5. Resilience versus pedigree for 50 Ecopath models. Grey dots (with resilience = 0) were considered as outliers and not used in the regression analysis. Using a leverage plot for this simple regression allow us to see with the observed versus predicted values if the relationship is significant (Figure 3.6). The points on this plot are actual data coordinates, and the horizontal line for the constrained model is the sample mean of the response. The distance from a point to the line of fit is the actual residual and the distance from the point to the mean is the residual error if the regressor is removed from the model. Because the confidence region between the curves cross the horizontal line, we can conclude that there is a significant relationship between resilience and pedigree (Figure 3.6). 66 I 1—• 1 p » " 1 ° 1 1 I I 0 10 20 30 40 50 60 70 80 90 Resilience Predicted Figure 3.6. Predicted versus observed values for the regression of resilience and pedigree. Grey dots (with resilience = 0) were considered as outliers and not used in the regression analysis. 3.3.2 Predicting ecosystem's resilience (multiple regression) All the optimally transformed values were standardized and then analysed together in a stepwise (backward) multiple regression, in order to select a subset of effects predicting resilience through a regression model. Different sets of predicting variables were tested and ultimately only those which increased the R were included in the regression model. The significant model emerging to predict resilience (F = 6.276, df = 49, p < 0.040) increased in R from 0.104 when it included only pedigree as a variable, to 0.189 when based on partial regression coefficients in Table 3.6. Table 3.6. Effect of the significant regressors for predicting resilience in a multiple regression for all ecosystem models (n = 50). Parameter Partial regression coefficient t Ratio F ratio P Intercept , 0.19 1.48 0.147 Pedigree (log) 0.39 2.92 8.50 0.006 Relative ascendancy 0.22 1.50 2.26 0.141 SOI (log) 0.10 0.69 0.48 0.495 67 When the leverage plot represents a multiple regression, the points are no longer actual data values. The horizontal line then represents a partially constrained model instead of a model fully constrained to one mean value. However, the intuitive interpretation of the plot is the same as for the simple regression. Here again, because the confidence region between the curves does not include the horizontal line, we can say that there is a significant relationship between resilience and the partial regression coefficients (Figure 3.7). -2.0-I 1 1 1 1 1 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Resilience Predicted Figure 3.7. Predicted versus observed values for the model predicting models' resilience for the regression of resilience versus pedigree, ascendancy and SOI. The variables which account for systems with a fishery versus systems with a pseudo-fishery did not show a significant difference in the ecosystem response to perturbation. This means that the 'shock' I gave to the uhfished models was the same as a fishery would do. 3.3.3 Complexity versus stability From my results, there is no direct evidence of a relationship between the stability of ecosystems (expressed here by their resilience) and their complexity (SOI, connectance, Finn's cycling index, or ascendancy). However, when a multiple regression is performed, 68 a significant model emerges predicting that stability is linked to relative ascendancy and SOI (Table 3.6), only when the pedigree is taken into account as regression factor. This results in a positive relationship between these three variables and the resilience of ecosystems. Discussion This analysis represents the first large-scale analysis of the potential relationship between diversity and stability of marine ecosystems, dealing with the possible influence of the quality of the data used to construct these ecosystem models. 3.4.1 About the modelling approach The modelling approach presented here may have some weaknesses, but is innovative in the sense that it suggests new approaches to the analysis of the diversity- stability relationship. With a sample size of 50 ecosystem models from all around the world, we can assume that the outcomes of this modelling effort represent global responses for marine food webs. Studies addressing theoretical issues such as the diversity-stability relationship are usually based on fewer models, and on models from a limited area of the globe. For some particular models, running simulated perturbations (here: strong increase in fishing mortality) in Ecosim allowed us to discover some interesting features. For example, it seems that fishing on large predators (e.g., fishing on tuna in the Azores system) does not destabilize as much the biomass structure of other species as it would do by fishing on lower trophic levels. However, when fishing is targeting large predators, the systems do not return to their original structure, even after 100-year simulation. The loss of higher trophic levels might affect ecosystems' resistance, as opposed, for example, to the loss of keystone species (as modeled by Vasconcellos et al. in 1997), which seems to affect more the resilience of the ecosystem. Most published work on the complexity-stability relationship base their analysis on connectance as an indicator of ecosystem complexity (Rozdilsky and Stone 2001; Eklof and Ebenman 2006). Other authors examined resistance versus species richness (Caldeira 69 et al. 2005). However, such studies can include questionable assumptions about structure and dynamics (Brose et al. 2003; Green et al. 2005), and might not use the best indicator of ecosystem's complexity. My choice of analyzing relative ascendancy and SOI rather than using connectance proved to be relevant. Ascendancy was already identified by Heymans and Baird (2000) as an important indicator of system complexity. The SOI is also considered as a better descriptor of trophic interactions, as proposed by Christensen et al. (2000). Indeed, my database contained some models that were built for the same ecosystems but at different time periods (e.g., the Bay of Biscay, by Ainsworth et al. 2001). In these ecosystems, the trophic structure often remained the same, but the diet changed. The connectance index could not capture these changes, and was thus the same for both ecosystems, while SOI changed, as it was linked to the diet of interacting species. 3.4.2 Data quality and associated modelling outcomes Since it is practically impossible to make controlled laboratory experiments of the dynamics of marine ecosystems, modelling has become an efficient way of addressing global food web issues. The Ecopath modelling approach became more and more important over time, and the size of models increased as well. With one exception, a model constructed in 1996 with 50 trophic groups (Opitz 1996), most of larger models appeared after 2000. We now see some Ecopath models with more than 50 trophic groups (Guenette and Morato 2001; Okey 2001; Heymans and Pitcher 2002b). Obviously, there is no direct correlation between the number of groups in a model and its overall pedigree. Consequently, it is superfluous to waiste time trying to construct models with a very large number of trophic groups, if in the end we have no data to describe these groups. It is instead better to focus on fewer trophic groups backed by good ecological data, from local studies, and thus have a high pedigree. My results indicate that there is a significant relationship between the pedigree of an ecosystem model and its resilience. This means that models with better input data (high pedigree) would sustain their structure intact for a longer time before to go through important changes in terms of species biomass (persistence), and would also return faster to their previous equilibrium after the perturbation (resilience). In other words, this means 70 that models with high pedigree would predict a faster recovery after perturbations. In ecosystem modeling, the quality of input data has often been questioned, but was never analysed as a factor that could change the diversity-stability relationship. With this new approach, my results provide a counter example to the complexity/stability debate by showing that the quality of data used in ecosystems models, as well as the realism of the models used, are really important factors to consider while attempting to describe the relation between complexity and stability in ecosystems. Models of higher quality tend to perform better than poor ones in predicting changes in biomass after a given perturbation. Actual ecosystems have species whose vital parameters (growth, mortality, food consumption, etc.) have co-evolved, allowing for the living part of the system to remain roughly similar for longer periods of time, not withstanding fluctuations of the abiotic environment. Ecosystem models with high pedigree will tend to have parameter values close to the co-evolved values of the actual ecosystem. In contrast, ecosystem models assembled from random parameter sets - even if they exhibit onward stability when not disturbed, will lack such ability to withstand disturbances (environmental, or as a result of fishing), and this is likely to be worse with large (complex) model, which have more parameters. Thus, it is easy to imagine how ecosystem models with random parameters could have misled May (1972), Pimm (1982) and other into inferring that complexity begets instability, that flew in the face of much field ecology. My findings show not only that this is erroneous, but also why. 3.4.3 Cost-effectiveness of ecosystem models The subject of data quality is consistently addressed in many international forums and workshops, and new approaches stem from a growing consensus that management decisions are increasingly reliant on better input data. In 2003, the FAO has defined "cost- effectiveness" as a measure of the relationship between data costs and data benefits. Evaluations of the cost-effectiveness of ecosystem models in relation to improved science and better management have almost never been analysed (FAO 2003). A study by the National Research Council (2000) also states that they: 71 "could find no existing analyses of the costs and benefits of data collection and management for specific fisheries, particularly of the ratio of marginal costs and marginal benefits for each additional dollar spent on data collection". Of course, the approach presented here is only an attempt to evaluate construction costs of ecosystem models, but there is definitely a need for more in-depth analyses on the cost-effectiveness of ecosystem models in order to have a better idea on what part or a management budget should go on collecting good data and construction of reliable models. My results have shown that ecosystem models that are to be used for management ought to be based on high quality data. Modelling is an expensive process. Unfortunately, many fisheries that would greatly benefit from high-quality ecosystem model management are in developing countries, where financial resources are scarce, and not necessarily directed to address these kinds of issues (Pauly 1995). Nevertheless, when comparing the embodied costs of ecosystem models with their pedigree, I discovered that there is an optimal modelling effort (in terms of costs), above which quality reaches a plateau. This result could be seen as a reason, when a pedigree is high, to stop spending money on the data gathering, given the low increase of quality that would be gained. This finding could become very useful to managers in order to focus on the best way of getting a certain amount of data for an ecosystem model, and then being able to maximize its use. On the other hand, if biodiversity is an issue, adding more data on less commercially important species in an ecosystem can become very important and thus worth spending money on. Finally, it is important to point out that as a rule, the data used for Ecopath models were also collected for other purposes. Thus, Ecopath makes use of the expensive data that are already available. I do not know of any datasets that are collected especially in the aim of constructing an Ecopath model. Generally, Ecopath data are used in collaboration with other research projects and do not involve any direct sampling. 72 3.4.4 On complexity-stability relationship My results indicate that two ecosystem complexity factors -ascendancy and omnivory index- can be expected, generally, to correlate wih stability. This corroborates findings that propose that complex ecosystems present more possible trophic links and energy flows between species, and if a perturbation breaks some of these links, other pathways will still be possible to maintain the food web structure and thus make it more stable. If system's complexity has a stabilizing role in the resilience of marine ecosystems, as it is proposed here, management and protection of selected trophic groups (key species of the food webs) can be very important for the overall richness of an ecosystem, and thus its stability. This is not only true for commercially important species or species with larger biomasses in the system. Indeed, non-target species may also play a crucial role in structuring the foodweb, habitats and ecosystem processes (Coleman and Williams 2002). Thus, if management choices have to be taken to protect marine species, they should be based on the entire structure of marine ecosystem, and not only on a single-species point of view. When the complexity-stability relationship is analyzed in a context of foodweb (and not from statistical procedures), most studies are based on simplified food webs with few trophic groups (e.g., Eklof and Ebenman 2006; Thebault and Loreau 2006). Here, my relationships are based on complex models representing more trophic groups, and for all parts of the world, making it more global and representative of a variety of marine ecosystems. The results also show that unlike other theoretical studies about the diversity- stability relationship (see Wilmers et al. 2002 for a review), the present study supports the notion that as community complexity (here presented as ascendancy and omnivory index) increases, stability increases too. Most of the time, this notion was supported by empirical studies conducted in controlled microcosms. This may suggest that my models were more representative of real ecosystems. However, we should note that there is a wide range of interests in analyzing ecosystem performance, and different measures serve different interests (see Steele 2006 for a review). While conservationists want to preserve diversity by creating marine reserves, fisheries scientists and managers would prefer to preserve productivity by imposing stricter fishing quotas. However, diversity, productivity and stability provide 73 unrelated measures of ecosystem performance (Steele 2006). There is thus a need to address these measures separately and understand them thoroughly to preserve marine ecosystems as much as possible, whatever the interest is. Ecosystem modelling is certainly an important tool to reach that goal. "Scientists have broken down many kinds of systems. They think they know most of the elements and forces. The next task is to reassemble them, at least in mathematical models that capture the key properties of the entire ensembles. " Wilson 1998 Chapter summary For many years, there has been a debate focusing on the question whether simple ecosystems are less stable than complex ones. To date, studies addressing the "diversity- stability debate" (McCann 2000) have yielded ambiguous results. Moreover, few studies devoted to this topic have been focusing on marine ecosystems, even if they cover a very important surface of the biosphere. I used fifty empirical models assembled from real food-webs relationships, with plausible interaction strengths, based on the Ecopath modelling approach. The relationship between complexity and stability was analyzed, along with the effect of the quality of data used in ecosystem models. Cost of constructing Ecopath models was discussed, based on a bibliometric analysis. My results indicate that there is a significant relationship between the pedigree of an ecosystem model and its resilience. With this new approach, I provide a counter example to the complexity/stability debate by showing that the quality of data used in ecosystems models, as well as the realism of the models used, are really important factors to consider while attempting to describe the relation between complexity and stability in ecosystems. Models of higher quality tend to perform better than poor ones in predicting changes in biomass after a given perturbation. Thus, ecosystem models that are to be used for management or prediction ought to be based on high quality data. However, modelling is an expensive process: when comparing the embodied costs of ecosystem models with their pedigree, I discovered that there is a direct relationship and an optimal modelling effort (in terms of 7 4 costs), above which quality reaches a plateau. Finally, my results also show that unlike other theoretical studies about the diversity-stability relationship, the present study supports the notion that as community complexity (here presented as ascendancy and omnivory index) increases, stability increases too. References Abarca-Arenas, L.G. and E. Valero-Pachenco. 1993. Toward a trophic model of Tamiahua, a coastal lagoon in Mexico, pp. 181-185 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Ainsworth, C , Ferris, B., Leblond, E., and Guenette, S., 2001. The Bay of Biscay, France; 1998 and 1970 models, pp. 271-313 In Guenette, S., Christensen, V., and D. Pauly (eds.). Fisheries impacts on North Atlantic Ecosystems: Models and analyses. Fisheries Centre Research Reports 9(4). Ainsworth, C , Heymans, J.J., Pitcher, T.J. and M. Vasconcellos (eds.). 2002. Ecosystem models of Northern British Columbia for the time periods 2000, 1950, 1900 and 1750. Fisheries Centre Research Reports 10(4). Arias- Gonzalez, J.E., Delesalle, B., Salvat, B., and R. Galzin. 1997. Trophic functioning of the Tiahura reef sector, Moorea Island, French Polynesia. Coral Reefs 16:231- 246. Beattie, A. 2001. A new model for evaluating the optimal size, placement and configuration of marine protected areas. M.Sc. thesis, University of British Columbia, Vancouver, BC. Broombridge, B. and M.D. Jenkins. 2000. Global biodiversity: Earth's living resources in the 21st century. World Conservation Press, Cambridge. Brose, U., Martinez, N.D., Williams, R.J. 2003. Estimating species richness: sensitivity to sample coverage and insensitivity to spatial patterns. Ecology 84: 2364-2377. Buchary, E.A. 2001. Preliminary reconstruction of the Icelandic marine ecosystem in 1950 and some predictions with time series data, pp 296-306 In Guenette, S., Christensen, V., and D. Pauly (eds.). Fisheries impacts on the North Atlantic ecosystems: Models and analyses. Fisheries Centre Research Reports 9(4). 75 Bulman, C , Condie, S, Furlani, D , Cahill, M , Klaer, N , Goldsworthy, S. and I. Knuckey. 2006. Trophic dynamics of the eastern shelf and slope of the south east fishery: impacts of and on the fishery. Final Report for the Fisheries Research and Development Corporation, Project 2002/028. Bundy, A. 1997. Assessment and management of multispecies multigear fisheries: a case study from San Miguel Bay, the Philippines. Ph.D. thesis, University of British Columbia, Vancouver, Canada. Bundy, A. 2001. Fishing on ecosystems: interplay of fishing and predation in Newfoundland-Labrador. Canadian Journal of Fisheries and Aquatic Sciences 5: 1153-1167. Bundy, A. 2004. Mass balance models of the eastern Scotian Shelf before and after the cod collapse and other ecosystem changes. Canadian Technical Report of Fisheries and Aquatic Sciences 2520. Bundy, A , Lilly, G.R, and PA. Shelton. 2000. A mass balance model of the Newfoundland-Labrador shelf. Canadian Technical Report of Fisheries and Aquatic Sciences 2310. Caldeira, M . C , Loreau, H . M , and J.S. Pereira. 2005. Species richness, temporal variability and resistance of biomass production in a Mediterranean grassland. Oikos 110: 115-123. Chabanet, P, Adjeroud, M., Andrefouet, S, Bozec, J .M, Ferraris, J , Garcia-Charton, J , and M. Schrimm. 2005. Human-induced physical disturbances and their indicators on coral reef habitats: A multi-scale approach. Aquatic Living Resources 18: 215- 230. Chavez, E .A, Garduno, M , and F. Arreguin-Sanchez. 1993. Trophic dynamic structure of Celestun Lagoon, Southern Gulf of Mexico, pp. 186-192 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Christensen, V. 1995. Ecosystem maturity - towards quantification. Ecological Modelling 77: 3-32. Christensen, V. and D. Pauly. 1992. The ECOPATH II - a software for balancing steady- state ecosystem models and calculating network characteristics. Ecological Modelling 61: 169-185. 76 Christensen, V., Walters, C.J., and D. Pauly. 2000. Ecopath with Ecosim: a user's guide. October 2000 Edition, Fisheries Centre, The University of British Columbia, Vancouver, B.C. and ICLARM, Penang, Malaysia. Coleman, F.C. and S.L. Williams. 2002. Overxploiting marine ecosystems engineers: potential consequences of biodiversity. Trends in Ecology and Evolution 17: 40-44. Dalsgaard, J. and D. Pauly (eds.). 1997. Preliminary mass-balance model of Prince William Sound, Alaska, for the pre-spill period 1980-1989. Fisheries Centre Research Reports 5(2). del Monte-Luna, P. 2004. Characterization of the biological activity center of the Gulf of Ulloa, Mexico, under an ecological modelling context. Ph.D. thesis, Instituto Politecnico Nacional, Centro Interdisciplinario de Ciencias Marinas, Mexico. Delos Reyes, M. and R. Martens. 1993. Geoecology of Laguna de Bay, Philippines. I. Techno-commercial impact on the trophic level structure of the Laguna de Bay aquatic ecosystem 1968-1980. Ecological modelling 75/76: 497-509. Dizon, L.B. and M.S.M. Sadorra. 1995. Patterns of publication by the staff of an international fisheries research center. Scientometrics 32: 67-75. Downing, J. A. 1991. Comparing apples with oranges: Methods of interecosystem comparison. 24-45 In Cole, J., Lovett, G., and S. Findlay. Comparative analyses of ecosystems. Patterns, mechanisms, and theories. Springer-Verlag, New York. Eklof, A. and B.O. Ebenman. 2006. Species loss and secondary extinctions in simple and complex model communities. Journal of animal ecology 75: 239-246. Finn, J.T. 1976. Measures of ecosystem structure and function derived from analysis of flows. Journal of Theoretical Biology 56: 363-380. Gilpin, M.E. 1975. Stability of feasible predator-prey systems. Nature 254: 137-139. Godinot, O. and V. Allain. 2003. A preliminary Ecopath model of the warm pool pelagic ecosystem. Working paper BBRG-5. 16th meeting on the standing committee on tuna and billfish, 9-16 July 2003, Mooloolaba, Australia. Green, J.L., Hastings, A., Arzberger, P., Ayala, F.J., Cottingham, K.L., Cuddington, K., Davis, F., Dunne, J.A., Fortin, M.-J., Gerber, L., and M. Neubert. 2005. Complexity in ecology and conservation: mathematical, statistical and computational challenges. Bioscience 55: 501-510. 77 Guenette, S. and I. Diallo. 2004. Modeles de la cote Guineenne, 1985 et 1998. pp 124-159 In Palomares, M.L. and D. Pauly (eds.). West African marine ecosystems : models and fisheries impacts. Fisheries Centre Research Reports 12(7). Guenette, S. and T. Morato. 2001. The Azores archipelago in 1997. pp 241-270 In Guenette, S, Christensen, V , and D. Pauly (eds.). Fisheries impacts on the North Atlantic ecosystems: models and analyses. Fisheries Centre Research Reports 9(4). Halfon, E. and N. Schito. 1993. Lake Ontario food web, an energetic mass balance, pp. 29-39 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Herbertz, H. and B. Muller-Hill. 1995. Quality and efficiency of basic research in molecular biology: a bibliometric analysis of thirteen excellent research institutes. Research Policy 24: 959-979. Heymans, J.J, and D. Baird. 2000. Network analysis of the northern Benguela ecosystem by means of NETWRK and ECOPATH. Ecological Modelling 131:97-119. Heymans, J.J, and T.J. Pitcher. 2002a. A Picasso-esque view of the marine ecosystem of Newfoundland and Southern Labrador: Models for the time periods 1450 and 1900. pp 44-71 In Pitcher, T.J, Heymans, J.J. and M. Vasconcellos (eds.). Ecosystem models of Newfoundland for the time periods 1995, 1985, 1900 and 1450. Fisheries Centre Research Reports 10(5). Heymans, J.J, and T.J. Pitcher. 2002b. A model of the marine ecosystem of Newfoundland and Southern Labrador (2J3KLNO) in the time periods 1985-1987 and 1995-1997. pp 5-43 In Pitcher, T.J, Heymans, J.J. and M. Vasconcellos (eds.). Ecosystem models of Newfoundland for the time periods 1995,1985, 1900 and 1450. Fisheries Centre Research Reports 10(5). Hogg, T , Huberman, B.A, and J.M. McGlade. 1989. The stability of ecosystems. Proceedings of the Royal Society of London Series B - Biological Sciences 237: 43- 51. Jarre-Teichmann, A , Brey, T , Bathmann, U.V, Dahm, C , Dieckmann, G.S, Gorny, M , Klages, M , Pages, F , Plotz, J , Schnack-Schiel, S.B, Stiller, M , and WE. Arntz. 1997. Trophic flows in the benthic shelf community of the eastern Weddell Sea, 78 Antarctica, pp. 118-134 In Valencia, J., Walton, D.W.H., and B. Battaglia (eds.). Antarctic communities: species, structure and survival. Cambridge University Press. Lawlor, L.R. 1978. A comment on randomly constructed model ecosystems. American Naturalist 112:445-447. Lightfoot, C , Roger, P.A., Cagauan, A.G. and C.R. dela Cruz. 1993. Preliminary steady- state nitrogen models of a wetland ricefield ecosystem with and without fish. pp. 56- 64 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Lin, H.-J., Shao, K.-T., Kuo, S.-R., Hsieh, H.-L., Wong, S.-I., Chen, I.-M., Lo, W.-T., and J.-J. Hung. 1999. A trophic model of a sandy barrier lagoon at Chiku in southwestern Taiwan. Estuarine, Coastal and Shelf Science 48: 575-588. Machena, C , Kolding, J., and R.A. Sanyanga. 1992. Preliminary assessment of the trophic structure of Lake Kariba, Africa, pp. 130-137 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Martin, J.M. 2005. The distribution, abundance and trophic ecology of the fishes of Darwin Harbour mangrove habitats. Ph.D. thesis, Charles Darwin University, Darwin, Australia. Mathisen, O.A. and N.J. Sands. 1999. Ecosystem modeling of Becharof Lake, a Sockeye salmon nursery lake in Southwestern Alaska, pp. 685-703 In Ecosystem approaches for fisheries management. Alaska Sea Grant College Program AK-SG-99-01. May, R. M. 1972. Will a large complex system be stable? Nature 238: 413-414. McCann, K. S. 2000. The diversity-stability debate. Nature 405: 228-233. Mendy, A.R. and E. Buchary. 2001. Constructing the Iceland marine ecosystem model for 1997 using a mass balance modeling approach, pp. 182-197 In Guenette, S., Christensen, V., and D. Pauly (eds.). Fisheries impacts on North Atlantic ecosystems: Models and analyses. Fisheries Centre Research Reports 9(4). Monaco, M. E. and R. E. Ulanowicz. 1997. Comparative ecosystem trophic structure of three U.S. Mid-Atlantic estuaries. Marine Ecology Progress Series 161: 239-254. Moreau, J., Nyakageni, B., Pearce, M., and P. Petit. 1993a. Trophic relationships in the pelagic zone of Lake Tanganyika (Burundi Sector), pp. 138-143 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. 79 Moreau, J. Ligtvoet, W , and M.L.D. Palomares. 1993b. Trophic relationship in the fish community of Lake Victoria, Kenya, with emphasis on the impact of nile perch (hates niloticus). pp. 144-152 In Pauly, D. and V. Christensen (eds.). Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Moreau, J , Christensen, V , and D. Pauly. 1993c. A trophic ecosystem model of Lake George, Uganda, pp. 124-129 In Pauly, D. and V. Christensen (eds.) Trophic models of aquatic ecosystems. ICLARM, Manila. Moreau, J , Mavuti, K , Daufresne, T. 2001a. A synoptic Ecopath model of biomass flows during two different static ecological situations in Lake Nakuru (Kenya). Hydrobiologia458: 63-74. Moreau J , Villanueva, M . C , Amarasinghe, U.S. and F. Schiemer. 2001b. Trophic relationships and possible evolution of the production under various fisheries management strategies in a Sri Lankan reservoir, pp. 201-214 In De Silva, S. (ed.). Reservoir and culture-based fisheries: Biology and management. ACIAR, Canberra, Australia. Morissette, L. and T. Pitcher. 2005. Model structure and balancing, pp. 5-24 In Haggan, N. and T.J. Pitcher (eds.). Ecosystem simulation models of Scotland's West coast and sea lochs. Fisheries Centre Research Reports 13(4). Morissette, L , S.-P. Despatie, C. Savenkoff, M.O. Hammill, H. Bourdages, and D. Chabot. 2003. Data gathering and input parameters to construct ecosystem models for the northern Gulf of St. Lawrence (mid-1980s). Canadian Technical Report of Fisheries and Aquatic Sciences 2497. Naeem, S. 2006. Expanding scales in biodiversity-based research: challenges and solutions for marine systems. Marine Ecology Progress Series 311: 273-283. National Research Council, Ocean Studies Board. 2000. Improving the collection, management and use of marine fisheries data. National Academy Press, Washington, D.C. Nederhof, A.J. and E.C.M. Noyons. 1992. Assessment of the international standing of university departments' research: A comparison of bibliometric methods. Scientometrics 24: 393-404. 80 Neutel, A.-M., Heesterbeek, J.A.P., and P.C. de Ruiter. 2002. Stability in real food webs: weak links in long loops. Science 296: 1120-1123. Nixon, A. 1997. World fisheries: the current crisis. Library of Parliament, Science and Technology Division. Background Paper BP-434E. Odum, E.P. 1969. The strategy of ecosystem development. Science 104: 262-270. Odum, E. P. 1971. Fundamentals of Ecology. Philadelphia, PA, W.B. Saunders. Okey, T.A. 2001. A 'straw-man' Ecopath model of the Middle Atlantic Bight continental shelf, United States, pp. 151-166 In Guenette, S., Christensen, V., and D. Pauly (eds.). Fisheries impacts on North Atlantic ecosystems: Models and analyses. Fisheries Centre Research Reports 9(4). Okey, T.A. and B. Mahmoudi (eds.). 2002. An ecosystem model of the West Florida Shelf for use in fisheries management and ecological research: Volume II. Model construction. Fish and Wildlife Conservation Commission, Florida Marine Research Institute, St. Petersburg. Okey, T. and D. Pauly (eds.). 1999. A trophic mass balance model of Alaska's Prince William Sound ecosystem for the post-spill period 1994-1996. Fisheries Centre Research Reports 7(4). Okey, T.A. and R. Pugliese. 2001. A preliminary Ecopath model of the Atlantic continental shelf adjacent to the Southeastern United States, pp. 167-181 In Guenette, S., Christensen, V., and D. Pauly (eds.). Fisheries impacts on North Atlantic ecosystems: Models and analyses. Fisheries Centre Research Reports 9(4). Okey, T.A., Banks, S., Born, A.F., Bustamante, R.H., Calvopina, M., Edgar, G.J., Espinoza, E., Farina, J.M., Garske, L.E., Reck, G.K., Salazar, S., Shepherd, S., Toral-Granda, V., and P. Wallem. 2004. A trophic model of Galapagos subtidal rocky reef for evaluating fisheries and conservation strategies. Ecological Modelling 172: 383-401. Opitz, S. 1996. Trophic interactions in Caribbean Coral Reefs. ICLARM Technical Report 43. Pauly, D. 1995. When is fisheries management needed? Keynote presented at the Joint FF A/SPC Workshop on the Management of South Pacific Inshore Fisheries. Noumea, 81 New Caledonia, June 26 -July 7, 1995. (available online at http://www.spc.int/coastfish/Reports/ICFMAP/pauly.pdf) Pauly, D. and V. Christensen. 1995. Primary production required to sustain global fisheries. Nature 374: 255-257. Pauly, D , Christensen, V , Dalsgaard, J , Froese, R. and F. Torres Jr. 1998. Fishing down marine food webs. Science 279: 860-863. Pauly, D , V. Christensen, and C. Walters. 2000. Ecopath, Ecosim, and Ecospace as tools for evaluating ecosystem impact of fisheries. ICES Journal of Marine Science 57: 697-706. Pedersen, S.A. and D. Zeller. 2001. A mass balance model for the West Greenland marine ecosystem, pp 111-127 In Guenette, S., Christensen, V. and D. Pauly (eds.). Fisheries impacts on North Atlantic Ecosystems: Models and Analyses. Fisheries Centre Research Reports 9(4). Pimm, S.L. 1982. Food webs. London, Chapman and Hall. Pimm, S.L. 1991. The balance of nature? Ecological issues in the conservation of species and communities. Chicago, University of Chicago Press. Pimm, S.L. and J.H. Lawton. 1978. Feeding on more than one trophic level. Nature 275: 542-544. Pinnegar, J.K, Blanchard, J.L, Mackinson, S, Scott, R.D, and D.E. Duplisea. Aggregation and removal of weak-links in food-web models: system stability and recovery from disturbance. Ecological Modelling 184: 229-248. Reyes-Marchant, P. Jamet, J.L, Lair, N , Taleb, H , and M.L.D. Palomares. 1993. A preliminary model of a eutrophic lake (Lake Aydat, France), pp. 95-101 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Roberts, C M . and J.P. Hawkins. 1999. Extinction risk in the sea. Trends in Ecology and Evolution 14: 241-246. Rounsefell, G.A. 1961. How can research production be measured? Proceedings of the Gulf and Caribbean Fisheries Institute 13: 139-150. Rozdilsky, I.D. and L. Stone. 2001. Complexity can enhance stability in competitive systems. Ecology letters 4: 397-400. 82 Savenkoff, C , Bourdages, H , Swain, D.P, Despatie, S.-P, Hanson, J .M, Methot, R, Morissette, L. and M.O. Hammill. 2004. Input data and parameter estimates for ecosystem models of the southern Gulf of St. Lawrence (mid-1980s and mid-1990s). Canadian Technical Report of Fisheries and Aquatic Sciences 2529. Savenkoff, C , Swain, D.P, Hanson, J .M, Castonguay, M , Hammill, M.O, Bourdages, H , Morissette, L , and D. Chabot. (in press). Effects of fishing and predation in a heavily exploited ecosystem: Comparing pre- and post-groundfish collapse periods in the southern Gulf of St. Lawrence (Canada). Ecological Modelling. Schlesinger, W.H. 1997. Biogeochemistry, 2nd edition. Academic Press, San Diego, CA. Sidi, M.T. and S. Guenette. 2004. Modele trophique de la ZEE mauritanienne: comparaison de deux periodes (1987 et 1998). pp 12-38 In Palomares, M.L.D. and D. Pauly (eds.). West African marine ecosystems : models and fisheries impacts. Fisheries Centre Research Reports 12(7). Stanford, R, Lunn, K , and S. Guenette. 2001. A preliminary ecosystem model for the Atlantic coast of Morocco in the mid-1980s, pp 314-344 In Guenette, S, Christensen, V , and D. Pauly (eds.). Fisheries impacts on North Atlantic Ecosystems: models and analyses. Fisheries Centre Research Reports 9(4). Steele, J.H. 2006. Are there eco-metrics for fisheries? Fisheries Research 77: 1-3. Thebault, E. and M. Loreau. 2006. The relationship between biodiversity and ecosystem functioning in food webs. Ecological Research 21: 17-25. Tibshirani, R. 1988. Estimating transformations for regression via additivity and variance stabilization. Journal of the American Statastical Association 83: 394-405. Ulanowicz, R.E (ed.). 1986. Growth and development: Ecosystem phenomenology. Springer, New York. Vasconcellos, V , S. Mackinson, K. Sloman and D. Pauly. 1997. The stability of trophic mass-balance models of marine ecosystems: a comparative analysis. Ecological Modelling 100: 125-143. Walline, P.D, Pisanty, S, Gophen, M , and T. Berman. 1993. The ecosystem of Lake Kinneret, Israel, pp. 103-109 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. 83 Wilmers, C.C., Sinha, S., and M. Brede. 2002. Examining the effects of species richness on community stability: an assembly model approach. Oikos 99: 363-367. Wilson, E.O. 1998. Consilience: the unity of knowledge. Knopf, New York. Yodzis, P. 1981. The stability of real ecosystems. Nature 289: 674-676. Zeller, D. and J. Reinert. 2004. Modelling spatial closures and effort restrictions in the Faroe Islands marine ecosystem. Ecological Modelling 172: 403-420. Zeller, D. and K.M.F. Freire. 2001. A north-east Atlantic marine ecosystem model for the Faroe Islands (ICES area Vb): input data, pp 207-212 In Guenette, S., Christensen, V. and D. Pauly (eds.). Fisheries impacts on North Atlantic ecosystems: models and analyses. Fisheries Centre Research Reports 9(4). 84 Chapter 4 Modelling the indirect positive feedback of marine mammal predation: outcomes for competition with fishery and ecosystem structure. Introduction The interaction between marine mammals and fisheries has received growing attention during the last 20 years (Yodzis 1998; Gales et al. 2003; Mackinson et al. 2003). Numerous studies and investigations of interactions between marine mammals and fisheries have been implemented around the world. Even if many studies have examined how fisheries may impact marine mammal populations (Dans et al. 2003; Roman and Palumbi 2003; DeMaster et al. 2001), the degree to which marine mammals compete for food with fisheries has been studied somewhat less broadly (Northridge 1991; Smith 1995; Bowen 1997; DeMaster et al. 2001; Kaschner et al. 2001; Kaschner and Pauly 2005). Nevertheless, this competition for fish resources between marine mammals and fisheries may be the primary source of current and future conflicts (Trites et al. 1997; Kaschner 2004; Dillingham et al. 2006). Besides the fact that many authors now document a growing concern about the widespread decline of many marine mammal populations on the planet (Bowen 1997; Trites et al. 1997; Springer et al. 2003; Estes et al. 2005; Kraus et al. 2005), there is a serious need to address the issue of their competition with fishery for the same food resources. Understanding the ways in which marine mammals and fisheries compete, and where and when they do, is not an easy task. First of all, detailed information on the dynamics of predation rates and how these relate to either fluctuation in fish availability or fluctuations in marine mammal population size is lacking (Northridge 1991, MacKenzie et al. 2002). Furthermore, it is usually very difficult to observe marine mammal feeding and/or interacting with fisheries (Williams et al. 2004; Kaneko and Lawler 2006). Quantifying their diets with estimation models (scats, stomach contents, fatty acids, etc.) 85 is also problematic as diets can vary substantially over time (Lawson and Stenson 1995; Bowen and Siniff 1999; Laake et al. 2002; Hume et al. 2004). Finally, even on the fisheries side, exhaustive data on yield and precise estimates on catches, bycatch (especially the commercially less important species), or discards are known to be relatively hard to get (National Research Council 1999; Zeller and Pauly, 2005). The use of models that are validated for ecological reliability becomes an interesting tool to address the trophic role of marine mammals in ecosystems, and their competition with fisheries. Until now, Ecopath with Ecosim (EwE) models have been considered inadequate to provide reliable answers in the context of competition between marine mammals and fisheries, being more used to generate hypotheses about possible impacts of fisheries on marine ecosystems (Kaschner 2004). However, even if Ecopath models are generally constructed in a fisheries context, and merely consider the commercially important species, in some cases marine mammal compartments are included in the models, making a better representation of all trophic interactions in the entire ecosystem (Morissette et al. 2006). In order to investigate trophic interactions between marine mammals and fisheries, the present analysis has the following objectives: ' 1. to calculate the resource overlap between marine mammals and fisheries using ecosystem models; 2. to examine the global trophic impacts of marine mammals and fisheries on the key species of each ecosystem; and 3. to simulate the extirpation of marine mammal populations in these ecosystems and analyze the resulting changes on the structure of the foodweb. The principal hypothesis of this chapter is that the presence of marine mammals in ecosystems and their overlap with fisheries for food resources does not have a significant effect on the overall productivity of the ecosystem, but that their removal by humans can decrease this productivity and alter the structure of the food web. 86 Materials and methods 4.2.1 Global ecosystem representation The Ecopath approach was developed in the early 1980s by J.J. Polovina and co- workers at the NMFS Laboratory in Honolulu. It was first applied to a coral reef system, as a tool to allow quantitative analyses of biological interactions within aquatic systems (Polovina 1984). Ecopath was originally developed by J.J. Polovina (1984) as a tool to quantitavely analyse the biological interactions within aquatic systems. Modified and updated since then by Christensen and Pauly (1992, 1993), Ecopath with Ecosim is now a widespread software package widely used for the analysis of exploited aquatic ecosystems (Christensen and Walters 2004). Currently, the software counts more than 2800 registered users from approximately 120 countries (see Chapter 1 of this thesis). Because a good coverage based on the same modelling methodology is available throughout the world's oceans, I could take models constructed using this approach to quantify and analyze the impact of marine mammals' consumption in marine food webs. The seven ecosystem models selected for this analysis were chosen in terms of their geographical distribution and the quality of their documentation. The quality of input data used in the models matters, as it has been shown that model performance is influenced by the quality of input data (see Chapter 3 of this thesis), as well as by the aggregation of trophic groups (Pinnegar et al. 2005). Particular effort was made to cover both northern and southern hemispheres, in an attempt to have a representative coverage of the world's oceans for the global extrapolation. Kaschner and Pauly (2005) have shown that the prominent hotspots of competition between marine mammals and fisheries include the Bering Sea where the potential negative impacts of the US groundfish fisheries on the endangered western population of Steller sea lions (Eumetopias jubatus) have been a great concern (Fritz et al. 1995, Loughlin and York 2000) and the east coast of North America where the largest annual marine mammal cull worldwide is in part being justified based on the perception that the growing harp seal (Phoca groenlandica) population impedes the recovery of the northwest Atlantic cod (Gadus morhua) stocks (see review in Yodzis 2001). In addition, their model identified areas of potential conflict in the Benguela system off southwest Africa with the potential impacts of the increasing population of South 87 African fur seals (Arctocephalus pusillus) on the hake stocks has been an issue of much debate (Wickens et al. 1992, Punt and Butterworth 1995). To these three ecosystems I added other hotspots identified in the Kaschner and Pauly (2005) analysis (see their Figure 7): the eastern tropical Pacific Ocean; the North Sea; the Gulf of Thailand and the Strait of Georgia (Figure 4.1). Figure 4.1. Location of the ecosystem modeled with Ecopath and used for this analysis. For each of the ecosystem selected, an Ecopath with Ecosim model was obtained from the scientists who created it, and verified for its ecological reliability. By this, I mean models reliable enough to re-create time series of biomass changes that are similar to official data. The general information on each ecosystem model used for my analysis is described in Table 4.1. The eastern Bering Sea model (National Research Council, 2003) was constructed to examine interactions between Alaska groundfish fisheries and Steller sea lions and the role of these fisheries on the evolving status of the sea lion population. A total of 26 functional groups (representing hundreds of species that make up the Bering Sea ecosystem) are included in this model. Data on marine mammals mainly came from local studies and are quite precise. Six groups of marine mammals are represented in that model, covering whale and seal species (Table 4.1). In the northwest Atlantic, Ecopath models were developed for the Northern and the Southern Gulf of St. Lawrence ecosystems, for the periods before (mid-1980s) and 88 after (mid-1990s) the collapse of demersal fish stocks in the two areas. The objective was to determine if significant changes in ecosystem structure had occurred between the two time periods that might contribute to the failure of the cod stocks to recover in the 1990s (see Chapter 5 of this thesis). The northern Gulf of St. Lawrence model was here used to represent a typical Northwest Atlantic ecosystem. This model contains five marine mammals groups, covering whale and seal species (Table 4.1). An EwE model of the southern Benguela was fitted to available time-series data to explore how changes in target fish populations could be attributed to feeding interaction terms and population control patterns, the impact of fishing, and environmental forcing (Shannon et al. 2004). Two marine mammal groups are present in the ecosystem: cetaceans and seals (Table 4.1). An EwE model was constructed by Olson and Watters (2003) to represent the tropical region of the eastern Pacific Ocean, aiming to get insights into the relationships among the various species in the system. The model was also used to explore ecological implications of alternative methods of harvesting tunas. Even if the model was primarily built around the latter species, four groups of marine mammals were nevertheless represented: baleen whales; toothed whales; spotted dolphin (Stenella attenuata) and mesopelagic dolphins (Table 4.1). Many ecosystems of northern Europe have been modeled over the past years (Harvey et al. 2003; Booth and Zeller 2005; Morissette and Pitcher 2005; Skaret and Pitcher, in press). I chose to use the model of the North Sea ecosystem (Christensen et al., 2002) as it shows a better representation of marine mammals in the foodweb, and is also recent. This model was built to represent the period from 1963 to 1999, as part of a comparative analysis with MSVPA. Two marine mammals groups are present in the model, representing cetacean and seal species (Table 4.1). In the Gulf of Thailand, an Ecopath model was developed by Thai scientists and Dr. Villy Christensen at an FAO workshop for the implementation of a Code of Conduct for Responsible Fisheries (FISHCODE). This model was used to address the impact on the Gulf of Thailand ecosystem and on biodiversity of a continuation of the very high levels of mostly indiscriminate fishing effort, especially bottom trawling (FAO 2001). Marine 89 mammals are present in the model, but they were aggregated in only one trophic group (Table 4.1). Finally, the Strait of Georgia Ecopath model was developed to simulate various management policies and analyze their consequences (Martell et al. 2002). The model builds on another one created for the 'Back to the Future' project (Pauly et al. 1998b). It has four marine mammal groups out of a total of 27 trophic groups. The marine mammal species represented are transient killer whales (Orcinus orca), dolphins & resident killer whales, seals, and sea lions (Table 4.1). Table 4.1. Ecopath models used for analyses of marine mammals consumption. Ecosystems No. of trophic groups Marine mammals groups Average pedigree of marine mammals' diets Reference ( Eastern Bering Sea 26 1) Baleen whales 2) Toothed whales 3) Sperm whales 4) Beaked whales 5) Walrus and bearded seals 6) Other seals 7) Steller sea lions 0.7 National Research Council, 2003 Northwest Atlantic (Gulf of St. Lawrence) 32 1) Cetaceans 2) Harp seals 3) Hooded seals 4) Grey seals 5) Harbour seals 0.6 Morissette et al. 2003 Benguela system 32 1) Seals 2) Cetaceans 0.7 Shannon et al. 2004 Eastern tropical Pacific Ocean 39 1) Baleen whales 2) Toothed whales 3) Spotted dolphins 4) Meso. Dolphins 0.8 Olson and Watters 2003 North Sea 32 1) Seals 0.7 Christensen et al. 2002 Gulf of Thailand 40 1) Marine mammals 0.5 FAO/FISHCODE 2001 Strait of Georgia 27 1) Transient orcas 2) Dolphins & Resident orcas 3) Seals 4) Sea lions 0.6 Martell et al. 2002 Since most models represent the major hotspots of conflict between fishery and marine mammals, and because they cover very well both northern and the southern 90 hemispheres (Figure 4.1), I argue that it is reasonable to extrapolate these results to the whole world. 4.2.2 Modelling approach Our modelling approach used the Ecopath with Ecosim software package (Christensen and Pauly, 1992). In Ecopath, several system's indices are computed to describe the food web, its complexity, and the way trophic groups interact with one another. Ecopath models used here will first serve as comparison points for the global resource overlap analysis done by Kaschner and Pauly (2005). Based on different modelling techniques and offering an in-depth analysis of the entire foodweb in some of the 'hotspots' that were identified, this approach gives a closer look at the trophic interactions (especially consumption by marine mammals and catch by fisheries) in the zones where there is potential competition for marine resources. The software also allows making dynamic simulations based on Ecosim, a dynamic modelling application for exploring past and future impacts of fishing and environmental disturbances (Christensen and Walters 2004). Ecosim converts the trophic flows of Ecopath into dynamic, time-dependent predictions (full details of the EwE modelling approach and equations are available from http://www.ecopath.org). Each model has different complexity levels (in terms of number of trophic groups). However, all are rather complex, with at least more than 25 compartments (see Chapter 3 of that thesis for comparative information). Obviously, all models that were selected include one or more marine mammal groups. 4.2.3 Static ecosystem indices {Ecopath outputs) For each model, a comparison of the Ecopath outputs for food consumption by marine mammals versus the catch by fisheries was performed. The estimated annual catch (i.e., "food consumption") of fisheries and marine mammals calculated by each ecosystem modeled is presented, as well as the estimated composition of the catch and the diet, 91 respectively, of fisheries and marine mammals. Finally, the primary production required (PPR) to sustain fisheries is compared with the PPR to sustain marine mammals groups. The mean trophic level of marine mammals' consumption (TLQ) and of fisheries catch (TLc) were derived from Ecopath outputs. The TLc is a dimensionless index computed by Ecopath to measure ecosystem health (Christensen et al. 2005), and is an indicator of the ecosystem health and the state of the fisheries (Pauly and Watson 2005). The TLc is based on Lindeman's (1942) concept of trophic levels, but can be fractional (e.g., 1.3, 2.7, etc.) as suggested by Odum and Heald (1975). A routine assigns definitional trophic levels (TL) of one to producers and detritus. For the remaining species of the ecosystem, TL is calculated as: f \ TL = \ + Eq. 4.1 where pj is the proportion (in weight) of each prey group in the diet of its predator. TL, is the trophic level of each prey group j. The mean trophic level of the catch is calculated from: Eq. 4.2 where Y, is the total landings of species i (in tons), SY is the sum of landings for all species, and TL, is the trophic level for species i. Similarly, I developed an equation to calculate the trophic level of consumption (TLQ) by marine mammals using an equation derived from the TLc, and modified to represent the consumption by marine mammals and make it comparable with the fisheries catch: TL: * M n V ; - i J) Eq. 4.3 where Q,y is the consumption of prey i (in tons) by marine mammal j, Oj is the total consumption of all species by marine mammal j, and TL, is the trophic level for species /. Eq. 4.3 represents the average trophic level on which marine mammals feed, i.e., the average TL of each species, multiplied by their proportion in the consumption matrix (tonnes per km"2 per year that marine mammals consume). 92 Patterns of marine mammal consumption versus fisheries catches were based on categories that were first described in Pauly et al. (1998c). Table 4.2 describes the prey categories used to apportion dietary information for marine mammals. Table 4.2. Eight prey categories used to apportion dietary information for marine mammals (adapted from Pauly etal. 1998c). Group Description BI Benthic invertebrates. Mainly molluscs (notably bivalves and gastropods, but also including octopus, echinoderms and crustaceans. LZ Large zooplankton. Mainly small crustaceans, especially euphausiids (krill) such as Euphausia superba in Antarctic waters. SS Small squids. Families with mantle lengths of up to 50 cm, such as Gonatidae, (see Roper et al., 1984). LS Large squids. Families with mantle lengths above 50 cm, such as Onychoteuthidae. SP Small pelagic fishes. Consisting of clupeoids, small scombroids and allied groups. MP Mesopelagic fishes. Predominantly fish of the family Myctophidae and other groups of the Deep Scattering Layer. MF Miscellaneous fishes. A diverse group consisting mainly of demersal round fish such as gadoids and perciforms, but also including anadromous fishes such as salmon. HV High vertebrates. Marine mammals, seabirds, plus the occasional turtle. NM Non-marine mammal food. Includes all species not taken by marine mammals, e.g., fish larger than 150 cm. For the purposes of my analysis, the miscellaneous fish group is particularly broad and imprecise, as it was originally used for more global analyses. For the present study, when addressing the overlap between marine mammals consumption and fisheries catch for this group, the particular species components will be discussed in more details for each ecosystem. Using Ecopath estimates of fisheries catches and marine mammals consumption, the assessment of overlap between marine mammal and fisheries for each ecosystem was performed using a modified version of the resource overlap index of Kaschner (2004). This index originally included a weighting factor to provide a measure of the importance of each cell (from a global model of all oceans) for either fisheries or marine mammals, based on the overall quantity of catch or food taken by either consumer in this cell. The resource overlap index presented here (a/>) uses a different weighting factor representing the relative importance of fisheries versus marine mammals as consumers within a given ecosystem: 93 a f m '^Pm.kPf.k k Z 2 2 Pm,k +Pf,k V * J *(pQm*pCf) Eq.4.4 where afm is the quantitative overlap between a fishery/and a marine mammal group m in each ecosystem, and the first term expresses the qualitative similarity in diet/catch composition between the marine mammal group m and fisheries / sharing the resource or food type k, with pmk and pjk representing the proportions of each food type in the diet of marine mammals or the catch by fisheries. This term is multiplied by the product of the proportion of total food consumption by marine mammals in the ecosystem Q and the total fisheries catches C in the ecosystem. Two versions of this index were calculated for the purpose of my study: firstly, the overlap index by food type, using the same categories than Kaschner (2004), will allow global comparisons. Secondly, the index calculated by trophic group will give a closer look at the overlap at the species level, giving the fact that species falling into each food categories may change from one ecosystem to another, and that fisheries and marine mammals may target different species from the same food type. The mixed trophic impact (TT) analysis of Ecopath was used to compare the 'with/without' impact of predation by marine mammals on the whole ecosystem. This analysis allows the estimation of the relative impact of a change in the biomass of one group on other components of the ecosystem, under the assumption that the diet composition remains constant (Ulanowicz and Puccia 1990). Beneficial predation was calculated as the percentage of the overall trophic impact by marine mammals that is positive for any prey group of this predator. 4.2,4 Dynamic simulations (Ecosim) The Ecosim model behaviour is based on a 'foraging arena' theory (Walters and Martell, 2004), which assumes that predator and prey behaviours cause partitioning of prey populations, which are either available or unavailable to predators. There is continuous change between these two stages for any given potential prey, whether it is hiding from predation in some refuge, or it is out to feed. This availability of prey to predators is called 'vulnerability' in Ecosim. Mackinson et al. (2003) demonstrated the 94 importance of setting the vulnerabilities to fit model predictions to time-series data, as Ecosim predictions are very sensitive to this parameter. Using default values for v has strong implications for assumptions about species abundance relative to their carrying capacity (V. Christensen, Fisheries Centre, UBC, personal communication). Basically, it assumes that each group can at most increase the predation mortality they impose on their prey with a factor of 2.0 (the default v value). A lower value implies a donor driven density-dependant interaction. On the other hand, a higher value involves a predator driven density-independent interaction, in which predation mortality is proportional to the product of prey and predator abundance (i.e., Lotka-Volterra). This implies a high flux rate for prey species in and out of vulnerable biomass pools (Ainsworth 2006). Instead of using default vulnerability settings across the predation matrix, some authors chose to use values scaled to trophic level (e.g., Pitcher and Cochrane 2002), representing the hypothesis that higher trophic level groups are less risk-averse than lower one. Another way to adjust v's is to fit them to time series, using an automated procedure in Ecosim. Here, I used a set of models that were already adjusted for vulnerabilities and fitted to historical data following Walters et al. (2005). Vulnerabilities were adjusted based on the specific ecology of each species or trophic groups (if their behaviour, niche, or diet make them more or less vulnerable to predators), rather than assumed to scale to trophic levels. Using credible models that can reproduce observed historical response to disturbances such as fishing allow us to gain confidence when analyzing the possible impact of removing marine mammals in the ecosystem. The first thing that was analyzed based on Ecosim runs was the overall behaviour of trophic groups driven by original time series of fishing effort. Increases or decreases in abundance were noted for the major impacted groups. Some Japanese studies asserts that fisheries would globally benefit from a major marine mammals cull (Anonymous, 2001). In order to quantify the potential impact of marine mammal predation on the ecosystem, and to examine if there really is strong competition with the fisheries, Ecosim simulations were done over 22 to 89 years time periods, depending on the model's available time series. A first simulation was done with the original ecosystem structure (and original catch of fish and marine mammals), another was performed with a very high catch of marine mammals species, with the purpose of 95 driving down their biomass close to zero. Both scenarios were analyzed to see which groups and/or fisheries would suffer or benefit (in terms of increase or decrease in biomass) from the extirpation of marine mammals in the ecosystem. A hunting pattern was chosen which generated an important increase in the mortality on the marine mammal mortality, in order to drive their populations close to extinction. Vasconcellos et al. (1997) showed that for fish species, a 5-fold increase leads to serious depletion in a group. Also, such an extreme scenario is routinely applied to many fish populations and often associated with stock collapse (Patterson, 1992). This has never been done for marine mammals, but my tests have shown that the same kind of increase in anthropogenic mortality is needed to simulate a crash in marine mammal biomass. However, because some marine mammal species are not targeted at all by any fishery, I could not only multiply fishing mortality (F) by five. Thus, an F of 1.0 (representing an average five-fold increase for marine mammal species that were already hunted) was applied to each marine mammal groups in the models. Since each model was fitted to different time series, covering different periods, I did not perform a 100 years simulation as it was done by Vasconcellos et al. (1997). Here, the higher values of F were kept constant for the first 20% of the time series, and then returned to the baseline, with the model running for the remaining 80% of the time. Results 4.3.1 Total annual food consumption of marine mammals and fisheries catches The global pattern of marine mammal consumption versus fisheries catches is similar to what was found by Kaschner (2004), indicating that a high overlap occurs between marine mammals and fisheries in the predicted 'hotspots' of potential conflict. When all the models' results are combined (Figure 4.2), I find that the major source of competition between marine mammals and fisheries in the principal 'hotspots' is for 'miscellaneous fish', with attributes such as: demersal; benthic; benthopelagic; bathydemersal; reef-associated habitat & common length <150 cm; or pelagic habitat & common length >60 cm and <150 cm (see Pauly et al. [1998c] for detailed food types categories) and 'small pelagic' (FishBase attributes: pelagic habitat & common length <60 96 cm; see Pauly et al. [1998c] for detailed food types categories). Marine mammals consumption is diversified and represents a great array of marine organisms, while the catches are much more concentrated on 'miscellaneous fish', which account for about 75% of all the fish biomass taken. While marine mammals can consume different prey groups (mainly large zooplankton, cephalopods, small pelagic fish and macrobenthos), fisheries in the seven studied ecosystems combined are mainly focused on small crustaceans such as shrimp, pelagic fish (redeye, Etrumeus whiteheads, redfish, Sebastes spp.; anchovy, Eugraulis capensis; sprat, Sprattus sprattus) and demersal species such as hake species, lingcod (Ophiodon elongates), and sandeel {Ammodytes tobianus). Figure 4.2. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km"2) for all ecosystems combined. Note the greater proportion of miscellaneous fish taken by the fisheries. While a global comparison is useful, it is also important to compare marine mammals and fisheries on an ecosystem-by-ecosystem basis. In the Eastern Bering Sea system (Table 4.3), all fisheries catches fell into three types: miscellaneous fishes; mesopelagic fishes and higher vertebrates. In contrast, these food types accounted for less than a third of marine mammal consumption, which was more diverse and principally composed of benthic invertebrates, large zooplankton and miscellaneous fish. 97 Marine mammals Fisheries Miscellanous fishes Figure 4.3. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km"2) in the Eastern Bering Sea ecosystem. In the Northern Gulf of St. Lawrence (Figure 4.4), miscellaneous fish are the main target, accounting for 32% and 74% for marine mammal consumption and fisheries, respectively. However, the remaining marine mammal consumption is shared between three important groups (small pelagics, benthic invertebrates, and large zooplankton), while the fishery mainly catch miscellaneous fish (cod, redfish, and large Greenland halibut), benthic invertebrates (shrimp, crab, and molluscs) and small pelagics (herring). Marine mammal catch (mainly seal hunt) also occurs in the Gulf of St. Lawrence, accounting for about 1% of the total catch (high vertebrates). Figure 4.4. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km"2) in the northern Gulf of St. Lawrence ecosystem. 98 In the Benguela system (Figure 4.5), more than 95% of all fisheries fell into three types: miscellaneous; mesopelagic and small pelagic fishes. Similarly, these food types were also the most important for marine mammals of this ecosystem, whose diets also included an important part of small squids. Marine mammals , Fisheries Benthic Figure 4.5. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km"2) in the Benguela ecosystem. The Eastern tropical Pacific model (Figure 4.6) shows that most resources taken by fisheries are composed of only two food types: miscellaneous fish and non-mammal food. On the other hand, marine mammals feed on a variety of food types in this ecosystem, mainly small squids, mesopelagic, and small pelagic fish. 99 Figure 4.6. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km'2) in the Eastern Tropical Pacific ecosystem. The North Sea model (Figure 4.7) shows that about 75% of resources taken by- marine mammals or by fisheries is composed of miscellaneous fish. However, the difference between marine mammals and fisheries is in the kind of miscellaneous fish they exploit. The main fishes eaten by marine mammals are dab and cod, while fisheries mostly target Norway pout {Trisopterus esmarkii), sprat and sandeel. In this model, both marine mammals and fisheries catch fell in three categories: miscellaneous fish and small pelagic for both, as well as benthic invertebrates for marine mammals and mesopelagic fish for fisheries. Marine mammals Fisheries Small pelagic fishes Figure 4.7. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km'2) in the North Sea ecosystem. 100 In the Gulf of Thailand (Figure 4.8), marine mammals feed on a great variety of groups, while fisheries mainly catch benthic invertebrates and miscellaneous fish. These two food types represent more than 75% of the catch by fisheries and about a third of consumption by marine mammals. 'Trash fish' (bycatch catches that are used in the production of fishmeal) is one of the most important miscellaneous fish to be taken by fisheries and marine mammals, but then the two competing groups differ as marine mammals consume more small pelagic and benthos, and fisheries catch more shellfish and shrimp. Figure 4.8. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km-2) in the Gulf of Thailand ecosystem. In the Strait of Georgia ecosystem, almost all fish caught by the fisheries fell into one food group: miscellaneous fish (Figure 4.9). The same food type represents about 70% of the consumption by marine mammal, which is completed by mesopelagic and small pelagic fishes. Here again, fish taken by marine mammals are different than the one targeted by fisheries, even if they are all in the miscellaneous fish group: marine mammal consume more hake (Merluccius capensis) and demersal fish, while fisheries mostly catch herring (Clupea harengus), chinook salmon (Oncorhynchus tshawytscha) and lingcod. 101 Marine mammals Benthic — r . — j _ invertebrates High vertebrates Fisheries High vertebrates Small pelagic fishes Miscellanous fishes Miscellanous fishes Figure 4.9. Estimated mean annual catch and food consumption by food types expressed as proportions of total amounts taken (t-km"2) in the Strait of Georgia ecosystem. Thus, overall, I find that the degree of overlap depends largely on the resolution of the marine mammals prey and fisheries catches. Instances of direct overlap at the species level are observed in the Eastern Bering Sea (mainly for small flatfish, large flatfish, adult pollock [Pollachius virens], and other demersal fish), Gulf of St. Lawrence (planktivorous small pelagics, piscivorous small pelagics, shrimps and large crustaceans), Benguela (anchovy, redeye, and sardine [Sardinops sagax]), Eastern Tropical Pacific (small yellowfin tuna [Thunnus albacares], skipjack [Katsuwonus pelamis], and Auxis sp.), Gulf of Thailand ("trashfish", Rastrelliger spp., cephalopods, small demersals, and small pelagics), and Strait of Georgia (resident coho salmon [Oncorhynchus kisutch], resident chinook salmon, and lingcod). However, even at the species level, few overlap occurs between marine mammals prey and fisheries catches in the North Sea. 4.3.2 Mean trophic level of marine mammals' consumption and fisheries catches In most ecosystems, marine mammals feed on lower trophic level species (Table 4.3). Except for the northern Gulf of St. Lawrence and the Strait of Georgia, T L Q has lower values than TLc. The largest discrepancy between these two values is observed in the Eastern Tropical Pacific, where T L C is about 25% larger than the T L Q (4.70 versus 3.76). 102 Table 4.3. Mean trophic level of marine mammals' consumption and fisheries catches. Ecosystem model T L q T L C (Marine mammals) (Fisheries catch) Eastern Bering Sea 2.83 3.42 Gulf of St. Lawrence 3.24 3.71 Benguela 3.65 3.73 Eastern Tropical Pacific 3.76 4.70 North Sea 3.25 3.44 Gulf of Thailand 2.08 2.46 Strait of Georgia 3.36 3.25 Mean 2.88 3.42 4.3.3 Primary production required to sustain marine mammals' consumption and fisheries catch The primary production required (PPR) to sustain marine mammal consumption is always lower than PPR to sustain the fisheries (Table 4.4). Globally, PPR for fisheries is 50% higher than PPR for marine mammals. In terms of percentage, the PPR for fisheries is twice as high as PPR for marine mammals' consumption (20% versus 10%). Marine mammals of the Benguela system have the lowest PPR (they require only 2.2% of total primary production of the system), compared to marine mammals in the Eastern Bering Sea, where the PPR is 31.8% of total primary production. For fisheries, the lowest value is seen in the Benguela system (3.2%), while the highest PPR for fisheries catch is in the Eastern Bering sea (53.9%), closely followed by the North Sea (50.1%). Table 4.4. Primary production required to sustain marine mammals' consumption and fisheries catches. Numbers in parentheses are PPR in percentage of total primary production in the system. Ecosystem models PPR PPR PPR PPR Marine Marine Fisheries Fisheries mammals' Q mammals' Q catch catch (t'km'̂ year1) (% of total PP) (t'km -̂year"1) (% of total PP) Eastern Bering Sea 949.48 31.8 1607.97 53.9 Gulf of St. Lawrence 296.61 9.8 559.11 18.4 Benguela 510.68 2.2 751.98 3.2 Eastern Tropical Pacific 121.62 14.1 54.19 6.3 North Sea 212.69 3.0 3512.05 50.1 Gulf of Thailand 191.20 1.8 245.63 2.3 Strait of Georgia 1339.65 6.0 1488.29 6.7 Mean 517.42 9.7 784.53 20.1 103 4.3.4 Resource overlap between marine mammals and fisheries When marine mammals are considered as a whole group, their resource overlap with the fisheries varies a lot within the seven studied ecosystems (Table 4.5). The calculated resource overlap index, a, indeed varied from low (0.005 in the Eastern Tropical Pacific Ocean) to high (0.890 in the North Sea) values. Ecosystems with higher resources overlap seem to be the ones with lower diversity of food groups caught/eaten by fisheries/marine mammals. Models with a very high proportion of miscellaneous fish tend to have a higher resource overlap than systems where other food types are more important. When analysed per trophic group instead than per food type, the overlap is always lower. Highest overlap value is seen in the Benguela system, while lowest overlap happens in the Eastern Tropical Pacific. Interestingly, the North Sea ecosystem, which had the highest overlap per food type, has the third lowest value when overlap is calculated by trophic group. The number of trophic links in the ecosystem also has an effect on the resource overlap between marine mammal and fisheries. Indeed, food webs with lower connectance (less trophic links) tend to have higher overlap values (Table 4.5). Table 4.5. Estimated resource overlap index between marine mammals and fisheries and connectance of the related ecosystem models. Ecosystem model per food type per trophic group Connectance Eastern Bering Sea 0.031 0.006 0.274 Gulf of St. Lawrence 0.161 0.034 0.298 Benguela 0.714 0.120 0.231 Eastern Tropical Pacific 0.005 0.0003 0.218 North Sea 0.890 0.020 0.219 Gulf of Thailand 0.468 0.100 0.139 Strait of Georgia 0.163 0.024 0.250 GLOBAL (average) 0.149 0.043 - 4.3.5 Marine mammals and their impact on the trophic structure The effect of a change in marine mammals' biomass can be analyzed in two ways in EwE. First, from the Ecopath model, the mixed trophic impact shows that an increase in marine mammals' biomass would impact negatively or positively the biomass of other 104 groups in the foodweb. Second, Ecosim analyses allow simulating the effect of the removal of marine mammals populations on the rest of the foodweb over time. 4.3.5.1 Eastern Bering Sea The mixed trophic impact of marine mammals and fishery in the Eastern Bering Sea model shows that both marine mammals and fishery have an overall negative impact on the entire ecosystem (MM = -2.98; fishery = -3.04). The groups that are mostly impacted by marine mammal consumption are deepwater fish, large flatfish and other demersal fish. Conversely, small flatfish, deep pelagics and flatfish trawl seem to benefit from the presence of marine mammals in the ecosystem (Figure 4.10). The same 77 analysis for the fishery shows that the main negatively impacted groups of the Eastern Bering Sea ecosystem are most marine mammals, as well as flatfish (large and small). Deepwater fish, juvenile pollock and beaked whales (Mesoplodon stejnegeri) seem to benefit from the fishery (Figure 4.10). 0.4 u m a. E a. a IT "8 -0.2 \ -0.4 -0.6 I I 1 I * * i <* 1 1 1 2 3 I 1 8- 1 l a .1 I I I i Impacted groups i 3 i I i s j =5 w .3 a: « 5 Figure 4.10. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the Eastern Bering Sea. 105 In Ecosim, the Eastern Bering Sea model was first analyzed with time series of fishing mortality (F) to see which groups' biomass decline or increase over time. The model covers four sources of fishing mortality (pollock trawl, cod trawl, flatfish trawl, herring roe, deepwater pelagics, and whaling & sea lion cull), which were combined to see the overall effect on the whole ecosystem. Time series cover a period of 51 years, from 1950 to 2000. At the end of the simulation with the "true" fishing effort, some groups stabilized at higher biomass (deep pelagics, benthic pelagic feeders, small flatfish, large zooplankton, phytoplankton and cephalopods). Other groups stabilized after a decrease of biomass (deep-water fish, juvenile and adult pollock, shallow pelagics, sperm whales [Physeter macrocephalus], baleen whales, Steller sea lions, and epifauna). However, the biomasses of jellyfish and large flatfish exploded, while other demersal fish and piscivorous birds suffered from a substantial decline, or near extinction, respectively (Figure 4.11). After simulating marine mammal extirpation in the Bering Sea ecosystem (Figure 4.12), there was an increase in biomass (compared to the simulation without marine mammal extirpation) in the other demersal fish, deep pelagics, deepwater fish, jellyfish and cephalopods. All other groups showed a decrease in biomass if marine mammals were removed from the ecosystem (Figure 4.12). Over a period of 51 years, there is an overall decrease of 6% of total biomass if marine mammals are eradicated (Btot = 316 t'km" with marine mammals, and 298 t-km"2 without them; Table 4.6). This represents of course the complete extirpation of marine mammals biomass itself, but also seabirds, and other fish species such as small flatfish, as well as critical biomass decrease for shallow pelagics (- 99%), or epifauna (-97%). 106 • Baleen whales • Toothed whales Sperm whales • Beaked whales • Walrus&Bearded seals •Seals • Steller sea Hons - Adult poIlock2+ Juv. poUockO-1 Small Flatfish Shallow Pelagics Deep Pelagics Deepwater fish Cephalopods Benth.P.Feeders Epifauna Large Zoops Phytoplankton Years Figure 4.11. Ecosim results for the Eastern Bering Sea model with original fishing effort. Baleen whales Toothed whales Sperm whales Beaked whales Walrus&Bearded seals Seals SteDer sea Hons Adult polk>ck2+ Juv. poUockO-1 Large Flatfish Small Flatfish Shallow Pelagics Deep Pelagics Deepwater fish Cephalopods Benth.P.Feeders Epifauna Large Zoops Phytoplankton Years Figure 4.12. Ecosim results for the Eastern Bering Sea model with extirpation of marine mammals. o oo The main species targeted by the Eastern Bering Sea fisheries, all fleets combined, are adult pollock and shallow pelagics. When marine mammals are absent of the system, Ecosim predicts that there are less fish to catch for both fisheries than if marine mammals were present in the system (a decrease of 16% for adult pollock, and of 99% for shallow pelagics) (Figure 4.13). Out of eight fisheries (excluding whaling and sea lion cull), five suffer from a decrease in the biomass of their target species if there is no marine mammals in the ecosystem (Table 4.6). 90 -i n 6 1 a r-i r n ag ic s ag ic s rf is h yf is h § el l 1 Species N 1 X i Figure 4.13. Biomass change after a 51-year simulation in the Eastern Bering Sea ecosystem, with (white) and without (grey) marine mammals. 109 Table 4.6. Biomass (t-km"2-year"') change from the start (S) to the end (E) of the 51-year Ecosim simulation in the Eastern Bering Sea ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species. With M M Without M M B change after M M Biomass Biomass Biomass Biomass Biomass Biomass extirpation Group name (Start) (End) (E/S) (Start) (End) (E/S) (%) Baleen whales 0.694 0.432 0.620 0.547 0.000 0.000 -100.0 Toothed whales 0.009 0.009 1.020 0.007 0.000 0.000 -100.0 Sperm whales 0.438 0.259 0.590 0.345 0.000 0.000 -100.0 Beaked whales 0.001 0.001 1.170 0.001 0.000 0.000 -100.0 Walrus & Bearded seals 0.054 0.075 1.400 0.043 0.000 0.000 -100.0 Seals 0.106 0.112 1.050 0.084 0.000 0.010 -100.0 Steller sea lions 0.029 0.028 0.950 0.023 0.000 0.000 -100.0 Piscivorous birds 0.006 0.005 0.850 0.006 0.000 0.000 -100.0 Adult pollock 2+ 5.063 1.985 0.390 5.494 1.658 0.300 -16.5 Juvenile pollock 0-1 0.901 0.349 0.390 0.903 0.185 0.200 -47.0 O. Demersal Fish 8.918 9.851 1.100 8.956 7.503 0.840 -23.8 Large Flatfish 1.155 0.473 0.410 1.158 0.146 0.130 -69.1 Small Flatfish 8.519 6.954 0.820 8.519 0.000 0.000 -100.0 Shallow Pelagics 20.818 20.142 0.970 20.914 0.216 0.010 -98.9 Deep Pelagics 7.814 8.992 1.150 7.811 30.506 3.910 239.3 Deepwater fish 1.010 2.175 2.150 1.016 12.198 12.010 460.8 Jellyfish 0.047 2.428 51.280 0.047 0.171 3.610 -93.0 Cephalopods 3.406 3.945 1.160 3.485 2.478 0.710 -37.2 Benth.Pel. feeders 28.758 23.957 0.830 28.772 56.987 1.980 137.9 Infauna 75.058 82.789 1.100 75.056 49.053 0.650 -40.7 Epifauna 7.995 14.074 1.760 8.003 0.418 0.050 -97.0 Large Zooplankton 38.937 45.625 1.170 38.914 47.986 1.230 5.2 Herb. Zooplankton 48.909 55.863 1.140 48.944 51.094 1.040 -8.5 Phytoplankton 27.493 35.565 1.290 27.486 37.892 1.380 6.5 Discards 0.000 0.000 1.000 0.000 0.000 1.000 0.0 Detritus 0.000 0.000 1.040 0.000 0.000 1.340 0.0 Total 286.137 316.088 1.100 286.535 298.494 1.040 -5.6 4.3.5.2 Northern Gulf of St. Lawrence In the Northern Gulf of St. Lawrence, the mixed trophic impact of marine mammals and fishery shows that fisheries have an overall negative impact (-4.93) that is much higher than that of marine mammals (-2.93). The groups that are the most negatively impacted by marine mammal consumption are large demersals, large pelagics and Greenland halibut (Reinhardtius hippoglossoides; large and small). In contrast, skates, small demersals, shrimp and most benthic invertebrates seem to benefit from marine mammals in the ecosystem (Figure 4.14). 110 All marine mammals and seabirds, large cod, shrimp, small demersals and most benthic invertebrates are negatively impacted by fisheries in the Gulf of St. Lawrence. Small Greenland halibut, large demersals, and small cod seem to benefit from fisheries (Figure 4.14). 0.4 ] Impacted groups Figure 4.14. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the Gulf of St. Lawrence. When analyzed with time series of fishing mortality in Ecosim, the Gulf of St. Lawrence model was analyzed to see which groups' biomass decline or increase over time. The model covers one source of combined fishing mortality that represents all fishing fleets operating in the system. Time series cover a period of 42 years, from 1960 to 2002, but only the last 22 years are documented in terms of fish biomass (the first 20 years are time series of marine mammal biomasses and catches). At the end of the simulation with original fishing effort, some groups show a relatively constant increase in biomass (Greenland halibut, harbour seals, grey seals, and hooded seals), while most groups stabilized after a slight increase of biomass. No groups show severe depletion (Figure 4.15). I l l 3.0 2.5 2.0 3 1-5 a c 3 1.0 0.5 0.0 0\ Os o o r- £v r> *\ C N °0 P> O N K K fN.rs.oo oo ON ON 0\ 0\ Os ON ON oo ec oo oo oo oo o ON o\ o\ a a » o\ ^ O N 0 \ O \ O \ O \ O N O \ 0 N C > 0 \ O \ 0 N O \ O \ 0 \ 0 \ Years Cetacea Harp seals Hooded seals Grey seals Harbour seals Seabirds Juvenile cod Adult cod Juvenile Greenland halibut Adult Greenland halibut American plaice Flounders Redfish L. demersals S. demersals Capelin Sand lance Arctic cod L. pelagics S. pisciv. pelagics S. plankt. pelagics Large Crustacea Figure 4.15. Ecosim results for the Gulf of St. Lawrence model with original fishing effort. Years • Cetacea •Harp seals • Hooded seals • Grey seals Harbour seals Seabirds • Juvenile cod Adult cod Juvenile Greenland halibut Adult Greenland halibut American plaice Flounders Redfish - L. demersals S. demersals Capelin • Sand lance Arctic cod L. pelagics S. pisciv. pelagics - S. plankt. pelagics Large Crustacea Figure 4.16. Ecosim results for the Gulf of St. Lawrence model after the extirpation of marine mammals. When all seal and cetacean species are removed from the ecosystem, its structure changes (Figure 4.16). There is an explosion in the biomass of Greenland halibut (adult and juveniles), and an increase in large pelagic and demersal groups. The remaining groups of that simulated system without marine mammals stabilized around a similar equilibrium, and there is no significant change in total biomass (439 t*km") over the 43- year simulation (Table 4.7). Most groups which originally had lower biomasses in the Gulf of St. Lawrence system appeared to increase their biomass after the extirpation of marine mammals. However, for groups that were important in the system (e.g., adult cod, capelin, and small planktivorous pelagics), the increase was limited (Table 4.7). The most important fishery in the Gulf of St. Lawrence in terms of landings in the 1980s was cod, redfish, small planktivorous pelagics (herring) and shrimp. In a scenario without marine mammals, most of these target species show no significant change, or decrease slightly as in the case of shrimp (Figure 4.17). 114 Table 4.7. Biomass (t-km'2-year'') change from the start (S) to the end (E) of the 43-year Ecosim simulation in the Gulf of St. Lawrence ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species. With M M Without M M B change after M M Biomass Biomass Biomass Biomass Biomass Biomass extirpation Group name (Start) (End) (E/S) (Start) (End) (E/S) (%) Cetacea 0.104 0.107 1.030 0.082 0.000 0.000 -100.0 Harp seals 0.048 0.051 1.060 0.025 0.002 0.070 -96.1 Hooded seals 0.003 0.004 1.100 0.003 0.000 0.000 -100.0 Grey seals 0.007 0.008 1.060 0.006 0.000 0.000 -100.0 Harbour seals 0.002 0.002 1.150 0.001 0.000 0.000 -100.0 Seabirds 0.002 0.002 1.010 0.002 0.003 1.050 50.0 Juvenile cod 0.805 0.911 1.130 0.805 0.904 1.120 -0.8 Adult cod 1.813 1.986 1.100 1.813 2.027 1.120 2.1 Juv. Greenland halibut 0.144 0.209 1.450 0.144 0.385 2.680 84.2 Ad. Greenland halibut 0.280 0.350 1.250 0.280 0.615 2.190 75.7 American plaice 0.773 0.790 1.020 0.773 0.806 1.040 2.0 Flounders 0.498 0.527 1.060 0.498 0.577 1.160 9.5 Skates 0.190 0.204 1.080 0.190 0.214 1.130 4.9 Redfish 7.840 8.045 1.030 7.840 8.046 1.030 0.0 L . demersals 0.746 0.790 1.060 0.746 0.930 1.250 17.7 S. demersals 1.508 1.503 1.000 1.509 1.480 0.980 -1.5 Capelin 11.765 11.741 1.000 11.768 11.782 1.000 0.3 Sand lance 2.230 2.229 1.000 2.230 2.238 1.000 0.4 Arctic cod 0.036 0.036 1.000 0.036 0.036 1.020 0.0 L. pelagics 0.051 0.054 1.060 0.051 0.075 1.480 38.9 S. pisciv. pelagics 2.213 2.217 1.000 2.214 2.319 1.050 4.6 S. plankt. pelagics 1.651 1.688 1.020 1.651 1.696 1.030 0.5 Shrimp 0.919 0.910 0.990 0.919 0.899 0.980 -1.2 Large Crustacea 0.951 0.970 1.020 0.951 0.968 1.020 -0.2 Echinoderms 104.490 104.507 1.000 104.490 104.503 1.000 0.0 Molluscs 57.757 57.770 1.000 57.757 57.769 1.000 0.0 Polychaetes 13.380 13.379 1.000 13.380 13.375 1.000 0.0 0. benthic invertebrates 6.750 6.750 1.000 6.750 6.749 1.000 0.0 Large zooplankton 12.850 12.849 1.000 12.850 12.847 1.000 0.0 Small zooplankton 61.069 61.072 1.000 61.069 61.072 1.000 0.0 Phytoplankton 14.850 14.850 1.000 14.850 14.850 1.000 0.0 Detritus 132.609 132.636 1.000 132.608 132.638 1.000 0.0 Total 438.334 439.146 1.000 438.291 439.805 1.000 0.2 115 16 l,r-n r-n ,m ,m, m 1 < S 5 1 1 1 "B 1 1 5 < 12 1 0! 1 1 0 ™ ™ n a > 1 I Species Figure 4.17. Biomass change after a 43-year simulation in the Gulf of St. Lawrence ecosystem, with (white) and without (grey) marine mammals (echinoderms, polychaetes and small zooplankton (not shown) have very high and nearly identical biomasses with and with out marine mammals in the system, i.e., 105, 58, and 51 t-km"2, respectively). 4.3.5.3 Benguela system The mixed trophic impact of marine mammals and fishery in the Benguela model shows that both marine mammals and fishery have an overall negative impact on the entire ecosystem (-0.011 and -0.105, respectively). However, fisheries' negative impact on the groups' biomass in the Benguela ecosystem is larger by an order of magnitude. Fish groups that are the most negatively impacted by marine mammal consumption are cape hake (Merluccius capensis), horse mackerel (Trachurus capensis) and cephalopods. Conversely, apex chondrichthyans, mesopelagics and redeye seem to benefit from marine mammals presence in the ecosystem. The fleet "other fisheries" would also benefit from an increase in marine mammal biomass in terms of mixed trophic impacts (Figure 4.18). The same analysis of mixed trophic impacts for the fishery shows that the main negatively impacted groups in the Benguela ecosystem are snoek (Thyrsites atun), sardine, 116 other large pelagics, shallow-water cape hake, and deep-water cape hake (Merluccius paradoxus) (Figure 4.18). 0.4 4 is 7 .6 -J •3 » *2 f i l l 1 el i ° -8 ? I --g«-g 3 3 O Impacted groups Figure 4.18. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the Benguela system. With Ecosim, the Benguela model was first analyzed with fishing mortality to check which groups decline or increase. The total 15 fishing fleets (purse seine, midwater trawl, demersal trawl, line fishery, long line, inshore shallow Cape hake capensis trawl, squid jig, South offshore Cape hake capensis trawl, West offshore trawl, squid demersal trawl, midwater horse mackerel trawl, demersal horse mackerel trawl, snoek trawls, South offshore Cape hake paradoxus trawl, and other) were combined to analyze their overall effect on the entire ecosystem. Time series cover a period of 25 years, from 1978 to 2002. At the end of the simulation with the original fishing effort, some groups stabilized at higher biomass (seals, benthic chondrichthyans, pelagic chondrichthyans, benthic demersals, large shallow-water Cape hake, macrobenthos, large deep-water Cape hake, cephalopods and apex chondrichthyans). Other groups slightly decreased in biomass (Chub mackerel [Scomber japonicus], pelagic demersals, juvenile horse mackerel, small Cape hake M. paradoxus, mesopelagics, macrozooplankton, and mesozooplankton). 117 Sardine, anchovy, other large pelagics, snoek, seabirds and cetaceans showed an important explosion in biomass (Figure 4.19). When marine mammals are removed from the ecosystem, the dynamics change (Figure 4.20). The removal of the cetaceans and seals groups seems to generate an explosion of sardine, as well as anchovy, snoek, other pelagics and seabirds. On the other end, mesopelagics, redeye and mesozooplankton groups decrease more than in the basic simulation. The remaining groups seem to stabilize around equilibrium after the 25-year simulation. There is no significant decrease in total biomass (258 versus 256 t-km"'), even if marine mammals are severely depleted. Other groups that showed important decrease in biomass are mesopelagics (-64%), other small pelagics (-27%), redeye (-20%), and pelagic demersals (-19%). An important increase in biomass is observed for snoek (47%), other large pelagics (28%), seabirds (27%), and large M. capensis (20%). No group except from marine mammals seems to be totally depleted (Table 4.8). 118 • Cetaceans • Seals - Mesozooplank. - Macrozooplank. Anchovy Sardine Redeye Juv.hmack. Ad.hmack. Mesopelagic Snoek Otherlargepel Cephalopods - SmaDM.capens LargeM.capens SmaBM.parad - LargeM.parad Pelagicdemers Benthicdemers - Pelagicchond - Benthic chond - Apexchond - Seabirds - Macrobenthos Years Figure 4.19. Ecosim results for the Benguela model with original fishing effort. Cetaceans Seals Mesozooplank. Macrozooplank. - Anchovy Sardine Redeye Juv.hmack. Ad.hmack. Mesopelagic Snoek Otherlargepel Cephalopods SmaHMxapens LargeMxapens SmaHM.parad LargeM.parad Pelagicdemers Benthkdemers Pelagicchond Benthicchond Apexchond Seabirds Macrobenthos Years Figure 4.20. Ecosim results for the Benguela model after the extirpation of marine mammals. to o Table 4.8. Biomass (t-km'2-year') change from the start (S) to the end (E) of the 25-year Ecosim simulation in the Benguela ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species. Ecosim with M M Ecosim without M M B change after M M Biomass Biomass Biomass Biomass Biomass Biomass extirpation Group name (Start) (End) (E/S) (Start) (End) (E/S) (%) Phytoplankton 61.330 84.115 1.370 61.330 83.973 1.370 -0.2 Benthic producers 6.606 6.111 0.920 6.606 6.119 0.930 0.1 Microzooplankton 1.296 1.830 1.410 1.296 1.891 1.460 3.3 Mesozooplankton 5.682 6.158 1.080 5.682 5.297 0.930 -14.0 Macrozooplankton 9.533 11.063 1.160 9.531 10.393 1.090 -6.1 Gelatinous zooplank. 4.504 5.444 1.210 4.504 4.687 1.040 -13.9 Anchovy 4.380 14.973 3.420 4.380 22.752 5.190 52.0 Sardine 0.788 8.021 10.170 0.788 7.964 10.100 -0.7 Redeye 5.429 4.904 0.900 5.436 3.922 0.720 -20.0 Other small pelagics 0.357 0.376 1.050 0.358 0.275 0.770 -26.9 Chub mackerel 0.280 0.285 1.020 0.281 0.253 0.900 -11.2 Juvenile h. mackerel 0.193 0.234 1.210 0.193 0.217 1.130 -7.3 Adult horse mackerel 1.611 1.688 1.050 1.618 1.645 1.020 -2.5 Mesopelagic 7.820 7.283 0.930 7.821 2.648 0.340 -63.6 Snoek 0.148 0.383 2.580 0.149 0.564 3.800 47.3 Other large pelagics 0.131 0.379 2.890 0.131 0.485 3.700 28.0 Cephalopods 1.344 1.640 1.220 1.354 1.652 1.220 0.7 Small M.capensis 0.420 0.448 1.070 0.422 0.463 1.100 3.3 Large M.capensis 0.822 1.083 1.320 0.823 1.301 1.580 20.1 Small M.paradoxus 1.231 1.230 1.000 1.233 1.070 0.870 -13.0 Large M.paradoxus 0.840 1.033 1.230 0.841 0.959 1.140 -7.2 Pelagic demerals 2.924 2.941 1.010 2.925 2.386 0.820 -18.9 Benthic demersals 3.494 4.579 1.310 3.496 4.497 1.290 -1.8 Pelagic chondrichth. 0.582 0.749 1.290 0.582 0.728 1.250 -2.8 Benthic chondrichth. 0.873 1.166 1.340 0.873 1.156 1320 -0.9 Apex chondrichth. 0.045 0.047 1.050 0.045 0.044 0.990 -6.4 Seals 0.133 0.168 1.270 0.084 0.005 0.060 -97.0 Cetaceans 0.074 0.128 1.740 0.047 0.006 0.130 -95.3 Seabirds 0.015 0.030 1.980 0.015 0.038 2.510 26.7 Meiobenthos 12.310 15.056 1.220 12.310 15.055 1.220 0.0 Macrobenthos 58.475 74.005 1.270 58.473 73.549 1.260 -0.6 Total 193.672 257.549 1.330 193.627 255.995 1.320 -0.6 The main targeted species in the Benguela ecosystem are anchovy, sardine, redeye (all caught by purse seine) and large deep-water Cape hake (caught by west offshore trawl). When marine mammals are removed from the ecosystem, most of these commercially important fish end up with less biomass than in the initial ecosystem, except for anchovy (Figure 4.21). Out of 15 different fisheries in the Benguela ecosystem, eight 121 underwent an important loss in the biomass of their target species after 25-year without marine mammals. Moreover, four of these were in the top-five fisheries with the most important catch at the beginning of the simulation. 25 1 Specii Figure 4.21. Biomass change after a 25-year simulation in the Benguela ecosystem, with (white) and without (grey) marine mammals (phytoplankton and macrobenthos [not shown] have very high and nearly identical biomasses with and with out marine mammals in the system, i.e., 84 and 74 t-km'2, respectively). 4.3.5.4 Eastern tropical Pacific Ocean In the Eastern tropical Ocean, the mixed trophic impact of marine mammals and fisheries shows that fishery (-1.869) and marine mammals (-2.334) have an overall negative impact on the entire ecosystem. Nevertheless, the negative impact of marine mammals on other groups of the foodweb is larger than the impact of fisheries. Fish groups that are the most negatively impacted by marine mammal consumption are small bigeye (Priacanihus arenatus), small wahoo (Acanthocybium solandri), albacore (Thunnus alalunga) and skipjack. On the other hand, large bigeye, large dorado 122 {Coryphaena hippurus), large wahoo, large sharks, rays and flying fish (Cypselurus naresii) all benefit benefit from the presence of marine mammals presence in the ecosystem, while they are negatively affected by fisheries. Some fisheries also benefit from the presence of marine mammals in the system: the horse mackerel fisheries (midwater and demersal fleets) and snoek trawls (Figure 4.22). The mixed trophic impact analysis also shows that groups that are the most positively impacted by the combined effect of all fisheries in the Eastern tropical Pacific Ocean are mainly the juveniles of important commercial fish (wahoo, dorado, swordfish [Siphias gladius], sailfish [Istiophorus platypterus], marlins, bigeye). When all the fisheries are put together, their combined effect on the individual fishing fleets is mostly negative (Figure 4.22). 0.8 -i Impacted groups Figure 4.22. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the Eastern tropical Pacific system. The Eastern tropical Pacific model was first analyzed in Ecosim with the current fishing mortality to see which groups decline or increase over time. A total of five fishing fleets (school sets, longliners, floating, dolphin sets, and bait boats) were combined to see their overall effect on the entire ecosystem. Time series cover a period of 89 years, from 123 1910 to 1998. At the end of the simulation with the original fishing effort, some groups showed large variation and increased a lot (large swordfish, large and small marlins), and other stabilized at higher biomass (rays, small swordfish, albacore, small sailfish, and small wahoo). Most remaining groups stabilized after a slight change in biomass, but large yellowfin tuna and large phytoplankton strongly declined (Figure 4.23). When marine mammals are removed from the ecosystem, its dynamics is altered (Figure 4.24). The absence of marine mammals species seems to generate large oscillations in the biomass of commercially important fish (mainly large bigeye, but also yellowfin tuna, marlins, swordfish, small sharks, and small bigeye). Over a period of 89 years, there is no significant change in total biomass, even if marine mammals are totally removed from the ecosystem. Marine mammals predation seems to have been replaced by increase in the biomass and hence predation of large fish such as wahoo (strongest increase in biomass; 248%), yellowfin tuna, bigeye, sharks, skipjack, albacore, bluefin tuna (Thunnus orientalis), and dorado (Table 4.9). 124 0.5 4 Baleen Whales Toothed Whales Spotted Dolphin - Meso Dolphin Lg Yellow fin Lg Bigeye Lg Marlins Lg Swordfish Lg Dorado Lg Wahoo Lg Sharks Rays Skipjack Albacore Bluefin Sm Yellow fin Sm Bigeye Sm Marlins Sm Sailfish Sm Swordfish Sm Dorado — Sm Wahoo Sm Sharks 0.0 J O N O \ C > O N O \ C > O N C > C > 0 \ C > C > C > C > O N O N C ^ C > C > C > C > C > C > 0 \ C ^ Ye are Figure 4.23. Ecosim results for the Eastern tropical Pacific model with original fishing effort. 4.0 3.5 3.0 « 2.5 JD Id 2.0 1.5 1.0 0.5 - Baleen Whales - Toothed Whales Spotted Dolphin Meso Dolphin Lg Yellowfin Lg Bigeye Lg Marlins Lg Swordfish - Lg Dorado Lg Wahoo Lg Sharks Rays Skipjack - Albacore - Bluefm - Sm Yellowfin Sm Bigeye - Sm Marlins - Sm Sailfish - Sm Swordfish Sm Dorado Sm Wahoo Sm Sharks \ 0.0 to °o o rv O? O? O? 0\ o? # ? £ 0\ 0\ 0\ O 'V- 'o Oo O »N lr> (rV (o ^ <0 « « Oi Oi a c> c> o\ Q\ Years Figure 4.24. Ecosim results for the Eastern tropical Pacific model after the extirpation of marine mammals. Table 4.9. Biomass (t-km -̂year'1) change from the start (S) to the end (E) of the 89-year Ecosim simulation in the Eastern Tropical Pacific ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species. With M M Without M M Group name Biomass (Start) Biomass (End) Biomass (E/S) Biomass (Start) Biomass (End) Biomass (E/S) aiier M M extirpation (%) Pursuit Birds 0.001 0.001 0.840 0.001 0.001 0.840 0.2 Grazing Birds 0.000 0.000 0.880 0.000 0.000 0.860 -2.9 Baleen Whales 0.009 0.009 1.000 0.007 0.000 0.000 -100.0 Toothed Whales 0.031 0.031 1.000 0.024 0.000 0.000 -100.0 Spotted Dolphin 0.004 0.003 0.870 0.003 0.000 0.000 -100.0 Meso Dolphin 0.017 0.015 0.870 0.013 0.000 0.000 -100.0 Sea Turtles 0.000 0.000 0.840 0.000 0.000 0.810 -3.6 Large Yellowfin 0.008 0.005 0.660 0.008 0.005 0.660 -0.4 Large Bigeye 0.010 0.008 0.860 0.010 0.016 1.650 93.0 Large Marlins 0.001 0.001 1.250 0.001 0.001 1.760 40.6 Large Sailfish 0.000 0.000 1.060 0.000 0.000 1.090 2.2 Large Swordfish 0.000 0.000 1.540 0.000 0.000 1.540 -0.3 Large Dorado 0.000 0.000 1.010 0.000 0.000 1.150 14.5 Large Wahoo 0.001 0.001 0.920 0.001 0.004 3.190 247.5 Large Sharks 0.000 0.000 0.900 0.000 0.000 1.100 22.3 Rays 0.000 0.000 1.300 0.000 0.000 1.310 0.7 Skipjack 0.027 0.025 0.900 0.027 0.030 1.100 23.3 Albacore 0.003 0.003 1.110 0.003 0.005 1.730 57.3 Auxis sp. 0.144 0.149 1.040 0.144 0.141 0.980 -5.0 Bluefin 0.001 0.001 1.050 0.001 0.002 1.490 41.7 Small Yellowfin 0.009 0.007 0.760 0.009 0.007 0.760 0.1 Small Bigeye 0.010 0.009 0.840 0.010 0.015 1.450 73.2 Small Marlins 0.000 0.000 1.070 0.000 0.000 1.530 43.3 Small Sailfish 0.000 0.000 0.960 0.000 0.000 0.990 3.1 Small Swordfish 0.000 0.000 1.210 0.000 0.000 1.220 0.9 Small Dorado 0.002 0.002 1.050 0.002 0.003 1.270 23.0 Small Wahoo 0.003 0.003 0.950 0.003 0.005 1.600 70.6 Small Sharks 0.000 0.000 0.900 0.000 0.000 1.090 21.7 Misc. Piscivores 0.017 0.018 1.080 0.017 0.019 1.110 3.7 Flying fish 0.162 0.162 1.000 0.162 0.158 0.980 -2.0 Misc. Epi. Fish 2.283 2.266 0.990 2.284 2.253 0.990 -0.6 Misc. Meso. Fish 2.013 2.020 1.000 2.015 2.003 0.990 -0.9 Cephalopods 1.105 1.170 1.060 1.109 1.229 1.110 5.1 Crabs 0.122 0.119 0.980 0.122 0.117 0.960 -1.8 Mesozooplankton 0.735 0.663 0.900 0.735 0.670 0.910 1.0 Microzooplankton 0.818 0.840 1.030 0.818 0.839 1.030 -0.2 Lg Phytoplankton 0.485 0.335 0.690 0.485 0.333 0.690 -0.6 Sm Producers 3.013 2.986 0.990 3.013 2.990 0.990 0.1 Detritus 2.022 1.977 0.980 2.022 1.979 0.980 0.1 Total 13.056 12.831 0.980 13.051 12.827 0.980 0.0 127 Figure 4.25 shows that the most important fish in terms of biomass (Auxis sp., flying fish, and miscellaneous epipelagic fish) would decline if marine mammals were removed from the ecosystem. However, commercially important species (especially bigeye, wahoo and skipjack), which tend to have lower biomasses, would benefit from the extirpation of marine mammals in the Eastern tropical Pacific system. After 89 years of simulation, the total biomass of the system is not changed, even if marine mammals are extirpated. Their original biomasses appear to be replaced by other trophic groups in the system. 0.16 e 0.12 "J» 0.08 9\ 00 u ^ 0.06 -I g 0.04 o £ 0.02 ] raD 1 1 I | 1 a $ 1 1 E O Species Figure 4.25. Biomass change after a 89-year simulation in the Eastern tropical Pacific ecosystem, with and without marine mammals. Note that the miscellaneous epipelagic fish, miscellaneous mesopelagic fish, cephalopods, mesozooplankton, microzooplankton, large phytoplankton, small producers and detritus groups were excluded from the graph due to their large biomass. No significant difference was observed with or without marine mammals for these groups (see Table 4.9). 4.3.5.5 Gulf of Thailand The mixed trophic impact analysis for shows that most groups in the Gulf of Thailand model are negatively impacted by fisheries. This is especially true for Scianidae, sharks, cephalopods, large piscivores, Saurida spp., and Lutjianidae. When all the 128 fisheries are analyzed as a whole, their combined effect on each and every fishing fleet is mostly negative on the different fleets in the ecosystem (Figure 4.26). In the Gulf of Thailand, the mixed trophic impact of the fisheries is overall negative (-7.735), and much higher than that of marine mammals (-1.899). Marine mammals have their larger negative impact on small pelagics, Carangidae and Rastrelliger spp., and on the purse seine fishery as well (Figure 4.26). 0.75 -, r, 0.55 - O ft Impacted groups Figure 4.26. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the Gulf of Thailand system. The Gulf of Thailand model was first analyzed, using Ecosim, with the time series of fishing mortality to see which groups decline or increase over time. The model covers six sources of fishing mortality: pair trawls; beam trawls; pushnets; purse seines; otter board trawls and other gears. All fisheries were combined to see their general effect on the entire ecosystem. Time series cover a period of 24 years, from 1973 to 1996. At the end of the simulation with the original fishing effort, some groups stabilized at higher biomass, such as false trevally (Lactarius lactarius), demersal piscivores, Sillago spp. and small demersals. Most of the remaining groups stabilized after a slight decrease of biomass. Nevertheless, Plectorhynchidae, shellfish, coastal tuna, pelagic species (large and small), 129 Scomberomus and Priacanthus spp. collapsed, while demersal benthivores' exploded after approximately 15 years of simulation (Figure 4.27). When all marine mammals of the Gulf of Thailand are extirpated, ecosystem structure changes (Figure 4.28). Their extirpation still leads to an explosion in demersal benthivore species, but also to large variation in the biomass of small demersals, small pelagics and Rastrelliger spp. The remaining groups seem to stabilize around equilibrium after the 24-year simulation. Over that period, there is no significant decrease in total biomass (91 versus 92 t-km"2), even if marine mammals are severely depleted (Table 4.10). 130 • Mammals - Rastrelliger spp. Scomberomorus Carangidae Small pelagics False rrevally L. piscivores Scianidae - Saurida spp. Plectorhynchidae Sfllago Sharks - Shrimps Small demersal - Dem.piscivore - Dem.benthivore Years Figure 4.27. Ecosim results for the Gulf of Thailand model with original fishing effort. — Mammals Rastrelliger spp. Scomberomorus Carangidae Small pelagics False trevally L. piscivores Scianidae Saurida spp. Plectorhynchidae Sillago Sharks Shrimps Small demersal Dem.piscivore Dem.benthivore Years Figure 4.28. Ecosim results for the Gulf of Thailand model after the extirpation of marine mammals. to Many groups that had small biomass in the Gulf of Thailand ecosystem increased considerably: large piscivores (750%); small pelagics (724%); coastal tuna (700%); Scomberomorus spp. (225%) and Sciaenidae (200%). However, except for small pelagics, these groups have small biomasses (all less than 0.5 t*km"2), and remain minor in the ecosystem. In addition to marine mammals, Plectorhynchidae show a total depletion in an ecosystem without marine mammals (Table 4.10). 133 Table 4.10. Biomass (t-km~2-year"') change from the start (S) to the end (E) of the 24-year Ecosim simulation in the Gulf of Thailand ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species. Ecosim with MM Ecosim without MM B change after MM Biomass Biomass Biomass Biomass Biomass Biomass extirpation Group name (Start) (End) (E/S) (Start) (End) (E/S) (%) Rastrelliger spp. 0.190 0.134 0.710 0.195 0.308 1.580 129.9 Scomberomorus spp. 0.015 0.004 0.290 0.015 0.013 0.870 225.0 Carangidae 0.082 0.054 0.660 0.082 0.120 1.460 122.2 Pomfrets 0.007 0.006 0.790 0.007 0.006 0.830 0.0 Small pelagics 0.455 0.107 0.240 0.466 0.882 1.890 724.3 False trevally 0.003 0.004 1.130 0.003 0.004 1.080 0.0 Large piscivores 0.053 0.002 0.030 0.054 0.017 0.320 750.0 Sciaenidae 0.031 0.001 0.040 0.031 0.003 0.090 200.0 Saurida spp. 0.054 0.000 0.000 0.054 0.000 0.000 - Lutianidae 0.016 0.000 0.010 0.016 0.003 0.210 - Plectorhynchidae 0.008 0.001 0.100 0.008 0.000 0.040 -100.0 Priacanthus spp. 0.071 0.032 0.450 0.071 0.037 0.520 15.6 Sillago spp. 0.033 0.071 2.160 0.033 0.055 1.660 -22.5 Nemipterus spp. 0.093 0.000 0.000 0.093 0.000 0.000 - Ariidae 0.018 0.011 0.600 0.018 0.010 0.580 -9.1 Rays 0.048 0.004 0.080 0.048 0.004 0.070 0.0 Sharks 0.020 0.000 0.020 0.020 0.003 0.150 - Cephalopods 0.400 0.309 0.770 0.403 0.505 1.250 63.4 Shrimps 0.181 0.123 0.680 0.181 0.110 0.610 -10.6 Crabs & lobsters 3.518 3.434 0.980 3.518 3.441 0.980 0.2 'Trash fish' 0.524 0.336 0.640 0.531 0.429 0.810 27.7 Small demersals 0.181 0.367 2.030 0.183 0.556 3.030 51.5 Demersal piscivores 0.058 0.072 1.250 0.058 0.070 1.210 -2.8 Dem. Benthivores 0.090 0.338 3.760 0.090 0.316 3.520 -6.5 Shellfish 0.170 0.025 0.150 0.170 0.025 0.150 0.0 Jellyfish 2.000 2.038 1.020 2.000 1.999 1.000 -1.9 Sea cucumbers 1.000 0.997 1.000 1.000 1.001 1.000 0.4 Seaweeds 1.000 1.000 1.000 1.000 1.000 1.000 0.0 Coastal tuna 0.021 0.003 0.140 0.021 0.024 1.150 700.0 Sergestid shrimps 0.058 0.049 0.850 0.058 0.034 0.590 -30.6 Mammals 0.100 0.087 0.870 0.078 0.007 0.090 -92.0 Pony fishes 0.054 0.085 1.570 0.054 0.071 1.310 -16.5 Benthos 33.001 33.074 1.000 33.008 33.193 1.010 0.4 Zooplankton 17.300 17.461 1.010 17.299 17.296 1.000 -0.9 Juvenile pelagic 0.076 0.026 0.340 0.076 0.090 1.190 246.2 Juvenile Caraiix 0.027 0.020 0.740 0.027 0.035 1.300 75.0 Juvenile Saurida 0.019 0.000 0.000 0.019 0.000 0.000 - Juvenile Nemipterus 0.023 0.000 0.000 0.023 0.000 0.000 - Phytoplankton 30.000 29.906 1.000 30.000 30.014 1.000 0.4 Detritus 9999.924 9984.550 1.000 9999.787 10002.780 1.000 0.2 Total 10090.920 10074.730 1.000 10090.790 10094.460 1.000 0.2 134 The most important groups in the Gulf of Thailand in terms of landings are 'trash fish', shellfish, shrimps, Rastrelliger spp., cephalopods, small pelagics, and crabs & lobsters. In a scenario without marine mammals in the ecosystem, most of these target species increase, except for shrimp, which is commercially very important (V. Christensen, Fisheries Centre, UBC, personal communication) (Figure 4.29). 4 1 Species Figure 4.29. Biomass change after a 24-year simulation in the Gulf of Thailand ecosystem, with (white) and without (grey) marine mammals. 4.3.5.6 North Sea Seals' and fisheries' mixed trophic impacts in the North Sea model shows that both marine mammals and fishery have an overall negative impact on the entire ecosystem. However, the negative impact of fisheries (-7.79) is three times larger than the impact of seals (-2.22). Many of the groups that are seriously impacted by fisheries present a positive impact by seals; this is the case, for instance, for gurnards (Lepidotrigla spp.), horse mackerel (Trachurus trachurus), sole (Solea soled), juvenile saithe (Pollachius 135 virens), rays and herring. Some groups are positively impacted by fisheries, such as dab (Limanda limanda), sandeel, juvenile haddock {Melanogrammus aeglefinus), juvenile cod and birds. All of the groups (except for dab) are also positively impacted by seals. Negative impacts from seals are mainly observed for cod, saithe, plaice and dab (Figure 4.30). Here again, the overall impact of all fisheries grouped together is damaging for each single fishing fleets. Finally, seals have a slight positive impact on seiners (Figure 4.30). 0.4 -\ Impacted groups Figure 4.30. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the North Sea system. In Ecosim, the North Sea model was first analyzed with time series of fishing mortality to see which group decline or increase over time. The model includes four sources of fishing mortality (trawlers, gill nets, seiners and industrial fisheries), which were combined to quantify overall effects on the ecosystem. Time series cover a period of 22 years, from 1974 to 1995. At the end of the simulation with the original fishing effort, horse mackerel are strongly increasing, while other groups seems to stabilize at slightly higher biomass (juvenile cod, sole, birds, herring, haddock, juvenile saithe, Norway pout, 136 whiting [Merlangius merlangus], and gurnards). Other groups stabilized after a decrease of biomass (juvenile haddock, cod, plaice [Pleuronectes platessa]), and mackerel (Scomber scombrus) biomass is in constant decrease, reaching near extirpation at the end of the simulated period (Figure 4.31). Finally, the simulated seal extirpation in the North Sea ecosystem (Figure 4.32), the dynamics of the rest of the groups remains approximately the same. No fish species shows a clear increase in biomass linked to the removal of seals. Over a period of 22 years, the biomass remains approximately the same (Btot = 290 t*km" with marine mammals, and 293 t*km"2 without; Table 4.11, Figure 4.32), but seals are completely extirpated from the ecosystem. Their biomass may have been replaced by cod and saithe, which show an increase in biomass following the decline of seals (Table 4.11). 137 Figure 4.31. Ecosim results for the North Sea model with original fishing effort. Figure 4.32. Ecosim results for the North Sea model after the extirpation of seals. o Table 4.11. Biomass (t-km"2-year') change from the start (S) to the end (E) of the 22-year Ecosim simulation in the North Sea ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species. Group name Ecosim with MM Ecosim without MM B change after MM extirpation Biomass (Start) Biomass (End) Biomass (E/S) Biomass (Start) Biomass Biomass (End) (E/S) Cod 0.426 0.248 0.580 0.428 0.357 0.840 44.2 Haddock 0.512 0.597 1.170 0.513 0.627 1.220 4.9 Herring 0.738 1.178 1.600 0.738 1.135 1.540 -3.6 Mackerel 1.906 0.132 0.070 1.906 0.133 0.070 0.3 Norway pout 3.243 3.679 1.130 3.243 3.671 1.130 -0.2 Plaice 0.713 0.375 0.530 0.713 0.541 0.760 44.2 Saithe 1.299 1.318 1.010 1.300 1.499 1.150 13.8 Sandeel 9.829 9.413 0.960 9.829 9.515 0.970 1.1 Sole 0.094 0.145 1.540 0.094 0.170 1.810 17.3 Whiting 0.492 0.578 1.180 0.493 0.639 1.300 10.5 Birds 0.003 0.004 1.300 0.003 0.004 1.300 0.0 Gurnards 0.530 0.538 1.020 0.530 0.540 1.020 0.3 Horse mackerel 0.291 1.343 4.620 0.291 1.314 4.520 -2.2 Other predators 0.543 0.559 1.030 0.543 0.551 1.020 -1.3 Skates & rays 0.090 0.092 1.030 0.090 0.096 1.070 4.0 Seals 0.002 0.005 2.520 0.002 0.000 0.030 -100.0 West mackerel 0.677 0.761 1.120 0.677 0.764 1.130 0.4 Other invertebrates 24.387 23.545 0.970 24.388 23.471 0.960 -0.3 Juvenile cod 0.084 0.129 1.540 0.084 0.128 1.520 -1.0 Juvenile haddock 1.708 1.249 0.730 1.708 1.159 0.680 -7.1 Juvenile saithe 0.040 0.046 1.130 0.040 0.046 1.130 0.0 Juvenile whiting 1.114 1.163 1.040 1.115 1.172 1.050 0.8 Sprat 0.513 0.501 0.980 0.514 0.525 1.020 4.8 Dab 4.401 4.728 1.070 4.404 5.110 1.160 8.1 Copepods 25.400 25.158 0.990 25.400 25.150 0.990 0.0 Euphausiids 20.329 21.583 1.060 20.329 21.624 1.060 0.2 Other crustaceans 13.241 12.807 0.970 13.241 12.795 0.970 -0.1 Echinoderms 24.000 24.128 1.010 24.000 23.779 0.990 -1.4 Polychaetes 32.000 32.906 1.030 32.000 32.802 1.030 -0.3 Other macrobenthos 49.002 51.030 1.040 49.004 51.363 1.050 0.7 Phytoplankton 46.999 47.358 1.010 46.999 47.366 1.010 0.0 Detritus 24.997 24.798 0.990 24.997 24.817 0.990 0.1 Total 0.426 0.248 0.580 0.428 0.357 0.840 44.2 The main species targeted by the North Sea fisheries are adult Norway pout, sandeel, sprat, saithe, herring, whiting, and haddock. When marine mammals are absent from the system, Ecosim predicts that these fisheries are catching about the same amount of fish than if marine mammals were present in the system (the largest changes are increases of 11% and 14% for whiting and saithe, respectively; Table 4.11). Out of four fisheries, one (seiners, catching herring and mackerel) is definitely decreasing in terms of 140 catch if there are no marine mammals in the ecosystem. In general, all trophic groups stay at approximately the same level of biomass, with or without marine mammals in the ecosystem (Figure 4.33). 25 1 Species Figure 4.33. Biomass change after a 22-year simulation in the North Sea ecosystem, with (white) and without (grey) seals (polychaetes, other macrobenthos and phytoplankton [not snown] have very high and nearly identical biomasses with and with out marine mammals in the system, i.e., 32, 51, and 47 tkm" , respectively). 4.3.5.7 Strait of Georgia The mixed trophic impact analysis for the Strait of Georgia shows that while the overall trophic impact of marine mammals is near neutral (-0.83), there is a strong negative impact by fisheries on the foodweb (-7.79). This negative impact of fisheries affects almost all fish species in the ecosystem (Figure 4.34), except for dogfish (Squalus acanthias), juvenile herring and juvenile coho salmon, which seem to benefit from the presence of fisheries. The strongest positive impact of the fisheries is seen on benthic invertebrates, jellyfish and plankton. 141 When marine mammals have a negative impact on fish in the Strait of Georgia ecosystem, this is always smaller than fisheries' impact on the same groups. Interestingly, the overall effect of marine mammal on fishing is positive for trawlers, gill nets, seiners and industrial fleets. Total marine mammals impact is also strongly negative for resident killer whales, seals and sea lions (Figure 4.34). Impacted groups Figure 4.34. Mixed trophic impact of marine mammals (white) and fisheries (grey) on fish and fleet groups of the Strait of Georgia system. The Strait of Georgia model was first analyzed with Ecosim with time series of fishing mortality to see which group decline or increase over time. The model covers eight sources of fishing mortality: longline; salmon fishers; herring gill/seine; 'seal control'; trawls; gill nets; seiners and industrial fleets. All fisheries were combined to test for their effect on the whole ecosystem. Time series cover a period of 50 years, from 1950 to 1999. At the end of the simulation with the original fishing effort, the juvenile resident coho salmon's biomass appears to explode; the zooplankton appears to show great variations, as well (Figure 4.35). Most groups stabilized after a slight decrease of biomass, and seals & sea lions, dolphins, juvenile resident coho salmon and adult herring stabilized after a slight increase. No groups had collapsed at the end of the simulation. 142 When all marine mammals were removed from the Strait of Georgia ecosystem, the ecosystem structure was altered (Figure 4.36). The extirpation of marine mammals lead to strong variation in halibut biomass, and drove the biomass of small pelagics, jellyfish, eulachon (Thaleichthys pacificus) and adult hake (Urophycis tenuis) close to zero. All other groups appeared to stabilize at lower levels at the end of the 50-years simulation. Over that period, there is no significant decrease in total biomass (543 versus 542 t*m"2; Table 4.12). 143 Figure 435. Ecosim results for the Strait of Georgia model with original fishing effort. Figure 4.36. Ecosim results for the Strait of Georgia model after the extirpation of marine mammals. Table 4.12. Biomass (t-km~2-year"') change from the start (S) to the end (E) of the 50-year Ecosim simulation in the Strait of Georgia ecosystem, with and without marine mammals (MM). Groups in bold represent the commercially important species. Group name Ecosim with MM Ecosim without MM B change after MM extirpation Biomass (Start) Biomass (End) Biomass (E/S) Biomass (Start) Biomass Biomass (End) (E/S) Transient Orcas 0.003 0.002 0.670 0.002 0.000 0.000 -100.0 Dolphins (Res. Orca) 0.060 0.072 1.190 0.047 0.002 0.030 -97.2 Seals & Sealions 0.381 0.842 2.210 0.322 0.007 0.020 -99.2 Halibut 0.004 0.002 0.470 0.004 0.006 1.450 200.0 Lingcod 5.415 1.708 0.320 5.416 2.517 0.460 47.4 Dogfish shark 6.528 4.888 0.750 6.528 5.957 0.910 21.9 Adult hake 7.807 1.865 0.240 7.814 2.224 0.280 19.2 Juvenile hake 2.546 0.586 0.230 2.548 0.640 0.250 9.2 Adult resident coho 0.186 0.047 0.250 0.187 0.211 1.130 348.9 Juv. Resident coho 0.910 0.865 0.950 0.911 0.909 1.000 5.1 Ad. Resident chinook 0.335 0.076 0.230 0.336 0.537 1.600 606.6 Juv. Resident chinook 1.283 0.518 0.400 1.286 0.766 0.600 47.9 Demersal fishes 12.781 4.099 0.320 12.791 5.024 0.390 22.6 Seabirds 0.020 0.008 0.390 0.020 0.008 0.410 0.0 Small pelagics 3.331 0.584 0.180 3.335 0.369 0.110 -36.8 Eulachon 2.495 0.464 0.190 2.495 0.308 0.120 -33.6 Adult herring 15.402 16.894 1.100 15.404 14.668 0.950 -13.2 Juvenile herring 3.635 1.944 0.530 3.635 1.956 0.540 0.6 Jellyfish 18.815 4.250 0.230 18.815 4.455 0.240 4.8 Pred. invertebrates 9.510 4.608 0.480 9.511 3.834 0.400 -16.8 Shellfish 231.457 123.610 0.530 231.457 126.475 0.550 2.3 Grazing invertebrates 491.271 228.769 0.470 491.273 224.623 0.460 -1.8 Cam. zoolplankton 22.162 16.966 0.770 22.157 16.664 0.750 -1.8 Herb, zoolplankton 39.683 33.429 0.840 39.684 34.289 0.860 2.6 Kelp / seagrass 29.465 13.151 0.450 29.465 13.154 0.450 0.0 Phytoplankton 101.806 81.893 0.800 101.807 81.079 0.800 -1.0 Detritus 1.382 1.127 0.820 1.382 1.119 0.810 -0.7 Total 1008.672 543.266 0.540 1008.631 541.800 0.540 -0.3 The main targeted species in the Strait of Georgia ecosystem are herring, resident Chinook salmon, lingcod and resident Coho salmon. When marine mammals are removed from the ecosystem, most of these commercially important fish end up with more biomass than in the initial ecosystem, except for herring, which decreases by 13% (Figure 4.37). 146 20 , 1 - n ^ m . - n . r f l , ,l i—I.I i—• I 8 1 ha ke  Li n fis h s du ll Do g < fir ds  ig ic s :h on  Se at  ill pe l. Species Figure 4.37. Biomass change after a 50-year simulation in the Strait of Georgia ecosystem, with (white) and without (grey) marine mammals (shelfish, grazing invertebrates, and phytoplankton [not shown] have very high and nearly identical biomasses with and with out marine mammals in the system, i.e., 124, 229, and 82 t-km"2, with marine mammals, and 126, 224, and 81 t-km"2 without marine mammals, respectively). 4.3.6 Mixed trophic impact versus catch or consumption When mixed trophic impacts are plotted against consumption (for marine mammals) or catch (for fisheries), we see that marine mammals consume generally less than fisheries catch, and that their 77 is less negative than that of fisheries for the same consumption or catch level (Figure 4.38). Moreover, the overall mixed trophic impacts of the marine mammals on the whole ecosystem becomes less negative with increasing consumption. This is a rather surprising result, to be discussed further below. Because of the exceptionally high level of consumption by large marine mammals (killer whales) in the Eastern Bering Sea, this ecosystem was considered as an outlier in the analysis. 147 -8 -7 -6 -5 -4 -3 -2 -1 0 Mixed trophic impact Figure 4.38. Total consumption by marine mammals (black dots) or total catch by fisheries (open dots) versus their respective overall mixed trophic impact for each studied ecosystem. Density ellipses represent the 90% confidence intervals. Discussion The trophic impact of marine mammals on food webs is generally seen as a direct relationship between the predators and their prey. Many studies (Parsons 1992; Kenney et al. 1995; Lesage et al. 2001; Tjelmeland and Lindstram 2005) have addressed the question of the trophic role of top predators without taking into account the indirect trophic effects that they can have on their prey. The results presented here suggest that marine mammals can have important indirect effects on trophic structure. Therefore, my analysis offers a new perspective on the function of these predators in marine food webs, and their interaction with fisheries. We clearly see from results presented above that a change in marine mammal biomass can lead to important alterations in the structure of the ecosystem. In a time where marine ecosystems are overexploited (Pauly et al. 1998a; Lotze et al. 2006), polluted (Angel 1995; Sindermann 1995; Clark 2001) and subject to climate change 148 (Harley et al. 2006), improving our ability to understand ecological processes involving marine mammals and fisheries becomes crucial. 4.4.1 Resource overlap and trophic levels On a global scale, most food consumed by marine mammals consists of prey types that are not the main target of fisheries (Figure 4.2), and whales seem to consume most of their food in areas where commercial vessels do not fish (Kaschner and Pauly 2005). In areas where competition between marine mammals and fisheries is evident (identified as hotspots of resource overlap by Kaschner and Pauly [2005]), my results show that the resource overlap is indeed higher than the global average presented in Kaschner and Pauly (2005). However, most overlap appears to occur between fisheries and larger, deep-diving toothed whales (Kaschner 2004), so when all marine mammals are analyzed as a whole, their overlap is not as strong as may be expected. Depending on the ecosystem, the overlap between marine mammals and fisheries index involves different food types. In the North Sea, the Benguela, and the Strait of Georgia systems, marine mammals and fisheries compete mainly for the 'miscellaneous fish'. This group includes demersal, benthic, benthopelagic and bathydemersal fish that are less than 150 cm, and pelagic fish that are between 60 cm and 150 cm (Pauly et al. 1995; Kaschner 2004). For the purpose of this analysis, this prey group is clearly overaggregated; it represents too many different species. Thus, it is important to look at the different species composing the 'miscellaneous fish' group in different ecosystems. In the North Sea, the Benguela and the Strait of Georgia ecosystems, (and for the time periods they represent), there were important fisheries for larger fish (Christensen et al. 2002; Shannon et al. 2004; Wallace 1998), and the marine mammals species in these ecosystems are higher trophic-level predators, who mostly feed on these large fish (Christensen et al. 2002; Shannon et al. 2004; Dalsgaard et al. 1998). These ecosystems are quite different in terms of structure from the Gulf of St. Lawrence, where the intense fishing activity in the 1980s has lead to the depletion of most groundfish stocks, leaving mainly smaller planktivorous fish and crustaceans for fisheries and marine mammals (Morissette et al. 2006; Savenkoff et al. in press; and see Chapter 5). Consequently, in this 149 ecosystem, small pelagics are the main overlapping resource. In the Gulf of Thailand, benthic invertebrates and miscellaneous fish are the food types that are most overlapping. Here again, the major development of trawl fishery in the 1960s has lead to a shift to lower trophic levels such as 'trash fish' and shrimps (Gulland 1983), and most marine mammals are dolphins and whales that mainly eat smaller fish (Perrin et al. 2005). There may have been a time when fisheries and marine mammals did not overlap much in terms of food resources, but now that fisheries has moved down the food web (Pauly et al. 1998), the target food types might have become very similar. Finally, in the eastern tropical Pacific ocean, where the resource overlap is the lowest, there is not much competition between fisheries and marine mammal. Fisheries target mainly miscellaneous fish and large species such as tuna, and marine mammals feed mostly on small squids, mesopelagics and small pelagic fish. The number of trophic links in the ecosystem also has an effect on the resource overlap between marine mammal and fisheries. Indeed, food webs with lower connectance (less trophic links) tend to have higher overlap values. Some ecosystems show an important overlap between marine mammals and fisheries when analyzed by food types: the North Sea, Benguela and Gulf of Thailand are the most important. When the same analysis is done per trophic group (using the complete structure of the catch and the consumption matrix of marine mammals), the overlap index is reduced, but most ecosystems showing important overlap between marine mammals and fisheries remain the same. However, in the case of the North Sea, the overlap index calculated per trophic group is very low compared to the index calculated by food type, suggesting that there is no overlap at the species level. Overall, landings from global fisheries have shifted from large piscivorous fish toward smaller invertebrates and planktivorous fish (Pauly et al. 1998). The mean trophic level of fisheries catches calculated here is thus lower than it was 50 years ago. By contrast, nothing is known about the potential decline in trophic level of marine mammals' consumption over the last century. However, my results show that globally, the trophic level of marine mammals prey is lower than the trophic level of the catch (2.88 versus 3.42). As fisheries continue to move further down in terms of the trophic level of species caught, the competition for food resources with marine mammals might become more important. In ecosystems such as the Eastern tropical Pacific where the mean trophic level 150 of the catch is still high, the overlap with marine mammals is negligible. Interestingly, the Strait of Georgia represent a higher trophic level of marine mammals consumption than trophic level of the catch. This is due to the fact that in this ecosystem, an important marine mammal predator is the killer whale, who mainly feed on pinnipeds, at a very high trophic level. Their presence in the Strait of Georgia definitely increases the average trophic level of marine mammal consumption. Lately, Japanese scientists have stressed that whales of the world consume annually some 3 to 5 times more marine fish and invertebrates than are fished for direct human consumption or for reduction into fish meal and oil (Anonymous 2001). This situation, it is alleged, is not "in balance" with the world's increasing need for a stable food supply. Such arguments are used extensively to justify whaling activity, as it is shown in this citation of Mr. Masayuki Komatsu, executive director of the Japanese Marine Fisheries Research and Development Department, to BBC on "the forces that drive Japanese whaling", 15 June 2006: "Whale fare] abundant. The number of fish is falling while the number of whales is rising. Surely, the rapid increase in the whale population influences the level of fish stocks? We need to know more about it". , my results show that there is no clear and direct relationship between marine mammals' predation and the potential fish catch in the world's oceans. Many whales do eat fish, but the species that they eat are not necessarily targeted by fisheries. In fact the global overlap of food resources, representing the main 'hotspots' of competition between marine mammals and fisheries (Kaschner 2004; Kaschner and Pauly 2005), is relatively low and similar to ecosystems such as the Gulf of St. Lawrence or the Strait of Georgia. Moreover, as the simulation results showed, it is not that clear whether the extirpation of marine mammals in ecosystems would even increase the biomass of the fish targeted by most fisheries. 151 4.4.2 Primary production required to sustain marine mammals and fisheries In the ecosystems studied, the primary production required to sustain marine mammals represents an average of 9.7% of total primary production, less than half the PPR to sustain fisheries catch (20.1%). The latter value represents nearly three times the estimate by Pauly and Christensen (2002) for global fisheries. This is probably due to the fact that my analysis focuses on zones where fishing activity is intense. Accounting for other ecosystems that are less fished would reduce that average PPR presented here. In most ecosystems, PPR for marine mammals is smaller than PPR for the catch. However, this is not the case in the Eastern tropical Pacific Ocean. All my results show quite different results for this ecosystem. This is likely due to the fact that, indeed, marine mammals are more important than fisheries in the Eastern tropical Pacific Ocean. A second order effect may also be that this model represents a very large area, and that information about marine mammals (biomass, consumption rates, diet, production, etc.) is applied directly on the populations known to be within this area, while fisheries' effects might be more 'diluted' and less important in high sea. The highest absolute PPR for fisheries catch is seen in the North Sea, and this is driven by very high PPR for sustaining the mackerel catch (1500 t4̂ m"2 for this single species early in the simulation). Fisheries in this ecosystem require 50% of the total primary production of the system. An even worse case is seen in the Eastern Bering Sea, where 54% of primary production is required to sustain fisheries. PPR required by marine mammals is also the highest in this ecosystem. On the opposite end of the spectrum, the Benguela ecosystem, with the lowest PPR values for both marine mammals and fisheries, is known to be a very productive ecosystem, due to the seasonal, wind-driven upwelling, which enrich the shelf waters and result in a very large productivity (Shannon et al. 2003). 4.4.3 Comparing the mixed trophic impact of marine mammals and fisheries The effect of marine mammals on their prey and consequently on available resources for fisheries is not only a direct predator-prey relationship. Rather their effect is also indirect, for example through feeding both on a prey and the competitors of the prey. 152 Even if negative for all studied ecosystems, the overall trophic impact of marine mammals on the different trophic groups of the ecosystem was always less strong than that of fisheries, except in the Eastern tropical Pacific Ocean where fisheries target mainly large tunas (important predators of many trophic groups in the ecosystem), while marine mammals feed on a larger array of smaller prey (Olson and Watters 2003). For marine mammals, there is a paradoxical trend suggesting that the more they consume, the less they tend to reduce overall biomass. This is possible due to the fact that if marine mammals have to increase their consumption, they will feed on a wider array of prey and induce beneficial predation (see below). In contrast, the mixed trophic impact of fisheries on the entire ecosystem is always strongly negative. Finally, my results show that overall, when marine mammals have a positive impact on some trophic group, the impact of fisheries on the same group is usually negative. 4.4.4 What if marine mammals were not present in these ecosystems? When the extirpation of marine mammals is simulated in the studied ecosystems, the biomass of other species of the food web also changes. In some ecosystems, commercially important species increase significantly after the eradication of marine mammals (e.g., halibut and large pelagics in the Gulf of St. Lawrence, tuna species in the Eastern tropical Pacific, tuna and pelagic species in the Gulf of Thailand, and cod and plaice in the North Sea). However, when all commercial species are considered, there is no obvious benefit for the fisheries. Indeed, total biomass, with no marine mammals in the ecosystem, remains generally and surprisingly similar, or even decrease (as it is the case with the Eastern Bering Sea and the Gulf of St. Lawrence). Indeed, the extirpation of marine mammals may lead to reduced abundances of commercial important fish in some ecosystems. Cape hake, sardine, redeye in the Benguela upwelling, and herring in the North Sea and in the Strait of Georgia decrease if marine mammals are removed from these systems. In the case of the Gulf of Thailand, the Plectorhynchidae group becomes totally depleted when marine mammals are absent. On the other hand, when species or groups increases as a result of the extirpation of marine mammals in the ecosystem, these 153 species or groups are not necessarily the most important commercially (deepwater fish, jellyfish and cephalopods in the Eastern Bering Sea; cephalopods, juvenile pelagics, or juvenile carangids in the Gulf of Thailand). Finally, when commercially important species increase following the extirpation of marine mammals, the instability that this new situation might lead to is not necessarily a stable equilibrium (Berkeley et al. 2004). There might be more fish to catch, but once overfished, these ecosystems could become unstable and at risk of severe losses in biodiversity. In certain ecosystems (e.g., in Eastern Canada), there have been an important debate on culling marine mammals in an attempt to rebuild stocks of once commercially important fish species (e.g., cod) (Dwyer 1999, Fisheries Resource Conservation Council 2002) . In that particular case, at least for the Gulf of St. Lawrence ecosystem, my results suggest that culling of marine mammals would not have led to recovery of the stocks of cods, nor otherwise benefited the commercial fishery. This corroborates the findings of Trzcinski et al. (2006), who suggested that even the complete removal of grey seal predation in the eastern Scotian Shelf (Northwest Atlantic) would not assure the recovery of the cod population, given the high levels of other sources of natural mortality. 4.4.5 Marine mammals and beneficial predation The view that fewer marine mammals would mean more fish in the ocean and more fish for human consumption is often said to be based on "common sense" (Lavigne 2003) . However, this view does not account for other components of the ecosystem, and ecological effects such as prey switching, competition, or cascading effects that may overall result in unintended outcomes, including beneficial predation. I propose in the following conceptual model three types of beneficial predation that can occur between marine mammals and their prey (Figure 4.39). First, a predator (A) can have a positive impact on a prey (2) by also relying on another prey (3), which is a competitor of prey 2 (beneficial predation type 1 in Figure 4.39). An increase of predation from A on 3 would reduce its biomass and thus the level of competition with prey 2. If prey 3 is a preferred prey of predator A and that competition between prey 2 and prey 3 is strong, such a change in the biomass of prey 3 can create a positive effect that 154 encompasses or else exceed the effect of predation from A on prey 2, creating an overall positive effect of predator A on prey 2. This type of beneficial predation may occur for instance in the Gulf of St. Lawrence, where harp seals are major predators for capelin, but also for Arctic cod, a species that is potentially competing with capelin for the same resources. The overall mixed trophic impact of harp seals on capelin is a positive one (Morissette et al. 2006). When all marine mammal species are grouped, this effect is not visible, however. The same kind of phenomenon happens in the Strait of Georgia ecosystem with marine mammals (dolphins, resident killer whales, seals and sea lions) feeding on small pelagic fish and on the competing herring group (competition for zooplankton). The mixed trophic impact analysis clearly shows that even if marine mammals are predators of small pelagics, their presence in the ecosystem is beneficial to this group because they also feed on competing species. In that case, the Ecosim scenario with no marine mammal results in a decrease of small pelagic fish, which corroborates that beneficial predation was occurring. Predator A can also have a positive effect on prey 2 by also feeding on prey 1, which is a predator of prey 2 (beneficial predation type 2 in Figure 4.39). When predator A feeds on prey 1, the predation level on prey 2 is decreased (less predation from prey 1 on prey 2). Once again, if the predation of predator on prey 1 is larger than on prey 2, the overall effect of having Predator A in the ecosystem would be beneficial for prey 2. This is the case, for example, in the Benguela ecosystem, where marine mammals (cetaceans and seals) are feeding on mesopelagic fish, but also on cephalopods, which are major predators of mesopelagic fish. The overall mixed trophic impact of marine mammals on mesopelagic fish is positive, even if they represent a source of predation for that trophic group. Furthermore, the Ecosim scenario with no marine mammal results in an important decrease of mesopelagic fish, which indicates that there may be an indirect effect of marine mammals on this trophic group. Finally, predator A can have a positive effect on prey 3 by being itself a prey for another, larger marine mammal: predator B (type 3 in Figure 4.39). If predator A and prey 3 are both present in an ecosystem, even if predator A is eating prey 3, its presence can be beneficial if predator B has a diet preference for predator A. In this case, the presence of predator A redirected predation by predator B on prey 3. If the predator A becomes less 155 abundant, predator B can switch to prey 3 and thus this results in higher predation on prey 3. This seems to be the case in the Strait of Georgia ecosystem. Indeed, transient killer whales are predators of small pelagics, but also (and mainly) feed on seals and sea lions, who themselves are important predators of small pelagic fish. The overall mixed trophic impact of marine mammals on small pelagic fish is positive. In Ecosim a scenario with no marine mammal would result in a decrease of small pelagic fish. Figure 4.39. Trophic interactions (predation in black, competition in grey) between predators and their prey that can lead to beneficial predation. The numbers above arrows represent the different types of beneficial predation (see text). 4.4.6 Strengths and weaknesses of the modelling approach My approach emphasized the use of ecosystem models to examine the outcomes of trophic interactions between marine mammals and fisheries. Studying this ecological problem through a modelling approach can enhance our understanding of effect of relationship that would otherwise be very difficult to study. The fact that I could compare and generalize from different ecosystems constructed using the EwE approach was one advantage of having comparable models, based on the same framework. The level of detail included in Ecopath model structure and equations is a real asset and represent a rigorous analytical framework (Plaganyi and Butterworth 2004). 156 Some authors warn modelers that caution may be taken in applying EwE models to marine mammal populations, because their life history is very different from most fish (Plaganyi and Butterworth 2004). However here, the simulations I performed consisted in removing all marine mammals, and thus releasing their predation on other trophic groups in the ecosystem. The temporal variation of marine mammals abundance itself was not addressed here (the simulated scenarios always presented how the system would react to the extirpation of marine mammals species, not with the possible causes and modalities of such extirpation). In Ecosim runs with original fishing effort (where marine mammals are present in ecosystems), I assume the vulnerabilities and the ecological reliability of each model is representative of the species dynamics and life cycles. These models have all been published and peer-reviewed, and the marine mammals groups present in the models are assumed to have realistic ecological parameters. However, the analysis done with these seven Ecopath models did not included any environmental effects that may affect ecosystem dynamics. I thus assumed that everything was explained by species interactions, which is definitely not the case. There are indeed strong evidences that climate change, for example, have an important impact on the availability of marine mammals' prey (Harwood 2001), as well as on their distribution and abundance (Simmonds and Isaac 2007). One inconvenience of my approach is that I grouped all marine mammals together instead of analyzing pinnipeds, toothed whales, and baleen whales separately. Doing so, I masked the major differences in their feeding ecology. Also, analyzing marine mammals as a whole group could not show what effect their relative abundance have on the trophic impacts. The relative proportions of different marine mammals species would also affect the trophic level. For example, a large biomass of baleen whales, who would mainly consume krill, would significantly lower my estimate of trophic level of consumption. The calculated overlap index has the disadvantage of representing very large categories of species. Aggregation into large functional groups does not well represents the dynamics of the ecosystem. Thus, while the overlap index represent a good way to have a global and simple representation of the interaction between fisheries and marine mammals, the in-depth analysis of the structure of ecosystems remains crucial. The overlap index calculated at the species level showed that in the case of the North Sea, this 157 can lead to opposite conclusions. Finally, the fact that the overlap index was calculated for marine mammals as a whole group present, once again, a difficulty. Indeed, as most overlap with fisheries occurs among pinnipeds and dolphins, our estimates of the global overlap between fisheries and marine mammals may be underestimated. The use of mixed trophic impact analysis is not sufficient when analyzing the global impact of marine mammals on the foodweb. It is never enough to evaluate only the impact from an ecological snapshot that is fixed in time. Dynamic simulations, validated to reproduce reasonably well the past patterns of change in relative abundance of major species (Walters et al. 2005) remains the best option. Conclusions Our analysis identified that marine mammals are important top predators in marine ecosystems, and that they play an important role in structuring the trophic relationships among food webs. My results show that even in hotspots of competition between marine mammals and fisheries, the overlap for food resources is lower than earlier thought. My results confirm Kaschner and Pauly (2005), who suggested that even the complete eradication of all marine mammals, from all oceans, would likely not increase fisheries catches. Hence, large-scale cullings, as advocated by the Japanese studies (Anonymous 2001) would not increase catches. This study has focused on the top-down influences of marine mammals and fisheries on the fish species in different ecosystems. Although 'bottom-up' changes were not investigated here, their effects might just be additive and alter even more the structure of the ecosystem. This is particularly true in the actual context of climate change, which can affect the productivity of the world's oceans (Loreng 2004; Sanae et al. 2006). There is still much debate about this idea and it will be important to find different ways of addressing this issue. The analysis presented here provided an insight into the problem, but further work on this needs to be pursued. Chapter summary The competition between marine mammals and fisheries for marine resources is a major concern. I examined trophic interactions between marine mammals and fisheries 158 with a resource overlap index developed by Kaschner (2004), using seven Ecopath models including marine mammal groups. On a global scale, most food consumed by marine mammals consisted of prey types that were not the main target of fisheries. For each ecosystem, the primary production required (PPR) to sustain fisheries was compared to the PPR to sustain marine mammals groups. The PPR to sustain marine mammals was less than half the PPR to sustain fisheries catch. I also developed an index representing the mean trophic level of marine mammal's consumption (TLQ) and compared it with the mean trophic level of fisheries catch. My results showed that globally, the trophic level of marine mammals prey was lower than the trophic level of the catch (2.88 versus 3.42). As fisheries continue to move further down in terms of the trophic level of species caught (Pauly et al. 1998), this competition for food resources with marine mammals might become more important. I also observed patterns of marine mammal consumption versus fisheries catches. I estimated the relative mix trophic impacts of a change in marine mammal's biomass on all other components of the ecosystem, including impacts on fisheries. This allowed me to show that marine mammals can have important indirect effects on trophic structure, even beneficial for some of their prey. Within the Ecosim framework, I assessed the change in the trophic structure of an ecosystem after the simulated extirpation of marine mammal populations in these seven ecosystems. Changes in marine mammal biomass lead to important alterations in the structure of the ecosystem. According to what I found, there was no clear and direct relationship between marine mammals' predation and the potential fish catch in the world's oceans. Many whales do eat fish, but the species that they eat are not necessarily targeted by fisheries. As the simulation results showed, it is not that clear whether the extirpation of marine mammals in ecosystems would even increase the biomass of the fish targeted by most fisheries. Indeed, total biomass, with no marine mammals in the ecosystem, remained generally and surprisingly similar, or even decreased. References Abrams, PA. 1992. Why don't predators have positive effects on prey populations? Evolutionary Ecology 6: 449-457. 159 Ainsworth, C.H. 2006. Strategic marine ecosystem restoration in Northern British Columbia. PhD thesis, University of British Columbia, Canada. Angel, M.V. 1995. Deep sea pollution - who cares? LEE Colloquium on "Pollution of Land, Sea and Air: An Overview for Engineers" 165: 9/1-9/5. Anonymous. 2001. Increasing competition between fisheries and whales. Japan's whale research in the Western North Pacific (JARPNII). Fisheries Agency. Baker, C.S. and P.J. Clapham. 2004. Modelling the past and future of whales and whaling. Trends in Ecology and Evolution 19: 365-371. Bax, N.J. 1998. The significance and prediction of predation in marine fisheries. ICES Journal of Marine Sciences 55: 997-1030. Berkeley, S.A., Hixon, M.A., Larson, R.J., and M.S. Love. 2004. Fisheries sustainability via protection of age structure and spatial distribution of fish populations. Fisheries 29: 23-32. Booth, S. and D. Zeller. 2005. Mercury, food webs, and marine mammals: implications of diet and climate change for human health. Environmental Health Perspectives 113:521-526. Bowen, W.D. 1997. Role of marine mammals in aquatic ecosystems. Marine Ecology Progress Series 158: 267-274. Bowen, W.D., Tully, D., Boness, D.J., Bulheier B., and G. Marshall. 2002. Prey- dependent foraging tactics and prey profitability in a marine mammal. Marine Ecology Progress Series 244: 235-245. Bowen, W.D. and D.B. Siniff. 1999. Distribution, population biology, and feeding ecology of marine mammals, pp 423-484 In Reynolds, J.E. and S.A. Rommels (eds.). Biology of marine mammals. Smithsonian Institution Press, Washington, D.C. Brown, D.H., Ferris, H., Fu S., and R. Plant. 2004. Modeling direct positive feedback between predators and prey. Theoretical Population Biology 65: 143-152. Christensen, V. 1998. Fishery-induced changes in a marine ecosystem: insights for models of the Gulf of Thailand. Journal of Fish Biology 53(Supplement A): 128-142. 160 Christensen, V. and D. Pauly. 1992. E C O P A T H II - a software for balancing steady-state ecosystem models and calculating network characteristics. Ecological Modelling 61: 417-442. Christensen, V. and C.J. Walters. 2004. Ecopath with Ecosim: methods, capabilities and limitations. Ecological Modelling 172: 109-139. Christensen, V., Beyer, J.E., Gislason, H., and M. Vinter. 2002. A comparative analysis of the North Sea based on Ecopath with Ecosim and multispecies virtual population analysis, pp 39 In Christensen, V., Reck, G., and J.L. Maclean. Proceedings of the INCO-DC Conference Placing Fisheries in their Ecosystem Context, Galapagos Islands, Ecuador, 4-8 December 2000. ACP-EU Fisheries Research Report. Christensen, V., Walters C.J., and D. Pauly. 2005. Ecopath with Ecosim: a user's guide. Fisheries Centre, University of British Columbia, Vancouver. November 2005 edition. Clapham, P.J., Young, S.B., and R.L. Brownell Jr. 1999. Baleen whales: conservation issues and the status of the most endangered populations. Mammal Review 29: 35- 60. Clark, R.B. (ed.). 2001. Marine pollution. Oxford University Press, USA. Dalsgaard, J., Wallace, S.S., Salas, S., and D. Preikshot. 1998. Mass-balance model reconstructions of the Strait of Georgia: the present, one hundred and five hundred years ago. pp. 72-86 In Pauly, D., Pitcher, T. Preikshot, D, and J. Hearne (eds.). Back to the future: reconstructing the Strait of Georgia ecosystem. Fisheries Centre Research Reports 6(5). Dans, S.L., Alonso, M.K., Pedraza, N.S., and E.A. Crespo. 2003. Incidental catch of dolphins in trawling fisheries off Patagonia, Argentina: Can population persist? Ecological Applications 13: 754-762. DeMaster, D.P., Fowler, C.W., Perry, S.L., and M.F. Richlen. 2001. Predation and competition: the impact of fisheries on marine-mammals populations over the next one hundred years. Journal of Mammalogy 82: 641-651. Dillingham, P.W., Skalski, J.R., and K.E. Ryding. 2006. Fine-scale geographic interactions between Steller sea lion (Eumetopias jubatus) trends and local fisheries. Canadian Journal of Fisheries and Aquatic Sciences 63: 107-119. 161 Dwyer, E. 1999. Empty nets: Shad fisherman say catches have fallen off dramatically and they're blaming the harbour seals. Saint John Times Globe, June 14 1999. Estes, J.A. and D.O. Duggins. 1995. Sea otters and kelp forests in Alaska: generality and variation in a community ecological paradigm. Ecological Monographs 65: 75-100. Estes, J.A., Tinker, M.T., Williams, T.M., and D.F. Doak. 1998. Killer whale predation on sea otters linking oceanic and nearshore ecosystems. Science 282: 473-476. Estes, J.A., Tinker, M.T., Doroff, A.M., and D.M. Burn. 2005. Continuing Sea Otter population declines in the Aleutian Archipelago. Marine Mammal Science 21: 169- 172. FAO/FISHCODE. 2001. Report of a bio-economic modeling workshop and a policy dialogue meeting on the Thai demersal fisheries in the Gulf of Thailand held at Hua Hin, Thailand, 31 May - 9 June 2000. FI: GCP/INT/648NOR: Field Report F-16 (En). Rome, FAO. Fisheries Resource Conservation Council. 2002. 2002/2003 conservation requirements for groundfish stocks in the Gulf of St. Lawrence. Report to the Minister of Fisheries and Oceans, April 2002. FRCC.2002.R.2. Forgaty, M.J. and S.A Murawski. 1998. Large-scale disturbance and the structure of marine systems: fishery impacts on Georges Bank. Ecological Applications 8(Supplement): 6-22. Gales, N., Hindell, M., and R. Kirkwood (eds.). 2003. Marine mammals: fisheries, tourism and management issues. CSLRO publishing, Melbourne, Australia. Gulland, J.A. 1983.World resources of fisheries and their management, pp. 839-1061 In Kinne, O. (ed.). Marine Ecology vol. V, part 2. J. Wiley & sons Ltd. Hammill, M.O., and G.B. Stenson. 2000. Estimated prey consumption by harp seals (Phoca grownlandica), hooded seals (Cystophora cristata), grey seals (Halichoerus grypus) and harbour seals (Phoca vitulina) in Atlantic Canada. Journal of Northwest Atlantic Fishery Science 26: 1-23. Harley, C.D.G., Randall, H.A., Hultgren, K.M., Miner, B.G., Sorte, C.J.B., Thornber, C.S., Rodrigiez, L.F., Tomanek, L., and S.L. Williams. 2006. The impactos of climate change in coastal marine systems. Ecology Letters 9: 228-241. 162 Harvey, C.J., Cox, S.P., Essington, T.E., Hansson, S. and J.F. Kitchell. 2003. An ecosystem model of food web interactions and fisheries effects in the Baltic Sea. ICES Journal of Marine Science 60:1-12. Hardwood, J.. 2001. Marine mammals and their environment in the twenty-first century. Journal of Mammology 82: 630-640. Heymans, J.J. Shannon L.J. and A. Jarre. 2004. The Northern Benguela Ecosystem: Changes over three decades (1970, 1980 and 1990). Ecological Modelling 172: 175- 195. Holt, S.J. 2004. Counting whales in the North Atlantic. Science 303: 39-40. Hume, F., Hindell, M.A., Pemberton, D., and R. Gales. 2004. Spatial and temporal variation in the diet of a high trophic level predator, the Australian fur seal (Arctocephaluspusillus doriferus). Marine Biology 144: 407-415. Huntley, M.E., Lopez, M.D.G., and D.M. Karl. 1991. Top predators in the Southern Ocean: A major leak in the biological carbon pump. Science 253: 64-66. Kaneko, H. and I.R. Lawler. 2006. Can near infrared spectroscopy be used to improve assessment of marine mammal diets via fecal analysis? Marine Mammal Science 22: 261-275. Kaschner, K. 2004. Modelling and mapping resource overlap between marine mammals and fisheries on a global scale. Ph.D. thesis, University of British Columbia. Kaschner, K. and D. Pauly. 2005. Competition between marine mammals and fisheries: Food for thought, pp. 95-117 In Salem, D. J. and A. N. Rowan (eds.). The state of animals III: 2005. Humane Society of the United States Press, Washington, D.C. Kaschner, K., Watson, R., Christensen, V., Trites, A.W., and D. Pauly. 2001. Modelling and maping trophic overlap between marine mammals and commercial fisheries in the North Atlantic, pp 35-45 In Zeller, D., Watson, R., and D. Pauly (eds.). Fisheries impacts on North Atnaltic ecosystems: catch, effort and national/regional data sets. Fisheries Centre Research Reports 9(3). Kenney, R.D., Scott, G.P., Thompson, T.J., and H.E. Winn. 1997. Estimates of prey consumption and trophic impacts of cetaceans in the USA northeast continental shelf ecosystem. Journal of Northwest Atlantic Fishery Science 22: 155-171. 163 Kitchell, J.F., Eby, L.A., He, X., Schindler, D.E., and R.A. Wright. 1994. Predator-prey dynamics in an ecosystem context. Journal of Fish Biology 45(Supplement A): 209- 226. Kraus, S.D., Brown, M.W., Caswell, H., Clark, C.W., Fujiwara, M., Hamilton, P.K., Kenney, R.D., Knowlton, A.R., Landry, S., Mayo, C.A., McLellan, W.A., Moore, M.J., Nowacek, D.P., Pabst, D.A., Read, A.J., and R.M. Rolland. 2005. North Atlantic Right Whales in crisis. Science 309: 561-562. Laake, J.L., Browne, P., DeLong, R.L., and H.R. Huber. 2002. Pinniped diet composition: a comparison of estimation models. Fishery Bulletin 100: 434-447. Lavigne, D.M. 2003. Marine mammals and fisheries: the role of science in the culling debate, pp. 31-47 In Gales, N., Hindell, M., and R. Kirkwood (eds.). Marine mammals: fisheries, tourism and management issues. CSLRO publishing, Melbourne, Australia. Lawson, J.W. and G.B. Stenson. 1995. Historic variation in the diet of harp seals (Phoca groenlandica) in the northwest Atlantic, pp. 261-269 In Blix, A.S., Ultang, 0., and L. Walloe (eds.). Seals, whales, fish and man. Elsevier Science. Lesage, V., Hammill, M.O. and K.M. Kovacs. 2001. Marine mammals and the community structure of the Estuary and Gulf of St. Lawrence, Canada: evidence from stable isotope analysis. Marine Ecology Progress Series 210: 203-221. Lindeman, R.L. 1942. The trophic-dynamic aspect of ecology. Ecology 23:399—418. Livingston, P.A. (ed.). 2000. Ecosystem considerations for 2001. Appendix: Stock Assessment and Fishery Evaluation Report for the Groundfish Resources of the EBS/AI and GOA. North Pacific Fishery Management Council, Anchorage, AK, USA. Loreng, H. 2004. Marine Systems: The impact of climate change (ACIA Chapter 9). ACIA International Scientific Symposium on Climate Change in the Arctic, Reykjavik (Iceland), 9-12 Nov 2004. Lotze, H.K., Lenihan, H.S., Bourque, B.J., Bradbury, R.H., Cooke, R.G., Kay, M.C., Kidwell, S.M., Kirby, M.X., Peterson, C.H., and J.B.C. Jackson. 2006. Depletion, degradation, and recovery potential of estuaries and coastal seas. Science 312: 1806- 1809. 164 Lubick, N. 2003. New count of old whales adds up to big debate. Science 301: 451. MacKenzie, B.R., Alheit, J., Conley, D.J., Holm, P., and C.C. Kinze. 2002. Ecological hypotheses for a historical reconstruction of upper trophic level biomass in the Baltic Sea and Skagerrak. Canadian Journal of Fisheries and Aquatic Sciences 59: 173-190. Mackinson, S., Blanchard, J.L.j Pinnegar, J.K., and R. Scott. 2003. Consequences of alternative functional response formulations in models exploring whale-fishery interactions. Marine Mammal Science 19: 661-681. Martell, S.J., Beattie, A.I., Walters, C.J., Nayar, T., and R. Briese. 2002. Simulating fisheries management strategies in the Gulf of Georgia ecosystem using Ecopath with Ecosim. pp. 16-23 In Pitcher. T. and K. Cochrane (eds.). The use of ecosystem models to investigate multispecies management strategies for capture fisheries. Fisheries Centre Research Reports 10(2). Mitchell, E. 1973. Draft report on humpback whales taken under specific scientific permit by eastern Canadian land stations, 1969-1971. Report of the International Whaling Commission 23: 138-154. Mitchell, E. 1974. Present status of northwest Atlantic fin and other whale stocks, pp 108- 169 In W.E. Schevill (ed.). The whale problem: A status report. Cambridge. Harvard University Press. Mitchell, E. 1975. Trophic relationships and competition for food in the northwest Atlantic whales. Proceedings of the Canadian Zoological Society 1974: 123-133. Mitchell, E. 2004. Counting whales in the North Atlantic. Science 303: 39-40. Mohn, R. and W.D. Bowen. 1996. Grey seal predation on the eastern Scotian Shelf: Modelling the impact of Atlantic cod. Canadian Journal of Fisheries and Aquatic Sciences 53: 2722-2738. Molenaar, E.J. 2002. Ecosystem-based fisheries management, commercial fisheries, marine mammals and the 2001 Reykjavik declaration in the context of international law. International Journal of Marine and Coastal Law 17: 561-595. Morissette, L. and T. Pitcher. 2005. Model structure and balancing, pp. 5-24 In Haggan, N. and T.J. Pitcher (eds.). Ecosystem Simulation Models of Scotland's West Coast and Sea Lochs. Fisheries Centre Research Reports 13(4). 165 Morissette, L., Hamrnill, M.O. and C. Savenkoff. 2006. The trophic role of marine mammals in the Northern Gulf of St. Lawrence. Marine Mammal Science. 22: 74- 103. Myers, R.A., Hutchings, J.A., and N.J. Barrowman. 1996. Hypotheses for the decline of cod in the North Atlantic. Marine Ecology Progress Series 138: 1-3. Myers, R.A., Hutchings, J.A., and N.J. Barrowman. 1997. Why do fish stocks collapse? The example of cod in Atlantic Canada. Ecological Applications 7: 91-106. National Research Council. 1999. Sustaining marine fisheries. Committee on ecosystem management for sustainable marine ecosystems. National Academy Press, Washington, D.C. National Research Council. 2003. The decline of Steller Sea Lion in Alaskan waters: Untangling food webs and fishing nets. Committee on the Alaska groundfish fishery and Steller Sea lions. National Academies Press, Washington, D.C. Odum, W.E. and E.J. Heald. 1975. The detritus-based food web of an estuarine mangrove community, pp 265-286 In Wiley, M. (ed.). Estuarine research, Vol. 1. Academic Press, New York. Olson, RJ. and G.M. Watters. 2003. A model of the pelagic ecosystem in the eastern tropical Pacific ocean. Inter-American Tropical Tuna Commission 22: 133-218. Overholtz, W.J., Murawski, S.A., and K.L. Foster. 1991. Impact of predatory fish, marine mammals, and seabirds on the pelagic fish ecosystem of the northeastern USA. ICES Marine Science Symposia 193: 198-208. Pace, M.L., Cole, J.J., Carpenter, S.R., and J.F. Kitchell. 1999. Trophic cascades revealed in diverse ecosystems. Trends in Ecology and Evolution 14: 483-488. Patterson, K. 1992. Fisheries for small pelagic species: an empirical approach to management targets. Reviews in Fish Biology and Fisheries 2: 321-338. Parsons, T.R. 1992. The removal of marine predators by fisheries and the impact of trophic structure. Marine Pollution Bulletin 25: 51-53. Pauly, D. and V. Christensen. Primary production required to sustain global fisheries. Nagure 374: 255-257. 166 Pauly, D. and R. Watson. 2005. Background and interpretation of the 'Marine Trophic Index' as a measure of biodiversity. Philosophical transactions of the royal society: Biological sciences 360: 415-423. Pauly, D., Trites, A.W., Capuli, E., and V. Christensen. 1995. Diet composition and trophic levels of marine mammals. ICES Council Meeting Papers 1995/N:13. Pauly, D., Christensen, V., Dalsgaard, J., Froese, R., and Toreres, FJr. 1998a. Fishing down marine food webs. Science 279: 860-863. Pauly, D., Pitcher, T., Preikshot, D., and J. Hearne (eds.). 1998b. Back to the future: Reconstructing the Strait of Georgia ecosystem. Fisheries Centre Research Reports 6(5). Pauly, D., Trites, A.W., Capuli, E., and V. Christensen. 1998c. Diet composition and trophic levels of marine mammals. ICES Journal of Marine Science 55: 467-481. Perrin, W.F., Feeves, R.R., Dolar, M.L.L., Jefferson, T.A., Marsh, H., Wang, J.Y., and J. Estacion. 2005. Report of the second workshop on the biology and conservation of smnall cetaceans and dugongs of Southeast Asia. CMS Technical series publication No. 9. UNEP/CMS secretariat, Bonn, Germany. Plaganyi, E E . and D.S. Butterworth. 2002. Competition with fisheries, pp 268-273 In Perrin, W.F., Wtirsig, B., and H.G.M. Thewissen (eds.). Encyclopedia of Marine Mammals. Academic Press, San Diego. Plaganyi, E E . and D.S. Butterworth. 2004. A critical look at the potential of Ecopath with Ecosim to assist in practical fisheries management. African Journal of Marine Science 26: 261-287. Punt, A.E. and D.S. Butterworth. 1995. The effects of future consumption by the Cape fur seal on catches and catch rates of the Cape hakes. 4. Modelling the biological interaction between Cape fur seals Arctocephalus pusillus pusillus and the Cape hakes Merluccius capensis and M. paradoxus. South African Journal of Marine Science 16: 255-285. Roman, J. and S.R. Palumbi. 2003. Whales before whaling in the North Atlantic. Science 301: 508-510. 167 Sanae, C , Tadokoro, K., Sugisaki., and T. Saino. 2006. Effects of decadal climate change on zooplankton over the last 50 years in the western subarctic North Pacific. Global Change Biology 12: 907-920. Savenkoff, C , Swain, D.P., Hanson, J.M., Castonguay, M., Hammill, M.O., Bourdages, H., Morissette, L., and D. Chabot. (in press). Effects of fishing and predation in a heavily exploited ecosystem II: Comparing pre- and post-groundfish collapse periods in the southern Gulf of St. Lawrence (Canada). Ecological Modelling. Schiermeier, Q. 2002. How many more fish in the sea? Nature 419: 662-665. Schindler, D.E., Essington, T.E., Kitchell, J.F., Boggs, C , and R. Hilborn. 2002. Sharks and tunas: Fisheries impacts on predators with contrasting life histories. Ecological Applications 12: 735-748. Sea Around Us. 2006. A global database on marine fisheries and ecosystems. World Wide Web site www.seaaroundus.org. Fisheries Centre, University of British Columbia, Vancouver (BC, Canada). Visited November 2nd 2006. Shannon, L.J., Moloney, C.L., Jarre, A., and J.G. Field. 2003. Trophic flows in the southern Benguela during the 1980s and 1990s. Journal of Marine Systems 39: 83- 116. Shannon, L.J., Christensen, V., and C.J. Walters. 2004. Modelling stock dynamics in the southern Benguela ecosystem for the period 1978 - 2002. Afrian Journal of Marine Science 26:179-196. Shelton, P.A., Warren, W.G., Stenson, G.B., and J.W. Lawson. 1997. Quantifying some of the major sources of uncertainty associated with estimates of harp seal prey consumption. Part II: Uncertainty in consumption estimates associated with population size, residency, energy requirements and diet. Journal of Northwest Atlantic Fishery Science 22: 303-315 Sigurjonsson, J. 1997. Whale resources in the North Atlantic and the concept of sustainability. pp 17-32 In Petursdottir, G. (ed.). Whaling in the North Atlantic - Economic and Political Perspectives. Proceedings of a conference held in Reykjavik on March 1st, 1997, organized by the Fisheries Research Institute and the High North Alliance. 168 Simmonds, M.P. and S.J. Issac. 2007. The impacts of climate change on marine mammals: Early signs of significant problems. Oryx 41: 19-26. Sindermann, C.J. (ed.). 1996. Ocean pollution: effects on living resources and humans. CRC press, Boca Raton, FL, USA. Skaret, G. and T.J. Pitcher, (in press). An ecosystem model of the Norwegian Sea and Barents Sea fitted to data. Fisheries Centre Research Reports. Smith, T.D. 1995. Interactions between marine mammals and fisheries: an unresolved problem for fisheries research, pp 527-536 In Bilx, A.S., Walloe, L. and 0. Ultang (eds.) Whales, seals, fish and man. Elsevier Science. Springer, A.M., Estes, J.A., van Vliet, G.B., Williams, T.M., Doak, D.F., Danner, E.M., Forney, K.A., and B. Pfister. 2003. Sequential megafaunal collapse in the North Pacific Ocean; an ongoing legacy of industrial whaling? Proceedings of the National Academy of Sciences 100: 12 223-12 228. Stenson, G.B., Rivest, L.-P., Hammill, M.O., Gosselin J.-F., and B. Sjare. 2003. Estimating pup production of harp seals, Phoca groenlandica, in the Northwest Atlantic. Marine Mammal Science 19: 141-160. Stenson, G.B., Hammill, M.O., and J.W. Lawson. 1997. Predation by harp seals in Atlantic Canada: Preliminary consumption estimates for Arctic cod, capelin and Atlantic cod. Journal of Northwest Atlantic Fishery Science 22: 137-154. Tjelmeland, S. and U.Lindstrem. 2005. An ecosystem element added to the assessment of Norwegian spring-spawning herring: implementing predation by minke whales. ICES Journal of Marine Science 62: 285-294. Trites, A.W. 2001. Food webs in the ocean: who eats whom and how much? pp 125-141 In Sinclair, M. and G. Valdimarsson (eds.). Responsible fisheries in the marine ecosystem. CAB International, Wallingford. Trites, A.W. 2002. Predator-prey relationships, pp 994-997 In Perrin, W.F., Wursig, B., and H.G.M. Thewissen (eds.). Encyclopedia of Marine Mammals. Academic Press, San Diego. Trites, A.W., Christensen, V., and D. Pauly. 1997. Competition between fisheries and marine mammals for prey and primary production in the Pacific Ocean. Journal of Northwest Atlantic Fishery Science 22: 173-187. 169 Trites, A.W., Livingston, P., Mackinson, S., Vasconcellos, M., Springer, A., and D. Pauly (eds.). 1999. Ecosystem change and the decline of marine mammals in the Eastern Bering Sea: Testing the ecosystem shift and commercial whaling hypotheses. Fisheries Centre Research Reports 7(1). Trzcinski, M.K., Mohn, R., and W.D. Bowen. 2006. Continued decline of an Atlantic cod population: How important is gray seal predation? Ecological Applications 16: 2276- 2292. Ulanowicz, R.E., and C.J. Puccia. 1990. Mixed trophic impacts in ecosystems. Coenoses 5: 7-16. Wallace, S.S. 1998. Changes in human exploitation of marine resources in British Columbia (pre-contact to present day), pp 58-64 In Pauly, D., Pitcher, T., Preikshot, D. and J. Hearne (eds.). Back to the future: reconstructing the strait of Geortia ecosystem. Fisheries Centre Research Reports 6(5). Walters, C. and S. Martell. 2004. Fisheries ecology and management. Princeton University Press, Princeton, New Jersey. Walters, C.J., Christensen, V., Martell, S.J., and J.F. Kitchell. 2005. Possible ecosystem impacts of applying MSY policies from single-species assessments. ICES Journal of Marine Science 62: 558-568. Whipple, S.J., Link, J.S., Garrison, L.P., and M.J. Fogarty. 2000. Models of predation and fishing mortality in aquatic ecosystems. Fish and Fisheries 1: 22-40. Whitehead, H. 2002. Estimates of the current global population size and historical trajectory for sperm whales. Marine Ecology Progress Series 242: 295-304. Williams, T.M., Estes, J.A., Doak, D.F., and A.M. Springer. 2004. Killer appetites: assessing the role of predators in ecological communities. Ecology 85: 3373-3384. Worm, B. and R.A. Myers. 2003. Meta-analysis of cod-shrimp interactions reveals top- down control in oceanic food webs. Ecology 84: 162-173. Yodzis, P. 1994. Predator-prey theory and management of multispecies fisheries. Ecological Applications 4: 51-58. Yodzis, P. 1998. Local trophodynamics and the interaction of marine mammals and fisheries in the Benguela ecosystem. Journal of Animal Ecology 67: 635-658. 170 Yodzis, P. 2001. Must top predators be culled for the sake of fisheries? Trends in Ecology & Evolution 16: 78-84. Zeller, D. and D. Pauly. 2005. Good news, bad news: global fisheries discards are declining, but so are total catches. Fish and Fisheries 6: 156-159. 171 C h a p t e r 5 A case s t u d y i n e c o s y s t e m m o d e l l i n g : C o n t r a s t i n g c h a n g e s b e t w e e n t h e n o r t h e r n a n d s o u t h e r n G u l f o f S t . L a w r e n c e e c o s y s t e m s a s s o c i a t e d w i t h t h e c o l l a p s e o f g r o u n d f i s h s t o c k s . 2 5.1 Introduction In response to a steep decline in Atlantic cod (Gadus morhud) abundance in the late 1980s and early 1990s, a moratorium on commercial fishing in a number of stock management areas was imposed by the Canadian government in order to promote recovery of the depleted populations. The southern Gulf of St. Lawrence (SGSL) was closed to directed cod fishing in September 1993, followed by the closure of the northern Gulf of St. Lawrence (NGSL) fishery in January 1994 (Anonymous 1994). The NGSL moratorium was lifted in May 1997, and for the SGSL in 1999. During the three-year period that the moratorium was in place, differing effects were observed on cod stocks in the two regions. For example, a modest improvement in mature cod abundance was observed in the NGSL (Frechet et al. 2005), while spawner abundance remained unchanged in the SGSL (Chouinard et al. 2003). Since the re-opening of directed cod fisheries in the two ecosystems, mature cod biomass has remained roughly constant in the NGSL (Frechet et al. 2005), but has declined by nearly 20% in the SGSL (Chouinard et al. 2003). In this study, mass-balance models (Ecopath) were developed for NGSL and SGSL ecosystems, for the periods before (mid-1980s) and after (mid-1990s) the collapse 2 A version of this chapter has been submitted to Deep Sea Research II as Morissette, L . , Castonguay, M . . Savenkoff, C , Swain, D.P., Chabot, D., Bourdages, H . , Hammill, M . O . and J . M . Hanson. Contrasting changes between the northern and southern Gulf of" St. Lawrence ecosystems associated with the collapse of groundfish stocks. 172 of demersal fish stocks in the two areas. The objective was to determine if significant changes in ecosystem structure had occurred between the two time periods that might have contributed to the failure of the cod stocks to recover in the 1990s. A major change in abundance of a species inevitably has consequences for its predators and prey. There is a need to integrate interactions between species (competition and predation) and the exploitation by fishing in a multispecific approach (Christensen et al. 2000; Morissette 2001). Moreover, changes in ecosystems are often studied by comparing models representing two different periods, without including parameter uncertainty and the ensuing uncertainty in the interpretation of the results. Here, we use a new approach towards analyzing the changes between two ecosystems over a period of 10 years by combining a multispecies approach with an in-depth analysis of uncertainty. Materials and Methods 5.2.1 Study areas The Gulf of St. Lawrence forms one of the most important estuarine shelves in the world (Therriault 1991). It is the outlet of the St. Lawrence River into the Atlantic Ocean via the Strait of Belle Isle, in the north between Newfoundland and Labrador, and Cabot Strait in the south (Figure 5.1). At its widest, the Gulf extends roughly 500 km from north to south. The northern and southern parts have very different bathymetric characteristics and, to some degree, different faunas. The SGSL (Northwest Atlantic Fishery Organization [NAFO] division 4T) is a relatively shallow shelf (generally < 60 m deep; maximum depth 130 m), with a total area of 64,075 km2, and a permanent cold water layer that is in contact with the sediments in water 35 to 100 m deep (Gilbert and Pettigrew 1997) (Figure 5.1). In contrast, the NGSL (NAFO divisions 4RS) is characterized by channels as deep as 500 m: the Laurentian Channel, which extends nearly 1,000 km from the St. Lawrence Estuary to the Atlantic Ocean, in addition to the Esquiman Channel and the Anticosti Channel (Figure 5.1). In the NGSL, the study area covered 103,812 km2. Depths shallower than 37 m were not included in the NGSL model. In the SGSL model, we defined the study area to include depths between 15 and the 200 m depth contour, where the deeper water included part of the sharp drop into the Laurentian Channel. The nearshore region was not included in the models because exchanges between infra-littoral 173 and pelagic zones are not well sampled by government scientific survey trawlers; consequently, the community structure of the shallow depth zones is poorly understood. Figure 5.1. The regions of the northern (North Atlantic Fisheries Organisation divisions 4R and 4S) and southern (NAFO division 4T) Gulf of St. Lawrence. 5.2.2 Structure of the models NGSL and SGSL models, for both periods, were divided into 32 and 30 trophic groups, respectively (Table 5.1). Species were grouped based on their commercial significance and importance as predators or prey. We distinguished five marine mammal groups, one seabird group, 14 (SGSL) or 16 (NGSL) fish groups, six invertebrate groups, two zooplankton groups, one phytoplankton group, and one detritus group (Table 5.1). Some compartments such as large pelagic feeders and large demersal feeders were aggregated on the basis of similarity of size and ecological role of their species. Atlantic 174 cod, Greenland halibut (Reinhardtius hippoglossoides) (in NGSL), and American plaice (Hippoglossoides platessoides) (in SGSL) were, for some models, separated into large and small individuals based on diet, age/size at first capture, and age/size at maturity. Smaller animals prey mainly on invertebrates whereas larger animals prey mainly on fish. These changes tend to occur gradually with increasing length, but for these models it was assumed that the change occurs at 35 cm for Atlantic cod (Lilly 1991), at 40 cm for Greenland halibut (Bowering and Lilly 1992), and at 35 cm for American plaice (Pitt 1973). Bacteria were assumed part of the detritus compartment. 175 Table 5.1. Trophic groups used in the models of the northern (NGSL) and southern (SGSL) Gulf of St. Lawrence ecosystems, for the 1985-1987 and the 1994-1996 periods. Group Name Main species Comments Cetaceans Balaenoptera physalus, Balaenoptera acutorostrata, Megaptera novaeangliae, Phocoena phocoena, Lagenorhynchus acutus, Lagenorhynchus albirostris Harp seals Pagophilus groenlandica Hooded seals Cystophora cristata Grey seals Halichoerus grypus Harbour seals Phoca vitulina Seabirds Phalacrocorax carbo, Phalacrocorax auritus, Larus delwarensis, L. argentatus, L. marinus, Sterna hirundo, S. paradisaea, Cepphus grylle, Oceanodroma leucorhoa, Moras bassanus, Rissa tridactyla, Uria aalge, Alca torda, Fratercula arctica Large Atlantic cod (> 35 cm) Gadus morhua Small Atlantic cod (< 35 cm) Gadus morhua Large Greenland halibut (> 40 cm) Reinhardtius hippoglossoides Large and small Greenland halibuts were aggregated as Greenland halibut for the SGSL models Small Greenland halibut (< 40 cm) Reinhardtius hippoglossoides Large American plaice (> 35 cm) Hippoglossoides platessoides Large and small plaice were aggregated as American plaice for the NGSL models Small American plaice (< 35 cm) Hippoglossoides platessoides Flounders Limanda ferruginea, Glyptocephalus cynoglossus, Pseudopleuronectes americanus Skates Amblyraja radiata, Malacoraja senta, Leucoraja ocellata Redfish Sebastes mentella, Sebastes fasciatus Large demersal feeders Urophycis tenuis, Melanogrammus aegelfinnus, Centroscyllium fabricii, Anarhichas spp., Cyclopterus lumpus, lycodes spp., Macrouridae, Zoarcidae, Lophius americanus, Hippoglossus hippoglossus Small demersal feeders Myoxocephallus sp., Tautogolabrus adspersus, Macrozoarces americanus, juvenile large demersal feeders Capelin Mallotus villosus Sand lance Ammodytes spp. Included in the planktivorous small pelagic feeder group for the SGSL models Arctic cod Boreogadus saida Included in the capelin group for the SGSL models Large pelagic feeders Squalus acanthias, Pollachius virens, Merluccius bilinearis, Cetorhinus maximus, Thunnus thynnus 176 Table 5.1:Cont. Group Name Main species Comments Piscivorous small pelagic feeders Scomber scombrus, piscivorous myctophids and other mesopelagic species, Illex illecebrosus, piscivorous juvenile large pelagic feeders Planktivorous small pelagic feeders Clupea harengus harengus, planktivorous myctophids and other mesopelagic species, Scomberesox saurus, Gonatus sp., planktivorous juvenile large pelagic feeders Shrimp Pandalus borealis, P. montagui, Argis dentata, Eualus macilentus and E. gaimardi Large crustaceans Chionoecetes opUio, other non-commercial crab species (Hyas spp) Echinoderms Echinarachnius parma, Stronglyocentrotus pallidus, Ophiura robusta Molluscs Mesodesma deauratum, Cyrtodaria siliqua Polychaetes Exogene hebes, and other polychaetes species Other benthic invertebrates Miscellaneous crustaceans, nematodes, other meiofauna Large zooplankton (> 5 mm) Euphausiids, chaetognaths, hyperiid amphipods, cnidarians and ctenophores (jellyfish), mysids, tunicates >5 mm, icthyoplankton Small zooplankton (<5 mm) Copepods (mainly Calanus finmarchicus, C. hyperboreus and Oithona similis), tunicates < 5 mm, meroplankton Phytoplankton Diatom species, and a mixture of autotrophic and mixotrophic organisms including Cryptophytes, dinoflagellates, Prasinophytes, mixotrophic Stombidium spp., and Prymnesiophytes Detritus 5.2.3 Data used for the models Information on species, biomass, production, consumption, diet and catch for each trophic group of NGSL arid SGSL models was obtained from various sources given in Morissette et al. (2003) and Savenkoff et al. (2004 a, b). Biomass estimates for demersal fishes were obtained from annual bottom trawl surveys in the two ecosystems. For both NGSL and SGSL models, biomass estimates from these surveys were adjusted to total biomass based on the catchability coefficients given by Harley and Myers (2001) and Savenkoff et al. (2004 a, b). Overall, considerable effort was expended to obtain biomass, production, consumption, diet, and catch data that came from the study areas during the periods of interest. However, information on several groups (e.g., forage species, benthic invertebrates, and zooplankton) was sparse or non-existent for the areas and periods studied; in these situations, data were obtained either from the literature or from different period of the same area (Morissette et al. 2003; Savenkoff et al. 2004 a, b). Ill 5.2.4 Modelling approaches We used 31 balanced solutions obtained using inverse methodology for each ecosystem and period. These solutions corresponded to 31 random perturbations (including a response without perturbation) on each model input to a maximum of its standard deviation. The inverse approach was useful to obtain a first balanced solution by finding the solution that minimizes (objective least-square criterion) both the sum of squared flows (thus the total sum of flows through the food web) and the sum of squared residual errors (minimizes the imbalances between inputs and outputs) consistent with the constraints (Vezina and Piatt 1988). A more complete description of these balanced scenarios is given in Savenkoff et al. (in press a) for the NGSL and in Savenkoff et al. (in press b) for the SGSL. Each of these 31 balanced inverse solutions were then transposed into Ecopath software to estimate mortality (due to fishing, predators, and other sources), the basic emergent properties and network analysis indices for the two time periods, and estimates of the associated uncertainties. The emergent properties of the two systems were compared using a two-tailed t- test, with a significance level of 0.05. These t-tests verified the hypothesis that there was a significant difference, for a given emergent property and ecosystem, between the 31 balanced solutions obtained for the 1980s and the 1990s. Trophic levels of catches were also compared with a two-tailed t-test. Each value is presented with 95% confidence intervals around the mean (CI); if these CI do not overlap, the difference between the two values of a parameter is significant. The t-tests are fairly robust to lack of normality or to heteroscedasticity (Zar 1998). However, because our data sometimes departed markedly from the basic assumptions of the t-test, we also used the non-parametric test of Wilcoxon (also known as Mann-Whitney U test). The results were the same for most comparisons, but in case of conflict, we retained the non-parametric result. Only the latter are reported in the text or tables. Assuming an equilibrium or steady state, model estimates of biomass, production, catch in fisheries and consumption by predators were used to derive estimates of total mortality (Z), and its components (fishing mortality F, predation mortality M2, and other mortality MO). Note that, in contrast to the usual notation in fisheries science (e.g., Ricker 178 1975), we use these symbols to refer to annual rates rather than instantaneous rates. These indices were compared for some important groups, and between similar trophic groups of each ecosystem. Four main trophic groups were studied here: (1) marine mammals and seabirds; (2) piscivorous fish including all fish species except forage species; (3) forage fish including capelin, sand lance, Arctic cod, and small pelagic fish (mostly herring and mackerel); and (4) shrimp and large crustaceans. The systems' emergent properties were also compared using emergent properties estimates and network analysis indices of the two models for the two time periods calculated by the Ecopath with Ecosim software (Christensen et al. 2000). The indices used include the sum of all consumption, exports, respiration, production, and flow to detritus (reported as t-km'̂ year"1), the mean trophic level of the catch, gross efficiency of the catch (the catch divided by the net primary production; dimensionless), and the net system production (see Christensen [1995] and Christensen et al. [2000] for a description of the emergent properties indices). Network analysis indices were also analysed in terms of ascendancy (A), which is the product of the total systems throughput and the average mutual information of the flow structure (Ulanowicz and Puccia 1990), development capacity (Q, which is a measure of the network's potential for competitive advantage over other real or presumed network configurations and is the upper bound to the ascendancy (Ulanowicz and Puccia 1990), and overhead (O), which is complementary to ascendancy and indicates the inefficiency and redundant degrees of freedom in the system (Ulanowicz 2000). Ecopath methods also allow the calculation of many indices that can describe the complexity in the community structure (Christensen 1995). Attributes were selected according to a previous study by Vasconcellos et al. (1997), who identified ascendancy, connectance index, and system omnivory index as representative of Odum's attributes of ecosystem complexity (Odum 1971). Within these many indices, the system omnivory index (SOI) is the one that we favoured in our analyses. Pimm (1982) defined an omnivore as a species that feeds on more than one trophic level. The System Omnivory Index (SOT) is defined as the average Omnivory Index (OT) of all consumers weighted by the logarithm of the biomass consumed by each consumer (Christensen et al. 2000). In other words, the system omnivory index is a measure of how the feeding interactions are distributed 179 between trophic levels. This index is used to characterize the extent to which a system displays web-like features (Christensen et al. 2000). Results 5.3.1 Biomass The total catchability-corrected biomass of exploited fish and invertebrate species decreased by about 50% in NGSL (Figure 5.2) and about 30% in SGSL between the mid- 1980s and the mid-1990s (Figure 5.3); on the other hand there was a significant increase in marine mammals biomass between the 1985-1987 and the 1994-1996 periods. There was a significant decrease in the biomass of piscivorous fish in both NGSL and SGSL. Forage fish biomass also showed a downward trend, though the decline in their biomass was only slight in the SGSL. The total biomass of the shrimp and large crustaceans group changed little from the mid-1980s to the mid-1990s, except for shrimps, which doubled their biomass between these two periods in the SGSL. 40 n 30 H Marine mammals and seabirds Piscivorous fish Forage fish Shrimp & Large crustaceans Figure 5.2. Biomass change for marine mammals and seabirds, piscivorous fish, forage fish, and shrimp and large crustaceans in the northern Gulf of St. Lawrence (NGSL) from the 1985-1987 (grey histogram; B = 48±24 t-km"2) to the 1994-1996 (white histogram; B = 23±23 t-km"2) periods. B = total biomass for the four 180 groups, or all commercial species. Error bars indicate the 95% confidence intervals of biomass for the 31 balanced scenarios. c S r-^-rk o Marine mammals and seabirds Piscivorous fish Forage fish Shrimp & Large crustaceans Figure 5.3. Biomass change for marine mammals and seabirds, piscivorous fish, forage fish, and shrimp and large crustaceans in the southern Gulf of St. Lawrence (SGSL) from the 1985-1987 (B = 32±3 t-km"2) to the 1994-1996 (B = 27±6 t-km"2) periods. B = total biomass for the four groups, or all commercial species. Error bars indicate the 95% confidence intervals of biomass for the 31 balanced scenarios. 5.3.2 Consumption The major changes observed in the biomass structure also affected the total consumption (Q) of the different trophic groups on all their prey. In the NGSL, the total annual consumption by marine mammals significantly increased from the 1980s to the 1990s (Figure 5.4), while total consumption by piscivorous and forage fish, shrimp and large crustaceans showed a sharp significant decline (overall decrease ~ 80%). The most important proportional decrease in consumption was attributed to the piscivorous fish group, for which Q decreased from 12.7 to 2.0 Mcm"2'year"' (84%), although the absolute 2 1 decrease was much greater for forage fishes (a decrease of 27.4 t4xm" -year"). In the SGSL, total consumption by piscivorous fish also decreased significantly from 11.3 to 6.3 t-knf̂ year"1 (44%) (Figure 5.5). The consumption by marine mammals changed little 181 between the two time periods, while the consumption by both forage fish and shrimp and large crustaceans groups increased significantly. In both ecosystems, even in the 1990s, total consumption (i.e., predation) by marine mammals was always much smaller than consumption by any other trophic group examined (Figures 5.4 and 5.5). 50 i 40 E u 41 a. s o 30 .2 2 0 H a. E 3 f2 10 Marine mammals & seabirds Piscivorous fish Forage fish Shrimp & Large crustaceans Figure 5.4. Total consumption (Q) on all prey by marine mammals, piscivorous fish, forage fish, shrimp and large crustaceans in the northern Gulf of St. Lawrence (NGSL) in the 1980s (grey) and the 1990s (white). Error bars indicate the 95% confidence intervals of total consumption for the 31 balanced scenarios. 182 20 -i Marine mammals & seabirds Piscivorous fish Forage fish Shrimp & Large crustaceans Figure 5.5. Total consumption (Q) on all prey by marine mammals, piscivorous fish, forage fish, shrimp and large crustaceans in the southern Gulf of St. Lawrence (SGSL) in the 1980s (grey) and the 1990s (white). Error bars indicate the 95% confidence intervals of total consumption for the 31 balanced scenarios. In the NGSL, the overall consumption by all seal species increased significantly from the 1980s to the 1990s on all prey species except small cod for which consumption by seals significantly decreased, and for flounders and large crustaceans, where no significant difference was observed (Figure 5.6). In the SGSL, seal consumption significantly increased for half the prey groups (Greenland halibut, flounders, redfish, and capelin) (Figure 5.6). The increased consumption of capelin in the SGSL greatly exceeded the change in consumption of any other group. Overall, seals consumed more fish prey in the SGSL model than in the NGSL model, except for deep-water species such as Greenland halibut and redfish, consistent with the fact that they are not a major component of the SGSL ecosystem. Furthermore, seals consumed much more large cod in the SGSL than in the NGSL during both periods. In contrast, there was a sharp and significant decline in the consumption of small cod by seals in the SGSL in the 1990s model. Despite 183 increasing shrimp biomass, consumption of shrimp by seals actually dropped significantly in the SGSL model for the mid-1990s. • 4RS 1980s • 4RS 1990s • 4T 1980s • 4T 1990s Large cod Small cod Greenland American Flounders Redfish Capelin Shrimp Halibut Plaice Large crustaceans Figure 5.6. Seal consumption on various prey in the northern (NGSL) and southern (SGSL) Gulf of St. Lawrence in the 1980s and the 1990s. Error bars indicate the 95% confidence intervals of seals consumption for the 31 balanced scenarios. Q on capelin in N G S L 1990s = 0.47 ±0.12 t-km^-year"1, Q on capelin in SGSL 1990s = 1.27 ± 0.13 t-km"2-year''. 5.3.3 Mortality Three sources of mortality are distinguished by the Ecopath models: predation mortality (M2), fishing mortality (F), and other mortality (disease, other natural causes of death, and unexplained mortality, MO). At equilibrium, total mortality (Z) should be equivalent to the production to biomass ratio of each group. In the NGSL, small Greenland halibut had the highest significant increase in total mortality from the 1980s to the 1990s, while shrimp had the largest significant decrease (Table 5.2, Figure 5.7). 184 Table 5.2. Changes in predation mortality (M2), fishing mortality (F) and other mortality (MO) from the 1980s to the 1990s, for all trophic groups in the northern Gulf of St. Lawrence model. M2 F M0 1980s 1990s P (t/test) 1980s 1990s p (t/test) 1980s 1990s P ({/test) Species mean mean a = 0.05 mean mean a = 0.05 mean mean a = 0.05 Cetaceans - - - 0.003 0.002 <0.01 0.067 0.062 <0.01 Harp seals - - - 0.078 0.053 <0.01 0.024 0.010 <0.01 Hooded seals - - - 0.030 0.005 <0.01 0.042 0.085 <0.01 Grey seals • - - - 0.068 0.000 <0.01 0.038 0.045 <0.01 Harbour seals - - - 0.000 0.000 - 0.069 0.051 <0.01 Seabirds - - - 0.118 0.085 <0.01 0.240 0.289 0.02 Large cod 0.018 0.341 <0.01 0.349 0.019 <0.01 0.391 0.136 <0.01 Small cod 0.619 0.747 <0.01 0.001 0.000 <0.01 0.198 0.030 <0.01 L. Greenland halibut 0.086 0.099 0.46 0.098 0.048 <0.01 0.037 0.025 0.09 S. Greenland halibut 0.236 0.592 <0.01 0.000 0.000 - 0.059 0.103 0.11 American plaice 0.267 0.285 0.27 0.026 0.003 <0.01 0.039 0.032 0.24 Flounders 0.243 0.252 0.07 0.014 0.026 <0.01 0.029 0.035 0.06 Skates 0.226 0.250 0.80 0.000 0.008 <0.01 0.045 0.029 0.52 Redfish 0.158 0.181 <0.01 0.029 0.023 0.13 0.060 0.031 <0.01 L. demersal fish 0.113 0.164 <0.0T 0.010 0.008 <0.01 0.028 0.013 0.03 S. demersal fish 0.423 0.278 <0.01 0.000 0.000 - 0.029 0.021 <0.01 Capelin 0.730 0.950 <0.01 0.002 0.013 <0.01 0.178 0.100 0.40 Sand lance 1.043 0.588 <0.01 0.000 0.000 - 0.250 0.035 <0.01 Arctic cod 0.431 0.628 <0.01 0.000 0.000 - 0.053 0.127 <0.01 L. pelagic fish 0.131 0.235 <0.01 0.056 0.001 <0.01 0.059 0.020 <0.01 S. pisciv. pel. fish 0.240 0.318 <0.01 0.007 0.060 <0.01 0.051 0.018 <0.01 S. plankt. pel. fish 0.315 0.316 0.21 0.111 0.144 <0.01 0.040 0.026 .0.18 Shrimp 1.628 0.611 <0.01 0.116 0.166 <0.01 0.084 0.055 <0.01 Large crustaceans 0.194 0.192 0.04 0.053 0.083 <0.01 0.065 0.021 <0.01 Echinoderms 0.020 0.003 <0.01 0.000 0.000 - 0.321 0.390 <0.01 Molluscs 0.021 0.012 <0.01 0.001 0.000 <0.01 0.615 0.712 <0.01 Polychaetes 0.969 0.987 0.80 0.000 0.000 - 1.029 1.011 0.59 Other bent, invert. 0.401 0.220 <0.01 0.000 0.000 - 0.974 1.056 0.24 Large zooplankton 2.728 3.006 0.10 0.000 0.000 - 0.295 0.850 0.03 Small zooplankton 3.938 2.951 <0.01 0.000 0.000 - 1.011 2.496 <0.01 Phytoplankton 75.733 58.821 <0.01 0.000 0.000 - 29.302 34.530 0.96 185 Figure 5.7. Breakdown of total mortality (Z) into fishing (F), predation (M2) and unexplained (MO) mortalities, in annual rate (year'1), for commercial species of the northern Gulf of St. Lawrence (NGSL). Plain bars represent the 1980s while patterned bars represent the 1990s results. For clarity, 95% CI were excluded. In the SGSL, small demersal fish had the highest significant increase in total mortality, while large cod had the steepest significant decline in total mortality (Table 5.3, Figure 5.8). 186 Table 5.3. Changes in predation mortality (M2), fishing mortality (F) and other mortality (MO) from the 1980s to the 1990s, for all trophic groups in the southern Gulf of St. Lawrence model. M2 P F p M0 P 1980s 1990s (U test) 1980s 1990s (I/test) 1980s 1990s (f/test) Species mean mean a = 0.05 mean mean a = 0.05 mean mean a = 0.05 Cetaceans - - - 0.022 0.013 <0.01 0.048 0.075 <0.01 Harp seals - - - 0.004 0.034 <0.01 0.065 0.041 <0.01 Hooded seals - - . - 0.000 0.000 - 0.109 0.085 <0.01 Grey seals - - - 0.029 0.004 <0.01 0.101 0.112 <0.01 Harbour seals - - - 0.000 0.000 - 0.119 0.109 0.10 Seabirds - - - 0.097 0.119 0.12 0.246 0.258 0.85 Large cod 0.060 0.154 <0.01 0.318 0.024 <0.01 0.327 0.152 <0.01 Small cod 0.568 0.487 <0.01 0.000 0.000 - 0.051 0.063 0.16 Greenland Halibut 0.473 0.491 0.02 0.318 0.184 <0.01 0.189 0.120 0.03 S. American plaice 0.289 0.349 0.02 0.000 0.000 - 0.191 0.119 <0.01 L. American plaice 0.158 0.195 0.13 0.293 0.136 <0.01 0.066 0.028 <0.01 Flounders 0.236 0.211 0.01 0.028 0.012 <0.01 0.023 0.028 0.26 Skates 0.243 0.235 0.44 0.000 0.012 <0.01 0.031 0.046 <0.01 Redfish 0.209 0.223 0.36 0.104 0.049 <0.01 0.028 0.047 <0.01 L. demersal fish 0.188 0.401 <0.01 0.249 0.010 <0.01 0.020 0.026 0.01 S. demersal fish 0.298 0.779 <0.01 0.000 0.000 - 0.014 0.038 <0.01 Capelin 0.870 0.678 <0.01 0.005 0.001 <0.01 0.045 0.252 <0.01 L. pelagic fish 0.234 0.218 0.87 0.110 0.115 0.37 0.059 0.053 0.98 S. pisciv. pelagic fish 0.225 0.245 0.11 0.036 0.169 <0.01 0.072 0.057 0.44 S. plankt. pelagic fish 0.360 0.255 <0.01 0.157 0.270 <0.01 0.036 0.077 <0.01 Shrimp 3.172 2.639 0.27 0.005 0.010 0.01 0.527 0.685 0.97 Large crustaceans 0.204 0.191 0.03 0.137 0.140 0.85 0.019 0.034 0.30 Echinoderms 0.031 0.023 0.05 0.000 0.000 - 0.336 0.360 0.45 Molluscs 0.084 0.097 0.99 0.000 0.000 0.29 0.609 0.543 0.60 Polychaetes 1.838 1.728 0.11 0.000 0.000 - 0.242 0.241 0.18 Other bent, invert. 0.945 0.706 0.04 0.000 0.000 - 0.476 0.578 0.87 Large zooplankton 5.237 2.939 <0.01 0.000 0.000 - 0.687 0.970 0.22 Small zooplankton 3.214 3.994 <0.01 0.000 0.000 - 1.954 1.107 <0.01 Phytoplankton 74.568 62.756 0.13 0.000 0.000 - 23.227 35.040 0.13 187 1.20 A 1.00 A 0.80 A 0.60 0.40 0.20 0.00 9 F 1980's F 1990's M2 1980's M2 1990's MO 1980's MO 1990's Shrimp 1980's: Z=3.71 M2-3.17 M0=0.53 i s B s / / S3 I ft, 09' 6, to' Figure 5.8. Breakdown of total mortality (Z) into fishing (F), predation (M2) and unexplained (M0) mortalities, in annual rate (year"1), for commercial species of the southern Gulf of St. Lawrence (SGSL). Plain bars represent the 1980s while patterned bars represent the 1990s results. For clarity, 95% CI were excluded. For both models in the 1990s, fishing mortality of the major piscivorous fish was significantly reduced (Tables 5.2 and 5.3), while it increased for species of lower trophic levels such as flounders (only in the NGSL), skates, capelin (only in the NGSL), large pelagic fish (only in the NGSL), small pelagic fish, shrimp and large crustaceans (only in NGSL). Overall fishing mortality decreased significantly in both ecosystems (from 1.17 to 0.75 year"1 in NGSL and from 1.91 to 1.30 year"1 in the SGSL). Also, predation mortality (excluding predation from benthic invertebrates and plankton) significantly decreased from 7.10 to 7.03 year"1 in the NGSL, and remained approximately the same from 7.79 to 7.75 year"1 in the SGSL. Predation mortality significantly increased from the 1980s to the 1990s for some species such as cod (large and small), small Greenland halibut, redfish, large demersal fish, capelin, Arctic cod, large pelagic fish, and piscivorous small pelagic fish in the NGSL, and for large cod, Greenland halibut, small American plaice, and demersal fish (large and small) and small zooplankton in the SGSL. Overall, other 188 mortality (excluding MO on invertebrates and plankton) significantly decreased from the 1980s (MO = 2.17 year"1) to the 1990s (MO = 1.40 year"1) in the NGSL models, but was similar between the two periods in the SGSL models (2.39 year"1 in the 1980s and 2.51 year"1 in the 1990s). The most important level of MO in proportion to total mortality was found for large cod in NGSL (51% in the 1980s. However MO declined to 34% when 20% of misreported catches were included [Savenkoff et al. in press a]. In the 1990s, MO was 28%). In the SGSL, the most important level of MO in proportion to total mortality was also found for large cod (46% in the 1980s, 46% in the 1990s) (Tables 5.2 and 5.3, Figures 5.7 and 5.8). In the specific case of large cod, total mortality was similar in NGSL and SGSL in the 1980s (Z = 0.76 year"1 in the NGSL and 0.71 year"1 in the SGSL), and decreased to lower values in 1990s (Z = 0.50 year"1 in NGSL and 0.32 year"' in the SGSL) (Figures 5.7 and 5.8). Fishing mortality was also similar in both ecosystems in the 1980s (F = 0.35 year"1 in NGSL and 0.32 year"1 in the SGSL) and significantly decreased, to a value of 0.02 year"1 in the 1990s in both ecosystems. Predation mortality on large cod, M2, significantly increased in the NGSL, from 0.02 to 0.34 year"1 (Figure 5.7) but this increase was smaller (but still significant) in the SGSL (from 0.06 to 0.15 year"1) (Figure 5.8). Finally, other mortality M0 decreased significantly from the 1980s to the 1990s, in both ecosystems (from 0.39 to 0.14 year"1 in NGSL, and from 0.33 to 0.15 year"1 in the SGSL). 5.3.4 Trophic levels Among marine mammals, the trophic levels (TL) estimated by NGSL models significantly decreased from the 1980s to the 1990s for harp seals, but significantly increased for harbour seals. However, it remained about the same for cetaceans, hooded seals, and grey seals (Table 5.4). In the SGSL, trophic levels increased significantly for cetaceans, harp seals, and harbour seals but remained approximately the same for grey seals and hooded seals (Table 5.5). The most important decreases in TL were observed for large demersal fish in NGSL and skates in SGSL. The most important increases in TL were observed in piscivorous small pelagic fish in both ecosystems. Trophic levels did not change for skates, redfish, arctic cod, large crustaceans, benthic invertebrates, and small zooplankton in the NGSL, or for Greenland halibut, small American plaice, flounders, 189 redfish, demersal fish, capelin, shrimp, large crustaceans, and benthic invertebrates in the SGSL. Table 5.4. Average trophic level of the 30 balanced scenarios for the Northern Gulf of St. Lawrence in the 1980s and the 1990s. Species Average TL 1980s Average TL 1990s % change P Cetacea AM 4.14 0.4 0.202 Harp seals 4.37 4.14 -5.6 0.000 Hooded seals 4.71 4.72 0.2 0.702 Grey seals 4.55 4.55 -0.2 0.823 Harbour seals 4.28 4.37 2.0 0.000 Seabirds 4.15 4.22 1.6 0.001 Large cod 4.20 4.08 -2.9 0.000 Small cod 3.65 3.83 4.8 0.000 L. Greenland halibut 4.08 4.18 2.5 0.000 S. Greenland halibut 4.08 4.03 -1.3 0.012 American plaice 3.53 3.13 -12.6 0.000 Flounders 3.18 3.12 -1.7 0.005 Skates 4.07 4.02 -1.1 0.167 Redfish 3.61 3.65 0.9 0.012 L. demersals 3.73 3.29 -13.1 0.000 S. demersals 3.12 3.33 6.3 0.000 Capelin 3.27 3.33 1.7 0.000 Sand lance 3.15 3.21 1.9 0.000 Arctic cod 3.26 3.30 1.1 0.274 L. pelagics 3.94 3.66 -7.8 0.000 S. pisciv. pelagics 3.18 3.43 7.3 0.000 S. plankt. pelagics 3.14 3.32 5.3 0.000 Shrimp 2.45 2.58 4.9 0.006 Large Crustacea 3.04 3.02 -0.8 0.434 Echinoderms 2.00 2.00 0.0 - Molluscs 2.00 2.00 0.0 - Polychaetes 2.12 2.12 0.0 0.795 Other bent, invert. 2.00 2.00 0.0 - Large zooplankton 2.44 2.50 2.4 0.001 Small zooplankton 2.10 2.13 1.1 0.015 Phytoplankton 1.00 1.00 0.0 - Detritus 1.00 1.00 0.0 - 190 Table 5.5. Average trophic level of the 30 balanced scenarios for the southern Gulf of St. Lawrence in the 1980s and the 1990s. Average Average % Species TL 1980s TL 1990s change P Cetacea 4.23 4.35 2.8 0.000 Harp seal 4.01 4.15 3.3 0.000 Hooded seal 4.34 4.34 0.0 0.900 Grey seal 4.42 4.40 -0.4 0.223 Harbour seal 4.29 4.43 3.0 0.000 Seabirds 4.15 4.31 3.6 0.000 Large cod 3.50 3.69 5.2 0.000 Small cod 3.34 3.45 3.2 0.000 Greenland halibut 4.02 3.97 -1.4 0.053 Sm. american plaice 3.25 3.28 1.0 0.186 L. American plaice 3.14 3.22 2.4 0.034 Flounders 3.06 3.06 -0.3 0.244 Skates 4.25 4.01 -6.0 0.000 Redfish 3.62 3.62 -0.1 0.872 L. demersal feeders 4.20 4.23 0.6 0.191 Sm. demersal feeders 3.98 3.97 -0.1 0.887 Capelin 3.14 3.16 0.6 0.225 Large pelagic feeders 4.00 4.23 5.4 0.000 Pise. sm. pel. feeders 3.29 3.55 7.4 0.000 Plank, sm. pel. feed. 3.22 3.28 1.7 0.001 Shrimp 2.45 2.50 2.1 0.392 Large crustaceans 2.83 2.79 -1.7 0.448 Echinoderms 2.00 2.00 0.0 - Molluscs 2.00 2.00 0.0 - Polychaetes 2.25 2.22 -1.4 0.213 OBI 2.00 2.00 0.0 - Large zooplankton 2.43 2.50 2.6 0.001 Small zooplankton 2.12 2.14 1.0 0.218 Phytoplankton 1.00 1.00 0.0 - Detritus 1.00 1.00 0.0 - In both ecosystems, the mean trophic level of fishery landings over a 10-year period decreased significantly, from 3.30 to 3.08 in the NGSL, and from 3.33 to 3.21 in the SGSL (Figure 5.9). In the NGSL, landings changed from a predominance of large cod, redfish and planktivorous small pelagic fish (mainly herring) to a catch composition dominated by herring, shrimp and large crustaceans ten years later. In the SGSL, the fishery changed from a high proportion of large cod, planktivorous small pelagic fish (mainly herring), and large crustaceans to catches that were dominated by herring, large crustaceans, and piscivorous small pelagic fish (mainly mackerel) in the 1990s. Moreover, 191 the TL of landings in the 1980s was marginally higher in the SGSL than the NGSL (by about 1%), but this difference increased to 4% and became significant in the 1990s. 3.65 T 1980s 1990s Figure 5.9. Mean trophic level df the catch change over a decade in the southern (SGSL, in grey) and northern (NGSL, in white) Gulf of St. Lawrence ecosystems. Error bars indicate the 95% confidence intervals of trophic level for the 31 balanced scenarios. 5.3.5 Emergent properties of ecosystems Most of the emergent properties showed a significant difference from the 1980s to the 1990s (Table 5.6). The sum of all production, total system throughput, the ratio of total primary production to total production, ascendancy, overhead, capacity and throughput cycled (excluding detritus) significantly decreased in the NGSL, while they show a significant opposite change for the SGSL. In contrast, the ratio of total biomass to total throughput significantly increased in the NGSL while it decreased significantly in the SGSL. The mean trophic level of the catch, gross efficiency, total catches and connectance index decreased significantly for both ecosystems. There was also a significant change in the sum of all consumption in the NGSL, in addition to a significant change in the sum of all respiratory flows, sum of all flows into detritus, calculated net primary production, and 192 the ratio of total primary production to total respiration in the SGSL (Table 5.6). Most emergent properties that showed a significant change are indicators of a more productive ecosystem in the NGSL in the 1980s compared to the 1990s while, surprisingly, we see the opposite picture in SGSL (see Christensen et al. [2000] for a complete description of the indices). Table 5.6. Emergent properties of the 1980s and 1990s models for the northern and southern Gulf of St. Lawrence. NGSL NGSL P SGSL SGSL P 1980s 1990s (17 test) 1980s 1990s ([/test) Parameters Mean Mean a = 0.05 Mean Mean a = 0.05 Sum of all consumption3 1861 1553 0.01 1832 2072 0.06 Sum of all exports3 12 10 0.77 14 14 0.06 Sum of all respiratory flows3 1297 1118 0.07 1293 1544 0.02 Sum of all flows into detritus3 1206 1220 0.28 1239 1504 O.01 Sum of all productiona 1788 1571 0.01 1780 2059 0.02 Total system throughput3 4377 3901 0.01 4378 5134 0.01 Mean trophic level of the catch 3.303 3.076 <0.01 3.325 3.212 <0.01 Gross efficiency (catch/net p.p.) 0.0023 0.0004 <0.01 0.0023 0.0014 <0.01 Calculated total net primary production3 1310 1128 0.07 1307 1558 0.02 Total primary production/total respiration 1.010 1.009 0.96 1.011 1.009 0.01 Net system production3 12.4 10.5 0.77 14.1 13.7 0.06 Total primary production/total biomass 4.6 4.2 0.01 4.7 5.1 O.05 Total biomass/total throughput 0.066 0.071 <0.01 0.067 0.062 0.03 Total biomass (excluding detritus)a 285 276 0.74 284 308 0.16 Total catches a 2.794 0.471 O.01 2.811 1.996 <0.01 Connectance Index 0.254 0.243 <0.01 0.258 0.235 O.01 System Omnivory Index 0.121 0.126 0.35 0.130 0.126 0.09 Ascendancyd 3328 2958 0.02 3349 3873 0.03 Overhead*1 14246 12246 0.01 14159 16770 0.01 Capacity*1 17574 15204 0.01 17508 20642 0.01 Finn's cycling indexb 15.03 16.57 0.08 14.97 14.73 0.88 Predatory cycling indexc 7.32 7.56 0.25 7.20 8.41 0.55 Finn's mean path length 3.38 3.46 0.06 3.37 3.35 0.95 Throughput cycled excluding detritus3 142 113 0.04 137 173 0.01 Throughput cycled including detritus3 652 667 0.39 661 754 0.27 a = rkm"2,year"'; b = % or total throughput; c = % of throughput without detritus; d = flowbits (Ulanowicz and Puccia 1990) Discussion The comparison of SGSL and NGSL ecosystems and their changes over a 10-year period represents a step forward in the analysis of the Gulf of St. Lawrence ecosystem, 193 because of the areas covered, but also because of the thorough uncertainty analysis undertaken for the first time. All the results presented here were averaged from 31 balanced solutions obtained from inverse models, which represented possible stable states for the two ecosystems. Working with averages of many models (and their variation) is a much stronger approach than to only consider one balanced solution, and gives us more confidence in the results and their interpretation. Because it is the first time such an uncertainty analysis is coupled with Ecopath modelling, this work may represent an important step forward in such ecosystem investigations. 5.4.1 Biomass structure of ecosystems In both the NGSL and SGSL ecosystems, the structure shifted dramatically from one previously dominated by piscivorous demersal fishes during the mid-1980s to one dominated by small-bodied pelagic species in the mid-1990s, with a significantly greater marine mammal biomass during the mid-1990s. For both ecosystems, there was an important decline in abundance and the predatory role of piscivorous fish between the two periods modelled. Changes in the biomass of forage fish and large crustaceans groups were not significant in either ecosystem. For the SGSL, an increase in capelin biomass during the mid-1990s offsets decreases in the biomass of piscivorous and planktivorous small pelagic species, while shrimp biomass doubled. In the NGSL models, capelin biomass (the main forage species) was assumed to be the same in both periods, but this may reflect the lack of information on this and other forage species (e.g., sandlance and Arctic cod) in the NGSL (Morissette et al. 2003). For the SGSL, the model results are supported by the trawl survey data that showed substantial increases in capelin and shrimp biomass between the two periods - by an even greater amount than is indicated in the models (Benoit and Swain 2003). Furthermore, changes in ecosystem structure may have been even greater than indicated by our models, because information on abundance and diet of zooplankton and many benthic invertebrate groups was lacking. Finally, the decline in abundance of large piscivorous fish in both these ecosystems was similar to the declines reported in many other ecosystems of the world (Pauly and Maclean 2003), including other regions of the Northwest Atlantic (Rice and Rivard 2003). 194 5.4.2 Changes in consumption In the NGSL, the changes in the structure of the ecosystem were linked to a decrease in total consumption. Except for marine mammals, the consumption by all groups (excluding benthic invertebrates and plankton) decreased substantially. The main change in relative importance appears to be the decline in piscivorous fish, even though their importance relative to forage fish stayed about the same, i.e., they consumed about a third as much as forage fish in both periods. Interestingly, the consumption by shrimp and crabs decreased much more (about 50%, Figure 5.5) than their biomass (about 10%, Figure 5.2). This is mainly due to a substantial decrease in shrimp Q/B, (from 7.15 in the 1980s to 3.78 in the 1990s). This reduced Q/B was still within the range of possible values for shrimp, but was probably the maximum consumption the NGSL model could accept to reach a balanced solution. In contrast to the NGSL, there was an increase in consumption in the SGSL. The system changed from one dominated by piscivorous and forage fish to one dominated by forage fish. The decline in total consumption by piscivorous fish was compensated by an increased consumption by shrimp (whose biomass at least doubled), large crustaceans, and forage fish. The increase in consumption by forage fish in the SGSL was mostly due to a large increase in capelin biomass. Although total forage fish biomass changed little between models for the two periods (Figure 5.3), biomass of capelin had to be increased, relative to the 1980s estimate, in order to meet predator demands and obtain a balanced model for the 1990s. As stated above, the empirical evidence from trawl surveys supports this result. Furthermore, the models indicated a diet switch for the forage species as well, with an increase in their trophic levels in both ecosystems from the 1980s to the 1990s. The increased importance of capelin biomass in the 1990s in the SGSL induced an increase of its proportion in the diet of its predators (mackerel and herring), and thus, an increase in their respective trophic levels. These trophic level shifts could indicate either changes in prey eaten by the predator, or by one or more of its prey. If a species at a lower trophic level changes its diet (e.g., to benthic from pelagic prey, which results in increased trophic level of the prey species), then the predators feeding on this same prey would also show an increase in trophic level. For example, the observed decrease in the trophic level of marine mammals 195 (mainly harp seals) in the 1990s is consistent with the observed change in their diet. During the 1980s, small cod were important in the harp seal and grey seal diets but were replaced by a higher proportion of lower trophic level species in the 1990s. When planktonic prey were not considered, capelin was the most important prey during the 1990s and a large increase in the percent contribution of capelin to seal diets in the 1990s model was also observed. Thus, the increase in consumption by forage fish (i.e., capelin) could be explained by an increase in predation mortality, and then an increase in total mortality or production of capelin. 5.4.3 Predation by apex predators The reduced biomass of large predatory fish such as cod, American plaice, flounders, and redfish in the 1990s resulted in an increased relative mortality on these species due to predation by seals. In the NGSL, the consumption by seals increased, perhaps due to the important increase in their biomass. Indeed, seal populations increased with annual rates ranging from 3.9% year"1 (harp seal) to 9.7% year"1 (grey seal) in Atlantic Canada during the last two decades (Hammill and Stenson 2000), and they are now among the top three predators for many important fish species. However, despite their increase in biomass, total consumption by seals changed little in the SGSL, the increased consumption by grey seals being compensated by a substantial decline in consumption by harp seals. This decrease in consumption by seals coincides with a decrease in energy reserves (body condition) in the 1990s compared to the 1980s, and this has been documented for harp seals (Hammill et al. 1995; Chabot et al. 1996) and hooded seals off Newfoundland (Leblanc 2003). In the SGSL, seal consumption on large cod was much more important than in the NGSL. This is because grey seal abundance is higher in the SGSL (Hammill and Stenson 2000), and cod was its most common prey during the 1980s. Finally, the current level of consumption of cod and other prey by seals could actually represent an important source of competition for resources with the remaining fisheries (Trites et al. 1997). The growing populations of seals in the Gulf of St. Lawrence could have a direct negative effect on cod recruitment through predation on prerecruit cod; but they could also have an indirect positive effect through predation on pelagic fishes, which are possible 196 predators of early life history stages of cod (Swain and Sinclair 2000). For example, in the NGSL, harp seals may have a positive impact on skates and small demersal fish, because even though they prey on these species, they also feed on species that are potentially competing with them for the same resources (e.g., capelin, Arctic cod, and sand lance), or on species that are also predators of skates and small demersal fish (e.g., large demersal fish, Atlantic cod). This beneficial effect could be even greater than the predation itself, leading to an overall positive impact of the predator for its prey (Morissette et al. 2006, and see Chapter 4). 5.4.4 Impact of Ashing In both ecosystems, fishing mortality was intentionally reduced by a 3-4 year groundfish moratorium in the mid 1990s. As expected, fishing mortality decreased significantly from the mid-1980s to mid-1990s. However, for cod in both areas, this part of total mortality was replaced by predation (Figures 5.7 and 5.8). Top predators such as cetaceans and seals became the only large predators left, and ended up occupying the place that was vacated by the fishery and large piscivorous fish. The large decrease in consumption by predator fish was not entirely compensated by an increase in consumption by marine mammals, and total consumption decreased in the NGSL, but not in the SGSL. The most important commercial species in both ecosystems in the 1980s was cod. These stocks suffered from a commercial collapse in the late 1980s, a phenomenon that has been well documented (see Harris [1998] for a review). There is no consensus on the causes of the collapse of cod stocks. Some authors argue that it is due to intensive exploitation combined with a period of reduced productivity of cod stocks, as evidenced by poor condition and growth, and increased natural mortality (Dutil et al. 1999; Dutil and Lambert 2000). Others conclude that the collapse of cod stocks was solely due to overfishing (e.g., Hutchings and Myers 1994, Hutchings 1996). This is supported by our models relying on nominal catches, which suggest that a large part of cod mortality remains unexplained (M0). Savenkoff et al. (2004c) concluded that much of the other mortality in the NGSL ecosystem in the 1980s might be the result of under-reporting and discarding of catches. As a result, fishing mortality was substantially underestimated in the mid 1980s, just before the demise of the cod stock. Even though all the factors stated 197 above may have contributed to the decline in cod stocks in the late 1980s and early 1990s, the modelling approaches used in Savenkoff et al. (in press a) and here provide support the conclusion that overexploitation was by far the leading cause of the collapse for the NGSL and SGSL stocks. Pauly et al. (1998) called attention to the phenomenon of fishing down the food web. This occurs when large, relatively slow growing, piscivorous fishes are overfished and gradually replaced in fisheries landings by smaller, fast-growing, fish and invertebrates (Pauly and Maclean 2003). This phenomenon can be demonstrated through a decline in the mean trophic level of fisheries landings over time, a trend that has been frequently observed throughout the world. Slight declines in the mean trophic level of landings occurred between our models for the 1980s and 1990s, particularly in the NGSL. These decreases were mainly a reflection of the sharp declines in groundfish landings, i.e., high trophic level fishes, with little replacement by increased landings at lower trophic levels, as required for fishing "through" food webs (Essington et al. 2005). 5.4.5 Emergent properties of the models Ecopath modelling allowed us to analyse a series of parameters on the systems' emergent properties, which represented a useful way of comparing different ecosystems and to show important temporal or spatial trends. Even though the overall biomass and cycling (Finn's cycling index, predatory cycling, throughput cycled) did not change significantly over the 10-year period we examined, there was a strong indication that the sum of flows changed in both ecosystems (significant decrease in respiratory flows, total system throughput, and total production in the NGSL, but significant increase of the same parameters in the SGSL). All these indices represent the size of the entire system in terms of flow (Ulanowicz 1986), meaning that less energy is flowing in the entire ecosystem in the 1990s compared to the 1980s in the NGSL; however, there was greater flow of energy for the 1990s compared to 1980s in the SGSL. Moreover, the ascendancy, the overhead and the capacity are indices based on thermodynamics and information theory representing ecosystem growth and development (Christensen 1995). They are other strong indicators that the NGSL ecosystem was closer to maturity in the 1980s than in the 1990s (these three indices decreased significantly over the 10 years period), while the 198 opposite is true in the SGSL. This change of flows certainly indicates that the overall structure of the NGSL and SGSL ecosystems changed through time. However, these indices need to be studied in more ecosystems before their interpretation in terms of maturity can be confirmed. The removal of cod in the SGSL can hardly be a sign of maturity, considering that historically cod was the dominant piscivorous fish in the ecosystem. The lack of recovery of many important stocks of the GSL that underwent drastic declines leading to fishing moratoria clearly point to the need to understand how ecosystems react to perturbations. One argument is that more complex ecosystems will cope better with perturbations (McNaughton 1978). For example, major decreases in abundance of gadoid predators on the eastern shelf/Georges Bank ecosystem of the United States was offset by increased abundance of elasmobranchs and large demersal predators (i.e., various sculpins) such that total consumption of forage fishes changed remarkably little (Fogarty and Murawski 1998; Link and Garrison 2002). This concept of resilience is also linked to maturity, as ecosystems are thought to be more stable when they reach maturity. The connectance index, which reflects complexity, decreased significantly from the 1980s to the 1990s in both ecosystems. This may reflect that ecosystems were more perturbed in the 1990s than in the 1980s. In the NGSL, this loss in connectance is reflected by the loss in maturity (ascendancy, overhead and capacity), while in SGSL, even if the maturity seemed to increase over the 10-year period, the connectance decreased significantly. Also, since both ecosystems differed in their communities (NGSL was dominated by capelin and herring in the mid-1990s while SGSL was dominated by herring, capelin, mackerel, several flatfish species, and majiid crabs) and oceanographic features, we might predict that recovery from the current perturbation will also be different for each ecosystem. In NGSL, the emergent properties suggested a less productive system for the 1990s than for the 1980s. In contrast, in the SGSL, productivity was similar to that of the NGSL in the 1980s, but actually increased in the 1990s. In the SGSL, there were also some higher trophic level fish such as cod whose biomass had not completely collapsed, which might increase the productivity of the system. In addition, biomasses of groups such as capelin and shrimp increased substantially during the 1990s. Moreover, the fact that there 199 was a large increase in consumption by forage fish in the SGSL in the 1990s, when their biomass showed no significant decline since the 1980s could reflect an increased consumption of forage fish by other groups, and hence increased productivity of forage fish (to keep biomass the same despite increased predation). This could explain the higher consumption by forage fish, linked to a greater abundance of high trophic level groups to prey on forage fish in the SGSL than in the NGSL. In the latter ecosystem, many species showed severely depleted abundance, which would explain the reduced productivity of the system in the 1990s. Cury et al. (2003) defined a regime shift as a process of density compensation of some species after the removal of some other species in an ecosystem. As fisheries have removed extensive amounts of predatory fishes during the last decades, one must carefully consider the implications for the other components of these systems. However, our models suggest that in both NGSL and SGSL, fishing did not lead to clear compensatory of biomass or consumption of other species resulting from the removal of predatory fishes (except perhaps for capelin and shrimp, which increased in SGSL in the 1990s). In the SGSL case, the decrease in consumption by predatory fishes was more than offset by increases in consumption by forage fishes, shrimp and large crustaceans. For the SGSL, this increased importance of forage fishes, shrimp and crabs may reflect release from predation or competition, but it may also reflect other factors that are not examined here (e.g., favorable environmental conditions). Regardless of the cause, neither population of Atlantic cod has increased in abundance despite greatly reduced fishing pressure since 1993/94, and the failure of strong year-classes to develop is currently perplexing because both populations were able to increase due to strong pulses of recruitment during a similar population collapse in the 1970s (Chouinard and Frechet 1994; Sinclair and Murawski 1997). Seals have also continued to increase at a relatively constant rate, but no species has yet had sufficient time to move in to occupy the role of large predator that was previously occupied by cod. More recent models would be useful in attempting to explain why cod has not recovered further nearly ten years after the 1990s models. 200 Chapter summary In order to have a global view of ecosystem changes associated with the collapse of groundfish species in the Gulf of St. Lawrence during the early 1990s, four Ecopath mass-balance models were constructed that included an in-depth analysis of uncertainty. These models covered two ecosystems (northern and southern Gulf of St. Lawrence; NAFO divisions NGSL and SGSL), and two time periods (before the collapse, in the mid- 1980s, and after it, in the mid-1990s). Our analyses revealed that the ecosystem structure shifted dramatically from one previously dominated by piscivorous groundfish during the mid-1980s to one now dominated by small-bodied pelagic species during the mid-1990s in both southern and northern Gulf. The species structure in NGSL versus SGSL was different, and that may explain why these two ecosystems did not recover the same way from the collapse in the early 1990s. Productivity declined in the northern Gulf after the collapse but increased in the southern Gulf. The collapse of groundfish stocks resulted in declines in the mean trophic level of the landings in both the northern and the southern Gulf. Even though fishing mortality was then intentionally reduced, this part of the total mortality was taken up by predation. The temporal changes in the internal structure of both ecosystems are reflected in their overall emergent properties. References Anonymous 1994. Report on the status of groundfish stocks in the Canadian northwest Atlantic. DFO Atlantic Fish Stocks Status Report 94/4. Benoit, H. and D.P. Swain. 2003. Standardizing the southern Gulf of St. Lawrence bottom trawl survey time series adjusting for changes in research vessel, gear and survey protocol. Canadian Technical Report of Fisheries and Aquatic Sciences 2505. Bowering, W.R. and G.R. Lilly. 1992. Greenland halibut (Reinhardtius hippoglossoides) off southern Labrador and northeastern Newfoundland (northwest Atlantic) feed primarily on capelin (Mallotus villosus). Netherlands Journal of Sea Research 29: 211-222. Chabot, D., Stenson, G.B., and N.B. Cadigan. 1996. Short- and long-term fluctuations in the size and condition of harp seal (Phoca groenlandica) in the Northwest Atlantic. NAFO Scientific Council Studies 26: 15-32. 201 Chouinard, G. A. and A. Frechet. 1994. Fluctuations in the cod stocks of the Gulf of St. Lawrence. ICES Marine Science Symposia. 198: 121-139. Chouinard, G.A., Sinclair, A.F., and D.P. Swain. 2003. Factors implicated in the lack of recovery of southern Gulf of St Lawrence cod since the early 1990s. ICES CM 2003/U:04. Christensen, V. 1995. Ecosystem maturity - towards quantification. Ecological Modelling 77:3-32. Christensen, V. Walters, C.J., and D. Pauly. 2000. Ecopath with Ecosim: a user's guide. October 2000 Edition, Fisheries Centre, The University of British Columbia, Vancouver, B.C. and ICLARM, Penang, Malaysia. Cury, P., Shannon, L., and Y.-J. Shin. 2003. The functioning of marine ecosystems: A fisheries perspective, pp. 103-123 In Sinclair M. and G. Valdimarsson (eds.). Responsible fisheries in the marine ecosystem. CAB International, Wallingford. Dutil, J.-D., Castonguay, M., Gilbert, D., and D. Gascon. 1999. Growth, condition, and environmental relationships in Atlantic cod (Gadus morhua) in the northern Gulf of St. Lawrence and implications for management strategies in the Northwest Atlantic. Canadian Journal of Fisheries and Aquatic Sciences 56: 1818-1831. Dutil, J.-D. and Y. Lambert. 2000. Natural mortality from poor condition in Atlantic cod (Gadus morhua). Canadian Journal of Fisheries and Aquatic Sciences 57: 826-836. Essington, T.E., Beaudreau, A.H., and J. Wiedenmann. 2005. Fishing through marine food webs. Proceedings of the National Academy of Sciences of the United States of America 103: 3171-3175. Fogarty, M.J. and S.A. Murawski. 1998. Large-scale disturbance and the structure of marine ecosystems: fisheries impacts on Georges Bank. Ecological Applications 8: S6-S22. Frechet, A., and P. Schwab. 1998. Evaluation de la morue du nord du Golfe du Saint- Laurent (3Pn, 4RS) en 1997/Assessment of the northern Gulf of St. Lawrence cod stock (3Pn, 4RS) in 1997. DFO Canadian Stock Assessment Secretary Research Document 98/127. Frechet, A., Gauthier, J., Schwab, P., Pageau, L., Savenkoff, C , Castonguay, M., Chabot, D., Tournois, C , Lussier, J.-F., Spingle, J., and F. Collier. 2005. The status of cod 202 in the Northern Gulf of St. Lawrence (3Pn, 4RS) in 2004. DFO Canadian Stock Assessment Secretary Research Document 2005/60. Gilbert, D., and B. Pettigrew. 1997. Interannual variability (1948-1994) of the CIL core temperature in the Gulf of St. Lawrence. Canadian Journal of Fisheries and Aquatic Sciences 54(Supplement 1): 57-67. Hammill, M.O. and G.B. Stenson. 2000. Estimated prey consumption by harp seals (Phoca grownlandica), hooded seals (Cystophora cristatd), grey seals (Halichoerus grypus) and harbour seals (Phoca vitulina) in Atlantic Canada. Journal of Northwest Atlantic Fishery Science 26: 1-23. Hammill, M.O., Kingsley, M.C.S., Beck, G., and T.G. Smith. 1995. Growth and condition in the Northwest Atlantic harp seal. Canadian Journal of Fisheries and Aquatic Sciences 52:478-488. Harley, S.J. and R.A. Myers. 2001. Hierarchical Bayesian models of length-specific catchability of research trawl surveys. Canadian Journal of Fisheries and Aquatic Sciences 58: 1569-1584. Harris, M. 1998. Lament for an ocean: The Collapse of the Atlantic Cod Fishery: a True Crime Story. McClellan & Stewart, Toronto. Hutchings, J.A. 1996. Spatial and temporal variation in the density of northern cod and a review of hypotheses for the stock's collapse. Canadian Journal of Fisheries and Aquatic Sciences 53: 943-962. Hutchings, J.A. and R.A. Myers. 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. Leblanc, M.-J. 2003. Variations saisonnieres en taille et en condition du phoque a capuchon (Cystophora cristatd) dans le nord-ouest de l'Atlantique. M. Sc. thesis, Universite du Quebec, Rimouski, QC, Canada. Lilly, G.R. 1991. Interannual variability in predation by cod (Gadus morhua) on capelin (Mallotus villosus) and other prey off southern Labrador and northeastern Newfoundland. ICES Marine Science Symposia 193: 133-146. Link, J.S. and L.P. Garrison. 2002. Changes in piscivory associated with fishing induced changes to the finfish community on Georges Bank. Fisheries Research 55: 71-86. 203 McNaughton, S.J. 1978. Stability and diversity of ecological communities. Nature 274: 251-253. Morissette, L. 2001. Modelisation ecosystemique du nord du Golfe du Saint-Laurent. M. Sc. thesis, Universite du Quebec, Rimouski, QC, Canada. Morissette, L., Despatie, S.-P., Savenkoff, C , Hammill, M.O., Bourdages, H., and D. Chabot. 2003. Data gathering and input parameters to construct ecosystem models for the northern Gulf of St. Lawrence (mid-1980s). Canadian Technical Report of Fisheries and Aquatic Sciences 2497. Morissette, L., Hammill, M.O. and C. Savenkoff. 2006. The trophic role of marine mammals in the Northern Gulf of St. Lawrence. Marine Mammal Science. 22: 74- 103. Odum, E.P. 1971. Fundamentals of Ecology. W.B. Saunders, Philadelphia. Pauly, D., and J. Maclean. 2003. In a perfect ocean: the state of fisheries and ecosystems in the north Atlantic ocean. Island Press, Washington. Pauly, D., Christensen, V., Dalsgaard J., Froese, R., and F. Torres Jr. 1998. Fishing down marine food webs. Science 279: 860-863. Pimm, S.L. 1982. Food webs. Chapman and Hall, London. Pitt, T.K. 1973. Food of American plaice (Hippoglossoides platessoides) from the Grand Bank, Newfoundland. Journal of Fisheries Research Board of Canada 30: 1261- 1273. Power, M.E., Tilman, D., Estes, J.A., Menge, B.A., Bond, W.A., Mills, L.S., Daily, G., Castilla, J.C., Lubchenco, J., and R.T. Paine. 1996. Challenges in the quest for keystones. Biosciences 46: 609-620. Rice, J., and D. Rivard. 2003. Proceedings of the zonal assessment meeting - Atlantic cod. Canadian Stock Assessesment Secretariat Proceeding Series 2003/021. Ricker, W.E. 1975. Computation and interpretation of biological statistics of fish populations. Bulletin of Fisheries Research Board of Canada 191. Savenkoff, C , Bourdages, H., Castonguay, M., Morissette, L., Chabot, D., and M.O. Hammill. 2004a. Input data and parameter estimates for ecosystem models of the northern Gulf of St. Lawrence (mid-1990s). Canadian Technical Report of Fisheries and Aquatic Sciences 2531. 204 Savenkoff, C , Bourdages, H., Swain, D.P., Despatie, S.-P., Hanson, J.M., Methot, R., Morissette, L., and M.O. Hammill. 20046. Input data and parameter estimates for ecosystem models of the southern Gulf of St. Lawrence (mid-1980s and mid- 1990s), anadian Technical Report of Fisheries and Aquatic Sciences 2529. Savenkoff, C., Castonguay, M., Vezina, A.F., Despatie, S.-P., Chabot, D., Morissette, L., and M.O. Hammill. 2004c. Inverse modelling of trophic flows through an entire ecosystem: the northern Gulf of St. Lawrence in the mid-1980s. Canadian Journal of Fisheries and Aquatic Sciences 61: 2194-2214. Savenkoff, C , Castonguay, M., Chabot, D., Bourdages, H., Morissette, L., and M.O. Hammill. (in press a). Effects of fishing and predation in a heavily exploited ecosystem: comparing pre- and post-groundfish collapse periods in the northern Gulf of St. Lawrence (Canada). Estuarine, Coastal and Shelf Science. Savenkoff, C , Swain, D.P., Hanson, J.M., Castonguay, M., Hammill, M.O., Bourdages, H., Morissette, L., and D. Chabot. (in press b). Effects of fishing and predation in a heavily exploited ecosystem: Comparing pre- and post-groundfish collapse periods in the southern Gulf of St. Lawrence (Canada). Ecological Modelling. Sinclair, A.F. and S.A. Murawski. 1997. Why have groundfish stocks declined? pp. 71-93 In Boreman, J., Nakashima, B.S., Wilson, J.A., and R.L. Kendall. Northwest Atlantic groundfish: perspectives on a fishery collapse. American Fisheries Society, Bethesda, MD. Swain, D.P. and A.F. Sinclair. 2000. Pelagic fishes and the cod recruitment dilemma in the Northwest Atlantic. Canadian Journal of Fisheries and Aquatic Sciences 57: 1321- 1325. Therriault, J.-C. (ed.). 1991. The Gulf of St. Lawrence : small ocean or big estuary? Canadian Special Publication of Fisheries and Aquatic Sciences 113. Trites, A.W., Christensen, V., and D. Pauly. 1997. Competition between fisheries and marine mammals for prey and primary production in the Pacific Ocean. Journal of Northwest Atlantic Fishery Science 22: 173-187. Ulanowicz, R.E. 1986. Growth and development: ecosystem phenomenology. Springer Verlag, New York. 205 Ulanowicz, R.E. 2000. Toward the measurement of ecological integrity, pp. 99-113 In Pimentel, D., Westra, L., and R.F. Noss. Ecological integrity: integrating environment, conservation, and health. Island Press, Washington DC. Ulanowicz, R.E., and C.J. Puccia. 1990. Mixed trophic impacts in ecosystems. Coenoses 5:7-16. Vasconcellos, V., Mackinson, S., Sloman, K., and D. Pauly. 1997. The stability of trophic mass-balance models of marine ecosystems: a comparative analysis. Ecological Modelling 100: 125-143. Vezina, A.F., and T. Piatt. 1988. Food web dynamics in the ocean. I. Best-estimates of flow networks using inverse methods. Marine Ecology Progress Series 42: 269- 287. Zar, J.H. 1998. Biostatistical analysis (4th edition). Prentice Hall, Englewood Cliffs, NJ. 206 C h a p t e r 6 C o n c l u s i o n If the ecological status of the world's marine ecosystems is to be evaluated, it could be as follows: at the global scale, fish populations are declining in abundance (Myers and Worm, 2003), and crashes of many large-scale fisheries around the world are occurring (Pauly et al. 1998). What was perceived for a long time as an inexhaustible resource now seems quite limited (Rosenberg et al. 1993). To understand the nature and dynamics of exploited marine ecosystems, and more precisely the interactions of the species they are in, the development of an ecosystem approach is essential. In chapter 1, a database of Ecopath models of marine ecosystems was presented as a useful tool to investigate a wide range of scientific questions about the world's oceans. The database of models itself documents the great effort that is currently undertaken to describe the diversity of the world's ocean. Overall, this database assembled 393 Ecopath models that are now publicly available. As such, they can be out to more uses than the ones for which they were built. This thesis is one of these uses. It is based on a selection of ecosystem models representing the world's ocean, allowing three specific issues of marine ecology to be addressed. Thus, the use of several models simultaneously allowed me to address more general or global issues in marine ecology. Ecosystem modelling is a useful tool to understand marine ecological processes, but there is a lot of uncertainty related to the models' structure and outcomes. In Chapter 2,1 reviewed the different options available for uncertainty analyses in marine ecosystem models, and proposed some alternatives for a better investigation of uncertainty. Unfortunately, only few of the many plublished studies have gone so far as to examine uncertainty in their analyses. This chapter emphasized the importance of doing so. After discussing the strengths and weaknesses of the different ways to address uncertainty in the models, a new approach was proposed, combining Ecopath with other modelling techniques (inverse methods) in order to gain robustness. 207 Part of the uncertainty in ecosystem models is due to the quality of input data, and the trust we have in model parameters. Not only the quality of input data can influence model's uncertainty, but I also found out that it can affect importantly the way the model performs its simulations. In chapter 3,1 used 50 Ecopath models to test the hypothesis that the quality of ecosystem models' inputs plays a significant role in predicting the relationship between complexity and stability. My secondary hypothesis was that there is, in ecosystems, a direct relationship between complexity and stability. Quality of input data has sometimes been questioned, but was never analysed as a factor that could affect the diversity-stability relationship, or the outcomes of modeling results. My results have shown that the quality of data used in ecosystems models, as well as the realism of the models used, were really important (and significant) factors to consider while attempting to describe the relation between complexity and stability, or to model any trophic interaction in ecosystems. Models of higher quality tend to perform better in predicting changes in species' biomasses, and they corroborated the hypothesis that stability of ecosystems tend to increase with complexity. When changes in species biomass are analysed for a whole ecosystem, some of the most important groups to observe are the top-predators. Indeed, their overall impact on the ecosystem structure can be significant. In many ecosystems, marine mammals and fisheries represent the top-predators of the food web, and there is a growing concern to know if they compete for the same food resources. In chapter 4, a set of seven models presenting some competition between fisheries and marine mammal consumption was used to test the hypotheses that the presence of marine mammals in ecosystems and their overlap with fisheries for food resources does not have a significant effect on the overall productivity of the ecosystem, and that their extirpation can decrease this productivity because of beneficial predation effects marine mammals can have on their prey. Marine mammals are a very important component of marine ecosystems, and as top predators, they may compete with fishery for the same target species. However, in most ecosystems, the primary production required (PPR) to sustain marine mammals' consumption was smaller than the PPR for the catch. As for their mixed indirect trophic impact, it was always less important, even if also negative, than the impact of fishing on ecosystems compartments. Finally, when the extirpation of marine mammals was simulated, the 208 biomass of other species in the ecosystem was largely unaffected. Moreover, the impact of these top predators in marine ecosystems was not always negative. In many cases, there was a positive predation effect from marine mammals on their prey. Thus, the notion that hunting of marine mammals would solve the fisheries problems is too simplistic. These animals are part of a foodweb involving many species and their effect on these other species is not always direct. One ecosystem that suffered from such a major fisheries crisis is the Gulf of St. Lawrence. In that ecosystem, top-predators (mainly pinnipeds) are important, and so was the fisheries before the collapse of groundfish stocks in the early-1990s. Chapter 5 addressed this problem of collapsing fish stocks in the Northwest Atlantic. By comparing two ecosystems (Northern and Southern Gulf of St. Lawrence) and two time periods (pre- and post-collapse), I attempted to explain the collapse of important predatory fish with an ecosystem-based approach, implemented by environmental (climate variability) and anthropogenic (overfishing) factors. My analyses revealed that indeed, significant changes occurred in the ecosystem structure between the 1980s and the 1990s and have contributed to the failure of the cod stocks to recover in the 1990s. In both ecosystems, the structure shifted dramatically, from one previously dominated by piscivorous groundfish during the mid-1980s, to one now dominated by small-bodied pelagic species during the mid-1990s in both the southern and northern Gulf of St. Lawrence. The species structure was different between the northern Gulf and the southern Gulf, and this may explain why these two ecosystems did not collapse to the same degree in the early 1990s. Having a global representation of a food web using a modelling tool such as Ecopath is definitely an asset in trying to understand the structure and function of marine ecosystem. This is even more useful when many models are used simultaneously to address more general or global issues. Definitely, the use of models allowed me to test and examine major issues in marine ecology, and in some cases find some interesting answer to them. However, models are only possible representations of marine systems, and they cannot answer all our ecological questions, solve all our scientific concerns. 209 Strengths and weaknesses of this thesis research In order to address issues at the ecosystem level, scientists create models that they hope represent these ecosystems and simulate different scenarios to see how they will react to diverse situations. Ecosystem modelling has become very popular in applied ecology, and this is even more so for marine ecosystems, which are more difficult to investigate directly. Different modelling approaches are available to the scientific community. One of them is the Ecopath approach, which has its weaknesses, but most importantly has the great advantage of being an approach previously applied to many ecosystems around the world, leading to lots of replicates for comparative analyses, including at the global scale. The modelling approach presented in this thesis research is innovative in the sense that it proposed model quality as a new metric to consider in analyses of marine ecosystems. With the large database of Ecopath models that we created, we were able to create scenarios representing global responses of marine food webs to global issues. Only a minority of ecosystem studies address the uncertainty of model results. Uncertainty and variability are inherent in the very nature of ecosystem modelling. Therefore we need to use appropriate tools to define, represent and analyze this uncertainty. Ecopath is only one of the many approaches that are used worldwide for ecosystem modelling. The Ecopath software includes several tools that have been used - more or less successfully- to address uncertainty. The optimum solution seems to be the use of these tools in combination with the inverse modelling approach, in order to use the strengths of each approach. 210 Future improvements and applications Aquatic ecosystems are under enormous stress. A variety of disturbances, including overfishing, climate change and fish introductions, are threatening the species, their community and entire food webs. The ecosystem modelling approach, until now, dealt mainly with ecological issues such as predator-prey relationships, fisheries management, biodiversity, etc. Now that modelling is becoming more and more popular, there is a need to focus on merging different fields to better understand the structure and function of ecosystems. For example, it seems that the role of genetic diversity of populations is as important as the species diversity (Reusch et al. 2005). The evolution of the prey can also modify considerably predatory-prey relationships (Yoshida et al. 2003). In a context of climate change, oceanographic features are also to be considered when addressing ecosystem dynamics (Gilbert 2005). New approaches should integrate these different fields for even more representative models and analyses. Final conclusions and recommendations This thesis research ended up being a very collaborative approach, where Ecopath users helped by providing information about their models, and sharing their interest in understanding marine ecosystems. Having a large number of models to compare is definitely an asset. Comparing ecosystems can sometimes enlighten us on some ecological phenomena or management choices that are successful in one ecosystem, we can then apply it in another ecosystem. Now that these ecosystems are threatened by factors such as habitat degradation, climate change, invasive species, overexploitation, and pollution all around the world, such comparative analyses are urgently needed. References Gilbert, D., Sundby, B., Gobeil, C , Mucci, A., and G.-H. Tremblay. 2005. A seventy-two- year record of diminishing deep-water oxygen in the St. Lawrence estuary: the northwest Atlantic connection. Limnology and Oceanography 50: 1654-1666. Myers, R.A. and B. Worm. 2003. Rapid worldwide depletion of predatory fish communities. Nature 423, 280-283. 211 Pauly, D., Christensen, V., Dalsgaard, J., Froese, R., and F. Torres Jr. 1998. Fishing down marine food webs. Science 279: 860-863. Reusch, T.B.H., Ehlers, A., Hammerli, A., and B.Worm. 2005. Ecosystem recovery after climatic extremes enhanced by genotypic diversity. Proceedings of the National Academy of Sciences of the United States of America 102: 2826-2831. Rosenberg, A.A., Fogarty, M.J., Sissenwine, M.P., Beddington, J.R., and J.G. Shepherd. 2003. Achieving sustainable use of renewable resources. Science 262: 828-829. Yoshida, T., Jones, L.E., Ellner, S.P., Fussmann, G.F., and N.G. Hairston Jr. 2003. Rapid evolution drives ecological dynamics in a predator-prey system. Nature 424: 303- 306. 212 A p p e n d i c e s Appendix 1: Questionnaire sent to Ecopath users. Appendix 2: List of Ecopath models (models highlighted in grey are the ones that were already available on the Ecopath website) and their respective reference, when published. 213 Appendix 1 Questionnaire sent to Ecopath users Dear Ecopath user, My name is Lyne Morissette, I am a Ph.D. student at UBC Fisheries Centre with Dr. Daniel Pauly. Part of my project consists in a census of all the models that have been constructed over the years with the Ecopath with Ecosim software. Our database indicates that you downloaded the software. We would really appreciate if you could send the following information: 1. If a model was constructed or not with this software 2. Geographic area modeled 3. Period of time modeled (year) 4. Purpose of the model (management, description of the ecosystem, modelling protected areas, fishing policy, addressing theoretical ecology questions, etc.) 5 . Citations of published work based on the model (Authors, year, title, journal, pages) 6. Type of institution (university, government, industry, etc.) for which the model was constructed 7. Who constructed the model (student, post-doc,.research associate, faculty, etc.) Please send your reply to (email address) July 7th 2006. Thanks a lot, your help is much appreciated! as soon as possible, and hopefully before Lyne Morissette Ph.D. Cand idate Marine Ecosystem Modelling- Fisheries Centre, The University of British Columbia 214 Appendix 2 List of Ecopath models (models highlighted in grey are the one that were already available V on the Ecopath website) and their respective reference, when published. \ Geographic area Model Citation , Africa, Lake Chad Completed Palomares et al 1993a Alaska Gyre Completed Pauly etal. 1996 Aleutian Islands Completed Aydin et al. (submitted) Aleutian Islands Completed Heymans 2005b Antartica In progress Antartica, Antartic peninsula Completed Erfan and Pitcher 2005 Antartica, Falkland Islands Completed Cheung and Pitcher 2005 Antartica, Weddell Sea Completed Jarre-Teichmann et al. 1997 Arctic ocean Completed Okey 2003 Argentina, Coastal marine zone In progress Atlantic, Northeast (ICES area Via) In progress Atlantic, South Bight Completed Okey and Pugliese 2001 Australia, Darwin Harbour Completed Martin 2005 Australia, Great Barrier Reef Completed Gribble 2001 Australia, North West Shelf Completed Bulman et al. 2006 Australia, Northern South Wales, inshore Completed Forrest (PhD thesis in preparation) Australia, Northern South Wales, offshore Completed Forrest (PhD thesis in preparation) Australia, Port Phillip Bay Completed Fulton and Smith 2004 Australia, Southeast coast In progress Fulton (in preparation) Australia, Southeast, Eastern Bass Strait Completed Bulman et al. 2006 Australia, Southeast, Eastern Bass Strait Completed Goldsworthy et al. 2003 Australia, Southern Tasmanian midslope Completed Bulman 2002 Australia, Southern Tasmanian Seamoutns Reserve Completed Bulman et al. 2002 Bahamas, Bimini Completed Grant 2001 Baltic Proper Completed Mollmann and Eero 2004 Baltic Proper, open water Completed Cardinale and Casini 2004 Baltic Sea Completed Harvey et al. 2003 Baltic Sea Completed Zabele/a/. 2003 Baltic Sea, central Completed Jarre-Teichmann 1995 Baltic Sea, Curonian lagoon Completed Razinkovas 2004 Baltic Sea, Gulf of Riga, Dzeni Retion, Ainazi Completed Strake et al. 2004 215 Baltic Sea, Kaliningrad region In progress Baltic Sea, South (2) In progress In preparation Baltic Sea, Southern, Gulf of Gdansk, Puck Bay Completed Tomczak et al. 2005 Bangladesh, Bay of Bengal Completed Mustafa 2003 Barents Sea Completed Blanchard et al. 2002 Barents Sea Completed Skaret and Pitcher, in press Barents Sea Completed Falk-Petersen, (in preparation) Belize, Turneffe Atoll, Calabash Caye, Mangrove- seagrass-coral reef In progress Model completed, but no publication Benin, Lake Nokoue Completed Ph.D. thesis in preparation Black sea Completed Orek 2000 Brazil, Abrolhos Completed Telles 1998 Brazil, Amazon basin Completed Angelini et al. 2006 Brazil, Caete Mangrove estuary Completed Wolff et al. 2000 Brazil, East Completed Ph.D. thesis in preparation Brazil, Itamaraca Estuarine System In progress Model completed, but no publication Brazil, Rio de Janeiro, Cabo Frio In progress In preparation Brazil, Rio de Janeiro, Guanabara Bay In progress Model completed, but no publication Brazil, Rio de Janeiro, Itaipu coastal zone In progress Model completed, but no publication Brazil, Sao Paulo State, Broa Reservoir Completed Angelini and Petrere 1996; Angelini 2002 Brazil, South Bight Completed Gasalla and Rossi-Wongtschowski 2004 Brunei, Darussalam, Borneo coast Completed Silvestre et al. 1993 Burkina Faso, Bagre reservoir Completed Villanueva et al. (in press a) Burundi, Lake Tanganyika (2) Completed Moreau et al. 1993a Cambodia, Tonle Sap Lake In progress Model completed, but no publication Canada, Arctic (terrestrial) Completed Krebs et al. 2003 Canada, British Columbia, Northern coast (4) Completed Ainsworth et al. 2002 Canada, British Columbia, Strait of Georgia Completed Martell et al. 2002 Canada, Eastern Scotian Shelf Completed Bundy 2005; Bundy and Fanning 216 2005; Bundy 2004a; Bundy 2002 Canada, Eastern Scotian Shelf Completed Bundy 2005; Bundy and Fanning 2005; Bundy 2004a; Bundy 2003 Canada, Gwaii Haanas Completed Salomon et al. 2002 Canada, Hecate Strait Completed Beattie 2001 Canada, Hecate Strait Completed Hagganef a/. 1999 Canada, Kootenay Lake Completed Thompson 1999 Canada, Lancaster Sound Completed Mohammed 2001 Canada, Newfoundland (4) Completed Heymans and Pitcher 2002a,b; Heymans 2003 Canada, Newfounland-Labrador Shelf Completed Bundy 2001; Bundy 2002; Vasconcellos et al. 2002; Savenkoff et al. 2001; Bundy et al. 2000 Canada, Northern British Columbia Completed Ainsworth et al. 2002 Canada, Northern Gulf of St. Lawrence (3) Completed Morissette et al. 2003, Savenkoff etal. 2004a,6; Savenkoff and Morissette( in prep.) Canada, Puget Sound Completed Preikshop and Beattie 2001 Canada, Rivers Inlet of British Columbia Completed Watkinson and Pauly 1999 Canada, Southern Gulf of St. Lawrence (2) Completed Savenkoff et al. 2004c Canari Islands, Gran Canaria, Maspalomas Lagoon Completed Moreno and Castro 1998 Cap Verde Completed Stobberup et al. 2003 Carribean, British Virgin Island, Puerto Rico Completed Searles 2004 Caribbean, Grenada and Grenadines Completed Mohammed 2003 Caribbean, Southern Mexican, Coral reef Completed Alvarez-Hernandez 2003 Caribbean Sea, coral reefs Completed Opitz 1996 Caribbean, BVI Puerto Rico In progress Catalan Sea, South (2) Completed Coll et al. 2006; Coll et al., in press b Chile Completed Milessi 2005 Chile, Antofagasta Peninsula Completed Paves and Gonzalez, submitted Chile, Central Completed Neira et al. 2004 Chile, Central Completed Ortiz and Wolff 2002a,b Chile, central and south continental shelf In progress In preparation Chile, Central MPA Central In progress In preparation 217 Chile, Northern coast Completed Medina et al. (submitted) Chile, Tongoy Bay Completed Ortiz and Wolff 2002b Chile, Tongoy Bay Completed Wolff 1993 China Completed Song et al. 2007 China, Beibu Gulf Completed Chen et al. 2006 China, Bohai Sea Completed Tong et al. 2000 China, Guangdong Province, Zhujiang Delta Completed Ruddle and Christensen 1993 China, Hong Kong Completed Pitcher et al. 1998; Pitcher et al. 2002a,b China, Hong Kong Completed Sumalia et al. 2007 China, Yangtze River, Niushan Lake Completed Ph.D. thesis in preparation China, Zhengiian, Gougi Island, Seaweed beds In progress M.Sc thesis in preparation China Sea, deep shelf Completed Pauly and Christensen 1993 China Sea, East In progress China Sea, south, Cape Cambodia Completed Nguyen 1989; Pauly and Christensen 1993 China Sea, south, coral reefs Completed Pauly and Christensen 1993 China Sea, south, deep waters (50-200 m) Completed Pauly and Christensen 1993 China Sea, south, marine reserves Completed Pitcher et al. 2000 China Sea, south, ocean part (> 200 m) Completed Pauly and Christensen 1993 China Sea, south, shallow waters (< 10m) Completed Pauly and Christensen 1993 China Sea, southwest, eastern Malaysia and Completed Liew and Chan 1987; Pauly and southeastern Sumatra Christensen 1993 Colombia, Guarjira peninsula Completed Criales-Hernandez et al. 2006 Colombia, Gulf of Salamanca Completed Duarte and Garcia 2004 Congo, Lake Kivu Completed Villanueva et al. (in press b) Costa Rica, Golfo de Nicoya Completed Wolff etal. 1998 Costa Rica, Golfo Dulce Completed Wolff et al. 1996 Cuba, Batabano ecosystem In progress Wolff et al. (in preparation a) Denmark, coast In progress Denmark, Faroe Islands Completed Booth and Zeller 2005 Denmark, Faroe Islands (1961) Completed Zeller and Reinert 2004 Denmark, Faroe Islands (1997) Completed Zeller and Freire 2002 Ethiopia, Lake Awassa Completed Fetahi 2005 Fantasy wonderland with seadragons, mermaids, etc Completed B. Fulton, educational purposes CSIRO 218 France, Bay of Biscay (2) Completed Ainsworth et al. 2001 France, Bay of Somme, Eastern Channel Completed Rybarczyk et al. 2003 France, Bretagne, Iroise Sea Completed M.Sc. thesis in preparation France, Etang de Thau Completed Palomares et al. 19936 France, Gironde Estuary Completed Lobry 2004 France, Golfe de Gasgogne In progress France, La Seine Completed Boetef a/. 1999 France, Lake Aydat Completed Reyes-Marchant et al. 1993 France, Normandy, Seine Estuary Eastern Channel Completed Rybarczyk and Elkaim 2003 France, Toulouse, Garonne River Completed Palomares et al. 1993c French Frigate Shoals Completed Polovina 1984 French Polynesia, Moorea Island, Tiahura Reef (2) Completed Arias- Gonzalez et al. 1997 Galapagos, Floreana rocky reef Completed Okey et al. 2004a Gambia, coast Completed Mendy 2003 Gambia estauyy, Africa Completed Ph.D. thesis in preparation Generic demersal system Completed Christensen et al. 2005 Generic pelagic system Completed Christensen et al. 2005 Ghana, Sakumo Lagoon Completed Pauly 2002 Greece, Eastern central Ionian Sea In progress M.Sc. thesis in preparation Greenland, West coast Completed Pedersen 1994; Pedersen and Zeller 2001 Guinea Completed Dillao et al. 2003 Guinea (2) Completed Guenette and Dillao 2004; Guenette and Dillao (in press) Guinea Bissau, coast Completed Amorim et al. 2003 Gulf of Thailand Completed Christensen 1998 Gulf of Thailand Completed FAO/FISHCODE 2001 Gulf of Thailand, coast Completed Vibunpant et al. 2003 Iceland Completed Buchary 2001 Iceland Completed Natoumbi Mendy 1999 Iceland; Shelf Completed Mendy (unpublished data) Iceland, Shelf Completed Mendy and Buchary 2001 India, Arabian Sea off Karnataka (2) Completed Mohamed et al. 2006 India, Pullavali Brackishwater Completed Santhanam et al. 1993 India, Southwest coast Completed Vivekanandan et al. 2003 India, Veli Lake Completed Aravindan 1993 219 Indian Ocean, West-central Completed Ph.D. thesis in preparation Indonesia, Eastern In progress Indonesia, Papua, Raja Ampat, Kabui Bay Completed Bailey et al. 2007 Indonesia, Papua, Raja Ampat Islands Completed Ainsworth et al. 2007 Ireland, North, Lough Neagh In progress Ph.D. thesis in preparation Irish Sea Completed Lees and Mackinson 2007 Irish Sea Completed Ph.D. thesis in preparation Israel, Lake Kinneret Completed WallineeJa/. 1993 Italy, Adriatic Sea, North Completed Zucchetta et al. 2003 Italy, Adriatic Sea, North and Central In progress Coll et al. (in preparation a, b) Italy, Adriatic Sea, North and Central (2) Completed Coll et al. 2006; Coll et al. (in press a, b) Italy, Adriatic Sea, Upper, Miramare marine reserve Completed Libralato et al. (2006) Italy, Orbetello Lagoon Completed Ceccarelli et al. 2005 Italy, Orbetello Lagoon (2) Completed Brando et al. 2004 Italy, Venice Lagoon Completed Pranovi et al. 2003; Libralato et al. 2002; Granzotto et al. 2004 Ivory Coast, Lagoon Ebrie Completed McNamara 2004 Ivory Coast, Lagoon Ebrie Completed Ph.D. thesis in preparation Japan Completed Okamura et al. 2002 Japan, Hokkaido, Lake Toya Completed Matsuishi et al. 2005 Japan, Nagasaki, Lake Kawahara-oike Completed Matsuishi et al. 2005 Japan, Northern Pacific coast Completed Hamatsu, internal report (unpublished) Japan, Seto Inland Sea In progress Model completed, but no publication Kenya, Lake Naivasha Completed Mavuti etal. 1996 Kenya, Lake Nakuru (2) Completed Moreau et al. 2001a Kenya, Lake Turkana Completed Kolding 1993 Kenya, Lake Victoria Completed Moreau et al. 1993b Kenya, Lake Victoria Completed Kipkemboi 2006 Kerguelen Island Completed Pruvost et al. 2005 Korea, Gwangyang bay Completed Kang 2005 Lake Erie (2) In progress In preparation (Ecological Modelling) Lake Huron Completed Mason et al. 2004 220 Lake Michigan In progress In preparation Lake Ontario •  Completed Halfon and Schito 1993 Lake Ontario, Bay of Quinte (2) Completed Koops et al. 2006; Koops et al. (in press) Lake Superior Completed Uoffetal. 2000 Lake Superior Completed Kitchell et al. 2000 Laos, Nam Ngum Completed Jutagate et al. 2002 Lesser Antilles, Pelagic ecosystem In progress Lithuania, coastal zone Completed Bucas 2004 Madagascar, Antongil Bay Completed Ph.D. thesis in preparation Madagascar, Coastal stations Completed Ph.D. thesis in preparation Malawi, Lake Malawi Completed Degnbol 1993 Malawi, Lake Malawi Completed Nsiku 1999; Nsiku 2002 Malaysia, east coast Completed Alias 2003 Malaysia, Sabah Completed M.Sc. thesis in preparation Malaysia, Sabah Completed Garces et al. 2003 Malaysia, Sarawak Completed Garces et al. 2003 Malaysia, Terengganu Completed Liew and Chan 1987 Malaysia, west coast Completed Christensen et al. 2003 Mauritania, EEZ Completed Sidi and Guenette 2004 Mauritania, Banc d'Arguin Completed Sidi and Samba 2003 Mediterranean Sea, Corsica, Bay of Calvi Completed Pinnegar and Polunin 2004 Mediterranean Sea, Northwest (2) Completed Coll et al. (in press b) Mexico, Baja California Sur, La Paz Bay Completed Arreguin-Sanchez et al. 2004a Mexico, Baja California tip, Central zone In progress Mexico, Baja California, La Paz Bay Completed Ph.D. thesis in preparation Mexico, Central Gulf of California Completed Arreguin-Sanchez and Calderon- Aguilera 2002 Mexico, Continental shelf of Tabasco Completed M.Sc. thesis in preparation Mexico, Gulf of California, benthic ecosystem Completed Arreguin-Sanchez et al. 2002a, b Mexico, Gulf of Mexico, Ecosystem management Completed Walters et al. 2006 Mexico, Gulf of Mexico, North Continental Shelf Completed Arreguin-Sanchez and Manickchand-Heileman 1998 Mexico, Gulf of Mexico, Southwestern Completed Manickchand-Heileman et al. 1998a Mexico, Gulf of Mexico, Southwestern Campeche Completed Arreguin-Sanchez et al. 2004ft 221 Mexico, Gulf of Mexico, Terminos Lagoon Completed Manickchand-Heileman et al. 1998b Mexico, Gulf of Ulloa Completed del Monte-Luna 2004 Mexico, Huizache-Caimanero Lagoon Completed Zetina-Rejon et al. 2004 Mexico, Magdalena Bay, Channels zone In progress Model completed, but no publication Mexico, Mexican Caribbean fringing coral reefs Completed Ph.D. thesis in preparation Mexico, Northern Gulf of California Completed Morales-Zarate et al. 2004 Mexico, Oaxaca, Tonameca catchment Completed Avila-Foucat et al. 2004 Mexico, Quintana Roo, Bahia de la Ascension Completed Vidal and Basurto 2003 Mexico, South Caribbean Completed Arias- Gonzalez et al. 1998 Mexico, South Sinaloa In progress Salcido-Guevara (in preparation) Mexico, Southern Gulf of Mexico, Terminos Completed Rivera-Arriaga et al. 2003 Lagoon Mexico, Southwestern Gulf of Mexico, reef lagoon In progress Model completed, but no publication Mexico, Tamiahua Lagoon Completed Abarca-Arenas and Valero- Pacheco 1993 Mexico, Veracruz, Alvarado lagoon system Completed Abarca-Arenas and Valero- Pacheco 1993; Abarca-Arenas et al. 2004; Arreguin-Sanchez and Valero 1996 Mexico, Veracruz, Alvarado lagoon system In progress Model completed, but no publication Mexico, Veracruz, Mandinga Lagoon Completed De La Cruz-Aguero 1993 Mexico, Veracruz, Tampamachoco lagoon Completed Rosado-Solorzano and Guzman delProo 1998 Mexico, Yucatan Peninsula, Boca Paila-Tampalam- Completed Arias-Gonzalez et al. 2004 Mah Mexico, Yucatan Peninsula, Celestun Lagoon Completed Chavez et al. 1993 Mexico, Yucatan Peninsula, Mangroves (2) Completed Vega-Cendejas and Arreguin- Sanchez 2001; Vega-Cendejas 2003 Mexico, Yucatan, Northern Continental Shelf Completed Vega-Cendejas and Arreguin- Sanchez 2001 Micronesia, Enewetak Atoll Completed Dalsgaard 1998 222 Morocco, coast Completed Stanford et al. 2001 Mozambique, Maputo Bay Completed DePaulaefa/. 1993 Mozambique, Mozambique channel Completed Ph.D. thesis in preparation Netherlands, Ijsselmeer Lake Completed Buijseef al. 1993 New Caledonia, Coral reef lagoon Completed Bozec et al. 2004 New Caledonia, Loyalty Islands, Uvea Completed Bozec et al. 2004 New Caledonia, western tropical Pacific Completed Godinot and Allain 2003 New Zealand, Small coastal region In progress New Zealand, Southeast, Southern plateau Completed Bradford-Grieve et al. 2003 North Sea Completed Christensen 1995; Christensen et al. 2002 North Sea Completed Silvert et al, submitted Norway, Sorfjord Completed Pedersen et al. (submitted) Norwegian Sea Completed Skaret and pitcher, in press Norwegian Sea & Barents Sea Completed Dommasnes etal. 2001 Pacific Basin, Subarctic Completed Aydin et al. 2003 Pacific coast mangrove Completed Cunningham 2003 Pacific Ocean, Central Completed Cox et al. 2002 Pacific Ocean, Eastern Completed Preikshot 2005 Pacific Ocean, North Central Completed Kitchell et al. 2002 Pacific Ocean, Pelagic eastern tropical Completed Olson and Watters 2003; Olson and Watters 2003; Hinke et al. 2004; Olson et al. 2002; Watters et al. 2003 Peru, Independencia Bay In progress Wolff et al. (in preparation b) Philippines, Central Visayas, Bohol Island In progress Model completed, but no publication Philippines, Eastern Visayas, Leyte Gulf In progress Model completed, but no publication Philippines, Central Java, north coast Completed Nurhakim 2003 Philippines, Laguna de Bay (2) Completed Delos Reyes and Martens 1993; Delos Reyes 1995 Philippines, Leyte, San Pedro Bay Completed Campos 2003 Philippines, Northwest coast, Lingayen Gulf Completed Guarin 1991; Pauly and Christensen 1993 Philippines, Pangasinan, Bolinao Completed AMo etal. 1993 223 Philippines, San Miguel Bay Completed Bundy 1997; Bundy 2004b; Bundy and Pauly 2001 Philippines, Western Visayas, Panay Island, Sapian Completed Armada and Bacalso, 2004 Bay Philippines, wetland ricefield Completed Lightfoot etal. 1993 Poland, Gdansk basin Completed Wielgat 2004 Portugal, Azores archipelago Completed Guenette and Morato 2001 Portugal, South, Ria Formosa coastal lagoon, Water Completed Gamito and Erzini 2005 Reservoir Rwanda, Lake Ihema Completed Mavutietal. 1996 Sahel, artificial lakes "? Completed Model completed, but no publication Senega., coast Completed Samb and Mendy 2003 Senegal, Saloum estuary Completed Ph.D. thesis in preparation Shanghai In progress Sierra Leone Completed Heymans and Vakily 2002 South Africa, Benguela Completed Shannon et al. 2004a South Africa, Kromme estuary Completed Heymans and Baird 1995 South Africa, Northern Benguela Completed Heymans and Baird 2000 South Africa, Northern Benguela Completed Jarre-teichmann 1998 South Africa, Northern Benguela Completed Shannon and Jarre-teichmann 1999 South Africa, Saldanha Bay In progress South Africa, Southern Benguela Completed Shannon et al. 20046 South Africa, Sundays beach Completed Heymans and McLachlan 1996 South Georgia, South Orkney Islands Completed Bredesen 2003 Southern hemisphere (diatoms) In progress Model completed, but no publication Spain, Cantabrian Sea Completed Sanchez and Olaso 2004 Spain, Medes island Completed Ph.D. thesis in preparation Sri Lanka, coast Completed Ph.D. thesis in preparation Sri Lanka, Parakrama Samudra reservoir Completed Moreau etal. 2001b Taiwan, Kuo Sheng Bay Completed Lin et al. 2004 Taiwan, Southwestern Chiku Completed Lin etal. 1999 Taiwan, Tsengwen estuary Completed KupandShao 1999 Thailand, Pasak Dam In progress 224 Thailand, Sirindhorn reservoir Completed Jutagate et al. 2002 Thailand, Ubol Ratana Reservoir Completed Chookajorn et al. 1994 Trinidad / Venezuela, Gulf of Paria Completed Manickchand-Heileman et al. 2004 Uganda, Lake George Completed Moreau et al. 1993c United Kingdom, English Channel Completed Stanford and Pitcher 2004 United Kingdom, River Thames Completed Mathews 1993 United Kingdom, Scotland, East coast Completed Buchan 1997 United Kingdom, Scotland, West coast Completed Morissette and Pitcher 2005 Uruguay, Coastal lagoon (2) In progress USA, Alabama, Week Bay Completed Althauser 2003 USA, Alaska, Bercharof Lake Completed Mathisen and Sands 1999 USA, Alaska, Bering Sea Completed Aydin et al. 2002 USA, Alaska, Bering Sea Completed Trites et al. 1999 USA, Alaska, Bering Sea, East Completed Livingstone^/. 1999 USA, Alaska, Bering Sea, East Completed National Research Council 2003 USA, Alaska, Bering Sea, Southeast Completed Ciannelli et al. 2004 USA, Alaska, Bering Sea, West Completed Livingston et al. 1999 USA, Alaska, Lake Illiamma Completed Anonymous 2003 USA, Alaska, Prince William Sound Completed Okey and Pauly 1999 USA, Alaska, Prince William Sound (pre-oil spill) Completed Dalsgaard and Pauly 1997 USA, Alaska, South East (3) Completed Guenette 2005 USA, Atlantic continental shelf Completed Okey and Publiese 2001 USA, California, Monterey Bay Completed Olivieri et al. 1993 USA, California, Northern current Completed Field et al. 2001; Field and Francis 2005 USA, California, Pt. Reyes In progress USA, Chesapeake Bay In progress USA, Chesapeake Bay In progress USA, Chesapeake Bay (4) Completed Hagy 2002 USA, Columbia River Completed Harvey and Kareiva 2005 USA, Delaware Bay Completed Frisk 2007 USA, Everglades Completed Heymans et al. 2002 USA, Florida freshwater lakes In progress USA, Florida, Apalachicola Bay Completed Carlson, (in press) USA, Florida, Looe Key Completed Venier and Pauly 1997 225 USA, Florida, St. Marks Completed Baird et al. 1998; Christian and Luczkovich 1999; Luczkovich et al. 2002 USA, Florida, west shelf Completed Okey and Mahmoudi 2002 USA, Florida, west shelf Completed Okey et al. 2004b USA, Gulf of Alaska Completed Heymans 2005a USA, Gulf of California (2) Completed Ph.D. thesis in preparation USA, Gulf of California, North Completed Ph.D. thesis in preparation USA, Gulf of California, South, Coastal lagoon In progress M.Sc. thesis in preparation USA, Gulf of Maine Completed Heymans 2002 USA, Hawaiian Island chain Completed Papastamatiou et al. (in press) USA, Iowa lakes (3) In progress Model completed, but no publication USA, Iowa, Clear Lake In progress None (class project) USA, Lake Michigan, Green Bay In progress USA, Long Island Sound - nearshore In progress USA, Long Island Sound - offshore In progress USA, Mid-Atlantic bight Completed Okey 2001 USA, NC, Lower Neuse River Estuary Completed Christian et al. 2003; Dame and Christian 2006 USA, New York, Oneida Lake Completed Koops et al. 2006; Koops et al, in prep. USA, North Carolina Completed Butler 2007 USA, North Carolina Reservoir In progress USA, North Carolina, Neuse River, Contentnea Completed Pine and Kwak 2007 Creek USA, Northern Wisconsin freshwater lakes Completed Fayram 2005; Fayram et al, in press USA, Oneida Lake Completed Irwin et al. 2003 USA, Salt marsh ponds, Virginia Coast reserve Completed Dame 2005 LTER site USA, South Carolina, Charleston wathershed In progress USA, Texas, Laguna Madre In progress USA, Washington, Straight of Juan de Fuca In progress Venezuela, Northeastern Shelf Completed Mendoza 1993 Vietnam, coast Completed Pauly and Christensen 1993 226 Vietnam, Dong Nai Province, Tri A n Reservoir Completed Luong 2000 Vietnam, southwest Completed Christensen et al. 2003 Zimbabwe, Lake Kariba Completed Machenaefa/. 1993 References Abarca-Arenas, L.G. and E. Valero-Pacheco. 1993. Toward a trophic model of Tamiahua, a coastal lagoon in Mexico, pp. 181-185 In Christensen, V. and D. Pauly (eds.). Trophic models of aquatic ecosystems. ICLARM, Manila. Abarca-Arenas, L.G; Franco-Lopez, J; Chavez-Lopez, R; Arceo-Carranza, D., and A. Moran-Silva. 2004. Trophic analysis of the fish community taken as bycatch of shrimp trawls of the coast of Alvarado, Mexico. Proceedings of the Gulf and the Caribbean Fisheries Institute 55:384-394. Ainsworth, C , Ferris, B., Leblond, E., and Guenette, S., 2001. The Bay of Biscay, France; 1998 and 1970 models, pp. 271-313 In Guenette, S., Christensen, V., and D. Pauly (eds.). Fisheries impacts on North Atlantic Ecosystems: Models and analyses. Fisheries Centre Research Reports 9(4). Ainsworth, C , Heymans, J.J., Pitcher, T.J., and M. Vasconcellos (eds.). 2002. Ecosystem models of Northern British Columbia for the time periods 2000, 1950, 1900 and 1750. Fisheries Centre Research Reports 10(4). Ainsworth, C. Varkey D., and T. Pitcher. 2007. Ecosystem simulation models for the Bird's Head seascape, Papua, fitted to data, pp 6-174 In Pitcher, T.J., Ainsworth, C.H., and M. Bailey (eds.) Ecological and economic analyses of marine ecosystems in the Birtd's Head seascape, Papua, Indonesia: I. Fisheries Centre Research Reports 15(5). Alias, M. 2003. Trophic Model of the coastal fisheries ecosystem of the west coast of penisular Malaysia, pp. 313-332 In Silvestre, G., Garces, L., Stobutzki, I., Ahmed, M., Valmonte-Santos, R.A., Luna, C , Lachica-Alino, L., Munro, P., Christensen, V., and D. Pauly (eds.) Assessment, management and future directions for coastal fisheries in Asian countries. WorldFish Center Conference Proceedings 67. Alino, P.M., McManus, L.T., McManus, J.W., Nanola Jr., C.L. Fortes, M.D., Trono Jr., G.C., and G.S. Jacinto. 1993. Initial parameter estimations of a coral reef flat 227 ecosystem in Bolinao, Pangasinan, Northwestern Phillipines. pp. 252-258 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, > Manila. Althauser, L.L. 2003. An Ecopath/Ecosim Analysis of an Estuarine Food Web: Seasonal Energy Flow and Response to River-Flow Related Perturbations. Ph.D. thesis, Louisiana State University. Alvarez-Hernandez, J.H. 2003. Trophic model of a fringing coral reef in the Southern Mexican Caribbean, pp. 227-235 In Zeller, D., Booth, S., Mohammed, E., and D. Pauly (eds.) From Mexico to Brazil: Central Atlantic fisheries catch trends and ecosystem models. Fisheries Centre Research Reports 11(6). Amorim, P., Duarte, G., Guerra, M., Morato, T., and K.A. Stobberup. 2003. Preliminary Ecopath model fo the Guinea-Bissau continental shelf ecosystem (NW-Africa). pp. 73-90 In Pauly, D., Palomares, M.L.D., and J.M. Vakily. Trophic models of Northwest African ecosystems. Systeme d'Information et d'Analyses des Peches SIAP/EP/DT/03. Angelini, R. 2002. Avaliacao da capacidade-suporte da represa do Broa para a colocacao de tanques-rede. Revista Saude e Ambiente, Joinville (SC) 3: 42-48. Angelini, R. and M. Petrere Jr. 1996. The ecosystem of Broa Reservoir, Sao Paulo State, Brazil, as described using Ecopath. Naga 19: 36-41. Angelini, R., Fabre, N.N., and U.L. da Silva-JR. 2006. Trophic analysis and fishing simulation of the biggest Amazonian catfish. African Journal of Agricultural Research 1(5): 151-158. Anonymous. 2003. Kvichak River Sockeye Salmon stock status and action plan, 2003. A report to the Alaska Board of Fisheries. Regional Information Report No. 2A03-26. Aravindan, C M . 1993. Preliminary trophic model of Veli Lake, southern India, pp. 87-89 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Arias- Gonzalez, J.E., Delesalle, B:, Salvat, B., and R. Galzin. 1997. Trophic functioning of the Tiahura reef sector, Moorea Island, French Polynesia. Coral Reefs 16:231- 246. 228 Arias- Gonzalez, J. E. Delesalle, B. Salvat, B., and R. Galzin. 1998. Trophic models of protected and unprotected coral reef ecosystems in the South of the Mexican Caribbean. Journal of Fish Biology 53: 236-255. Arias-Gonzalez, J.E., Nunez-Lara, E., Gonzalez-Salas, C , and R. Galzin. 2004. Trophic models for investigation of fishing effect on coral reef ecosystems. Ecological Modelling 172: 197-212. Armada, N.B. and R.T.M. Bacalso. 2004. A preliminary trophic model of a shallow bay ecosystem: Sapian Bay Philippines. University of the Philippines Visayas (UPV) Journal of Natural Science. 9: 184-204 Arreguin-Sanchez, F. and E. Valero. 1996. Trophic role of the red grouper (Epinephelus morio) in the ecosystem of the northern continental shelf of Yucatan, Mexico, pp. 19-27. In Arreguin-Sanchez, F., Munro, J.L., Balgos, M.C., and D. Pauly (eds.) Biology, fisheries and culture of tropical groupers and snappers. ICLARM Conference Proceedings 48. Arreguin-Sanchez, F. and S. Manickchand-Heileman. 1998. The trophic role of lutjanid fish and impacts of their fisheries in two ecosystems in the Gulf of Mexico. Journal of Fish Biology 53:143-153. Arreguin-Sanchez, F. and L.E. Calderon-Aguilera. 2002a. Evaluating harvesting strategies for fisheries in the Gulf of California ecosystem, pp 135-141 In Pitcher, T.J. and K. Cochrane (eds.). The Use of Ecosystem Models to Investigate Ecosystem-based Management Strategies for Capture Fisheries; F AO/Fisheries centre workshop. Fisheries Centre Research Reports 10(2). Arreguin-Sanchez, F. Arcos, E., and E.A. Chavez. 20026. Flows of biomass and structure in an exploited benthic ecosystem in the Gulf of California, Mexico. Ecological Modelling 156:167-183. Arreguin-Sanchez, F., Hernandez-Herrera, A., Ramirez-Rodriguez, M., and H. Perez- Espana. 2004a. Optimal management scenarios for the artisanal fisheries in the ecosystem of La Paz Bay, Baja California Sur, Mexico. Ecological Modelling 172: 373-382. 229 Arreguin-Sanchez, F., Zetina-Rejon, M., Manickchand-Heileman, S., Ramirez-Rodriguez, M., and L. Vidal. 2004b. Simulated response to harvesting strategies in an exploited ecosystem in the southwestern Gulf of Mexico. Ecological Modelling 172: 421-432. Avila-Foucat, V.S., Raffaelli, D., and C. Perrings. 2004. Ecological economic modelling for integrating environmental services in the welfare of commons: a case study in Tonameca catchment, Oaxaca, Mexico. Oral presentation at "The Commons in an Age of Global Transition: Challenges, Risks and Opportunities," the Tenth Conference of the International Association for the Study of Common Property, Oaxaca, Mexico, August 9-13. Aydin, K., Gaichas, S., Ortiz, I., Kinzey, D., and N. Friday, (submitted). A comparison of the Bering Sea, Gulf of Alaska, and Aleutian Islands large marine ecosystems through food web modeling. NOAA Technical Memorandum. Aydin, K.Y., Lapko, V.V., Radchenko, V.I., and P.A. Livingston. 2002. A comparison of the eastern Bering and western Bering Sea shelf and slope ecosystems through the use of mass-balance food web models. NOAA Technical Memorandum NMFS- AFSC-130. Aydin, K.Y., McFarlane, G.A., King J.R., and B.A. Megrey. 2003. The BASS/MODEL report on trophic models of the subarctic Pacific basin ecosystems. North Pacific Marine Science Organization (PICES), Sidney, B.C., Canada. Bailey, M., Rotinsulu, C , and U.R. Sumaila 2007. The migrant anchovy fishery in Kabui Bay, Raja Ampat,Indonesia: catch, profitability and income distribution, pp 175-184 In Pitcher, T.J., Ainsworth, C.H., and M. Bailey (eds.) Ecological and economic analyses of marine ecosystems in the Birtd's Head seascape, Papua, Indonesia: I. Fisheries Centre Research Reports 15(5). Baird, D., Luczkovich, J., and R.R. Christian. 1998. Assessment of spatial and temporal variability in ecosystem attributes of St Marks National Wildlife Refuge, Apalachee Bay, Florida. Estuarine, Coastal and Shelf Science 47: 329-249. Beattie, A. 2001. A new model for evaluating the optimal size, placement and configuration of marine protected areas. M.Sc. thesis, University of British Columbia, Canada. 230 Blanchard, J.L., Pinnegar, J.K., and S. Mackinson. 2002. Exploring marine mammal - fishery interactions using 'Ecopath with Ecosim': modelling the Barents sea ecosystem. CEFAS Science Series Technical Report No. 117. Boet, P., Leger M., and C. Le Pichon. 1999. Modelisation du compartiment piscicole : Test d'un modele bioenergetique pour mieux center la biomasse piscicole d'un bief de la Seine. Programme Interdisciplinaire de Recherche sur l'Environnement de la Seine, rapport d'activite. Booth, S. and D. Zeller. 2005. Mercury, food webs, and marine mammals: implications of diet and climate change for human health. Environmental Health Perspectives 113:521-526. Bozec, Y.-M., Gascuel, D., and M. Kulbicki. 2004. Trophic model of lagoonal communities in a large open atoll (Uvea, Loyalty islands, New Caledonia). Aquatic Living Resources 17: 151-162. Bradford-Grieve, J.M., Probert, P.K., Baker, A.N., Best, H.A., Boyd, P., Broekhuizen, N., Childerhouse, S., Clark, M., Hadfield, M., Hall, J.A., Hanchet, S., Nodder, S.D., Safi, K., Thompson, D., Wilkinson, I., and J. Zeldis. 2003. Pilot trophic model for subantarctic water over the Southern Plateau, New Zealand: a low biomass, high efficiency system. Journal of Experimental Marine Biology and Ecology 289: 223- 262. Brando, V.E., Ceccarelli, R., Libralato, S., and G.P. Ravagnan. 2004. Assessment of environmental management effects in a shallow water basin using mass-balance models, Ecological Modelling 172: 213-232. Bredesen, E.L. 2003. Krill and the Antarctic: finding the balance. M. Sc., University of British Columbia, Canada. Bucas, M. 2004. Trophic model of economical coastal zone fo Lithuania, pp. 19-20 In Baltic Sea Regional Project (eds.) Workshop on Ecopath modelling of the Baltic Sea carbon and nutrient networks, Oct. 18-22, 2004, urmala, Latvia. Buchan, R.G. 1997. A comparative study of two East Scottish estuaries, the Lower Forth and Ythan, using the ECOPATH trophic interaction model. M.Sc. Thesis, Napier University. 231 Buchary, E.A. 2001. Preliminary reconstruction of the Icelandic marine ecosystem in 1950 and some predictions with time series data. pp. 296-306 In Guenette, S., Christensen, V., and D. Pauly (eds.). Fisheries impacts ofh the North Atlantic ecosystems: Models and analyses. Fisheries Centre Research Reports 9(4). Buijse, A.D., Van Eerden, M.R., Dekker, W., and W.L.T. Van Densen. 1993. Elements of a trophic model for Usselmeer (The Netherlands), a shallow eutrophic lake. pp. 90- 94 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Bulman, C M . 2002. Trophic ecology and food web modelling of mid-slope demersal fishes off Southern Tasmnia, Australia. Ph.D. thesis. University of Tasmania, Australia. Bulman, C M . 2006. Trophic webs and modelling of the North West Shelf. North West Shelf Joint Environmental Management Study Technical Report No 9. Bulman, C M . , Butler, A.J., Condie, S., Ridgway, K., Koslow, J.A., He, X., Williams, A., Bravington, M., Stevens, J.D. and J.W. Young. 2002. A trophodynamic model for the Tasmanian Seamounts Marine Reserve: links between pelagic and deepwater ecosystems. Report to CSIRO Marine Research Environment Australia and the Fisheries Research and Development Corporation: Hobart and Canberra. Bulman, C , Condie, S., Furlani, D., Cahill, M., Klaer, N., Goldsworthy, S., and I. Knuckey. 2006. Trophic dynamics of the eastern shelf and slope of the south east fishery: impacts of and on the fishery. Final Report for the Fisheries Research and Development Corporation, Project 2002/028. Bundy, A. 1997. Assessment and management of multispecies, multigear fisheries: A case study from San Miguel Bay, the Philippines. Ph.D. Thesis. University of British Columbia, Canada. Bundy, A. 2001. Fishing on ecosystems: the interplay of fishing and predation in Newfoundland-Labrador. Canadian Journal of Fisheries and Aquatic Sciences 58: 1153-1167. Bundy, A. 2002. Exploring multispecies harvesting strategies on the eastern Scotian Shelf with Ecosim. pp. 112-117 In Pitcher, T.J.P. and K. Cochrane (eds.). The use of ecosystem models to investigate multispecies management strategies for capture 232 fisheries, FAO/UBC Fisheries Centre Workshop July 17-July 20 2000. Fisheries Centre Research Reports. 10(2). Bundy, A. 2004a. Mass balance models of the eastern Scotian Shelf before and after the cod collapse and other ecosystem changes. Canadian Technical Report on Fisheries and Aquatic Sciences 2520. Bundy, A. 2004b. The ecological effects of fishing and implications for coastal management in San Miguel Bay, the Philippines. Coastal Management 32: 25-38. Bundy, A. 2005. Structure and function of the eastern Scotian shelf Ecosystem before and after the groundfish collapse in the early 1990s. Canadian Journal of Fisheries and Aquatic Sciences 62: 1453-1473. Bundy, A. and D. Pauly 2001 Selective harvesting by small-scale fisheries: ecosystem analysis of San Miguel Bay, Philippines. Fisheries Research 53: 263-281. Bundy, A. and P. Fanning. 2005. Can Atlantic cod recover? Exploring trophic explanations for the non-recovery of cod on the eastern Scotian Shelf, Canada. Canadian Journal of Fisheries and Aquatic Sciences 62: 1474-1489. Bundy, A., Lilly, G., and P. Shelton, 2000 A mass balance model of the Newfoundland- Labrador shelf. Canadian Technical Report of Fisheries and Aquatic Sciences 2310. Butler, C M . 2007. Atlantic Bluefin Tuna (Thunnus thynnus) Feeding Ecology and Potential Ecosystem Effects During Winter in North Carolina. M. Sc. Thesis, North Carolina State University. Campos, W. L. 2003. An ecosystem model of San Pedro Bay, Leyte, Philippines: initial parameter estimates, pp. 353-364 In Silvestre, G., Garces, L., Stobutzki, I., Ahmed, M., Valmonte-Santos, R.A., Luna, C , Lachica-Alino, L., Munro, P., Christensen, V., and D. Pauly (eds.) Assessment, management and future directions for coastal fisheries in Asian countries. WorldFish Center Conference Proceedings 67. Cardinale, M. and M. Casini. 2004. Ecosim model of the open water of the Baltic Proper. pp. 10-14 In Baltic Sea Regional Project (eds.) Workshop on Ecopath modelling of the Baltic Sea carbon and nutrient networks, Oct. 18-22, 2004, urmala, Latvia. Carlson, J.K. (in press). Modeling the role of sharks in the trophic dynamics of Apalachicola Bay, Florida. American Fisheries Society Symposium Series. 233 Ceccarelli, R., Moretti, F., Bardi, A., Piazzi, A., and M. Giunti. 2005. Fishery management and cormorant impact in the Orbetello Lagoon (Central Italy). 5th European Vertebrate Pest Management Conference - Budapest September 2005. Chavez, E.A., Garduno, M., and F. Arreguin-Sanchez. 1993. Trophic dynamic structure of Celestun Lagoon, Southern Gulf of Mexico, pp. 186-192 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Chen, Z., Qiu, Y., and X. Jia. 2006. Mass-balance ecopath model of Beibu Gulf ecosystem. Chinese journal of applied ecology. 17(6): 1107-1111. Cheung, W.W.L. and T.J. Pitcher. 2005. A mass-balance model of the Falkland Islands fisheries and ecosystems, pp 65-84 In Palomares, M.L.D., Pruvost, P., Pitcher, T.J., and D. Pauly. Modeling Antarctic marine ecosystems. Fisheries Centre Research Reports 13(7). Chookajorn, T., Leenanond, Y., Moreau, J., and B. Sricharoendam. 1994. Evolution of trophic relationships in Ubolratana reservoir (Thailand) as described using a multispecies trophic model. Asian Fisheries Sciences 7: 201-213 Christensen, V. 1995. A model of trophic interactions in the North Sea in 1981, the Year of the Stomach. DANA 11(1):1-19. Christensen, V. 1998. Fishery-induced changes in a marine ecosystem: insight from models of the Gulf of Thailand. Journal of Fish Biology 53: 128 - 142. Christensen, V., Beyer, J.E., Gislason, H., and M. Vinter. 2002. A comparative analysis of the North Sea based on Ecopath with Ecosim and multispecies virtual population analysis, pp 39 In Christensen, V., Reck, G., and J.L. Maclean. Proceedings of the INCO-DC Conference Placing Fisheries in their Ecosystem Context, Galapagos Islands, Ecuador, 4-8 December 2000. ACP-EU Fisheries Research Report. Christensen, V.T., L.R. Garces, G.T. Silvestre and D. Pauly. 2003. Fisheries impact on the South China Sea large marine ecosystem: A preliminary analysis using spatially- explicit methodology, pp. 51-62. In Silvestre, G., Garces, L., Stobutzki, I., Ahmed, M., Valmonte-Santos, R.A., Luna, C , Lachica-Alino, L., Munro, P., Christensen, V., and D. Pauly (eds.) Assessment, management and future directions for coastal fisheries in Asian countries. WorldFish Center Conference Proceedings 67. 234 Christensen, V., Walters, C.J., and D. Pauly. 2005. Ecopath with Ecosim: a user's guide. Fisheries Centre, University of British Columbia, Vancouver. November 2005 edition (available online at http://www.ecopath.org). Christian, R.R. and J.J. Luczkovich. 1999. Organizing and understanding a winter's seagrass foodweb network through effective trophic levels. Ecological Modelling 117: 99-124. Christian, R.R., Dame, J.K., Johnson, G., Peterson, C.H., and D. Baird. 2003. Monitoring and modeling of the Neuse River Estuary, phase 2: Functional assessment of environmental phenomena through network analysis. University of North Carolina - Water Resources Research Institute Report 343-E. Ciannelli, L., Robson, B.W., Francis, R.C., Aydin, K., and R.D. Brodeur. 2004. Boundaries of open marine ecosystems: An application to the Pribilof Archipelago, Southeast Bering Sea. Ecological Applications 14(3): 942-953. Coll, M., Palomera, I., Tudela, S., and F. Sarda. 2006. Trophic flows, ecosystem structure and fishing impacts in the South Catalan Sea, Northwestern Mediterranean. Journal of Marine Systems 59: 63-96. Coll, M., Santojanni, A., Arneri, E., and I. Palomera. (in press a). An ecosystem model of the northern and central Adriatic Sea: analysis of ecosystem structure and fishing impacts. Biologia Marina Mediterranea. Coll, M., Shannon, L.J., Moloney, C.L., Palomera, I., and S. Tudela. (in press b). Comparing trophic flows and fishing impacts of a NW Mediterranean ecosystem with coastal upwellings by means of standardized ecological models and indicators. Ecological Modelling. Coll, M., Palomera I., Tudela, S., and F. Sarda. (in preparation a). Assessing the impact of fishing and environment on a Northwestern Mediterranean ecosystem for the period 1978-2003. Coll, M., Santojanni, A., Arneri, E., and I. Palomera. (in preparation b). An ecosystem model of the northern and central Adriatic Sea. Cox, S.P., Essington, T.E., Kitchell, J.F., Martell, S.J.D., Walters, C.J., Boggs, C , and I. Kaplan. 2002. Reconstructing ecosystem dynamics in the central Pacific Ocean, 1952-1998. II. A preliminary assessment of the trophic impacts of fishing and effect 235 on Tuna Dynamics. Canadian Journal of Fisheries and Aquatic Sciences 59: 1736- 1747. Criales-Hernandez, M.I., Duarte, L.O., Garcia, C.B., and L. Manjarres. 2006. Ecosystem impacts of the introduction of bycatch reduction devices in a tropical shrimp trawl fishery: Insights through simulation. Fisheries research 77: 333-342. Cunningham, L. 2003. Trophic structure and stability of a modelled mangrove ecosystem. M.Sc. Thesis, University of York. Dalsgaard, A.J.T. 1998. Modeling the trophic transfer of beta radioactivity in the marine food web of Enewetak Atoll, Micronesia. M. Sc. thesis, University of British Columbia. Dalsgaard, J. and D. Pauly (eds.). 1997. Preliminary mass-balance model of Prince William Sound, Alaska, for the pre-spill period 1980-1989. Fisheries Centre Research Reports 5(2). Dame, J.K. 2005. Evaluation of ecological network analysis for ecosystem-based management. Ph.D. Thesis. East Carolina University. Dame, J.K. and R.R. Christian. 2006. Uncertainty and the use of network analysis for ecosystem-based fishery management. Fisheries 31:331-341. De La Cruz-Aguero, G. 1993. A preliminary model of Mandinga Lagoon, Veracruz, Mexico, pp. 193-196 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. De Paula E Silva, R., Sousa, M.I., and A.M. Caramelo. 1993. The Maputo Bay ecosystem (Mozambique), pp. 214-223 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Degnbol, P. 1993. The pelagic zone of central Lake Malawi - A trophic box model, pp. 110-115 7« Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila, del Monte-Luna, P. 2004. Characterization of the biological activity center of the Gulf of Ulloa, Mexico, under an ecological modelling context. Ph.D. thesis, Instituto Politecnico Nacional, Centro Interdisciplinario de Ciencias Marinas, Mexico. Delos Reyes, M. 1995. Geoecology of Laguna de Bay, Philippines: long-term alterations of a tropical-aquatic ecosystem 1820-1992. Ph.D. thesis, University of Hamburg. 236 Delos Reyes, M. and R. Martens. 1993. Geoecology of Laguna de Bay, Philippines. I. Techno-commercial impact on the trophic level structure of the Laguna de Bay aquatic ecosystem 1968-1980. Ecological modelling 75/76: 497-509. Diallo, I., Cisse, I., and A. Bah. 2003. Modele trophique du systeme cotier du plateau continental Guineen. pp. 91-108 In Pauly, D., Palomares, M.L.D., and J.M. Vakily. Trophic models of Northwest African ecosystems. Systeme d'Information et d'Analyses des Peches SIAP/EP/DT/03. Dommasnes, A., Christensen, W., Ellertsen, B., Kvamme, C , Melle, W., Nattestad, L., Pedersen, T., Tjelmeland, S., and D. Zeller, D. 2001. An Ecopath model for the Norwegian Sea and Barents Sea. pp. 213-240 In Guenette, S., Christensen, V., Pitcher, T., and D. Pauly (eds.). Fisheries impacts on North Atlantic Ecosystems: Models and analyses. Fisheries Centre Research Reports 9(4). Duarte, L.O. and C.B. Garcia. 2004. Trophic role of small pelagic fishes in a tropical upwelling ecosystem. Ecological Modelling 172: 323-338. Erfan, A. and T.J. Pitcher. 2005. An ecosystem simulation model of the Antarctic peninsula, pp 5-20 In Palomares, M.L.D., Pruvost, P., Pitcher, T.J., and D. Pauly. Modeling Antarctic marine ecosystems. Fisheries Centre Research Reports 13(7). Falk-Petersen, J. (in preparation). Ecosystem effects of red king crab invasion - a modelling approach using Ecopath with Ecosim. M.Sc. Thesis, University of Tromsa. FAO/FISHCODE. 2001. Report of a bio-economic modeling workshop and a policy dialogue meeting on the Thai demersal fisheries in the Gulf of Thailand held at Hua Hin, Thailand, 31 May - 9 June 2000. FI: GCP/INT/648NOR: Field Report F-16 (En). Rome, FAO. Fayram, A. H. 2005. Walleye stocking in Wisconsin lakes: species interactions, changes in angler effort, optimal stocking rates, and effects on community maturity. Ph.D. Thesis, University of Wisconsin-Milwaukee, USA. Fayram, A.H., J Hansen, M., and T.J. Ehlinger. (in press). Characterizing changes in maturity of lakes resulting from supplementation of walleye populations. Ecological Modelling. 237 Fetahi, T. 2005. Trophic analysis of Lake Awassa using mass-balance Ecopath model. Ph.D. thesis, Addis Ababa University. Ethiopia. Field, J.C., Francis, R.C., and A. Strom. 2001. Towards a fisheries ecosystem plan for the Northern California Current CalCOFI Report 42: 74-87. Field, J. and R. Francis. 2005. Mass balance models of the Northern California Current, pp 207-216 In Guenette, S. and V. Christensen (eds.) Food web models and data for studying fisheries and environmental impacts on Eastern Pacific ecosystems. Fisheries Centre Research Reports 13(1). Forrest, R. (in preparation). Modelling the marine ecosystem of New South Wales. Ph.D. thesis, University of British Columbia, Canada. Frisk, M. 2007. Building an Ecosystem Model of Delaware Bay. Presentation at the 63rd Annual Northeast Fish & Wildlife Conference, April 22-25,2007, Connecticut. Fulton E.A. and A.D.M. Smith. 2004. Lessons learnt from the comparison of three ecosystem models for Port Phillip Bay, Australia. African Journal Marine Science 26:219-243. Gamito, S. and K. Erzini. 2005. Trophic food web and ecosystem attributes of a water reservoir of the Ria Formosa (south Portugal). Ecological Modelling 181: 509-520. Garces, L.R., Alias, M., Abu Talib, A., Mohamad-Norizam M., and G.T. Silvestre. 2003. A trophic model of the coastal fisheries ecosystem off the West Coast of Sabah and Sarawak, Malaysia, pp. 333-352 In Silvestre, G., Garces, L., Stobutzki, I., Ahmed, M., Valmonte-Santos, R.A., Luna, C , Lachica-Alino, L., Munro, P., Christensen, V., and D. Pauly (eds.) Assessment, management and future directions for coastal fisheries in Asian countries. WorldFish Center Conference Proceedings 67. Gasalla, M.A. and C.L.D.B. Rossi-Wongtschowski. 2004. Contribution of ecosystem analysis to investigating the effects of changes in fishing strategies in the South Brazil Bight coastal ecosystem. Ecological Modelling 172:283-306. Godinot, O. and V. Allain. 2003. A preliminary Ecopath model of the warm pool pelagic ecosystem. Working paper BBRG-5. 16th meeting on the standing committee on tuna and billfish, 9-16 July 2003, Mooloolaba, Australia. Goldsworthy S.D., Bulman C , He X., Larcombe J., and C. Littnan. 2003. Trophic interactions between marine mammals and Australian fisheries: an ecosystem 238 approach, pp 62-99 In Gales N., Hindell M., and R. Kirkwood (eds.). Marine Mammals and Humans: Fisheries, tourism and management. CSIRO Publications. Grant, A.T. 2001. Preliminary trophic model of a lemon shark (Negaprion brevirostris) nursery at Bimini, Bahamas. Presentation at the American Elasmobranch Society 2001 Annual Meeting, State College, Pennsylvania. Granzotto, A., Libralato, S., Pranovi, F., Raicevich, S., and O. Giovanardi. 2004. Comparison between artisanal and industrial fisheries using ecosystem indicators. Chemistry and Ecology 20 (Supplement 1): S435-S449. Gribble, N.A. 2001. A model of the ecosystem, and associated penaeid prawn community, in the far northern Great Barrier Reef. pp. 189-207 In Wolanski, E. (ed.). Oceanographic Processes of Coral Reefs: Physical and Biological Links in the Great Barrier Reef. CRC Press. Guarin, F.Y. 1991. A model fo the trophic structure of the soft-bottom community in Lingayen Gulf, Philippines: An application of the Ecopath II software and modeling approach. Ph.D. thesis, University of the Philippines. Guenette, S. 2005. Models of Southeast Alaska, pp 106-178 In Guenette, S. and V. Christensen (eds.) Food web models and data for studying fisheries and environmental impacts on Eastern Pacific ecosystems. Fisheries Centre Research Reports 13(1). Guenette, S. and I. Diallo. 2004. Modeles de la cote Guineenne, 1985 et 1998. pp 124-159 In Palomares, M.L. and D. Pauly (eds.). West African marine ecosystems : models and fisheries impacts. Fisheries Centre Research Reports 12(7). Guenette, S. and I. Diallo. (in press) Exploration d'un modele preliminaire de l'ecosysteme marine de la Guinee. pp. 329-346 In Chavance, P., Ba, M., Gascuel, D., Vakily, J.M. and D. Pauly (eds.). Pecheries, ecosystemes et societes en Afrique de l'Ouest: un demi-siecle de changement. Institut de Recherche pour le Developpement, Paris, France and European Comission, Burssels, Belgium. Guenette, S. and T. Morato. 2001. The Azores archipelago in 1997. pp 241-270 In Guenette, S., Christensen, V., and D. Pauly (eds.). Fisheries impacts on the North Atlantic ecosystems: models and analyses. Fisheries Centre Research Reports 9(4). 239 Haggan, N., Beattie, A., and D. Pauly (eds.). 1999. Back to the future: Reconstructing the Hecate Strait ecosystem. Fisheries Centre Research Reports 7(3). Hagy, James D. 2002 Eutrophication, hypoxia and trophic transfer efficiency in Chesapeake Bay. Ph.D. Thesis, University of Maryland. Halfon, E. and N. Schito. 1993. Lake Ontario food web, an energetic mass balance, pp. 29-39 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Harvey, C.J. and P.M. Kareiva. 2005. Community context and the influence of non- indigenous species on juvenile salmon survival in a Columbia River reservoir. Biological Invasions 7: 651-663. Harvey, C.J., Cox, S.P., Essington, T.E., Hansson, S., and J.F. Kitchell. 2003. An ecosystem model of food web interactions and fisheries effects in the Baltic Sea. ICES Journal of Marine Science 60:1-12 Heymans, J.J. 2002. The Gulf of Maine, 1977-1986. pp 128-150 In Guenette, S. Christensen, V., Pitcher, T.J., and D. Pauly (eds.). Fisheries Impacts on North Atlantic ecosystems: models and analyses. Fisheries Centre Research Reports 9(4). Heymans, J.J. 2003. Revised models for Newfoundland (2J3KLNO) for the time periods 1985-87 and 1995-97. pp 40-61 In Heymans, J.J. (ed.). Ecosystem models of Newfoundland and Southeastern Labrador (2J3KLNO): Additional information and analyses for 'Back to the Future'. Fisheries Centre Research Reports 11(5). Heymans, J.J. 2005a. Ecosystem model of the Eastern Aleutians and the Gulf of Alaska in 1963. pp 83-105 In Guenette, S. and V. Christensen (eds.). Food web models and data for studying fisheries and environmental impacts on Eastern Pacific ecosystems. Fisheries Centre Research Reports 13(1). Heymans, J.J. 2005b. Ecosystem model of the Western and Central Aleutian Islands in 1963, 1979 and 1991. pp 8-77 In Guenette, S. and V. Christensen (eds.). Food web models and data for studying fisheries and environmental impacts on Eastern Pacific ecosystems. Fisheries Centre Research Reports 13(1). Heymans, J.J. and D. Baird. 1995. Energy flow in the Kromme estuarine ecosystem, St Francis Bay, South Africa. Estuarine, Coastal and Shelf Science 41:39-59. 240 Heymans, J.J., and D. Baird. 2000. Network analysis of the northern Benguela ecosystem by means of NETWRK and ECOPATH. Ecological Modelling 131:97-119. Heymans, J.J. and A. McLachlan. 1996. Carbon budget and network analysis of a high- energy beach/surf-zone ecosystem. Estuarine, Coastal and Shelf Science 43:485-505. Heymans, J.J. and T.J. Pitcher. 2002a. A model of the marine ecosystem of Newfoundland and Southern Labrador (2J3KLNO) in the time periods 1985-1987 and 1995-1997. pp 5-43 In Pitcher, T.J., Heymans, J.J., and M. Vasconcellos (eds.). Ecosystem models of Newfoundland for the time periods 1995, 1985, 1900 and 1450. Fisheries Centre Research Reports 10(5). Heymans, J.J. and T.J. Pitcher. 2002b. A Picasso-esque view of the marine ecosystem of Newfoundland and Southern Labrador: Models for the time periods 1450 and 1900. pp 44-71 In Pitcher, T.J., Heymans, J.J., and M. Vasconcellos (eds.). Ecosystem models of Newfoundland for the time periods 1995, 1985, 1900 and 1450. Fisheries Centre Research Reports 10(5). Heymans, J. J., and J. M. Vakily. 2002. Ecosystem structure and dynamics of the marine system of Sierra Leone for three time periods: 1964, 1978 and 1990. pp. 109-120 In Pauly, D., Palomares, M.L.D., and J.M. Vakily. Trophic models of Northwest African ecosystems. Systeme d'Information et d'Analyses des Peches SIAP/EP/DT/03. Heymans, J. J., R. E. Ulanowicz, and C. Bondavalli. 2002. Network analysis of the South Florida Everglades graminoid marshes and comparison with nearby cypress ecosystems. Ecological Modelling 149:5-23. Hinke, J.T., Kaplan, I.C., Aydin, K., Watters, G.M., Olson, R.J., and J.F. Kitchell. 2004. Visualizing the food-web effects of fishing for tunas in the Pacific Ocean. Ecology and Society 9: 10. Hoff, M., Schram, S., Schreiner, D., and C.J. Walters. 2000. Sustainability of the Lake Superior fish community: interactions in a food web context. Ecosystems 3: 545- 560. Irwin, B., Coleman, J., Koops, M. and L. Rudstam. 2003. Initial development of an Oneida Lake Ecopath model. Bay of Quinte - Oneida Lake. Presentation at the Comparative Modeling Project Workshop II, Burlington, Ontario, Canada. 241 Jarre-Teichmann, A. 1995. Seasonal mass-balance models of carbon flow in the Central Baltic Sea with emphasis on the upper trophic levels. ICES C M . 1995/T: 06. Jarre-Teichmann, A. 1998. The potential role of mass balance models for the management of upwelling ecosystems. Ecological Applications 8: S93-S103. Jarre-teichmann, A., Brey, T., Bathmann, U.V., Dahm, C , Dieckmann, G.S., Gorny, M., Klages, M., Pages, F., Plotz, J., Schnack-Schiel, S.B., Stiller, M., and W.E. Arntz. 1997. Trophic flows in the benthic shelf community of the eastern Weddell Sea, Antarctica, pp. 118-134 In Valencia, J., Walton, D.W.H., and B. Battaglia (eds.). Antarctic communities: species, structure and survival. Cambridge University Press. Jutagate, T., Mattson, N., Moreau, J., Srichareondham, B., and M. Kumsri. 2002. Trophic models of Sirindhorn and Nam Ngum reservoirs; A comparison. Kasestart University Fisheries Research Bulletin. 24: 1-14. Kang, Yun Ho. 2005. A preliminary trophic flow model for Gwangyang Bay, Korea. Journal of Korean Fisheries Society 38: 184-195. Kipkemboi, J. 2006. Fingerponds : seasonal integrated aquaculture in East African freshwater wetlands, exploring their potential for wise-use strategies. Ph. D. thesis, Wageningen University. Kitchell, J.F., Cox, S.P., Harvey, C.J., Johnson, T.B., Mason, D.M., Schoen, K.K., Aydin, K., Bronte, C , Ebener, M., Hansen, M., Hoff, M., Schram, S., Schreiner, D., and CJ . Walters. 2000. Sustainability of the Lake Superior fish community: interactions in a food web context. Ecosystems 3: 545-560. Kitchell, J.F., Essington, T.E., Boggs, C.H., Schindler, D.E., and C J . Walters. 2002. The role of sharks and longline fisheries in a pelagic ecosystem of the central Pacific. Ecosystems 5: 202-216. Kolding, J. 1993. Trophic interrelationships and community structure at two different periods of Lake Turkana, Kenya: a comparison using the Ecopath II Box Model, pp. 116-123 In Pauly, D. and V. Christensen (eds.) Trophic Models of Aquatic Ecosystems. ICLARM, Manila. Koops, M.A., Irwin, B.J., MacNeil, J.E., Millard, E.S., and E.L. Mills. 2006. Comparative modelling of the ecosystem impacts of exotic invertebrates and productivity changes 242 on fisheries in the Bay of Quinte and Oneida Lake. Great Lakes Fishery Cornmission Project Completion Report. Ann Arbor, MI. Koops M.A., Millard, E.S., and E.L. Mills, (in press). Documentation for mass balance models of the Bay of Quinte (Lake Ontario) and Oneida Lake (New York). Canadian Technical Report of Fisheries and Aqautic Sciences 2616. Krebs, C.J., Danell, K., Angerbjorn, A., Agrell, J., Berteaux, D., Brathen, K.A., Danell, O., Erlinge, S., Fedorov, V., Fredga, K., Hjalten, J., Hogstedt, G., Jonsdottir, I.S., Kenney, A.J., Kjellen, N., Nordin, T., Roininen, H., Svensson, M., Tannerfeldt, M., and C. Wiklund. 2003. Terrestrial trophic dynamics in the Canadian Arctic. Canadian Journal of Zoology 81: 827-843. Kuo, S.-R. and K.-T. Shao. 1999. Species Composition of Fish in the Coastal Zones of the Tsengwen Estuary, with Descriptions of Five New Records from Taiwan. Zoological Studies 38(4): 391-404. Lees, K. and Mackinson, S. 2007. An Ecopath model of the Irish Sea: ecosystems properties and sensitivity analysis. CEFAS Lowestoft, Science Series Technical Report 138. Libralato, S., Pastres, R., Pranovi, F., Raicevich, S., Granzotto, A., Giovanardi, O., and P. Torricelli. 2002. Comparison between the energy flow networks of two habitats in the Venice Lagoon. Marine Ecolo