"Science, Faculty of"@en . "Resources, Environment and Sustainability (IRES), Institute for"@en . "DSpace"@en . "UBCV"@en . "Vasconcellos, Marcelo"@en . "2009-07-20T19:47:29Z"@en . "2000"@en . "Doctor of Philosophy - PhD"@en . "University of British Columbia"@en . "The ecosystem impacts of fisheries have become an important concern in fisheries management\r\nand one of the most important issues in fisheries science. The objectives of this thesis were i) to\r\nevaluate the use of trophic models in the analysis of the ecosystem responses to fishing small\r\npelagic forage fish; ii) to forecast the 'fishing down marine food web' phenomenon in Brazil;\r\nand iii) to conduct an in-depth analysis of the sardine, Sardinella brasiliensis, fishery off\r\nsoutheastern Brazil to examine hypotheses of ecosystem changes following stock collapse, and to\r\nevaluate how uncertainties of ecological processes influence the choice of harvest strategies and\r\ncontrols in this fishery. Dynamic simulations of mass-balance trophic models were used to\r\ncompare the ecosystem responses to fishing forage fishes across different types of marine\r\necosystems, and to examine the stability characteristics of ecosystems when impacted by\r\nfisheries. The comparative analysis among trophic models indicated that as 'wasp-waist' species\r\nin upwelling ecosystems, small pelagics will only sustain much more conservative exploitation\r\nrates than the levels that have brought about historical stock-collapses. In the upwelling\r\necosystem off southeastern Brazil, sardine has been historically the main target of commercial\r\nfisheries, although 'fishing down marine food webs' for small pelagic fish species is not yet an\r\noverall observed phenomenon in Brazil. The collapse of the sardine fishery in the late 1980s has\r\napparently favored a competing small pelagic fish (anchovy, Engraulis anchoitd) in the\r\necosystem. The combined effect of fishing and environmental effects makes it difficult,\r\nhowever, to characterize the changes in the sardine population and the ecosystem, and to predict\r\nthe results of rehabilitation measures for the stock. On the other hand, the explicit recognition of\r\nthese ecological uncertainties allows a sound choice of precautionary strategies for the fishery,\r\nand a better scrutiny of research programs to improve management. In this context, trophic\r\nmodels will have a complementary role to single-species models in the analysis of the broad\r\nconsequences of fishing policies, and in the test and formulation of hypotheses about the causes\r\nof observed changes in marine ecosystems."@en . "https://circle.library.ubc.ca/rest/handle/2429/10998?expand=metadata"@en . "18640239 bytes"@en . "application/pdf"@en . "E C O S Y S T E M I M P A C T S OF F I S H I N G F O R A G E F I S H E S : A N A N A L Y S I S OF H A R V E S T S T R A T E G I E S F O R T H E B R A Z I L I A N S A R D I N E by M A R C E L O V A S C O N C E L L O S B . Sc., University of Rio Grande, 1991 M . S c , University of Rio Grande, 1994 A T H E S I S S U B M I T T E D I N P A R T I A L F U L F I L M E N T OF T H E R E Q U I R E M E N T S F O R T H E D E G R E E OF D O C T O R OF P H I L O S O P H Y in T H E F A C U L T Y OF G R A D U A T E S T U D I E S (Resource Management and Environmental Studies) We accept this thesis as conforming to the required standard T H E U N I V E R S I T Y OF B R I T I S H C O L U M B I A February, 2000 \u00C2\u00A9 Marcelo Vasconcellos, 2000 In presenting t h i s thesis i n p a r t i a l f u l f i l m e n t of the requirements for an advanced degree at the U n i v e r s i t y of B r i t i s h Columbia, I agree that the Library s h a l l make i t f r e e l y a v a i l a b l e for reference and study. I further agree that permission f o r extensive copying of t h i s thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It i s understood that copying or p u b l i c a t i o n of t h i s thesis f o r f i n a n c i a l gain s h a l l not be allowed without my written permission. Department of The U n i v e r s i t y of B r i t i s h Columbia Vancouver, Canada Abstract The ecosystem impacts o f fisheries have become an important concern in fisheries management and one of the most important issues in fisheries science. The objectives of this thesis were i) to evaluate the use of trophic models in the analysis of the ecosystem responses to fishing small pelagic forage fish; ii) to forecast the 'fishing down marine food web' phenomenon in Brazi l ; and i i i) to conduct an in-depth analysis of the sardine, Sardinella brasiliensis, fishery off southeastern Brazi l to examine hypotheses of ecosystem changes following stock collapse, and to evaluate how uncertainties of ecological processes influence the choice of harvest strategies and controls in this fishery. Dynamic simulations of mass-balance trophic models were used to compare the ecosystem responses to fishing forage fishes across different types of marine ecosystems, and to examine the stability characteristics of ecosystems when impacted by fisheries. The comparative analysis among trophic models indicated that as 'wasp-waist' species in upwelling ecosystems, small pelagics w i l l only sustain much more conservative exploitation rates than the levels that have brought about historical stock-collapses. In the upwelling ecosystem off southeastern Brazi l , sardine has been historically the main target of commercial fisheries, although 'fishing down marine food webs' for small pelagic fish species is not yet an overall observed phenomenon in Brazi l . The collapse of the sardine fishery in the late 1980s has apparently favored a competing small pelagic fish (anchovy, Engraulis anchoitd) in the ecosystem. The combined effect o f fishing and environmental effects makes it difficult, however, to characterize the changes in the sardine population and the ecosystem, and to predict the results of rehabilitation measures for the stock. On the other hand, the explicit recognition of these ecological uncertainties allows a sound choice of precautionary strategies for the fishery, and a better scrutiny of research programs to improve management. In this context, trophic models w i l l have a complementary role to single-species models in the analysis of the broad consequences of fishing policies, and in the test and formulation of hypotheses about the causes of observed changes in marine ecosystems. ii Table of Contents Abstract i i List of Tables V List of Figures v i Acknowledgements \X Chapter 1. Introduction 1 Fisheries for small pelagic fish resources 1 Guidelines for Fisheries Management 4 1.1. Objectives 6 Problems in managing small pelagic fisheries 6 A framework for fisheries assessment 10 Chapter 2. Ecosystem impacts of harvesting small pelagic fishes. A comparative analysis using trophic mass-balance models 17 2.1. Introduction 17 2.2. Methods 21 General Predicted Impacts 23 Ecosystem Stability 25 2.3. Results and Discussion 30 Ecosystem responses to fishing 30 Ecosystem stability and resilience 43 Model limitations 48 Management implications 50 2.4. Summary 51 Chapter 3. Fishing down food webs and the carrying capacity of marine ecosystems in southern Brazil 52 3.1. Introduction 52 3.2. Methods 54 3.3. Results .' 59 PPR and Trophic Levels 59 Fishing Down the Food Web 63 3.4. Discussion 66 3.5. Summary 69 Chapter 4. The sardine fishery in the Southeastern Brazilian Bight 70 4.1. Introduction 70 Sardine fishery development 73 Overview of Management Context 75 4.2. Methods 79 Ecosystem structure and dynamics 79 Spawning Stock and Recruitment 84 Stock distribution, effort dynamics and catchability change 88 4.3. Results and Discussion 90 Ecosystem structure and dynamics 90 Stock distribution, effort dynamics and catchability change 114 4.4. Summary 125 Chapter 5. Analysis of harvest decisions and information needs in the management of the Brazilian sardine. Comparing multi-species and single-species modeling approaches 126 5.1. Introduction 126 5.2. Methods 127 Modeling approaches 127 5.3. Results and Discussion 137 Single-species approach 137 Expected Value of Perfect Information 143 Multi-species approach 144 Conclusions from model simulations 151 On the choice of a modeling approach 157 5.4. Summary 161 Chapter 6. Conclusions 162 6.1. Evaluation of ecosystem responses to fishing using dynamic trophic models 162 6.2. Fishing down marine food webs in Brazil 165 6.3. The case of the sardine fishery off southeastern Brazil 166 Notes 169 Bibliography 170 IV List of Tables Box 1.1. International cooperative programs on oceanography and productivity of marine fish populations 8 Table 2.1. Trophic models of upwellings used in the analysis of the impact of small pelagic fisheries.. 24 Table 2.2: Models used in the analysis of ecosystem stability 29 Table 2.3: Correlation among the ecosystem attributes defined in the text 44 Table 2.4. Trophic models used in the regression between quality rank and persistence rank 50 Table 3.1.Area, primary productivity and total primary production off southern Brazil 55 Table 3.2. Trophic level of the main species landed in Brazil 56 Table 3.3. Parameters of the trophic model of the pelagic ecosystem off southern Brazil 57 Table 3.4. Diet matrix of the model of the pelagic ecosystem off southern Brazil 57 Table 3.5. Trophic level, mean catch and PPR estimates for the southern shelf. 60 Table 3.6: Trophic level, mean catch and PPR estimates for the southeastern shelf. 60 Table 3.7. Summary statistics of the southern and southeastern shelves 60 Box 4.1. Ecopath models of the southeastern Brazil shelf ecosystem 82 Table 4.1. Parameters of the split pools in Ecopath used by the delay difference model in Ecosim 83 Table 4.2. Catch (millions of fish) at age data used in V P A 85 Box 4.2. Virtual Population Analysis procedure for reconstruct sardine biomass at age 86 Table 4.3. Parameters of the Ecopath trophic model of the Southeastern shelf ecosystem 90 Table 4.4. Diet matrix of the trophic model of the Southeastern shelf ecosystem 92 Table 4.5. Ecosystem attributes used in the comparison of trophic models of upwelling ecosystems 94 Table 4.6. Transfer efficiencies between trophic levels 95 Table 4.7. Trophic level, mean catch and PPR for the main species landed in southeastern Brazil 96 Table 4.8. Time-scale hierarchy of variables controlling the Southeastern shelf ecosystem 99 Table 4.9. Biomass at age estimated from V P A 109 Table 4.10. Parameters used in the representation of stock-recruitment relationships 112 Table 4.11. Effort, fishing mortality and catchability of sardine stock to purse seiners 115 Table 4.12. Results of partial correlation analysis between catchability, stock size and temperature.... 117 Table 5.1. Hypothesis, models and parameters used to predict sardine recruitment rates 129 Table 5.2. Percentage of individuals mature and and the reproductive output at age 134 Table 5.3. Decision table on the choice of harvest rate and effort level for the sardine fishery 143 Table 5.4. Results of the analysis of the expected value of perfect information 144 Table 5.5. Relative change in sardine biomass after 5 year under different fishing scenarios 145 Table 5.6. Relative change in sardine catches after 5 years under different fishing scenarios 146 Table 5.7. Relative change in sardine biomass after 5 years under two strategies for stock recovery.... 146 Table 5.8. Relative change in biomass of main harvested species in the Southeastern Brazilian Bight. 147 Table 5.9. Relative change in biomass of main harvested species in the Southeastern Brazilian Bight. 147 List of Figures Figure 1.1. Analytical framework for integrating fisheries stock assessment and management 11 Figure 1.2. Ecosystem models used to evaluate the effects of fishing small pelagic fish species 16 Figure 2.1. Schematic representation of marine food webs with waists at the mid-trophic level 21 Figure 2.2. Simulation of system recovery of the Venezuela shelf model 26 Figure 2.3. Predicted equilibrium biomass and catches of anchovy in the California System (65-72).... 31 Figure 2.4. Predicted equilibrium biomass and catches of anchovy in the California System (78-85).... 32 Figure 2.5. Predicted equilibrium biomass and catches of omnivorous fish in the Monterey System 33 Figure 2.6. Predicted equilibrium biomass and catches of sardine in the Namibia System (71-77 34 Figure 2.7. Predicted equilibrium biomass and catches of anchovy in the Namibia System (78-83) 35 Figure 2.8. Predicted equilibrium biomass and catches of sardine in the N W Africa System (70-79).... 36 Figure 2.9. Predicted equilibrium biomass and catches of anchoveta in the Peruvian System (64-71)... 37 Figure 2.10. Predicted equilibrium biomass and catches of anchoveta in the Peruvian System (73-81). 38 Figure 2.11. Predicted equilibrium biomass and catch of small pelagic fish in the Venezuela System... 39 Figure 2.12. Predicted fishing mortality at maximum yield under two trophic control hypotheses 42 Figure 2.13. Recovery time of the 23 models perturbed by pulse fishing wasp-waist populations 43 Figure 2.14. Relation between Finn's index of detritus recycling and system recovery 45 Figure 2.15. Effect of increasing detritus cycling on the stability of the Venezuela shelf model 46 Figure 2.16. Rank correlation between model persistence and quality 49 Figure 3.1. Shelf regions of Brazil 53 Figure 3.2. Flowchart of trophic relationships in the pelagic association off southern Brazil 58 Figure 3.3. Mean trophic level (A), and species composition (B and C)of total Brazilian landings 62 Figure 3.4. Equilibrium simulation of increasing fishing mortality for anchovy 64 Figure 3.5. Relationship between total catch and the mean trophic level of catches 65 Figure 4.1. Detail of the Southeastern Brazilian Bight 71 Figure 4.2. Sardine landings by the three main state fleets 74 Figure 4.3. Fishing areas (shaded) of the Santa Catarina fleet used in the analysis of effort allocation... 89 Figure 4.4. Trophic flows diagram of the Southeastern Brazilian shelf ecosystem 91 Figure 4.5. Holling's four phase model of ecosystem dynamics 97 Figure 4.6. Pattern of variation observed in many marine fish populations 100 Figure 4.7. Output predicted by Ecosim with a cyclic regime in primary productivity (PP) 103 Figure 4.8. Output predicted by Ecosim with selective predation mortality on juvenile sardine 107 vi i Figure 4.9. Reconstructed time-series of sardine biomass from V P A 110 Figure 4.10. Bayes marginal posterior probability distribution of natural mortality rate 110 Figure 4.11. Graphic representation of the two stock-recruitment hypothesis 112 Figure 4.12. Graphic representation of the \"dome-shaped\" regime hypothesis 114 Figure 4.13. Relationship between catchability to the purse seine fleet and sardine stock biomass 116 Figure 4.14. Changes in the catchability of sardine with stock size and the mean temperature 116 Figure 4.15. Sardine stock distribution area as estimated by acoustic surveys 119 Figure 4.16. Proportion of total sardine landings by fishing area 120 Figure 4.17. Fleet distribution by landing place 123 Figure 4.18. Test for the equalization of cpue among fishing areas off Santa Catarina 124 Figure 4.19. Monthly effort allocation by the Santa Catarina fleet 125 Figure 5.1. Marginal posterior probability distribution of catchability parameters 130 Figure 5.2. Conceptual structure of a decision analysis on the choice for fishing strategies for sardine..132 Figure 5.3. Monte Carlo procedure used in the evaluation of the outcomes of fishing strategies 134 Figure 5.4. Outcomes of fishing strategy that capture a constant catch 139 Figure 5.5. Outcomes of fishing strategies based on effort control 140 Figure 5.6. Outcomes of fishing strategies that harvest a constant proportion of the stock 141 Figure 5.7. Outcomes of fishing strategies based on a constant stock escapement 142 Figure 5.8. Dynamic simulation of fishing scenarios in 5 years 148 Figure 5.9. Predicted average yield of sardine under fishing mortality rates 149 Figure 5.10. Example of a pulse fishing strategy for sardine 151 Figure 5.11. Percentage spawning per recruit for equilibrium exploitation rate and age at first capture.. 154 vii J Acknowledgements I would like to thank the members of my supervisory committee, Les Lavkulich, Daniel Pauly, Carl Walters and Tony Pitcher for their guidance and invaluable inputs to my work. I would especially like to thank my supervisor, Tony Pitcher, for giving me the opportunity to come to the Fisheries Centre, U B C , to pursue the doctorate degree. I thank the Conselho Nacional de Desenvolvimento Cientifico e Tecnologico (CNPq/Brasil) for the financial support provided during the program, and for the assistance of their staff, Jessy Alves, Ilka Oliveira, Marisa Cassim, and others, which greatly facilitated my studies in Canada. A number of people have given me assistance during the course of this study. Data for the case study was kindly provided by Maria Cristina Cergole and by the I B A M A / C E P S U L , Itajai. I also thank Mar ia Gasalla, Jorge Castello, V i l l y Christensen, Kevern Cochrane, Katherine Sloman, Manoel Haimovici and Humber Andrade for providing data, suggestions and/or criticism to much of this study. I am especially grateful to friends and colleagues, Steve Mackinson, Ingrid Ross, Dave Preikshot, Trevor Hutton, Rashid Sumaila, and to the many 'coffee breaks', that helped create a friendly and exciting place to work at the Fisheries Centre. N o thanks would be too much to my wife, Daniela, for her continuous support and inspiration. IX Chapter 1. Introduction One of the major challenges of natural resources management centers on the science-policy interface (Cicin-Sain and Kenecht, 1998; Walters, 1998). The main reason to relate science and policy is to improve the information available for making management decisions. Improvements in the management of natural resources is largely dependent on improvements in the understanding of the processes involved in the resource problem. Many resource management situations have been aggravated because of inadequate information or lack of foresight in anticipating unexpected problems; yet, good information does not always guarantee reliable decisions, since very often information is available but no appropriate management choices are implemented. Sti l l , there is a perceived strong political benefit for grounding decisions on science (Jasanoff, 1990; Cicin-Sain and Kenecht, 1998). This thesis focuses on the use of scientific ecological information in the management of fisheries for small pelagic fish resources. Fisheries for small pelagic fish resources Since the advent of large-scale industrialized fisheries, small pelagic fishes have dominated world landings. Out of the five leading species in world catches between 1980 and 1990, four were either small or mid-size pelagics: the Peruvian anchoveta, Engraulis ringens, the Japanese and South American pilchards (Sardinops melanosticta, S. sagax) and the Chilean horse mackerel, Trachurus murphyi (Csirke, 1995). The term 'small pelagic' is used to refer to species such as anchovies, sardines, pilchards and herrings, which share common biological characteristics, such as strong shoaling behavior, planktonic feeders, fast growth and short lives. These characteristics lead to stocks with large biomasses, a key role in the transfer of energy in marine food webs, substantial temporal variability in abundance and to their potential as target of fisheries exploiting a cheap protein source for human use (Glantz, 1983; Csirke, 1995; Bakun, 1996). Biological , economic and technological factors have played a role in the development of small pelagic fisheries under an 'unlimited resources' paradigm, in which the stocks were considered not prone to be overfished. Shoaling pelagic fish have highly aggregated distributions which makes them relatively easy to capture with modern and efficient fishing technologies, depriving 1 stock assessment and management of early warnings of overfishing and stock collapse. On the other hand, their life history strategies often make small pelagic fish stocks appear to be among the most productive and abundant marine fisheries resources. The Peruvian anchoveta, for instance, once supported the largest fishery ever recorded, accounting alone for over 10% of total world fish production. Fishmeal production is the main objective o f pelagic fisheries worldwide. The increase in world demand for fishmeal to the feed industry following the World War II created economic conditions for the fast development of reduction fisheries for small pelagics. One of the best documented cases of rapid economically-driven development was that of the Peruvian anchoveta fishery in the early 1950s, when U . S . factories for fishmeal production were built in Peru following the collapse of the California Pacific sardine, Sardinops sagax (Glantz, 1983). Today, pelagic catches used worldwide by the reduction industry represents roughly one third of world marine catches and about 80% of total world pelagic catches (Durand, 1998). The history of many pelagic fisheries worldwide reveals, however, a different reality. Small pelagic stocks were depleted to the point of collapse, in some cases repeatedly, and often in alternating 'regimes' between anchovy and sardine species (Lluch-Belda et al., 1989). The depletions caused not only economic and social difficulties but also undesirable and unexpected ecological disruptions, such as massive mortality of guano birds after the collapse of the Peruvian anchoveta (Tovar et al, 1987), and an increasing dominance of jellyfishes after heavy exploitation of small pelagics and intense eutrophication in the Black Sea (Mee, 1992). Today, cyclic 'boom and bust' dynamics of pelagic fisheries is an widely recognized characteristic that is explained by alternative (but non-exclusive) hypotheses concerning climatic effects on recruitment, trophic interactions, and overfishing (Radovich, 1982; Caddy and Gulland, 1983; Lluch-Belda et al., 1989; Csirke, 1995; Bakun, 1996). In fact, vigorous debate on the extent that fisheries and environmental processes have contributed to cyclic regimes still persist (Radovich, 1982; Bakun, 1996; Steele, 1996). A metanalysis carried by Patterson (1992) showed, for instance, that most cases of small pelagic stock collapse occurred when F (fishing mortality) was higher than 0.6 M (natural mortality). The demise of many small pelagic fisheries rouses a more fundamental rationale - that stock assessment research has not been providing adequate advice to managers, specially through the design of fishery control systems that can respond to the types o f variability and uncertainties observed in nature. Scientific debate around the causes of collapses, recoveries and species shifts 2 in many ecosystems and the relatively short time series o f data for most fisheries, have been historically responsible for the lack of a comprehensive framework for scientific advice to fisheries management. Debate on the causes of fisheries collapse has always permeated the philosophies, research activities and conclusions of fisheries scientists on different agencies, as portrayed by Radovich (1982) for the case of the Pacific sardine: \"The Federal scientists, working for an agency whose fundamental charter was to assist the development and maintenance of U.S. commercial fisheries, looked for reasons other than fishing, for the sardine's declining condition, while the scientists employed by the State (whose basic role was protector of the State's resources) supported the premise that overfishing was having a detrimental effect on the standing stock. These were capable, competent scientists using the same data and coming up with different conclusions in part because they were employed by agencies whose fundamental goals were different.\" On the other hand, a perceived need for scientific consensus is present in many resource management occasions, and in some of the most disastrous examples of fisheries mismanagement. For instance, in the case of the Atlantic cod, Gadus morhua, the insistence on consensus has led a government-administered assessment framework that suppresses scientific uncertainties and fails to document important differences in scientific opinion regarding the risks of stock collapse (Hutchings et al., 1997). Two types of responses to the lack of scientific consensus are commonly encountered in fisheries management: first to completely ignore uncertainties when making decisions; and second to direct more research efforts to better understand the nature of processes affecting fish stocks before changes in the status quo policies take place. More often, however, uncertainties have impeded scientific inputs into management decisions, which end up obeying strictly economic and political forces. Ludwig et al. (1993) asserted that we shall never attain scientific consensus concerning the systems that are being exploited because controlled and replicated experiments are impossible to perform in large-scale systems. Therefore there w i l l be always ample scope for differing interpretations of natural events. In this context, fisheries management should acknowledge the uncertain nature of ecological processes and seek for robust strategies for exploiting populations 3 and ecosystems. Moreover, it is necessary to recognize that science and management must be coupled to better understand the nature of processes affecting fish populations (Walters, 1986). Guidelines for Fisheries Management Complications arising from the uncertain nature of fisheries management, and the common realization of the need for a broader, ecosystem perspective on natural resources management has lead to the development of many principles and guidelines for effective fisheries management throughout the world (e.g. F A O , 1997; G E S P E , 1997; Canadian Ocean Act) . With respect to uncertainties, the F A O Code of Conduct for Responsible Fisheries ( F A O , 1995) adopts the precautionary approach which states that \"...where there are threats of serious or irreversible damage, lack offull scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation\". The proper implementation of the precautionary approach w i l l require, among other issues identified by F A O Technical Guidelines for Responsible Fisheries ( F A O , 1997), \"-consideration of the needs of future generations and avoidance of changes that are not potentially reversible; - prior identification of undesirable outcomes and of measures that will avoid them or correct them promptly; - that where the likely impact of resource use is uncertain, priority should be given to conserving the productive capacity of the resource; - that harvestable and processing capacity should be commensurate with estimated sustainable levels of resource, and that increases in capacity should be further contained when resource productivity is highly uncertain\". The F A O Technical Guidelines for Responsible Fisheries ( F A O , 1997) also recognizes the need to place fisheries in an ecosystem context; \"....harvesting any one species is almost certain to impact others,...The impact of ecological linkages (e.g. through the trophic chain) between species, may lead to changes in species dominance and affect the dynamic equilibria of the resource system, potentially affecting future 4 options. These multi-species effects need to be considered in responsible fishing, which should aim to ensure that no species, whether targeted, by-catch or indirectly affected by fishing, is reduced to below sustainable levels. \" Moreover, \" Responsible fisheries management should consider the impact of fisheries on the ecosystem as a whole, including its biodiversity, and should strive for sustainable use of whole ecosystems and biological communities.\" Mangel et al. (1996) proposed an independent set of principles for the conservation of wi ld l iving resources which summarized the point of view of 42 scientists. Three out of seven principles explicitly recognize the need to account for human impacts on ecosystems, e.g. \"The goal of conservation should be to secure present and future options by maintaining biological diversity at genetic, species, population, and ecosystem levels; as a general rule neither the resource nor other components of the ecosystem should be perturbed beyond natural boundaries of variation.\" \"Assessment of the possible ecological and sociological effects of resource use should precede both proposed use and proposed restrictions or expansion of ongoing use of a resource.\" \"Regulation of the use of living resources must be based on understanding the structure and dynamics of the ecosystem of which the resource is part and must take into account the ecological and sociological influences that directly and indirectly affect resource use.\" The adoption of these principles w i l l challenge scientists and decision makers to consider the more diverse as possible consequences of fisheries to marine populations and ecosystems. Therefore, a critical re-examination of concepts, methods, tools, and regulatory measures commonly used in fisheries assessment and management is primarily needed. 5 1.1. Objectives It is the overall goal of this thesis to evaluate fishing strategies for small pelagic fish stocks considering the inherent variability of the ocean environment and the ecological role of these species in marine food webs. The specific objectives of the thesis are - to make specific predictions about ecosystem responses to fishing small pelagic species, using dynamic trophic models; - to analyze the characteristics of stability and resilience of exploited marine ecosystems when impacted by fisheries for mid-trophic level forage species; and - to carry out a case study with the fishery for sardine, Sardinella brasiliensis, in the Southeastern Brazilian Bight to examine how the combined effect of environmental, behavioral and ecological processes, as captured by different modeling approaches, influence the choice of harvest strategies and controls. Problems in managing small pelagic fisheries The inability to predict the effect of harvest policies given the lack of understanding of the processes affecting fish stocks has lead to major research initiatives into fisheries-oceanography, specially in the analysis of the effects of environmental variability on recruitment of commercially important marine populations. International initiatives were supported by the Intergovernmental Oceanographic Commission ( I O C - U N E S C O ) program on Ocean Sciences in Support of L iv ing Resources (OSRL) , which aimed to promote the development of plans for major oceanographic studies of the physical-ecological interactions of importance to fishery resource-related problems (Bakun et al, 1982). The rationale of the program was that many fisheries problems, such as recruitment, population availability and species interactions, could not be solved solely based on the analysis of fisheries data. The accumulated evidence derived 6 from deposits of fish scales in sediments (Soutar and Isaacs, 1974; Baumgartner et al., 1992), and the global synchrony in the raising and falling of the large populations of anchovies and sardines in the major upwelling regions of the world (Kawasaki, 1983; Lluch-Belda et al., 1989), indicated that radical fluctuations in abundance may be an intrinsic feature of many marine fish populations. Moreover it pointed to the existence of physical processes that teleconnect these widely separated systems. I f so, then the hope for successful fisheries management would appear to rely in gaining scientific understanding of the natural factors determining reproductive success and population dynamics of fishery stocks. The rationale would have two main implications to fisheries management: first, it would show the need to develop robust strategies for fisheries viability under conditions of radically varying resources; second, and most importantly, that fisheries management would directly benefit from the increased scientific ability to identify and predict transition periods or 'regime shifts', when distinct levels o f precautionary actions would be necessary (Bakun, 1998). A research framework was proposed involving a combination o f a rational (cause-effect) approach, based on the understanding of relevant physical and biological processes for the viability of marine populations, and an empirical approach, in which functional relationships between these processes were defined by analysis of available data. The difficulty in addressing some of the physical-biological processes experimentally called for the combination of a third approach based on the comparative method of analysis (Bakun, 1996). Results of the comparative studies of geographical climatology of fish reproductive habitats have tended to identify a triad of physical factors capable of exerting control on reproductive success of marine populations (Bakun, 1996; Bakun et al., 1998). These are enrichment processes that lead to the production of the zooplankton upon which the young stages depend for food; concentration processes that aggregate food products and therefore increase their availability for the growing larvae; and retention processes that keep the young in favorable nursery habitats. In coastal upwelling areas enrichment is basically driven by persistent alongshore winds that forces water to move offshore and, in order to balance the superficial deficit, deep nutrient-rich water is forced into the surface illuminated layers where primary production is possible. Frontal structures (vertical and horizontal) are the main features responsible for the concentration of plankton in thin stable layers and for creating optimal larval feeding conditions. Retention involves mechanisms such as Taylor column, Ekman divergence and zones o f stability (Bakun,- 1996). Some corollaries of triad theory have been tested by Cury and Roy (1989) who showed that the 7 reproductive success in upwelling ecosystem is tightly associated to an optimum window of moderate wind stress, thought weak winds are not sufficient to enrich the upper layers and winds too strong can otherwise disrupt stable structures in the ocean. The \"triad framework\" is being applied in diverse international programs aimed at identifying the links between climatic-oceanographic processes and marine fish populations (Box 1.1). Box 1.1. International cooperative programs on oceanography and productivity of marine fish populations. \u00E2\u0080\u00A2 I R E P / S A R P . The Sardine-Anchovy Recruitment Project (SARP) was established as the initial focus of the International Recruitment Project (IREP) component o f the O S R L . The main objective o f S A R P was to promote a greater understanding o f the biologic-oceanographic processes involved in controlling recruitment success in sardine and anchovy populations (Bakun et al., 1982). \u00E2\u0080\u00A2 C E O S . The Climate and Eastern Ocean Systems (CEOS) project was a joint effort of the National Marine Fisheries Service ( N M F S ) , the Institut Francais de Recherche Scientifique pour le Developpement en Cooperation ( O R S T O M ) , and the International Center for Liv ing Aquatic Resources Management ( I C L A R M ) . The C E O S project was an international collaborative study of the potential effects of global climate change on the l iving resources of the highly productive eastern ocean upwelling ecosystems and on the ecological and economic issues directly associated with such effects. A major focus o f the study were the clupeoid fishes (Durand et al., 1998). \u00E2\u0080\u00A2 G L O B E C / S P A C C . International Global Ocean Ecosystem Dynamics ( G L O B E C ) program on Small Pelagic Fishes and Climate Change ( S P A C C ) . The objective of S P A C C is to identify how physical forces are linked with the growth o f pelagic fish populations, which is believed to be mediated through the dynamics of zooplankton populations. The long-range goal o f S P A C C is to forecast how changes in ocean climate w i l l alter the productivity of small pelagic fish populations. To address this goal, S P A C C uses comparative retrospective and process studies (Hunter and Alheit, 1997). The proposed improvement in management practice using climatic/oceanographic process studies has been fiercely criticized (Walters and Coll ie , 1988; Walters, 1998). A s stated previously, one direct benefit o f these studies would appear to be the improvement in the quality of decisions on harvest levels by allowing the adjustment o f policies according with predicted environmental conditions. For instance, i f short term variations could be accurately forecast, the best policy would be to adjust the target stock size so as to increase catch when positive variation is anticipated and to reduce when negative variation is anticipated. Conversely, it may involve allowing escapement to increase when good environmental conditions are forecast so as to take the maximum advantage of the optimum conditions (Parma, 1990; Walters and Parma, 1996). Walters and Coll ie (1988) argued, however, that in most cases better understanding and predictive models are not that useful, since the same benefits could be met more cheaply by improved monitoring programs, i.e., recruitment and spawning biomass surveys, and existing feedback policies. Better understanding of the processes controlling recruitment success seems to be mostly needed for situations in which the effects of long-term environmental factors are confounded with the effect of stock size and fishing (Walters and Coll ie, 1988). Environmental cycles of intermediate periods superimposed on stock production dynamics can obscure any underlying relationship between recruitment and spawning biomass (Armstrong and Shelton, 1988). In fact, most accounts of stock collapses during the last decades have at their core the endless debate about whether it was the result of fishing or of environmental effects. Among the best documented examples are the sardine/anchovy collapses in coastal upwelling systems (Pauly and Palomares, 1989; Barnes et al., 1992), the decline in recruitment of groundfish stocks off east coast of North America (Walters and Maguire, 1996) and the \"Thompson-Burkenroad\" debate on the causes of recruitment fluctuations in the stock of Pacific Halibut (Parma and Deriso, 1990). In these cases the understanding of how environmental changes affect fish productivity would be very valuable to fishery management. Yet, it is argued that such understanding w i l l not be achieved by continued correlative and biological process studies and w i l l instead require sound management experiments in which environmental studies are coupled with deliberate manipulation of stock sizes through changes in harvest policies (Walters and Coll ie, 1988; Walters, 1998). A more fundamental criticism of the process-oriented studies is that they do not reflect necessarily the questions and information needs of managers. Walters (1998) stated that weak communication between scientists and managers have repeatedly resulted in the development of models based on variables and factors that do not consider policy options and values identified by managers, the allocation of scarce resources into research programs that are only vaguely related to policy questions, and, ultimately, the design of policies that most commonly fail when implemented because important details were not properly accounted for. Relating research and management w i l l require a more thorough analysis of information needs for making management decisions, and a better scrutiny of research programs in which to allocate resources. 9 A framework for fisheries assessment Fisheries stock assessment research has been recognized as a mean to provide some structured use of available information to estimate the nature of the tradeoffs when comparing fisheries management choices (Hilborn and Walters, 1992). Tradeoffs and choices are the core elements of fisheries assessment advice, and imply an active role of fisheries scientists in aiding managers understand the responses of fishery systems to alternative management choices, being the responses usually related to the impact of fishing on fish stock and ecosystem, social desirability of fishery regulations and economic returns to fishery sectors. The main scientific input to fisheries management is in the evaluation of the consequences of alternative harvest policies for marine resources. This advisory work should consider the consequences of fishing activities to marine populations and ecosystems, being ideally linked by a common framework (Mangel et al, 1996). A framework for fisheries stock assessment must be able to provide managers with information on (Butterworth et al, 1997; Cochrane et al, 1998): \u00E2\u0080\u00A2 tradeoffs among different regulatory mechanisms in the short, medium and long term; \u00E2\u0080\u00A2 risks associated with each regulatory mechanism; \u00E2\u0080\u00A2 outputs that could be easily understandable for all decision makers; \u00E2\u0080\u00A2 guidance on what type of data to collect for the purpose of the chosen regulatory mechanism; and \u00E2\u0080\u00A2 definition of a rank of research priorities, and a time frame for revision of both research and regulatory mechanisms, in light of any changes in understanding of resource/fishery. The achievement of a basis for informed judgment on the costs of postponing or attenuating decisions demands an integrated strategy for fisheries management and research. Figure 1.1 outlines an integrated framework for fisheries assessment based on a set of interrelated questions. 10 1 what are the decisions for which ecological information is required? policwrecommendations 2 which type of information would provide the basis for decisions? 3 how well cantTecisions be made with the available information? Figure 1.1. Analytical framework for integrating fisheries stock assessment and management (based on Dorcey and The first step in the framework comprise the analysis of the management or decision context. The achievement of objectives is the sole reason for being interested in any decision. Yet, unfortunately, objectives are not adequately articulated and explicit in many management situations because of either a lack of involvement of all fisheries interests in the decision making process, or for the constant political pressure put on managers to produce short-term \"tangible results\" (Keeney, 1992). The values relevant to a decision context w i l l indicate what type of information is important. In this respect, data w i l l have value only i f it w i l l help lead to better consequences, either through the creation of better alternatives or through the wiser choice of alternatives (Keeney, 1992). For fisheries management, decisions normally involve the choice of harvest strategies, and ecological information w i l l be used to estimate the impact of these regulatory mechanisms on the ecological sustainability of the fishery. The ecological criteria for sustainability may involve an array of indicators such as catch, the size (biomass) of the spawning stock, and/or indicators of fisheries impact on the ecosystem. Ecosystem impacts may include direct effects such as the catch of non-target species (by-catch) and the destruction of marine habitats, and indirect effects on non-target species due to trophic cascade effects through the food web. Hall, 1983). 11 Many possible combinations of harvest strategies and controls are possible, and it is the role of stock assessment to provide managers with advice on the possible consequences of each regulatory mechanism (management decisions). Once the decision context has been properly outlined it is then possible to analyze how well can the choice for management strategies be made, what are the sources of uncertainties and the opportunities for learning. A t this stage decision analysis methodology can be used to provide an \"heuristic aid\" (sensu Rowen, 1976) for relating means to ends, for thinking about ends, as well as for identifying new management alternatives (Morgan and Herion, 1990). Decision analysis involve processes subsequent to the analysis of objectives, such as the identification of alternative hypothesis about the system been managed, inference of uncertainties on processes and states relevant for the decision context, and the assessment of the consequences and risks associated to each strategy using criteria defined by the objectives. The completion of these steps w i l l ultimately lead to policy and research recommendations (Figure 1.1). Research recommendations can be oriented for improving both descriptive and functional knowledge (Dorcey and Hal l , 1983). Descriptive knowledge may include data on fish distribution and abundance, species diet compositions, fleet characteristics, oceanographic parameters, etc., and involves research activities such as inventorying and monitoring. Functional knowledge involves understanding system relationships such as how recruitment is affected by reducing the spawning stock, how catch rates change with oceanographic conditions, etc. New functional knowledge can only be developed by carrying out experiments to test hypotheses through either experimental research, experimental management or desk analysis (Dorcey and Hal l , 1983). The later should precede experimental management or experimental research in order to develop new hypothesis and guide experimental design. Very often, when experimentation is difficult or unfeasible, the comparative method of analysis has been employed to acquire new functional knowledge of ecological processes (Mayr, 1997). The two types of knowledge are linked by the fact that new functional knowledge is only obtained when the data produced by inventorying and monitoring are used to test hypothesis about system processes. Characteristically, research recommendations w i l l have a longer time frame for involving the understanding of processes operating at medium and long-term scales (e.g. oceanic regimes, recruitment variability, cycles of investment, technological changes, etc.). 12 Policy recommendations, on the other hand, require short-term responses from scientists who are compelled to explicitly confront and communicate uncertainties to decision makers. There are usually competing hypothesis about the dynamics of a natural resource, and the outcomes of a given management decision may differ considerably according to which o f the alternatives come out to be true. Therefore when decisions are made under uncertainty, managers are bound in effect to consider the risks associated with their actions and the impacts of their choices on the 'legacy of uncertainty' that their successors w i l l face. The emphasis on the need to account for uncertainties when making decisions is however not often transparent. Morgan and Herion (1990) describe situations in which accounting for uncertainties is crucial for a decision analysis, e.g., when people's attitude towards risk is likely to be important, such as in many health and environmental issues; when multiple information sources and uncertainties must be combined; when a decision must be made on the allocation of resources to obtain additional information on the problem being analyzed; and in situations where the process of decision making involve multiple actors making explicit and implicit decisions over an extended period. In these cases the analysis w i l l be more useful i f it treats the uncertainty explicitly allowing users to evaluate its conclusions and limitations better in the changing context of the ongoing decision process (Morgan and Herion, 1990). Taking account of uncertainties is on the other hand a fundamental requirement for the adoption of sound precautionary measures in fisheries management ( F A O , 1997). Fisheries stocks assessment is usually faced with uncertainties on the biological production, and on the expected outcomes of harvest strategies and controls. Uncertainties on the biological production, and hence in the relationship between current catches and the future state of the stock, frequently result from biases in fishery data as a descriptor of stock dynamics, sampling errors, and by process errors created by environmental effects on recruitment and production rates. Also , the ecosystems supporting small pelagic fish populations undergo productivity changes of decadal frequencies, often called 'regime shifts' (Steele, 1996; 1998). These systematic shifts in productivity can be driven by a variety of factors including climate changes, variability in the abundance of preys, competitors, predators as wel l as by changes in the internal structure of the stock (Walters, 1987). The lack of understanding of processes controlling biological production and the multi-interest nature of resources exploitation generates uncertainties in the choice for management strategies and in the expected outcomes of a chosen fishing strategy. Finally, the information required to achieve the objectives of a fishing strategy 13 may include data on spatial distribution, temporal distribution and abundance of the stock, and on fleet characteristics and movement on the fishing ground. The information requirements for harvest control are best remembered as the answers to the questions, \"where, when and how many ?\" (Mundy, 1985). Uncertainties in the outcomes of the implemented tactics arise whether the information on the state of the resource is poor or catch rates vary independently of stock size as a result of changes in catchability. Fisheries advisory work can be based on two types of modeling approaches: single-species and ecosystem or multi-species models. Multi-species models are in their infancy and have been little used to date. Multi-species models have been proposed as tools for guiding the implementation of ecosystem principles in fisheries management, but exactly how the approach could be used, and also what should be the role of single-species approaches in this new paradigm is still unclear. Such comparisons have not been performed explicitly before. This thesis is organized in four chapters. Chapter 2 uses trophic models to examine the ecosystem impacts of harvesting small pelagic fishes in selected marine ecosystems. Results are used to discuss three main questions: i) what are the general predicted ecosystem responses to fishing small pelagics species?; ii) what is the role of trophic cascades in the replacement of species at the pelagic planktonic niche? and; iii) what are the characteristics of stability of marine ecosystems when impacted by fishing at the mid-trophic level groups? The validity of the predicted results is evaluated with published observations and limitations of the modeling approach. Chapter 3 analyzes the carrying capacity of marine shelf ecosystems in southern Brazi l for harvestable species by (1) quantifying the amount of available primary production appropriated by fisheries catches, (2) evaluating the trend in the mean trophic level of fisheries, and (3) simulating the ecosystem effects of 'fishing down the food web' in an intensively exploited shelf region. The analysis aims to forecast the 'fishing down marine food web' effects in Brazi l , and to compare the footprint of fisheries in two of the most intensively exploited regions of Brazi l : the southern and southeastern shelf ecosystems. 14 Chapter 4 reviews the status of fisheries assessment of the Brazil ian sardine in the Southeastern shelf ecosystem. The fishery was chosen as a case study because it exhibits many similarities with other pelagic fisheries worldwide, notably for its history of development and collapse, as well as for its regional importance as one of the most productive and well studied marine fisheries resources in Brazi l . Ecological and fisheries data are used to describe the changes in recruitment, stock catchability, and the structure of the Southeastern Brazilian Bight ecosystem. This information is used to formulate hypotheses about the resilience of the population and ecosystem when impacted by fisheries. Chapter 5 centers on the assessment of risks of management decisions in the sardine fishery with currently available information and according to two modeling approaches: a single-species and a multi-species model. The analysis aims to evaluate the short and long term predictions of the impacts of harvest strategies and controls; to discuss the relative values of reducing current process uncertainties; to recommend the types of research that would most likely provide the information needed to improve the quality of decisions; and to suggest the precautionary measures that should be adopted in face of the ecological uncertainties. Finally, Chapter 6 summarizes the general conclusions o f this work. The map in figure 1.2 identifies the trophic models used in each chapter to evaluate the ecosystem responses to fishing small pelagic fish species. 15 California anchovy Chapter 2 \u00C2\u00AB Venezuela small pelagics Chapter 2 N W Africa sardine Chapter 2 Peru anchovy Chapter 2 Southern Brazil anchovy Chapter 3 SE Brazil sardine Chapter 4 & 5 Namibia sardine & anchovy Chapter 2 Figure 1.2. Ecosystems for which trophic models were used to evaluate the effects of fishing small pelagic fish species. The boxes indicate the species fished and the correspondent chapters where the models were used. 16 Chapter 2. Ecosystem impacts of harvesting small pelagic fishes. A comparative analysis using trophic mass-balance models1 2.1. Introduction Fisheries management has been urged to consider the ecosystem impacts of fishing activities, given the signs of human dominance and impact on the oceans. Recent assessments of the worldwide status of marine capture fisheries revealed for instance that, fisheries alone appropriate ca. 8% of the total marine primary production and up to one third of temperate shelf systems production (Pauly and Christensen, 1995); over 60% of the most important fish stocks are either overexploited or at the verge of becoming overexploited by current fishing intensity (Garcia and Newton, 1997); and that approximately 27 mil l ion tons o f nontarget animals are discarded annually as \"trash\" fish (Alverson et al., 1994). Also , present exploitation patterns are resulting in a \"fishing down marine food webs\" phenomenon, by which heavy commercial fisheries are causing a progressive simplification of ecosystems in favor of smaller, high-turnover, lower-trophic-level fish and invertebrate species that are adapted to withstand disturbance and habitat degradation (Pauly et al., 1998; Pitcher and Pauly, 1998). The effect of fisheries on ecosystems are usually classified in direct and indirect impacts (Botsford et al., 1997; Goni, 1998). Direct impacts include overfishing, by-catch and discard of non-target species, changes in genetic diversity of stocks, physical disturbances and habitat destruction by fishing gears. Fisheries, for instance, have the potential to affect the genetic diversity of populations by selectively removing older and larger individuals (Goni, 1998), and by depleting small reproductive stocks of metapopulations of species such as salmon and herring (Policansky and Magnuson, 1998). Ultimately, the decrease in genetic diversity by fishing can cause loss of resilience of fish populations to both human and natural impacts. Fisheries are also responsible for physically damaging important marine habitats, specially with towed gears (otter trawlers, beam trawlers, dredges). The effects of trawling on the sea bed vary from destruction of suitable habitats for the settlement of juvenile and adult phases of diverse marine organisms, changes in abundance and species composition of benthic communities, and concomitant changes in the fish species composition associated with the physical alterations of bottom structures (Goni, 1998). 17 Indirect ecosystem impacts of fisheries are at least partly mediated through the food web, where the'effects of fishing 'cascade' to other components of the system. The ecosystem-wide effects of catching fish attracted little attention and probably had minor importance before the advent of large-scale industrialized fisheries. But, today, the scale of fish production can considerably alter the structure of marine food webs. Heavy commercial fishing has been often associated with drastic changes in species composition in marine communities. Examples of major changes in marine ecosystems with intense fishing are present in almost every region of the world (Goni, 1998), although in many cases it has proved difficult to separate the natural and anthropogenic causes of changes (Steele, 1998). Among the best documented examples are the cases of switches in dominance between sardines and anchovies in coastal upwelling systems after the activity of reduction fisheries (Lluch-Belda et al., 1989; Bakun, 1996). In the Bering Sea, human exploitation of whales and other top predators is thought to be responsible for cascading effects on other components of the ecosystem, such as declines in sea lions and seals, and the dominance of groundfish species such as pollock, Theragra chalcogramma, and large flatfishes ( N R C , 1996; Trites et al., 1999). On Georges Bank, large scale disturbances caused by intense fishing and habitat destruction were also associated with apparent replacement of gadids and flounders by species of low commercial value, including dogfish sharks and skates (Fogarty and Murawski, 1998). Over the last 70 years a dramatic catalogue of stock collapses have involved small pelagic forage fish, but despite some pioneering analyses (Beddington and May , 1977) there have been few rigorous attempts to model and predict the potentially devastating ecosystem consequences of overfishing. In the case of the Peruvian ecosystem there is clear belief, but little evidence, that the loss of ca. 5 mil l ion guano-producing birds with the collapse o f the anchoveta stock w i l l have had profound ecosystem impacts (Pauly and Tsukayama, 1987). Beverton (1990) warned against the subtle consequences to the ecosystem that may result from the collapse of major small pelagic populations. He suggested that there was \"some inferential evidence that the disappearance of some 10 mil l ion tonnes of Norwegian spring spawning herring and 2 mil l ion tonnes from the North Sea may have resulted in reorientation o f the flow of production into alternative stable states\". Some believe this reorientation of flow was responsible for the \"gadoid outburst\" (Cushing, 1980). In a review of cases of species replacement, Daan (1980) concluded that for the North Sea some sort of replacement was likely. However, under his rather strict 18 criteria for replacement, only one out of nine candidate cases (Pacific sardine and anchovy) could be considered as true replacement rather than coincident changes. While evidence of ecosystem impacts of fisheries accumulates throughout the world, we see the development o f fisheries management principles and guidelines which acknowledge the need to place fisheries in an ecosystem context (e.g. Mangel et al., 1996; F A O , 1997). Scientific advisory work is now compelled to consider the more diverse as possible aspects or consequences of fishing activities to marine populations and ecosystems (Mangel et al., 1996). Although the need for ecosystem management has been widely recognized, scientific advice is still hampered by the lack of understanding of the complex dynamics of ecosystems, and the lack of consensus on which framework should be used to account for the ecosystem effects of fisheries. Despite the reality that fisheries are generally not restricted to affecting one species alone, the development of single species models for fisheries management has centered around that very assumption. And , due to our lack of ability to model complex systems, such methodology is still prevalent. Attention during the last decades has been given to the development of tools that describe patterns of trophic interactions in the food web, mainly represented by multispecies virtual population analysis, M S V P A (Sparre, 1991), and ecosystem mass-balance models such as Ecopath (Polovina, 1984; Christensen and Pauly, 1992). The implementation of the first type of modeling tool has been hampered in part by the need for extensive time series of catch-at-age data, difficult parameterization, the high degree of expertise required from the modeller, and the overall lack of transparency in the estimation procedure (Walters et al., 1997). Ecopath offers, on the other hand, a simpler approach for the reconstruction of trophic interactions in fished ecosystems, and has been widely applied to aquatic ecosystems (more than 80 Ecopath models have been published world-wide describing upwelling, shelf, lake, river, open ocean, and terrestrial farming systems; see the Ecopath homepage at http:Wwww.ecopath.org). The approach has some advantages over other existent trophic models viz. , it includes all trophic levels in the analysis (from primary producers to top predators) as opposed to focusing only on the commercially important fish species; the emphasis on ecological relationships makes it intuitively simple; it incorporates an standardize large amounts of scattered information, from data routinely collected by fisheries scientists and marine biologists to the traditional ecological knowledge (Pauly et al., 1998); and, more importantly, the widespread use of the approach 19 creates opportunity for comparative studies of ecosystem's response to fisheries impact. Moreover, further developments of the mass-balance model, which originally focused on describing systems at steady-state conditions, has resulted in a dynamic ecosystem model called Ecosim (Walters et al., 1997) that is capable of answering \"wha t - i f questions about policy and ecosystem changes that would cause shifts in the balance of trophic interactions. This chapter uses Ecosim to examine the ecosystem impacts of harvesting small pelagic fishes in selected models of marine ecosystems. Many marine ecosystems share a similar configuration of their biological community structure, characterized by a crucial intermediate trophic level often occupied by a small plankton-feeding pelagic species (Rice, 1995; Bakun, 1996). Unlike typical food webs where different types of predators feed upon different types of prey, in these ecosystems one prey type usually dominates as the primary channel of energy from lower to higher trophic levels (Fig. 2.1). Trophic dynamics in these 'wasp-waist' ecosystems (sensu Rice, 1995) is thought to be largely dependent on this mid-trophic level species, often represented by important fisheries resources of anchovy, sardines and herrings. Results of model simulations are discussed around three main questions: i) what are the general predicted ecosystem responses to fishing small pelagic species?; ii) what is the role of trophic cascades in the replacement of species in the pelagic planktivore niche? and; i i i) what are the characteristics of stability of exploited marine ecosystems when impacted at the 'wasp-waist' species? Finally, the validity of results is evaluated with published observations and the limitations of the model. The practical utility of the results and approach for defining ecosystem objectives in fisheries management is discussed. 20 Figure 2.1. Schematic representation of marine food webs with waists at the mid-trophic level. A waist is characterized by a small number of taxa that transfer most of the energy between lower and higher trophic levels. In 'wasp-waist' ecosystems only one or a few species at the intermediate trophic level transfer most of the production from all lower levels to predators at the top of the food web. The flow diagram on the right shows a pelagic system where a small pelagic species (anchovy) is the 'wasp-waist' species. The thickness of lines is proportional to the biomass flow between species; double-head arrows indicate exports from the system. 2.2. Methods Ecosystem impacts of harvesting small pelagic fishes were examined with a dynamic trophic model, Ecosim (Walters et al., 1997), structured from mass-balance assessments with Ecopath (Polovina, 1984; Christensen and Pauly, 1992). Ecopath provides a static picture of the ecosystem trophic structure by estimating trophic flows and biomasses which satisfy growth and mortality constraints. The model relies on the truism that for each group (i) in the system, and to any time period: Production by (i) = All predation on (i) - Fisheries catches - Other mortality - Losses to adjacent systems This can also be articulated as Bt.(^)i.EEl -J-Bj.^pj.DCji-(Y+EX)i = ABt (1) 21 where in a system of i=l,...,n functional groups; P/Bj is the production/biomass ratio of (i) (equal to the total mortality rate Zj under the assumption of equilibrium); E E ; is the ecotrophic efficiency, i.e. the fraction of the production that is accounted for by consumption within the system (predation) or harvested; Y ; is the yield of (i), in weight, with Y{ = Fj .B ; , where F ; is the fishing mortality; E X ; is other exports of (i) from the system; Bj is the biomass of the consumers or predators; (Q/B)j is food consumption per unit of biomass for consumer j , and DCjj is the fraction of i in the diet of j . A B i is biomass accumulation rate per time in cases where the analysis do not use data from an initial equilibrium situation. B y re-expressing the system of linear equations (1) as differential equations, Ecosim provides a dynamic model suitable for simulation of the effects of F varying in time on the biomass of each group in the system. The model provides dynamic biomass predictions of each (i) as affected directly by fishing and predation on (i), changes in food available to (i), and indirectly by fishing or predation on other groups with which (i) interacts (Walters et al., 1997). Constructing a dynamic model from equation (1) involves three changes; a) replace the left side with a rate of change of biomass; b) provide a functional relationship to predict changes in P / B ; with biomass Bj and consumption, and c) provide a functional relationships predicting how the consumption w i l l change with changes in the biomasses of B f and Bj (Walters et al., 1997). Thus equation (1) is re-expressed as where f ( B J is a function of Bj i f (i) is a primary producer or f(Bj) = g ; X Cjj (B;,Bj) i f (i) is a consumer, where gi is the net growth efficiency, and c^Bj.Bj) is the function used to predict consumption rates from B ; to Bj. Ecosim uses a function for Cy derived from assuming possible spatial/behavioral limitations in predation rates f(Bi)-M.Bt -Fi.Bi- X cij(Bj.Bj) n (2) dt (3) where Cy is the trophic flow, biomass per time, between prey (i) and predator (j) pools; 22 B ; and Bj are the biomasses of prey and predators, respectively; aSj is the rate of effective search for prey i by predator j ; and Vy and v'jj are prey vulnerability parameters Parameters Vy and v'y represent the rate of exchange of biomass between two prey behavioural states: a state vulnerable to predation and a state invulnerable to predation. The rationale of this representation is that at a given moment in time not all prey biomass is vulnerable to predators; predator-prey relationships in nature are often limited by behavioral and physical mechanisms, such as schooling behavior and diel vertical migration patterns in clupeid fish or spatial refuges used by many reef fish that considerably limit exposure to predation. The model is designed so that the user can specify the type of trophic control in the food web by hypothesizing the maximum consumption rates (and indirectly the rate of exchange of biomass Vy) that a predator can ever exert on food resources. For low predator biomass or high exchange rates (vy) the functional relationship approximates a mass-action flow, or Lotka-Volterra type of model c = a B ; Bj, implying a strong top-down effect. For high consumer biomass or low exchange rates the functional relationship approaches a donor-controlled (bottom-up) flow rate (c = VyBj), so v ; j can be interpreted as the maximum possible instantaneous mortality rate that j can cause on i . General Predicted Impacts The impact of fisheries on upwelling ecosystems was examined by comparing equilibrium biomass predictions provided by Ecosim over a range of fishing mortality rates for a small pelagic species in five different upwelling ecosystems off California, Namibia, Northwest Africa,, Peru and Venezuela. Details on each model are provided in table 2.1. Two contrasting trophic control hypotheses were tested by setting the maximum instantaneous mortality rate (vy) that consumer j could ever exert on food resource i . For a 'bottom-up' control Vy was fixed as 4 times the baseline mortality rate; for a 'top-down' control Vy was fixed as 20 times the baseline mortality rate. 23 Table 2.1. Trophic models of upwelling ecosystems used in the analysis of the impact of small pelagic fisheries. California: Horse mackerel, Trachurus symmetricus; Hake, Merluccius productus; Mackerel, Scomber japonicus. Namibia: Mackerel, Scomber japonicus; Horse mackerel, Trachurus capensis; Snoek, Thyrsites atun; Hake, Merluccius capensis, M. paradoxus. NWAfrica: Anchovy, Engraulis encrasiculus; Sardinella aurita; Mackerel, Scomber japonicus; Horse mackerel, Trachurus trachurus; Hake, Merluccius merluccius. Peru: Sardine, Sardinops sagax; Mackerel, Scomber japonicus; Horse mackerel, Trachurus murphyi; Bonito, Sarda chiliensis; Hake, Merluccius productus; Pelican, Pelecanus thagus; Sea lion, Otaria flavescens. Venezuela: Small pelagics, Engualidids and Clupeids; Groupers, Serranids; Croakers, Cynoscion spp and Micropogonias furnieri; Squids, \u00E2\u0080\u0094 r-r-System/ Fished group N\u00C2\u00B0 of Commercially exploited Top predators Reference Period groups groups California 1965-1972 Anchovy, Engraulis mordax 17 Anchovy, Horse Mackerel, Hake, Demersals Marine Mammals Marine Birds Jarre-Teichman and Christensen, 1998 California 1978-1985 Anchovy, Engraulis mordax 17 Anchovy, Mackerel, Horse Mackerel, Hake, Demersals Marine Mammals Marine Birds Jarre-Teichman and Christensen, 1998 Monterey Omniv. fish 16 Carnivorous nekton, Demersals, Omnivorous fish Micronekton Sea Mammals Sea Birds Oliviera et al., 1993 Namibia Sardine, 17 Anchovy, Sardine, Marine birds Jarre-Teichman 1971-1977 Sardinops ocellatus Mackerel, Horse Mackerel, Snoek /Tuna, Other pelagics, Hake, Other demersals Marine mammals and Christensen, 1998 Namibia 1978-1983 Anchovy, Engraulis capensis 17 Anchovy, Sardine Mackerel, Horse Mackerel, Snoek/ Tuna, Other pelagics, Hake, Other demersals Marine birds Marine mammals Jarre-Teichman and Christensen, 1998 NWAfrica 1970-1979 Sardine, Sardina pilchardus 18 Anchovy, Sadinella sp, Sardine, Mackerel, Horse mackerel, Larg. Scombrids, Other pelagics, Hake, Demersals Marine Birds Marine mammals Jarre-Teichman and Christensen, 1998 Peru 1964-1971 Anchovy, Engraulis ringens 20 Macrobenthos, Anchovy, Sardine, Mackerel, Horse Mackerel, Bonito, Hake, Other pelagics, Other demersals Other mammals Pelican Sea lion Jarre-Teichman, 1991 Peru 1973-1981 Anchovy, Engraulis ringens 20 Macrobenthos, Anchovy, Sardine, Mackerel, Horse Mackerel, Bonito, Hake, Other pelagics, Other demersals, Fur seals Other mammals Pelican Sea lion Jarre etal., 1991 Venezuela Small 16 Small pelagics, Small Small sharks Mendoza, 1993 pelagics sharks, Scombrids, Scombrids Carangids, Groupers, Barracudas Squid, Croakers, Shrimps Snappers Groupers 24 Ecosystem Stability Fisheries exploiting the 'wasp-waist' populations have a potential disrupting effect on marine ecosystems communities with direct impact on dynamic stability. There are three general views of ecosystem stability in the ecological literature (Holling, 1973; Peterman et al., 1978): one assumes that ecosystems are globally stable and tend to recover their original structure after disturbance; a second view is that ecosystems are highly unstable, and that disturbances w i l l lead to system collapse; a third and intermediate view between the two extremes believes that ecosystems may have more than one equilibrium state which are separated by boundaries in the processes controlling the ability of the system to respond to disturbance. In this section ecosystem models are used to analyze the first type of stability, i.e., the ability of system to return to an equilibrium state after a temporary disturbance. The presence of narrow waists in food webs make it possible to compare the impact of disturbances on functionally similar groups in different ecosystems, which them allow us to derive stability properties from inter ecosystems comparisons. Stability comparisons were carried out with 23 models of marine ecosystems (Table 2.2). In each model a group or species was selected using a set of criteria allowing direct comparison of results among ecosystems. The criteria define the characteristic role of wasp-waist species, which must: (i) occupy an intermediate trophic level; (ii) provide an important link between lower and higher trophic levels, indicated by a high energy throughput compared to other groups at similar trophic levels, as well as by the species importance in the diet of higher trophic level groups; and (iii) already be fished in the baseline Ecopath model. A fishing pattern was chosen which generated a 5 fold increase in fishing mortality on the wasp-waist group. Simulations were run with a bottom-up control setting (mean vij =4.0 over all i-\u00C2\u00BBj flows). Throughout the work we tested different levels of increase in F without noticing any substantial difference in results. We decided for a 5 fold increase to create a situation where the group is severely depleted or displaced from its original steady-state condition in a relatively short period of time. Such a extreme scenario is routinely applied to many small pelagic fish populations and very often associated with stock collapse (Patterson, 1992). The higher F values 25 were kept constant for 10 years, then returned to the baseline, with the model running for further 80 years (Fig. 2.2). This allows estimation of the time it took the system to return to its original state after having been impacted, that is, the time it takes the last impacted group to return to its baseline biomass level (Fig. 2.2). The recovery time was considered as a measure of the internal stability of the model and hence an index of ecosystem stability. Figure 2.2. Simulation of system recovery of the Venezuela shelf model after fishing impact imposing a 5 times increase in fishing mortality. Top graph shows changes in biomass (B) by group; fished group: small pelagics (Sardinella auritd); last group to recover: croaker (Cynoscion spp. and Micropogonias furnieri); ^ start of simulation with F baseline; interval t, -1 2 sets the time the system was kept under higher fishing mortality rates (here 10 years); interval t2 - t3 corresponds to the system recovery time or time it took the last impacted group to recover to its original biomass. Time t4 equals 100 years. To identify ecosystem attributes directly related to stability, a comparative analysis was carried out using a set o f ecosystem goal functions previously identified as representative of Odum's attributes of ecosystem maturity (Christensen, 1995a). The attributes were chosen to represent 26 three different aspects of ecosystem development: i) complexity in community structure, ii) community energetics and i i i) overall community homeostasis. A more detailed description of the indices listed below can be found in Christensen and Pauly (1992) and Christensen (1995a). i) Complexity in community structure: - Total number of ECOPATH pools (Npools); - Connectance Index (CI): The connectance index represents the ratio between the number of actual trophic links to the number of possible links in the system. Ecosystems are expected to evolve from linear to weblike trophic relationships as they mature; the more complex is the trophic structure the higher is the connectance index. - System Omnivory Index (SOI): The system omnivory is calculated as the average omnivory of all consumers in the system, omnivory being defined as the variance of the trophic level of preys of a consumer. The SOI measures how the feeding interactions are distributed between trophic levels and can be used along with the connectance index to describe the food web structure of a system. ii) Community energetics: During the development of ecosystems towards a climax stage the relationship between three energetic properties, namely production, respiration and biomass, are expected to change. In the early stages of development, and hence in immature systems, the total primary production is expected to exceed respiration, leaving an excess net production to accumulate as biomass in the system. A s the system matures biomass is expected to increase and production w i l l tend to balance respiration. Five indices are used here to describe these changes in ecosystem energetics: the ratio between total primary production and total system biomass (Pp/B); system biomass and production (B/P); system respiration and biomass (R/B); the system throughput (SThr), which measures the size or total activity of the system; and the ratio between biomass and system throughput (B/T). i i i) Community overall homeostasis: Ecosystems w i l l tend to attain during the different stages of succession large and diverse organic structures as a result of community actions such as symbiosis, nutrient conservation, and the increase in the information content of flows. These mechanisms also are thought to increase the ecosystem stability. Four indices are used in the analysis of community homeostasis. Ascendancy (Ascd) is a measure of the degree of development of an ecosystem, both in terms of size (throughput) and organization (mutual information) (Ulanowicz and Wulff, 1991; Baird et al., 1991). The system internal flow overhead (Iovh) measures the fraction of the development capacity that does not appear as organized structure due to the magnitude of an uncertainty in the imports, exports and 27 respiration flows, as well as to redundancies in the pathways between organisms in the system. The cycling index (FCI), developed by Finn (1976), measures the fraction of an ecosystem's throughput that is recycled. The degree of recycling of energy and nutrients in an ecosystem is assumed to increase as ecosystems mature and develop routes for nutrient conservation (Odum, 1969). The path length (Path) is defined as the average number of components or pools that a unit of flow passes through on its way from inflow to outflow (Finn, 1980). A s diversity of flows and cycling is expected to increase with maturity, and as these tendencies should result in an increased path length it is assumed that the path length w i l l be highest for more mature systems (Christensen, 1995a). 28 Table 2.2: Models used for analyses of ecosystem stability. Group trophic level is computed as l+(the weighted average of the prey's trophic level) (Christensen and Pauly, 1992). F Baseline (year1) refers to the wasp-waist group's fishing mortality estimated from the corresponding mass-balanced E C O P A T H model. Model Location Species/Group Fished Trophic F Source Code Level Baseline Bru Brunei, Pelagic fish 2.9 0.28 Silvestre et al. (1993) South China Sea Cmp Campeche Bank, Opisthonema oglinum 2.9 0.89 Vega-Cendejas et al. (1993) Gulf of Mexico Cal65 California system, Engraulis mordax 2.6 0.05 Jarre-Teichman and Christensen 1965 - 1972 (1998) Cal78 California system, Engraulis mordax 2.6 0.09 Jarre-Teichman and Christensen 1978 - 1985 (1998) Cel Celestun Lagoon, Eucinostomus spp 3.1 0.04 Chavez et al. (1993) Gulf of Mexico GMx Northern shelf, Pelagic fish 2.0 0.11 Browder(1993) Gulf of Mexico Lng Lingayen Gulf, Sergestids 2.5 0.31 Pauly and Christensen (1993) Philippines Mpt Maputo Bay, Small pelagics 2.1 0.35 Paula e Silva et al. (1993) Mozambique Mrey Monterey Bay, Omnivorous fish 2.5 0.16 Olivirei et al. (1993) California Nam71 Benguella system, Sardinops ocellatus 2.4 0.29 Jarre-Teichman and Christensen 1971 - 1977 (1998) Nam78 Benguella system, Engraulis capensis 2.4 0.56 Jarre-Teichman and Christensen 1978 - 1983 (1998) Nsea North Sea Sandeel, 3.7 0.45 Christensen (1995b) Ammodytes spp NWAfrica Northwest Africa, Sardina pdchardus 2.2 0.13 Jarre-Teichman and Christensen 1970- 1979 (1998) Peru 60 Peru upwelling, Engraulis ringens 2.2 0.78 Jarre etal.( 1991) 1960 - 1969 Peru 70 Peru upwelling, Engraulis ringens 2.4 0.54 Jarre et al. (1991) 1973 - 1979 Sch Schlei Fjord Planktivorous fish 3.1 0.28 Christensen and Pauly (1992) Germany Thai 10 Gulf of Thailand, Small demersal fish 3.3 3.86 Pauly and Christensen (1993) 10-50 m. Thai50 South China Sea Small demersal fish 3.2 3.70 Pauly and Christensen (1993) Thau Etang de Thau, Atherinids 2.8 0.02 Palomares et al. (1993) France Trg Kuala Trengganu, Pelagic fish 3.2 2.71 Christensen (1991) Malaysia Venz Venezuela shelf, Small pelagics 2.6 0.08 Mendoza(1993) 10-50 m WMx Western Gulf of Eucinostomus spp 3.0 0.06 Arreguin-Sanchez et al. (1993a) Mexico Yet Yucatan shelf, Herring, 3.2 0.01 Arreguin-Sanchez et al. (1993b) Mexico Opisthonema oglinum 29 2.3. Results and Discussion Ecosystem responses to fishing Figures 2.3 to 2.11 show the predicted equilibrium effects of fishing mortality rates on the biomass and yield of small pelagic species, and on the biomass o f low, medium and high trophic level groups. Altering the ecosystem control mechanism between assumptions of top-down and bottom-up control results in marked difference of the predicted biomass changes in terms of both rate and magnitude; biomass changes are more extreme under assumptions of top-down control of trophic interactions. For all systems, corresponding with depletion of small pelagics, biomass of food groups (specially zooplankton) tend to increase. A t the mid-trophic level, predictions of increase in biomass of a competing small pelagic species (e.g. sardine, anchovy or micronekton) occur in all but in the Peruvian system during the 1970s (Peru70 model; F ig . 2.10). The degree of replacement of one species by the other is more pronounced under assumptions of top-down control for the California system (Figures 2.3 and 2.4), the Namibian system (Figures 2.6 and 2.7), and the Northwest Africa system (Figure 2.8). In these cases it is the decrease in the biomass o f a common predator, following the depletion of a forage species, that leads to a burst in the production of the competing mid-trophic level species. This is less apparent under bottom-up control, when competition for food resources were limiting interactions. Some of the models presented small oscillations in the predicted equilibrium biomasses, specially in the top-down cases (Figs 2.6 and 2.8). These types of responses do not correspond to 'real' predictions, as they result from an approximation problem caused by not running the model long enough to reach the actual equilibrium. 30 Top-down Bottom-up Biomass ^ [ Figure 2.3. Predicted equilibrium biomass and catches of anchovy in the California System (1965-1972) under different fishing mortality rates (upper panel); lower panels show the predicted percent change in biomass (relative to the Ecopath baseline level) of all other groups in the system with the change in equilibrium fishing mortality rates for anchovy. Ecopath groups are organized according to their position in the food web (from lower to higher trophic levels). 31 Top-down Bottom-up et \u00C2\u00A3 o s cu OX) d a JS U 15 r 10 h E c \u00E2\u0080\u00A2 o Biomass Catch 0.0 0.5 F anchovy (year ) 1.0 0.5 H. mack. T.O cu OX) a 0.0 10 r Biomass Catch 0.5 F anchovy (year ) 1.0 Lg. Scomb. \"Willi I 300 200 100 0 l -100 -200 Lg. scomb. Hake/Whit. & jniiiiiiilllllllllllllllllllllllllll Mar. birds Figure 2.4. Predicted equilibrium biomass and catches of anchovy in the California System (1978-1985) under different fishing mortality rates (upper panel); lower panels show the predicted percent change in biomass (relative to the Ecopath baseline level) of all other groups in the system with the change in equilibrium fishing mortality rates for anchovy. Ecopath groups are organized according to their position in the food web (from lower to higher trophic levels). 32 Top-down Bottom-up \u00C2\u00AB5 CA CS E o CQ _g cu OX) fl es J= U -200 \"Biomass -Catch 1.0 1.5 2.0 F Omnv. fish (year ) Mcrozoop. mtlllllllllllllllllllllllllllliiiillllllllllllllllllllllllllll Cam. fish CA CA 05 s o S sa D\u00C2\u00A3 fl S3 JS u 0.0 100 50 S?0 i -50 -100 -200 Biomass Catch 0.5 1.0 1.5 2.0 F Omnv. fish (year ') 1.5 f 2.0 Sm phyt. Figure 2.5. Predicted equilibrium biomass and catches of omnivorous fish in the Monterey System under different fishing mortality rates (upper panel); lower panels show the predicted percent change in biomass (relative to the Ecopath baseline level) of all other groups in the system with the change in equilibrium fishing mortality rates for omnivorous fish. Ecopath groups are organized according to their position in the food web (from lower to higher trophic levels). 33 Top-down Bottom-up C\u00C2\u00AB ca S \u00C2\u00A9 S a CU DX> 8 ca u 200 Biomass Catch F sardine (year ) Zoop. \u00C2\u00BB mi H. mack. o pa CU OJD a 93 .a u Biomass Catch 0.0 20 15 10 5 S? 0 -5 i -10 -15 -20 200 100 i -100 -200 200 100 ^ 0 ( -100 -200 0.5 1.0 F sardine (year ) Zoop 0.5 R ty t T.0 0 5 Anchovy 1 0 0.5 Mar. mam Mar. birds Figure 2.6. Predicted equilibrium biomass and catches of sardine in the Namibia System (1971-1977) under different fishing mortality rates (upper panel); lower panels show the predicted percent change in biomass (relative to the Ecopath baseline level) of all other groups in the system with the change in equilibrium fishing mortality rates for sardine. Ecopath groups are organized according to their position in the food web (from lower to higher trophic levels). 34 Top-down Bottom-up Figure 2.7. Predicted equilibrium biomass and catches of anchovy in the Namibia System (1978-1983) under different fishing mortality rates (upper panel); lower panels show the predicted percent change in biomass (relative to the Ecopath baseline level) of all other groups in the system with the change in equilibrium fishing mortality rates for anchovy. Ecopath groups are organized according to their position in the food web (from lower to higher trophic levels). 35 Top-down Bottom-up Figure 2.8. Predicted equilibrium biomass and catches of sardine in the Northwest Africa System (1970-1979) under different fishing mortality rates (upper panel); lower panels show the predicted percent change in biomass (relative to the Ecopath baseline level) of all other groups in the system with the change in equilibrium fishing mortality rates for sardine. Ecopath groups are organized according to their position in the food web (from lower to higher trophic levels). 36 Top-down Bottom-up o s a es JS O Biomass Catch 0.0 0.5 1.0 1.5 2.0, F anchoveta (year ) 2.5 3.0 Zoop. 0 mi mi mi inn i 700 600 500 400 300 LOther. pelagics 2oo L I i H. Mack. 100 0 -100 Sardine mil mi Hill llll II a S o S a CU ec a ss JS U Biomass Catch 0.5 1.0 1.5 2.0 2.5 F anchoveta (year') Figure 2.9. Predicted equilibrium biomass and catches of anchoveta in the Peruvian System (1964-1971) under different fishing mortality rates (upper panel); lower panels show the predicted percent change in biomass (relative to the Ecopath baseline level) of all other groups in the system with the change in equilibrium fishing mortality rates for anchoveta. Ecopath groups are organized according to their position in the food web (from lower to higher trophic levels). 37 Top-down Bottom-up Figure 2.10. Predicted equilibrium biomass and catches of anchoveta in the Peruvian System (1973-1981) under different fishing mortality rates (upper panel); lower panels show the predicted percent change in biomass (relative to the Ecopath baseline level) of all other groups in the system with the change in equilibrium fishing mortality rates for anchoveta. Ecopath groups are organized according to their position in the food web (from lower to higher trophic levels). 38 Top-down Bottom-up \u00E2\u0080\u00A2Biomass \u00E2\u0080\u00A2Catch 0.5 F Small pelagics (year ) 1.0 Z o o i Benth. prod. _j i i 0.5 1.0 Phyt. VI d E o s d W) d cs JS U Biomass Catch 0.0 0.5 F Small pelagics (year ) 1.0 j i i i 0.5 Croackers i Q -100 200 200 r tfiiiiiiiiiiiiiiiiiiililnilimiiiii Scorrb. _J I L_ Sharks 0.5 1.0 Snappers Scomb Figure 2 .11 . Predicted equilibrium biomass and catch of small pelagic fish in the Venezuela System under different fishing mortality rates (upper panel); lower panels show the predicted percent change in biomass (relative to the Ecopath baseline level) of all other groups in the system with the change in equilibrium fishing mortality rates for small pelagic fish. Ecopath groups are organized according to their position in the food web (from lower to higher trophic levels). 39 Responding to the depletion of small pelagic species, top predators display a decrease in biomass in all but in the California system models (Figures 2.3, 2.4 and 2.5). The degree to which predators are impacted further depends upon the relative importance of the small pelagic fish in their diet. A s would be expected, those predators that rely solely upon small pelagics as their prey are impacted most heavily. On the other hand, predators that do not rely primarily on small pelagics, such as horse mackerel and mackerel in the Peruvian system o f 1970s (Figures 2.10) and pelicans in the Peruvian system of 1960s (Figure 2.9), tend not to be heavily impacted, the increase in biomass of other species compensating their diet loss. In some systems, competitors of small pelagics represent important food items in the diet of top predators and hence there are close links between changes in biomass. For instance, while most top predators are negatively impacted by the depletion of small pelagics in the Peru, Namibia and Northwest Africa models (Figs. 2.6 to 2.10), sea mammals and predatory fish in the California system models appear to benefit (Figs. 2.3 to 2.5), probably as a result of the increased biomass of other prey groups. Evidence does exist of top predators, specially seabirds, that have been able to successfully shift their diet between sardine and anchovy following collapse (Velarde et al., 1994). But, experience gained in the Peruvian and Northwest Africa systems show undoubting evidence of drastic impacts on predatory fish, marine mammals and birds following the collapse of forage resources (Pauly and Tsukayama, 1987; Cury et al., in press). Cury et al. (in press) used this information to infer that the dominant relationship between top-predator and prey in upwelling ecosystems is bottom-up. Small pelagics evolved life-history strategies in the form of schooling and shoaling behavior to reduce mortality risks by predation, so it is expected that aggregation in the pelagic environment does limit predation. In all cases analyzed here, model simulations suggest that changes in the biomass of top-predators following the collapse of their prey occurs under both assumptions, and are in fact more pronounced under assumptions of top-down control. M u c h scientific debate on the causes of shifts in species composition in many marine ecosystems has focused on a 'bottom-up' perspective in which the effect of physical forcing, mediated through climatic-oceanographic processes, leads to changes in primary production and reproductive success of fish populations which in turn w i l l trigger changes in the food web (Beamish, 1995; Bakun, 1996). A n alternative 'top-down' perspective, mostly applied in the study of lakes, asserts that predation affects directly and indirectly the structure o f populations and communities, and production processes at all trophic levels in the food web (Kitchell et al., 40 1994). Mode l simulations allowed us to explore the effects of both assumptions on predicting the changes in the food web accompanying the exploitation of small pelagic fish. Under top-down control the system responded with a marked increase in the biomass of a competitor species due to cascade effects up and down the food web. These results differ from the early emphasis placed on food competition as the driven force of species replacement (Cushing, 1980; Daan, 1980), and reinforce the potential role of predation mechanisms and trophic cascades effects in shaping the dynamics of mid-trophic level, forage species. The fact that in the four major upwelling ecosystems analyzed, natural mortality of small pelagic fish due to predation is much higher than their fishing mortality (Jarre-Teichman and Christensen, 1998) corroborates the above hypothesis. The observed dynamics demonstrate that small pelagics play a central role in upwelling ecosystems. Their high throughput of energy, their intermediate trophic level, and the high connectivity to other components in the ecosystem indicate that depleting stocks of small pelagics through fishing is expected to have important consequences for other ecosystem components and for fisheries on other commercially important species. Where these other species are dominant predators of the small pelagics the likely outcome is a reduction in their biomass and catch. The converse may be true when the target species is a competitor, the increased biomass of food sustaining greater biomass of the competitor. I f validated, these results indicate that one of the long term effects of intensive fisheries for small pelagics might be the intensification of the observed downward trend in the mean trophic level of fisheries (Pauly et al., 1998), by both accelerating the decline of species at the top of the food chain and favoring the outburst of other mid-trophic level forage species. Predicted equilibrium yields and fishing mortality at maximum sustainable yield (F m s y ) also vary significantly between the two control scenarios (upper panels in Figs. 2.3 to 2.11). Bottom-up control generally produces a catch curve that achieves an asymptote at higher Fs, predicting that stock can sustain much higher fishing pressure before it begins to decline. This occurs because under bottom-up control, predation mortality rate (My = Cy/Bj) tends to remain more stable, while the consumption rate (C/Bj) of small pelagics increases more, to make them more productive per biomass due to donor control of total food eaten (Cj). Strong predation control generally prevents higher yields under top-down control. F m s y values predicted under the two control scenarios are compared in Figure 2.12. A s stated before, using a bottom-up control predicted 41 higher F m s y values. Wi th the exception of two models (Namibia78-83 and Peru 64-71), predicted F^y values fall in a conservative range of fishing mortality rates, below the sustainable exploitation rate of 0.4 suggested for small pelagic fish stocks by Patterson (1992). o Top-down \u00E2\u0080\u00A2 Bottom-up anch. anch. omn. fish sard. anch. sard. anch. anch. s. pel. Cal65-72 Cal77-85 Monterey Nam71-77 Nam78-83 NWAfrica Peru64-71 Peru73-81 Venezuela 2.5 Peru 64-71 \u00E2\u0080\u009E 1.5 Namibia 78-83 CD \u00C2\u00B0 - * 0.5 \u00C2\u00A7 \u00E2\u0080\u00A2 % \u00E2\u0080\u009E o 0.5 1 1.5 M (year 1) Figure 2.12. Upper panel: Predicted fishing mortality at maximum yield (F m s y ) under two contrasting assumptions on the dominant type of trophic control in the food web. The x-axis indicates the species fished in the respective models. Lower panel: relationship between F m s y and the natural mortality rate of the small pelagic species fished in each model. The dotted line represent the exploitation rate (F/Z) of 0.4 suggested by Patterson (1992) as a sustainable fishing rate for small pelagic stocks. 42 Ecosystem stability and resilience O f the 23 models analyzed, one (Thai 10 model) did not recover the original stable state after being perturbed. Walters et al. (1997) suggested that such unstable behavior is expected when Ecosim is used to extrapolate to a state far away from that described by a given Ecopath model, or when diet composition data and group definitions are inadequate to represent actual trophic interactions among pools. Six out of the eight upwelling ecosystem models used in the analysis had not recovered completely their original biomass levels 80 years after the end of the perturbation (Figure 2.13). The low stability of these models must reflect some characteristics of upwelling ecosystems, either in terms of the importance of the impacted wasp-waist populations to other ecosystem components, or in terms of intrinsic properties of the food web captured in the models. One immediate conclusion from these results is that strategies of pulse fishing, historically applied to many small pelagics, have a potential destabilizing effect in upwelling ecosystems on time scale responses far from most projections customarily provided by conventional stock assessment research. On the other hand, results indicate an apparent lack of biological 'mechanisms' in these systems to cope with perturbations. That is explored below using ecosystem indexes calculated by Ecopath. 10 20 30 40 50 60 70 80 More Recovery time (years) Figure 2.13. Recovery time of the 23 models perturbed by pulse fishing at the mid-trophic level, wasp-waist populations. The recovery time of the models of upwelling ecosystems is highlighted. 43 System recovery time of the remaining 15 models that recovered the baseline equilibrium biomasses was correlated against the different ecosystem attributes (Table 2.3). Only Finn's indices of cycling and path length are statistically correlated to recovery time (r = -0.76 and -0.67, respectively). On the other hand, the assumption that ecosystem organization is a more important aspect of system stability (Ulanowicz, 1986) was not confirmed: none of the attributes of trophic web organization (Relative ascendancy and Internal flow Overhead) was correlated with recovery time. Community complexity and energetics attributes also showed low correlation, as may perhaps have been expected (Begon et al., 1990; Hansen and Jorgensen, 1990). M a y (1972) concluded that a simple relationship between stability and complexity may not be obtained, and in some cases, complex systems may fluctuate more than less complex ones. Here, we refute the hypothesis that stability and complexity (expressed in terms of connectance, web structure (SOI) and trophic groups richness) are correlated in the 15 models compared. Table 2.3: Correlation among the ecosystem attributes defined in the text. Numbers in bold indicate significance at the 5% level (d.f. = 13). Recovery time (Rec. time) is the measure of stability utilized in the analysis. CI: connectance index; SOI: system omnivory index; Npools: total number of system components; FCI: Finn's cycling index; Ascd: ascendancy; Iovh: system internal flow overhead; R/B: respiration/biomass ratio; Pp/B: primary production/biomass ratio; B/T: biomass/throughput ratio; B/P: biomass/production ratio; SThr: total system Attribute Complexity Homeostasis Energetics CI SOI Npools FCI Path Ascd lOvh R/B Pp/B B/T B/P SThr Rec. time -0.10 -0.15 0.12 -0.76 -0.67 0.09 -0.32 -0.10 0.33 -0.13 -0.14 -0.28 CI 0.28 -0.57 0.13 -0.09 -0.47 0.53 -0.25 -0.23 -0.13 -0.06 -0.28 SOI 0.29 0.23 0.48 -0.35 0.24 0.33 0.25 -0.45 -0.40 0.77 Npools -0.14 -0.02 0.22 -0.51 -0.19 -0.05 -0.37 -0.16 -0.01 FCI 0.80 0.27 -0.04 -0.26 -0.34 0.20 0.24 0.49 Ascd -0.83 -0.46 -0.36 0.44 0.50 -0.04 Iovh 0.29 0.38 -0.23 -0.32 0.27 R/B 0.63 -0.86 -0.83 -0.04 Pp/B -0.55 -0.65 0.16 B/T 0.94 0.06 B/P -0.07 Finn's cycling index showed a consistent negative relation with system recovery time (Fig. 2.14). Systems with a low percentage of energy recycled, simple recycling structures and short average path length (e.g. the upwelling systems off Venezuela and Monterey Bay) exhibit characteristically slow recovery time. On the other hand, systems where recycling is important, such as coastal lagoons and fjords (e.g. Schlei Fjord and Etang de Thau models), appeared more stable. To test the hypothesis that stability changes with recycling in a single system, we simulated the effect of an increase in the importance of detritus cycling in the Venezuela shelf model. The steps were as follows: first we gradually increased the diet components of detritus for heterotrophic benthos from 0 to 45% (thereby increasing the F C I from 0 to 2.8%); then we increased the diet component of detritus for zooplankton from 0 to 95% (the F C I thereby increased from 2.8 to 24.5%). Simulations were run for each level of recycling, leading to the results shown in F ig . 2.15. A negative relationship clearly emerges between the amount of detritus cycled and the recovery time of the model. Mode l stability does not appear very sensitive to increasing recycling up to a certain threshold from which small changes in recycling lead to substantial reductions in recovery time. 50 r 40 60 J3 30 o >. o 3 'B Q 20 10 Lng \u00E2\u0080\u00A2 Thau Sch NSea \u00E2\u0080\u00A2 Thai50 \u00E2\u0080\u00A2 Trg Bru. Cel Mrey Yet WMx \u00E2\u0080\u00A2 Cmp \u00E2\u0080\u00A2 Vnz GMx \u00E2\u0080\u00A2 Mpt 10 20 30 40 50 60 Recovery time (years) 70 80 Figure 2.14. Relation between Finn's index of detritus recycling and system recovery time (r=-0.76; p=0.001; d.f.=13). See Table 1 for models descriptions. Empty squares indicate the recovery time of the models of upwelling systems (Monterey and Venezuela). 45 30 e 2 0 o \u00E2\u0080\u00A2\u00E2\u0080\u00A25 10 .43 Q 20 \u00E2\u0080\u00A2 30 40 50 Recovery time (years) 60 Figure 2.15. Effect of increasing detritus cycling on the stability of the Venezuela shelf model (see text for procedure). The square corresponds to the system recovery time (t3 on Fig. l) for the baseline Finn cycling index value (2.7%). A view of the role of recycling different from the presented here was proposed by R. E . Ulanowicz and co-workers who argued that it is perturbed systems which tend to do more recycling (Ulanowicz, 1984; Ulanowicz and Wuff, 1991; Baird et al., 1991); from this view, they concluded that recycling is a measure of stress. Indeed, Ulanowicz and Wuf f (1991) suggested that in ecosystem development, the number and length of recycling routes is a more important attribute than the amount of recycling, since in some ecosystems recycling occur over short benthic cycles and do not involve loops through various trophic levels. Hence, the amount of energy and nutrients recycled may result in overevaluating the importance of recycling to the overall community. However, in the present study, recycling and path length were significantly correlated (r = 0.80) (Table 2.3), pointing to the substantial complexity and importance of the recycling routes for the marine ecosystem models analyzed. We believe that Ulanowicz et al. interpretation of the role of detritus recycling as a result (and a measure) of stress to be due to they having selected a priori ascendancy as a strong and positive correlate of maturity, then evaluating all other system properties relative to ascendancy. This approach is problematic because ascendancy (or at least that of its two component meant to express the 'information content' of an ecosystem) is in fact negatively correlated with maturity 46 (Christensen, 1995a). On the other hand, relative overheads, a measure of stability derived from Ulanowicz's theory does show a strong relationship with Finn's index (Christensen and Pauly, 1993b). Thus we conclude that recycling does indeed have an important role in the maintenance of ecosystem stability. Basically, the higher the recycling the more quickly w i l l the effects of perturbations be eliminated from the system (O 'Ne i l l , 1976; DeAngelis, 1980). Traditional views of ecosystem development consider that succession culminates in a stabilized ecosystem in which biomass and symbiotic functions between organisms are maintained per unit of available energy (Odum, 1969). Odum interpreted the strategy of succession as the one of increase control and or homeostasis with the physical environment in the sense of achieving maximum protection from its perturbations. The net result of community development would be symbiosis, nutrient conservation, stability, a decrease in entropy, and an increase in information. According to Odum, biotic control of grazing, population density and nutrient cycling provide the chief positive feedback mechanisms that contribute to stability in mature systems by preventing overshoots and destructive oscillations. Our results are in agreement with the above theory to the extent it shows that systems with higher capacity to recycle detritus are systems with a higher ability to recover from perturbations. Taking, therefore, stability and recycling as directly related attributes during ecosystem development the results o f this analysis provide support to Odum's theory and to previous comparative studies of marine food webs (Christensen and Pauly, 1993b; Christensen, 1995a). Very often ecosystems develop what Odum called pulse stability, where a more or less regular physical perturbation can maintain an ecosystem at some intermediate point in the development sequence toward maturity. Connell (1978) showed that in fact intermediate levels of disturbances are essential for the maintenance of certain ecosystem characteristics, such as species diversity, even in highly mature systems. Evidence gained in ecological studies, mainly of terrestrial ecosystems, has shown that change during ecosystem development is not continuous and gradual, but it is episodic with slow accumulation of biomass and nutrients punctuated by sudden releases and reorganization (Holling et al., 1995). According to Hol l ing, the structure of biological communities is therefore controlled through the balance of destabilizing and stabilizing forces. While destabilizing forces are important in maintaining diversity, resilience and opportunity, stabilizing forces, such as nutrient recycling, are important in maintaining productivity and biogeochemical cycles. 47 The role of destabilizing forces may be particularly important in pelagic marine ecosystems. In the sea, short-term physical variability is damped out by the very large heat capacity of the ocean. In turn, this large thermal capacity and the long period exchange rates between deep and near-surface waters lead to relatively large-amplitude changes at the long term scales (Steele, 1985). A s a result, less robust internal ecosystem processes are needed to handle the smaller amplitude variability at short periods. The possible absence of such mechanisms, combined with increase variance with period, can mean that pelagic marine populations or ecosystems have to continually adapt to physical variability in the short as well as the long term (Holling et al., 1995). The absence of well structured recycling routes, the low recycling and reduced stability of upwelling ecosystems (Figures 2.13 and 2.14) can be considered a result of a longer-term adaptation of biological community to the physical variability and transitory nature of these systems. In fact, taking fisheries yield as an indicator, upwelling ecosystems can be considered the most variable oceanic systems (Stergiou, 1998). Bakun (1996) considered variability itself a key asset for the massive small pelagic wasp-waist populations inhabiting upwelling systems, which must rely on pulsing its abundance to cope with the temporal and spatial patterns presented by their prey, while simultaneously presenting patterns to their predators that overcome growth of intolerable levels of predation. Model limitations Notes of caution relating to the interpretation of simulations were presented by Walters et al. (1997), but w i l l be summarized here. First, each Ecopath model may differ in terms of the time span which the data represent, the number of groups used, and the parameters which are estimated. It is difficult to determine what effect differences in data quality have on the simulations but it has been suggested that poor data may lead to models with a tendency for self-simplification (i.e., low persistence), through competition or predatory exclusion of some groups after disturbance by fishing (Walters et al., 1997). O f the 40 Ecopath models tested by Walters et al., those representing coastal and shelf marine ecosystems show a positive relationship between persistence and model quality (Pauly et al., in press; Figure 2.16; Table 2.4). In this analysis model persistence was ranked according to the maximum flow rate (mean v ; j) between predator-prey that led to persistence of all groups in the system after a disturbance. Mode l quality was 48 ranked according to the accuracy and quality of data used as input in the construction of an Ecopath model. Walters et al. (1997) suggested that positive relationships between persistence and quality have two interrelated consequences; first, it indicates the possibility that strong trophic interactions do occur in nature and lead to a selective process such that \"what persists to be studied as an 'equilibrium' in the field is a very peculiar or particular set of interaction parameter values\"; second, and consequently, one can only estimate these interaction values in models built from data accurately measured. These results differ from the studies of trophic interactions in aquatic ecosystems (McCarthy et al., 1995; Scheffer and de Boer, 1995) which show that the patchy distribution of organisms may lead to ratio dependence in predator-prey interactions, and consequently to relatively weak predator control (low Vy) in aquatic food webs. Ultimately, the relationship between model quality and persistence indicates that results from dynamic simulations with Ecosim w i l l depend in part on the quality of information used to construct the original Ecopath model. B at B 2 c \u00E2\u0080\u00A2 rs=0.452 _i i i i i i i 1 2 3 4 5 6 7 Quality ranking 6J3 e 2 a rs=0.557 1 2 3 4 5 6 7 Quality ranking Figure 2.16. Rank correlation between model persistence and quality, for Ecopath models of (A) coastal (r=0.452) and (B) shelf (r=0.557) ecosystems. Other limitations of the approach relate to its inherent simplifying assumptions, which ignores spatial processes, does not allow for numeric responses, as opposed to biomass predictions only, and does not take into account environmental variability, a factor frequently ignored in complex food web models (Hunter and Price, 1992). Some of these limitations, which may as well apply for many single species models, were recently addressed by incorporating into Ecosim a delay-differential model structure that is capable of tracking both numbers of individuals and biomasses of juvenile and adult stages of key species in the ecosystem (Walters et a l , in press a), and an interface that allow for the representation of spatial processes (Walters et al., in press b). Walters et al. (1997) also warned about the risk of using Ecosim to extrapolate to circumstances 49 far from the equilibrium for which Ecopath data are available. However, i f we only project a short time horizon and consider only short term dynamics, long term effects are easily missed. Despite the reservations expressed about long term simulation periods, they still can be useful in predicting the directions of biomass change, and serve as a warning to keep in mind potential delayed responses, a point that should be emphasized especially when considering complex and unstable ecosystems. Table 2.4. Trophic models used in the regression between quality rank and persistence rank. Shelf systems Quality rank Max vM persistence Brunei Darussalam 5 20 gulf of Thailand, 10-50 m 4 5 Kuala Terenganu, Malaysia 6 11 Northern Venezuela Shelf 2 40 North Sea 1 35 Northern Gulf of Mexico 7 8 Yucatan shelf 3 8 Coastal systems Campeche Bank, Mexico 8 6 Celestun lagoon, Mexico 5 40 Etang de Thau, France 3 4 Lingayen Gulf, Philippines 7 10 Maputo Bay, Mozambique 1 43 Schlei Fjord, Germany 2 46 Shallow areas, South China Sea 4 7 Southwestern Gulf of Mexico (coast) 6 8 Management implications The comparative analysis of the ecosystem impacts of harvesting small pelagic fish using multi-species trophic models showed that: i) small pelagic fish play a central role in upwelling ecosystems; changes in their abundance can have considerable consequences to species both at the top and the bottom of the food web; ii) observed patterns of species replacement at the pelagic niche can be explained by the effect o f fisheries and trophic interactions alone, being more evident when the system is dominated by predatory control, i.e., top-down; i i i) upwelling ecosystems have characteristics of unstable systems, lacking biological mechanisms (e.g. nutrient recycling) to damp the impact of external forces created by fishing and environmental factors; and iv) quantitative prediction of the ecosystem effects of fisheries using mass-balance ecosystem models is very sensitive to the type of dominant trophic control assumptions, and may be problematic for situations of poor data quality. However, these models are capable of output 50 robust qualitative responses which may guide the screening of fisheries policies under ecosystem principles. In this regard, results of the present work can have direct implications to the management of small pelagic fish with ecosystem goals. It indicates that as 'wasp-waist' species in upwelling ecosystems, small pelagic forage fish w i l l sustain much more conservative exploitation rates than what has been historically applied in the cases of stock collapse. Also , fisheries for small pelagics have a potential disrupting effect on ecosystem dynamics and may cause shifts in the species composition at the mid-trophic level. Ultimately, the characteristics of unstable systems w i l l mean that variability is a key asset for the resilience of upwelling ecosystems, and management should in turn be prepared to cope with the unexpected changes resulting from both fishing and environmental effects. 2.4. Summary This chapter analyzes the ecosystem impacts of harvesting small pelagic fish in upwelling systems comparing the results of simulations carried with trophic models. Results include the predicted responses of the system to fishing, in terms of changes in the biomass of small pelagic fish and other ecosystem components; the recovery time of marine ecosystems when disturbed at the mid-trophic level, wasp-waist populations; and the characteristics of stability of upwelling systems. These results are used to drawn conclusions on the effect o f fisheries and trophic cascades on species shift at the pelagic niche, and on the applications and limitations of multi-species trophic models in the design of ecosystem principles and precautionary measures for small pelagic fisheries. 51 Chapter 3. Fishing down food webs and the carrying capacity of marine ecosystems in southern Brazil2 3.1. Introduction Recent assessments of the worldwide status of marine capture fisheries revealed alarming signs of human dominance and impact on the oceans, such as overfishing of important stocks, by-catch and discard o f non-target species, and the fishing down o f marine food webs (Botsford et al., 1997; Pauly et al., 1998). In line with some of these global trends, marine capture fisheries of Brazi l are in a state of crisis caused by the scarcity of resources, over-capitalization of fisheries activities and the lack of sound fisheries management policies. The crisis in Brazilian fisheries occurs concurrent with major efforts to assess the potential production of fishery resources in the Exclusive Economic Zone, triggered by the country's ratification to the United Nations Convention on the Law of the Seas. Early assessments of the fisheries resources along the coast carried during the 1970s (Hempel, 1971; Neiva & Moura, 1977) pointed at a potential total catches over 1 mil l ion tons/year, when in fact, since the mid-1980s landings have stabilized around 700 thousand tons/year while many o f the traditional fish stocks have become either fully exploited or overexploited (Dias-Neto, 1991a; 1991b; I B A M A , 1994a; 1994b; 1994c; Reis et al., 1994; Cergole, 1995; Matsuura, 1995; Haimovici et al., 1997). The current situation raises concerns on the limits of marine ecosystems carrying capacity i.e., whether the level o f exploitation can be sustained without impairing the productivity and integrity of these ecosystems. Brazi l has an extensive coast line that extends from 5\u00C2\u00B0N to 34\u00C2\u00B0S, including regions of tropical and subtropical climate. Matsuura (1995) divided the Brazilian coast in five regions with distinct environmental characteristics and types of fishing activities (Fig. 3.1). In the north biological production is high as a result of the continental runoff from the Amazon river (Teixeira and Tundisi, 1967). The wide continental shelf and the rich benthic community favored the development o f trawling activities in this region, mostly for shrimps and large catfishes. The northeast and east regions present oligotrophic conditions due to the influence of tropical waters from the Brazi l Current. Rocky bottoms and a mostly narrow continental shelf induced the development of hook-and-line and longline fisheries for rockfishes, sharks and tunas. In the 52 southeast, primary production is mainly driven by seasonal upwelling o f nutrient-rich, cold subtropical waters pumped by alongshore winds and by cyclonic vortexes originated from the Brazi l Current (Bakun and Parrish, 1990; Matsuura, 1995). The southern part of the Brazilian coast is under the influence of the Subtropical Convergence between the southward and northward flowing Brazi l and Malvinas currents. The confluence o f water masses and the high volume of continental runoff provide physical and chemical conditions for high biological production on the shelf (Seeliger et al., 1997). Trawling is the main type of fishing activity in the southeastern and southern regions, although the presence of highly abundant pelagic stocks, mainly sardine, in the southeast has also lead to the development of an important purse seine fishery as early as 1950. Figure 3.1. Shelf regions of Brazil. The southeastern shelf is considered from Cabo de Sao Tome to Cabo de Santa Marta Grande, and the southern shelf from Cabo de Santa Marta Grande to Chui. The regions also differ in the type of fisheries production. While catches in the north, northeast and east regions are mainly artisanal (Diegues, 1995), in the southern regions it is the industrial fisheries that provide most of the fisheries production, accounting for approximately half of the 53 total Brazilian catches (IB A M A / I B G E , 1995). Historically, it was in the south and southeast that industrial fisheries were mostly developed through a series o f government incentives, and it is where fisheries are best documented. This chapter presents a comparative analysis of fisheries in the southeastern and southern regions of Brazi l which aims to assess the carrying capacity o f the marine shelf ecosystems for harvestable species. Carrying capacity has been defined as the maximum size of a population or activity that could be indefinitely sustained without degrading the ecosystem's future productivity or suitability for that use (Odum, 1997). In the oceans, carrying capacity is usually referred to as the upper limit of biomass of organisms that can be supported by a set of primary production and food web structure (Christensen and Pauly, 1998). Fisheries yield is directly related to the carrying capacity of marine ecosystems, since there is a maximum sustainable rate of fish production associated to the total fish biomass at the carrying capacity. Also , overfishing can directly affect the carrying capacity of marine ecosystems by altering the structure of food webs and changing their potential productivity. In this chapter I analyze the carrying capacity of marine ecosystems in three ways. Firstly by computing the total flux of energy from primary producers available to different trophic levels in the food web, and the total primary production required to sustain fisheries catches (Pauly and Christensen, 1995). The ratio of these two quantities provides a straightforward measure of the \"appropriated carrying capacity\" (sensu Rees, 1996) o f ecosystems, i.e. the amount of the available energy in a ecosystem already appropriated by fisheries catches. Secondly, I do a diagnosis of fisheries for the 'fishing down the food web' phenomenon using trophic level estimates and national and regional catch statistics. Finally, I analyze how fisheries are likely to alter the species composition in ecosystems structured by trophic relationships. I explore this effect by simulating a 'fishing-down-the-food-web' scenario in the southern shelf region, where traditional demersal fish stocks are overexploited and the prospects for increasing yield relies on exploiting abundant small pelagic forage fish. 3.2. Methods The method used here to quantify the appropriated carrying capacity follows the approach developed by Pauly and Christensen (1995) for the analysis of primary production required to sustain world fisheries. Primary production required by fisheries (PPR) is estimated based on the 54 trophic level of the species caught, the energy transfer efficiency between trophic levels, and on the primary productivity of the two shelf regions (Table 3.1). Primary production estimates for the southeast and south were obtained from Brandini (1990) and Odebrecht and Garcia (1997), respectively. Species trophic levels (Table 3.2) were computed according to Odum and Heald (1975) using available information on diet composition, and from trophic models. In this analysis primary producers have trophic level 1, and each higher order consumers has trophic level 1 + the weighted average trophic level of its preys. Rocha et al. (1998) constructed a simplified trophic model of the Ubatuba region in the southeastern shelf that is here used to estimate the mean trophic transfer efficiency for the region. Transfer efficiency for the southern shelf is calculated from the trophic model described below. P P R estimates are based on a conversion factor of 0.06 g Carbon = 1 g wet weight of catches (Walsh, 1981) and on the mean transfer efficiency per trophic level, that is PPR = catches a^1'^ where a = TE\" 1 , and T E is the mean trophic efficiency measured as the proportion of the energy transferred between consecutive trophic levels (TL). P P R is commonly expressed as a percentage of the total primary production (%PP). Table 3.1.Area, primary productivity and total primary production of southern and southeastern shelf regions of Brazil. Shelf areas were measured to the 200 m depth line using planimetry. Region Area PP Total PP m 2 \u00C2\u00AB10 1 0 gOm^year 1 gOyearMO 1 2 Southeast 17.14 100 17.14 South 11.40 160 18.25 Catch statistics obtained from Haimovici et al.(1997), Haimovici(1998) and from the Instituto de Pesca de Santos, Sao Paulo, were used in the analysis of P P R for the two shelf regions. F A O catch statistics of reported Brazilian catches were used to compute the mean trophic level of landings in Brazi l from 1950 to 1994. 55 Table 3.2. Trophic level of the main species landed in Brazil. Trophic level estimates are from model in figure 3.2, from diet composition studies (references in table footnote), and/or from other published trophic models (Christensen and Pauly, 1993). Group Species Trophic Level Shrimps Farfantepenaeus spp; Xiphopenaeus kroyeri 2.3 2.3 Lobsters Panulirus argus; 2.6 Panulirus spp 2.6 Small and mid-size pelagics Sardinella brasiliensis\" 2.8 Engraulididae 3.0 Scomber japonicus Scomberomorus spp 3.1 3.3 Common squids Loligo spp 3.4 Miscellaneous marine fishesb Osteichthyes 3.5 Micropogonias furnierf'd Umbrina canosai' 3.4-3.5 3.2 Cynoscion sppc^ Macrodon ancylodon? Trichiurus lepturush 3.9-4.0 4.3. 4.3 Batistes capriscus' Pomatomus saltatrix' 3.4 4.2 Pinguipes spp' Ariidae* 3.8 3.8 Mugilidae' 3.8 Sharks, rays, skates Elasmobranchs''' 3.4 -3.8 Groupers Epinephelus spp Mycteroperca spp 3.7 3.7 Snappers Lutjanidae Ocyurus chrysurus 3.8 3.8 Common dolphinfish Coryphaena hippurus 3.9 Skipjack tuna Katsuwonus pelamis\" 3.9 Large pelagic fish\"'\"''' Thunnus spp Xiphias gladius Other Scombroidei 3.9 3.9 3.9 a. Goitein, 1983,Gasalla & Oliveira, 1997; b. Vazzoler et. al. (in press); c. Gasalla, 1995; d. Vazzoler, 1975; e. Haimovici et al., 1989; f. Vieira, 1990; g. Juras and Yamaguti, 1985; h. Martins and Haimovici, 1997; i . Froese and Pauly, 1998; j . Haimovici and Krug, 1992; k. Araujo, 1984; 1. Soares et al., 1992; m. Vilela, 1990; n. Zavalla-Camin, 1982; o. Vaske, 1992; p. Vyalov & Ovchinnikov, 1980. The ecosystem effects of 'fishing down the food web' for anchovy in the Southern shelf was explored with a simplified mass-balance model ( E C O P A T H , Christensen and Pauly, 1992) of the trophic interactions in the pelagic ecosystem (Tables 3.3 and 3.4; F ig . 3.2). The model was constructed based on the pelagic species association described by Mel lo et al. (1992) for the winter and spring, and depicts anchovy as the dominant planktivorous fish species, being responsible for most o f the transfer o f energy from lower trophic levels to higher order consumers (Figure 3.2). The system is defined by the coordinates 32\u00C2\u00B0S - 43\u00C2\u00B030 'S and 51\u00C2\u00B0W -54\u00C2\u00B0W, with a total area of 28,661 K m 2 . 56 Table 3.3. Parameters of the trophic model of the pelagic ecosystem off southern Brazil. Underlined values, trophic levels (TL), and omnivory index (01) were estimated by the model. A n omnivory index equals zero indicates the predator feeds on a single trophic level. Cutlassfish, Trichiurus lepturus; Hake, Merluccius hubbsi, Weakfish, Cynoscion guatucupa; Mackerel, Scomber japonicus; Jack mackerel, Trachurus lathami; Anchovy, Engraulis anchoita. Species/Group T L OI B tons*Km2 P/B year'1 Q/B year' EE Yield tons\u00C2\u00BBKm'2 \u00E2\u0080\u00A2 year' Cutlassfish 4.25 0.259 0.240a 0.410c 2.050d 0.852 0.015b Hake 4.11 0.126 0.085 0.355\u00C2\u00B0 1.750d 0.950 0.004b Sharks 3.94 1.172 0.342 0.400e 4.000= 0.950 0.130b Other pelagics 4.02 1.080 0.748 0.570c 5.300e 0.950 0.405f Weakfish 3.59 0.302 2.000s 0.480c 2.340d 0.935 0.306b Mackerel 3.11 0.000 1.329 0.340e 2.710\" 0.950 0.033f Jack Mackerel 3.11 0.000 0.300h 0.350h 3.000d 0.958 0.054f Anchovy 3.00 0.111 13.710' 1.290\" 5.155d 0.240 \u00E2\u0080\u0094 Squids 3.32 0.185 0.200k 1.500c 3.230e 0.894 \u00E2\u0080\u0094 Marine shrimps 2.00 0.000 0.298 3.930' 19.130e 0.950 0.040f Zooplankton 2.11 0.111 8.000m 64.920m 324.600 0.619 \u00E2\u0080\u0094 Phytoplankton 1.00 0.000 16.700\" 100.000\" \u00E2\u0080\u0094 0.965 \u00E2\u0080\u0094 Detritus 1.00 0.179 150.000\u00C2\u00B0 \u00E2\u0080\u0094 \u00E2\u0080\u0094 1.148 \u00E2\u0080\u0094 a. Martins and Haimovici (1997); b. Haimovici (1998); c. based on Peterson and Wroblewski (1984); d. based on Palomares and Pauly (1989); e. based on other trophic models (Christensen and Pauly, 1993); f. Haimovici et al. (1997); g. I B A M A (1993a); h. Saccardo (1980); i . Lima and Castello (1995); j . Freire (unpubl.); k. Haimovici (1997); 1. DTncao (1991); m. Resgalla Jr. (unpubl.); n. Odebrecht and Garcia (1997); o.according to Pauly et al. (1993). Table 3.4. Diet matrix of the model of the pelagic ecosystem off southern Braz i l . Values represent the proportion of the diet of a predator (column) made of a given prey (row). Some of the groups (mainly sharks, weakfish, and other pelagics) have several feeding habitats such as the outer shelf and benfhic habitats. For these groups an Import was included as a \"prey\" in the diet composition. Prey \ Predator 1 2 3 4 5 6 7 8 9 10 11 1. Cutlassfish3 0.140 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 2. Hake b 0.020 0.100 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 3. Sharksc _ _ _ _ _ _ _ _ _ _ _ 4. Other pelagicsd \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 5. Weakfish\"e 0.120 \u00E2\u0080\u0094 0.050 0.050 0.050 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 0.050 \u00E2\u0080\u0094 \u00E2\u0080\u0094 6. Mackerel d \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 0.100 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 7. Jack Mackerel\" 0.030 0.030 0.020 _ _ _ _ _ _ _ _ 8. Anchovy f 0.570 0.800 0.330 0.300 0.450 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 0.150 \u00E2\u0080\u0094 \u00E2\u0080\u0094 9. Squids8 0.050 0.030 0.030 0.050 _ _ _ _ _ _ _ 10. Mar. shrimps'1 0.010 \u00E2\u0080\u0094 0.070 0.200 0.010 _ _ _ _ _ 11. Zooplankton 0.010 \u00E2\u0080\u0094 \u00E2\u0080\u0094 0.050 0.200 0.990 1.000 0.900 0.800 \u00E2\u0080\u0094 0.100 12. Phytoplankton _ _ _ _ _ _ _ 0.050 \u00E2\u0080\u0094 0.200 0.550 13. Detritus \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 0.050 \u00E2\u0080\u0094 0.800 0.350 Import 0.050 0.040 0.500 0.450 0.100 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 a. Martins (1992); b. Haimovici et al.(1993); c. Castello et al. (1997); d. Castello (1997); e. Vieira (1990); f. Schwingel and Castello (1995); g. Haimovici (1997); h. based on other trophic models (Christensen and Pauly, 1993). 57 pelagics T Mackerel X Hake 0 Cut fish 3 J Mack Weakfish M Anchovy ^shrimp j Sharks Squids Zoop Phyto \"* Flow \u00E2\u0080\u0094 Connector * Harvest Import Figure 3.2. Flowchart of trophic relationships in the pelagic association off southern Brazil. It describes the flows between groups (boxes), the biomass of each group (area of boxes proportional to the log of biomass), and the respective trophic levels. Only the consumption flows are shown. For full parameter descriptions see tables 3 and 4. 'Fishing down the food web' was simulated by increasing fishing mortality (F) for anchovy from 0 to 1 year\"1, while maintaining F constant for other exploited groups. A s some of the high trophic level species (e.g. sharks, weakfish and pelagics) are already fully exploited or overexploited (Haimovici et al., 1997), I considered fishing for anchovy as the most likely continuation of the 'fishing down the food web' effect in this system. Ecosim (Walters et a l , 1997) was used to calculate the predicted changes in equilibrium biomasses of species/group and the total catch from the system over the range of F values for anchovy. The model provides biomass predictions of each group in the system as affected directly by fishing and predation, changes in food availability, and indirectly by fishing or predation on other groups in the system (Walters et al., 1997). Simulations were run under two contrasting trophic control hypothesis, bottom-up and top-down, as described in Chapter 2. 58 3.3. Results PPR and Trophic Levels P P R estimates by shelf region and species landed are shown in tables 3.5, 3.6 and 3.7. Fisheries in southern Braz i l already use a large proportion of the productive capacity of the shelf ecosystems. In the south, primary production required to sustain catches has changed little from the 1970s to the 1990s, being in the order of 25% of the total primary production. Little change is also observed in the mean trophic level of fisheries in the south which have been targeting mostly high trophic level species (Table 3.5). A n increase in catches of tunas and sharks was observed in the southern shelf in the early 1990s accompanying the depletion of important demersal fish stocks, such as Umbrina canosai, Macrodon ancylodon and catfish species, Netuma spp (Table 3.5). However, this alternation of species in the catches did not result in major changes in the P P R and in the mean trophic level of landings between the two periods. Landings in the Southeast are on the other hand dominated by low trophic level species, sardine and marine shrimps being the most important stocks in terms of catch volume (Table 3.6). With the collapse of the Brazilian sardine during the late 1980s and early 1990s, and the increase in tuna and sharks catches, there was an increase in the mean trophic level of fisheries from 2.81 to 2.93. Despite the fact that catches were considerably lower in the latter period, the change in relative importance of the species landed resulted in an increase in P P R from 25.8 to 33.4 % of the total shelf primary production. Although substantially higher fishing yields are obtained in the southeast than in the south, the footprint of fisheries is relatively the same between the two regions as a result of differences in the mean trophic level of landings and the mean trophic efficiency (Table 3.7). 59 Table 3.5. Trophic level, mean catch and PPR estimates for the southern shelf. Species 1975 -1979 1990 -1994 Trophic Catches PPR Catches PPR Level tons gC \u00E2\u0080\u00A2 101 0 tons g C \u00C2\u00BB 1 0 1 0 Micropogonias furnieri 3.5 14,308 42.0 14,709 43.2 Umbrina canosai 3.2 16,900 25.7 9,629 14.7 Cynoscion guatacupa 3.9 6,439 65.2 8,785 88.9 Macrodon ancylodon 4.3 7,941 180.4 3,966 90.1 Miscellaneous teleosteans 3 3.5 4,052 15.9 4,143 13.9 Netuma spp 3.8 3,983 30.5 615 4.7 Trichiurus lepturus Demersal sharks\" 4.3 75 1.8 441 10.8 3.8 2,584 17.0 5,931 39.1 Rhinobatus horkelli 3.4 1,010 2.9 460 1.3 Rays and skates 3.6 116 0.5 746 3.1 Marine shrimps 2.3 \u00E2\u0080\u0094 \u00E2\u0080\u0094 1,148 0.2 Small and mid-size pelagics 0 3.2 ' 1,549 1.4 3,848 5.1 Pomatomus saltatrix 4.2 4,290 88.1 3,521 7.2 Mugil spp 3.8 2,081 14.4 1,524 10.6 Katsuwonus pelamis 3.9 \u00E2\u0080\u0094 \u00E2\u0080\u0094 8,088 70.4 Pelagic sharks 3.7 182 0.9 547 2.9 a. Pogonias cromis; Merluccius hubbsi; Paralichthys spp., Pagrus pagrus; Prionotus punctatus; Urophycis brasiliensis and Poliprion americanus. b. mostly Galeorhinus galeus; Mustelus schmitti and Squatina spp. c. Brevoortia pectinata; Scomber japonicus and Trachurus lathami. Table 3.6: Trophic level, mean catch and PPR estimates for the southeastern shelf. Species 1977 -1980 1990 -1995 Trophic Catches PPR Catches PPR Level (tons) g C \u00C2\u00AB 1 0 1 0 (tons) g C \u00C2\u00BB 1 0 1 0 Micropogonias furnieri 3.4 7,126 62.6 4,541 39.9 Macrodon ancylodon 3.7 2,053 35.9 1,870 32.7 Cynoscion jamaiscencis 4.0 1,921 90.4 2,245 105.6 Batistes capriscus 3.4 \u00E2\u0080\u0094 \u00E2\u0080\u0094 2,144 19.4 Sardinella brasiliensis 2.8 146,520 189.4 54,414 70.3 Rays and skates 3.4 \u00E2\u0080\u0094 504 4.6 Marine shrimps 2.3 17,371 5.0 13,997 4.0 Sharks 3.8 517 12.6 2,144 52.2 Katswonus pelamis 3.9 1,380 46.7 7,197 243.6 Table 3.7. Summary statistics of the mean catch, mean trophic level (TL), mean transfer efficiency (TE), the primary production required by fisheries catches (PPR), and the percentage of the total primary production appropriated by fisheries (% PP) in the southern and southeastern shelves. Region Catch T L T E P P R % P P (tons'year 1) % gCyear''10\"> Southern 1975-1979 65,510 3.63 8 487 26.7 1990-1994 68,101 3.64 471 25.8 Southeastern 1977-1980 176,888 2.81 5 443 25.8 1990-1995 89,055 2.93 572 33.4 The increasing trend in the mean trophic level o f catches observed in the southeastern shelf is also observed in the F A O fisheries statistics for Brazi l (Fig. 3.3). Fisheries in Brazi l had a 60 relatively constant mean trophic level of the species landed from 1950 to the early 1980s, but show a recent increase in mean trophic level caused by the combined effect of the collapse of small and mid-size pelagics (mostly sardine), and the increasing landings of large pelagic fishes (tunas and sharks) with the development of offshore fisheries. Another factor that contributes to the increase in the mean trophic level of landings is the steady increase in teleost catches. Teleosts are characterized by high trophic levels (from 3.2 to 4.3, Table 3.2), and represent the most important group in total Brazilian fisheries landings. However, the proper evaluation of the contribution of the group to the average trophic level w i l l require higher resolution in the species catch composition statistics, which currently aggregate a large part of teleosts species into Osteichthys (Table 3.2). 61 Figure 3.3. Mean trophic level (A), and species composition (B and C)of total Brazilian landings. Source FAO. 62 Fishing Down the Food Web A strategy very often proposed to increase catches in exploited ecosystem is to fish down the food web for highly abundant, small pelagic planktivorous fishes after larger species are depleted. Simulation results of a fishing down food web scenario for anchovy in the southern shelf are shown in figures 3.4 and 3.5. Figures 3.4 represents the predicted equilibrium yield and biomass of anchovy, and the percentage change in biomass of all other groups in the system under \"top-down\" and \"bottom-up\" control of trophic interactions. The model predicts considerably smaller yields and optimal fishing mortality rates for anchovy under top-down control (F m s y top-down - 0 . 1 year _ 1 ; F m s y bottom-up ~ 0.3 year _ 1) . Both hypotheses generate a similar pattern of decrease in biomass of higher trophic level species, increase in biomass of mid-trophic level groups and increase in zooplankton biomass with increasing F for anchovy. Predictions of biomass changes at the mid-trophic level are more pronounced under top-down control, where the release in predation mortality due to the depressed biomass of top predators leads to a sharp increase of jack mackerel abundance. 63 \"Top-down\" \"Bottom-up\" Cutlassf ish Hake 1 Sharks O. P e l a g i c s \u00E2\u0080\u0094 \u00E2\u0080\u0094 Weakfish -10 t -10 L F- anchovy (year-1) F- anchovy (year\"1) Zooplankton \u00E2\u0080\u0094 Riytoplankton Detritus Figure 3.4. Equilibrium simulation of increasing fishing mortality for anchovy. Upper panel represents the predicted equilibrium yield and absolute biomass of anchovy. Lower panels show the predicted relative change in biomass of all other groups in the system. 64 n OS o o Os SO r o SO r o in 00 SO in SO o s o so r o CO CN oo CN in O H in CN r- in OS r~ o r o Os oo so Os r - o o O \u00E2\u0080\u0094 CN in OS so CN oo o CN r - o r- CN in o CN o r- oo r o o t~- ^ m O o r- CN r o CN OS Os in in oo os ^ \u00C2\u00AB oo O in CN oo in oo oo \u00E2\u0080\u00A2~~' T* in oo in SO ' r o ox in in \u00E2\u0080\u00A2n in \u00E2\u0080\u009E < <=> O \u00E2\u0080\u0094 CN CN r o CO pa To simplify the calculations it was assumed that the fishery takes place in a relatively short period o f time during which the population is subjected only to fishing mortality. The basic reconstruction equation for any cohort in a discrete fishery is (Hilborn and Walters, 1992): Na+l,t+\ Na,t = \" + Ca>t where N a , accounts for the number of fish of age a alive at time t, s is the annual survival rate (= e~M), M is the natural mortality rate, and C a t is the observed catch o f age a fish at time t. It was also assumed that no fish are left alive after age 3.5. Fishing mortality is expressed in terms of t age-time specific exploitation rates uat =\u00E2\u0080\u0094\u00E2\u0080\u0094,which represent the fraction of individuals \u00E2\u0084\u00A2 a,t captured during the fishery. To initialize the backward recursion calculations, terminal exploitation rates were assigned for the last age in all years according to the procedure in Box 4.2. Box 4.2. Virtual Population Analysis procedure for reconstruct sardine biomass at age (Walters and Punt, 1994). 1. set u a T (the exploitation rate for the fully vulnerable ages in the last year T of data) equal uT; 2. calculate the fully vulnerable harvest rate (ut) in each year (1 to T- l ) as the average exploitation rate of the ages 2 and 3 (considered fully vulnerable to the fishery); 3. set N 3 5 t = C 3 j./u, for all t o o < o 1 o I S 2 ^ \u00E2\u0080\u0094i CN co i n so . O OO OS \u00E2\u0080\u0094 I I S on S3 ^ \u00E2\u0080\u00A25 3 55 J3 00 IS 43 in A -A \u00C2\u00AB 10 I S a a> .a 7? CJ 2 \u00C2\u00A3 B C N c o - r j - i n s o r \u00E2\u0080\u0094 0 0 0 s CN CN Three peculiar features of the ecosystem are represented in the model. First the partition of the system between pelagic and benthic food chains at lower trophic levels (1 to 3) representing energy pathways originated from phytoplankton and from detritus. Benthic and pelagic food chains are mainly linked at the top of the food web by predators, such as weakfish, that optimize foraging benefits by feeding from both systems, and by the juvenile stages of demersal species that are active pelagic planktonic feeders (e.g. weakfish and triggerfish). Second, the model depicts the partition of the mid-trophic level pelagic niche among Sardine, Anchovy, Other forage fish, and juveniles of top predator groups. Finally, the model represents the relatively high abundance of Salps noticed by Pires-Vanin et al. (1993) as major primary consumers that are apparently not utilized by higher order consumers in the system. Table 4.5 compares several ecosystem attributes obtained with network analysis in Ecopath. Comparisons are made with the original model of Rocha et al. (1998) and with 9 other trophic models of upwelling ecosystems listed on Chapter 2. Inter-ecosystem comparisons are particularly relevant for global attributes, considered common to all upwelling systems (Jarre-Teichman and Christensen, 1998). These are ecosystem characteristics relative to ascendency, recycling, path length, trophic transfer efficiency, and also fishery related attributes, such as the mean trophic level of catches. For instance, the Southeastern shelf ecosystem share common characteristics with other upwelling systems, such as the low recycling of nutrients, low relative ascendancy, low trophic transfer efficiency, and fisheries that target mostly low trophic level species. Local characteristics, such as the primary production and total catches, are more difficult to compare since almost all upwelling systems for which models were available represent very productive eastern boundary current ecosystems. Differences between the modified model of the Southeastern shelf ecosystem and the original model of Rocha et al. are mainly related to the size, activity and utilization of the system. The modified annual model reflects a smaller system (lower primary production and system throughput) with higher catches of lower trophic level groups. Also , recycling was relatively higher in Rocha's model and may possibly result from the authors including bacteria as one explicit group in the model. The models otherwise have similar attributes within the range of values expected for upwelling systems. 93 Table 4.5. Ecosystem attributes used in the comparison of trophic models of upwelling ecosystems. Upwelling models refer to the trophic models described in Chapter 2. Attributes Rocha et al. 1998 Modified model Upwelling models Global characteristics Average Range Ascendency 37.80 27.50 32.97 16.20-40.60 System Omnivory 0.10 0.14 0.16 0.11-0.32 Connectance Index 0.30 0.24 0.26 0.23-0.32 Finn's Cycling Index 14.51 7.33 8.11 1.58-13.60 Mean Path Length 3.54 3.16 2.78 2.45-3.01 Mean Transfer Efficiency 4.45 5.10 5.00 3.60-7.40 Fisheries Mean Trophic level 3.01 2.86 2.85 2.22-3.28 Local characteristics Total system throughput 6,373 5,314 24,256 8,058-59,677 Net Primary Production 1,802 1,680 7,836 3,290-22,059 Total Catches 0.10 0.60 19.99 0.95-91.67 Ecopath calculates the transfer efficiencies between the successive discrete trophic levels in an ecosystem as the ratio between the sum of the exports plus the flow that is transferred from one trophic level to the next, and the throughput on the trophic level. The transfer efficiencies between trophic levels II and V (herbivores and third level carnivores) are shown in table 4.6. A pattern emerges where lower trophic levels, specially herbivores/detritivores, have a higher transfer efficiency than groups at higher trophic levels. This pattern was also observed in other models of marine food webs (Christensen and Pauly, 1993) and it is consistent with higher respiration losses expected for long lived and more active groups at the top of the food chain. The mean trophic transfer efficiency of all flows in the system is 5.1 %, which is below the mean transfer efficiency of 10% estimated by Pauly and Christensen (1995) across different types of marine ecosystems, but in line with their estimates for upwelling systems. Upwelling ecosystems are considered relatively inefficient in transferring energy up the food web. Their energetic inefficiency seems to be related to the characteristics of primary productivity and food web organization. The high and variable new primary production characteristic of upwelling systems yields a higher export of carbon compared to more stable and less productive systems, where most of the primary production is regenerated (Berger et al., 1989). The variability of upwelling systems involves a continuous change in conditions that result in simpler food webs that lack fully co-adapted autotrophs, heterotrophs, and decomposers. Thus sporadic and seasonal nutrient injection stimulates rapid growth of phytoplankton which w i l l bloom not being grazed sufficiently to prevent exponential increase; a large part of this unutilized primary production w i l l end up being exported from the euphotic zone towards the seafloor. Part of the detritus exported from the pelagic system is recycled back into the food web by the activity of 94 benthos detritivores and omnivores. It is estimated that in the Southeastern shelf about 34% of the flows originates from detritus, although recycling (including detritus recycling) represents only 7.33% of the total system throughput. Table 4.6. Transfer efficiencies between trophic levels calculated using the trophic aggregation routine in E C O P A T H . Trophic level Source II III IV V Producers 11.5 3.9 2.4 4.6 Detritus 6.7 2.4 3.5 4.1 All flows 9.9 3.6 2.6 4.5 Estimates of the total primary production required to sustain catches (PPR) during the 1970s and 1990s are shown in table 4.7. P P R estimates for the 1970s were obtained by changing the catches in the trophic model with the values observed from 1977 to 1980 (as in Chapter 3). Notably low P P R values (-4%) are estimated for fisheries in the Southeastern shelf in both time periods, compared to values between 25 and 33% estimated in Chapter 3 using a different method. Ecopath calculates the P P R of catches of a given group by first identifying all the paths in the trophic network that lead primary production (or detritus) to the harvested group. For each path, the catches are then raised to the primary production equivalents using the product of the catch, the gross growth efficiency of each path element (=production/consumption), and the proportion the next element of the path contributes to the diet of a given path element (Christensen and Pauly, 1996). A t each consecutive step down the path, the calculated flow is divided by the ecotrophic efficiency of the path element (so that the resultant P P R reflects the fact that not all biological production is utilized at consecutive trophic levels). Ecopath thus offers a more rigorous method to calculate PPRs, compared to the approach used in Chapter 3 where the mean trophic efficiency between trophic levels is used to raise catch values to the primary production equivalents. Assuming an average trophic efficiency of 5% causes, for instance, an overestimation of the P P R of low trophic level species (that make up the bulk of the catches) for which transfer efficiencies are substantially higher (Table 4.6). The relatively low footprint of fisheries in the region is on the other hand expected from the low trophic level of catches, and may be conservative considering that discards were not included in the calculations, pelagic sharks were not included in the model, and that part of the catches may remain unreported in official fisheries statistics (Gasalla and Tomas, 1998). 95 Changes in the catch composition between the two periods are related to a modest increase in catches of triggerfish and rays/skates, decrease in catches of croaker, sardine and marine shrimps, and an increase in landings of bonito, Katswonus pelamis, which resulted in an increase in the mean trophic level of fisheries from 2.81 to 2.86. Table 4.7. Trophic level, mean catch and PPR estimates for the main species landed in southeastern Brazil. Species 1977 -1980 1990 -1995 Trophic Catches PPR Catches PPR Level tons tons\u00C2\u00BBKm\"2\u00C2\u00AByear\" tons tons'Km^year\"1 Bonito 3.73 1,380 4.35 7,197 23.09 White croaker 3.32 7,126 24.43 4,541 15.95 King weakfish 3.74 2,053 18.10 1,870 18.42 \u00E2\u0080\u00A2 Weakfish 3.80 1,921 13.26 2,245 16.68 Triggerfish 3.37 \u00E2\u0080\u0094 \u00E2\u0080\u0094 2,144 3.70 Sardine 2.84 146,520 20.79 54,414 7.72 Rays and skates 3.41 \u00E2\u0080\u0094 \u00E2\u0080\u0094 504 2.59 Marine shrimps 2.00 17,371 0.47 13,997 0.38 Total Catch and PPR 174,991 61.70 86,912 68.38 %PP 3.67 4.07 Fisheries Mean T L 2.81 2.86 In terms of ecosystem change between the two time periods a question remains on whether the energy previously available at lower trophic levels in the form of sardine biomass was loss due to stock overfishing, was reallocated to other species in the ecosystem (e.g. anchovy, triggerfish), or simply reflect a decrease in the productivity of the system. A n d also whether this new system configuration is in a stability domain that would impede the recovery of the sardine stock. Simply put, can the current system configuration support a larger biomass of sardine? The understanding of how ecosystems are structured and how they change in response to human impact has at its core the understanding of the dynamics of ecosystems succession and resilience. Holl ing (1986, 1992, 1995) suggested four primary phases in an ecosystem succession cycle which synthesize common features observed in both terrestrial and aquatic systems (Fig. 4.5). The phases are: exploitation, in which rapid colonization of recently disturbed areas by opportunist species (r-strategy) occur and lead to a growth in the size of the system (stored capital); conservation, in which the system slowly accumulates and stores energy and material, and develop a more complex structure (increase in connectedness) until a climax is attained. Characteristics o f this phase is the highly efficient processes o f energy utilization and conservation, and the presence of K-strategy species; release, in which the tightly bound 96 accumulation of biomass and nutrients becomes increasingly fragile (overconnected) until it is suddenly released by physical or biological disturbances and unexpected events such as forest fires, storms, insect pests, intense pulses of grazing, etc.; and reorganization, in which processes of nutrient mobilization minimize losses and reorganize nutrients to become available for the next phase of exploitation. The model depicts two aspects of stability. First the ability and speed with which systems recover from perturbations which, according to the inter-ecosystem comparison carried in Chapter 2, is mostly influenced by mechanisms o f nutrient recycling in the food web. The second aspect, that of the resilience of ecosystems, is represented by an arrow in the top left corner of the model which suggests that changes into another ecosystem structure following perturbation is most likely to occur during the release and reorganization phases. A Q. co O X} CD O W 2. Conservation 1. Exploitation 3. Release Weak Connectedness Strong Figure 4.5. Holling's four phase model of ecosystem dynamics. During the cycle of exploitation, conservation, release and reorganization, biological time flows unevenly. It is slow from the exploitation to the conservation phase, then very rapid to the release (when the overconnected system triggers sudden changes by agents such as fire, disease, etc.), rapidly to reorganization and back to the exploitation phase. Resilience and recovery are determined by the fast release and reorganization sequence, whereas stability and productivity are determined by the slow exploitation and conservation sequence. The arrow in the top left corner of the model suggests the phase where change into another ecosystem structure following perturbation is most likely to occur (source Holling, 1995). The four phases model considers that the dynamics of ecosystems is organized across scales in time and space around the operation of a small number of nested cycles of exploitation, conservation, release and reorganization, each driven by a few dominant variables (plants, animals and abiotic processes) (Holling, 1992). Examples of regional resource management (Gunderson et al. 1995) also suggest that institutions and societies achieve periodic advances in understanding and learning through similar cycles of growth, production (greatest efficiency), release (crisis) and renewal that shape the spatial and temporal dynamics of ecosystems. 97 Holling's four phases model thus provides a framework that is useful for understanding both the dynamics of ecosystems (its structure and change) and the functioning of institutions bound to resilience. The study of the structure and dynamics of the Southeastern Brazil ian shelf ecosystem (Pires-Vanin, 1993) indicates that this system is conditioned by cyclical physical events that control seasonal changes in the trophic structure and in the patterns of energy flow in the biological communities. The seasonal upwelling cycle is in fact nested in a hierarchy of cycles of different speeds, or time scales, as suggested in table 4.8-. A t faster speeds there are bacteria and phytoplankton organisms, which have characteristically higher turnover rates. Individual phytoplankton live for a few days and succession of populations in a patch or bloom can change significantly in the same time scale. Physical processes such as pulses o f favorable wind stress, storms, small scale vortexes of low residence time, and transient upwelling events caused by internal waves and fronts can all influence biological succession at this scale. Aidar et al. (1993) showed that phytoplankton biomass in the Southeastern bight is usually limited by the lack of nutrients, specially nitrogen, and it is mostly dominated by species of the nano-plankton which are better adapted, with a higher surface-volume ratio, to explore the general oligotrophic conditions. Once the system is disturbed by the intrusion of nutrient rich South Atlantic Central Water ( S A C W ) , species of the nano-plankton are substituted by other larger phytoplankton species better adapted to the new conditions. Physical processes operate at faster speeds to control small-scale water turbulence and influence the rate of encounter between planktonic predators and their prey (Mann and Lazier, 1991). Zooplankton and benthic invertebrates occupy an intermediate level in the time-scale hierarchy between phytoplankton and fishes. The structure of benthic communities in the southeastern shelf is mainly influenced by the seasonal cycles of upwelling which favor distinct types of organisms during summer and winter (Pires-Vanin, 1993). 98 Table 4.8. Time-scale hierarchy of physical and biological variables controlling the dynamics of the Southeastern shelf ecosystem. Speed of change Variables Time scale Equivalent physical force Fast Bacteria Phytoplankton Hours - Days Turbulent mixing, tides, breeze, storms, vortexes Intermediate Zooplankton Benthic organisms Days - Months Seasonal cycles of upwelling Slow Mid-trophic level fish Higher trophic level fish Marine mammals Management institutions Programs of economic incentives Scientific understanding Years - Decades Cycles in upwelling intensity and E l Nino (ENSO) Climatic-oceanographic regimes Variability in ocean circulation Very slow Large Marine Ecosystems Human myths of nature Decades -Centuries Global warming Most fish species fit into the category of slower variables, with turnover rates in the order of years and cycles in production in the order of years and decades. Cycles in marine fish populations bear close correspondence to the long term climatic-oceanographic regimes of the oceans. Characteristic of the sea is the \"red noise\" type o f variability spectrum, in which the variance of physical processes increases significantly with the time scale from hours to decades (Steele, 1985). This inherent dynamics of the oceans is thought to influence marine fish populations in two ways: first, the damped short and medium time scale variability create conditions that favor reproductive strategies such as high fecundity and absence of parental care, that rely on a relatively predictable combination of enrichment, concentration and dispersal mechanisms at these scales. Second, the absence of mechanisms to cope with the short-term variability may make marine fish populations susceptible and adaptable to variability at longer time scales. Examples of fish populations that show decadal oscillations in abundance are found in almost every marine system. Bakun (1996), for instance, compiled examples of synchronous oscillations in several fish populations geographically isolated from each other, but apparently driven by a common environmental property. These low-frequency population cycles varied among species, but were generally characterized by a period o f rapid population growth in the decade from the mid-1970s to the mid-1980s followed by stock declines after the mid-1980s (Fig. 4.6). According to Bakun the Brazilian sardine followed the pattern of \"crashing\" after the mid-1980s. 99 \"Dome-shaped\" population curve Sardines (Japan, Peru-Chile, California) Benguela anchovy North Pacific groundfish (Alaska pollock and other stocks) Lobsters, Sea birds, Seals, Reef fishes in tropical NP Newfoundland-spawning northern cod stock In opposite phase Anchovies (Japan, Peru-Chile, California) Benguela Sardine North Pacific Albacore Population expansion beginning in Mid-1970s Sardinella aurita (Gulf of Guinea) \"Crashing\" following the Mid-1980s Brazilian Sardine Northern Cod stocks Balistes (W. Africa) Population Size * Period of 'enhanced FJ Nino characteristics 70 72 74 76 78 80 82 84 86 88 90 92 Figure 4.6. Pattern of variation observed in many marine fish populations according to the \"Dome-shaped\" hypothesis (source Bakun, 1996). Management institutions, such as fishing policies, programs of economic incentives to fisheries development, and the scientific understanding of ecological phenomena, also present characteristics of slow change, as has been the case with the development of the sardine fishery in Brazi l . A t even slower speeds there are the changes in the configuration of Large Marine Ecosystems, both in terms of structure (number of species, trophic organization) and functioning (processes controlling production, recycling, etc.). That correspond to the speed of long term events such as those accompanying global warming effects. For instance, it is suspected that an increase in greenhouse effect would enhance the land-oceanic temperature contrast and intensify the upwelling along continental margins (Bakun, 1992). This phenomenon has been already reported for several regions of the world using multi-decade data on the intensity of upwelling-producing wind stress (Bakun, 1992), although its likely effects on small pelagic fish populations and marine ecosystems is still debatable. A t the human dimension, that is also the scale at which myths about nature endure, as stated by Light et al. (1995) in the case of water management in 100 the Everglades: \"[tjhe myth of \"quest for control\" has persevered through the past century and is characterized by a rational decision maker model that assumes that institutions can solve resource problems based on \"objective knowledge\" and the exploitation of technology in the name of progress\". Accompanying the increase in the time scale up the food web there is also an increase in area occupied by an individual species during its life cycle. For instance, certain fish species swim hundreds of kilometers during their life span, while most phytoplankton and zooplankton organisms have a distribution range confined to patches of a few kilometers. The combination of time and space hierarchies in the processes controlling biological variability result in a generic pattern in which larger organisms smooth out the effects of fluctuations in the smaller prey populations (Mann and Lazier, 1991). Connections do exist between processes at different scales . A critical feature of such hierarchies is that larger, slower levels usually maintain constraints within which faster levels operate, in other words, slower variables control faster ones (Holling, 1992). Hol l ing (1995) also suggests that at certain conditions slower and larger levels in ecosystems may become briefly vulnerable to small events and fast processes. In forest systems, for example, fast variables can dominate slow ones particularly at climax stages when the system is so overconnected that it lacks the resilience necessary to cope with forest fires or insect outbreaks (Holling, 1995). Also , during the phase of reorganization, when the system is underconnected and with weak regulation, there are opportunities for the establishment of a diversity of pioneer species which can result in completely different final ecosystem configurations (Holling, 1995). Therefore, sometimes fast variables can control slower ones. The analog of this process in the sea can be represented by the concept of Lasker windows. The success of a year class of a fish population depends on a diverse array of events that influence the survival of an individual fish from the egg stage to the recruitment size. Lasker (1975, 1980) and collaborators suggested that recruitment would be largely determined during a critical period in the early life history when the first feeding larvae have to find food of appropriate concentration and quality to avoid starvation. The formation of fine-scale patches of highly concentrated food particles depends on the existence of a temporal window, typically in the order of 5 days, during which turbulent mixing energy by the wind remains low so that particles can accumulate. The frequency of these calm periods of short time scale, called Lasker windows (Bakun, 1996), is thought to be responsible for most variability in 101 anchovy larvae survival and recruitment success in the California Current system, with potential consequences to the population size and to other species in the ecosystem. A central question for fisheries management is whether the switches in species composition observed in heavily exploited marine ecosystems represent a natural and reversible change caused by oceanic regimes or do they reflect a loss of resilience of ecosystems and a change of state caused by excessive exploitation. Several biological-oceanographic processes may be at play to cause the observed regime shifts in marine ecosystems following disturbance by fisheries and natural events. A review of the last 18 years of publications (referenced in the Aquatic Sciences & Fisheries Abstracts, A S F A ) on the causes of fluctuations in small pelagics showed that the majority of the papers (30 out of 52) associate changes to direct fishery and environmental effects. These include processes such as long term changes in temperature, wind intensity, and food enrichment mechanisms, such as the ones proposed by Bakun and co-authors, that directly affects recruitment success of fish populations. In the following section I used Ecosim to examine the effect of long term changes in food enrichment, food competition and predation on small pelagic forage fish (sardine) production. The ecosystem model was best suited to investigate possible relationships between changes in sardine recruitment rates and these ecosystem processes, specifically the interactions between sardine and anchovy. The simulation consisted of adding to the trophic model of the southeastern shelf a time forcing function that generates a 10 year cycle in primary productivity, as suggested in figure 4.6. During this time period, sardine fishing rate was maintained constant at twice the baseline fishing mortality. Figure 4.7 shows the predicted changes in the biomass of phytoplankton, sardine and anchovy, as well as in sardine recruitment. Results indicate that relatively small changes in primary production (+-50% the baseline primary productivity) can lead to larger changes in fish biomass. Sardine and anchovy are equally affected by changes in food enrichment, although sardine biomass declines faster than anchovy when food conditions worsen, due to the high fishing mortality. The combined effect of food enrichment and fishing generates a cyclic non-stationary pattern in sardine recruitment rates that culminates with the collapse of the stock. 102 Biomass/biomass baseline 1/3 1.5 1 0.5! PP/PP baseline Fishing rate (relative to baseline) years Figure 4.7. Output predicted by Ecosim with a cyclic regime in primary productivity (PP). Simulation run under a bottom-up trophic control. Not all groups in the system are represented. A fishing rate for sardine twice as high as the baseline value was applied during the simulation. The panel on the right represent the relationship between stock and recruitment for sardine obtained with the simulations on the left. Trophic mechanisms are also used to explain apparent regime shifts in the productivity of small pelagics. The most frequently visited hypotheses are food competition, and predation of adult anchovy/sardine on the early life stages of the competing species (Santander et al. 1983; Alheit, 1986; Butler and Pickett, 1988; Butler, 1991; Valdez Szeinfeld, 1991). Cury et al. (in press) also point at behavioral causes such as the \"school trap\" phenomenon commonly observed in small pelagics. The \"school trap\" is caused by behaviour mechanisms that drives sardines, anchovies and sardinellas species to school together with other species when their relative abundance is diminishing. B y doing so they subordinate their specific needs to a different set of environmental preferences, such as migration paths, habitat types, etc., not necessarily optimal for the species. Under these conditions stock productivity is expected to remain low for longer periods. The hypothesis of species replacement by food competition has been also used to explain shifts in species composition other than sardine-anchovy. Caverivieri (1991) and Bakun (1996), for instance, suggest that the collapse of sardine, Sardinella aurita, in the upwelling system of the 103 G u l f of Guinea led to a rapid increase in the abundance of triggerfish, Balistes capriscus, which is a semi-pelagic species as adult and pelagic planktonic feeders as juvenile. The apparent shift in dominance may result from climatic changes (more specifically to changes in continental runoff) but it is also considered that juvenile triggerfish effectively replaced the collapsed sardine at the mid-trophic level pelagic niche. Parallel changes have been observed in the Southeastern Brazilian Bight, where an inverse relationship between Balistes capriscus and Sardinella brasiliensis is noted in catch data from 1977 to 1995 (Table 4.6; Zavalla-Camin and Lemos, 1997). The hypothesis of replacement by food competition implies two interrelated assumptions; first that the mid-trophic level pelagic niche is occupied by a dominant species, and that a competing species may \" f i l l the environmental vo id\" created by the depletion of the dominant one (Turner and Bencherifi, 1983); second that food is limiting production, so that even in the absence of fishery the depleted population would not recover its biomass. In many cases, the hypothesis of food competition between sardine and anchovy has been refuted by considerable differences between the diets of the two species. In the Benguella system, for instance, competition between the two species seems to be limited by differences in the size composition of the diet, where sardine usually consume smaller prey that anchovy (Louw et al., 1998). Evidence of predation of adult sardine/anchovy on eggs and larvae o f the competing species would point to the existence of \"trophic triangles\" in the pelagic niche. In this case adults of sardine and anchovy share a common prey pool, while also feeding on the early life stages of the other. Thus reducing the biomass of one species of the pair would lead to an increase in prey availability and in the abundance of the competing species, which in turn would cause an increase in larvae and juvenile mortality due to predation. More common in marine and freshwater systems are however the \"trophic triangles\" among adult and juvenile stages of top predators and a third forage species that is both competitor/predator of juveniles and prey of adult stages (Walters and Kitchell , in press). A classic example is the cod-clupeid system in the Baltic sea (Rudstam et al., 1994 in Bax, 1998), where adult cod actively preys on adult and juvenile herring, while adult herring has cod eggs and larvae as a major food source. Therefore the depletion of adult cod may lead to planktivorous fish dominance that may prevent cod recruitment. Walters and Kitchel l (in press) suggest that this type o f trophic mechanism may be responsible for depensatory recruitment changes observed in marine fish populations that have been subjected to overfishing, including many clupeid species. The authors also noted the possibility of \"trophic quadrangles\" in zooplanktivorous fish, where adults feed selectively on 104 larger zooplankton than juveniles; reductions in adult fish biomass would cause an increase in large zooplankton that feed on smaller zooplankton groups also utilized by juvenile fish. There are two reasons to not support the hypothesis of predation relationships between sardine and anchovy in the Southeastern Brazilian shelf. First, the lack of evidence of eggs and larvae in stomach contents of both species (Goiten, 1983; Scwingel, 1996) . Second, which corroborates the first, is that although sardine and anchovy inhabit the same region, spawning behavior differs between species (Matsuura et al., 1992). While sardine spawning occurs in the surface warm mixed layer during spring and summer, anchovy spawns all year round mainly beneath the termocline, inside the cold S A C W which is present in the shelf bottom layers during spring and summer. Eggs and larvae of both species are found mainly above the termocline in the coastal regions. Therefore adults and early life stages seems to be spatially segregated during the peak spawning season. There is also no evidence of marked changes in food (zooplankton) selection with size between juvenile and adult stages of sardine (Saccardo and Rossi-Wongtschowski, 1991) which would make the case for \"trophic quadrangles\" at the bottom of the food web. Dynamic simulations with Ecosim were used to evaluate the effect of changing purse seine fishing rates on sardine biomass. In the following scenario it is assumed that most sardine predation mortality occurs during the species early life stages. Adult and juvenile sardine compete for food with other small forage fish as well as with juvenile stages of some apex predators (weakfish). Simulations were run under different hypotheses on the type of trophic control in the system and on the maximum relative foraging time of apex predators. The fishing scenario consisted of imposing high fishing rates on sardine for 4 years and then having the fishery closed to let the stock rebuild (Fig. 4.8). When the system is bottom-up controlled or when the maximum foraging time of apex predators is low ( F m a x =1) sardine promptly (within 2 years) recovers its baseline biomass once the fishery is closed. Consequently, food does not apparently limit production and any decrease in fishing mortality would be expected to result in the stock recovery. However when predation is controlling production at lower trophic levels and higher relative foraging time is assigned to apex predators (Fm a x=2) sardine biomass is maintained at a depressed level even after the fishery is closed; the population usually takes up to 10 years to recover its baseline biomass. Figure 4.8 represents the two types of responses obtained with Ecosim; the depensation scenario results from allowing apex predators (specially Pelagic feeding fish) to increase substantially its feeding time and consequently its consumption 105 rates to compensate for the decrease in sardine biomass; the high consumption rates in turn impede the prompt recover of the depressed population by imposing high mortality rates on juvenile fish. In the simulations increase in the biomass of competing forage fish with a decrease in sardine biomass was never observed. Results however imply that apparent regimes of low productivity in small pelagic fish could be explained as wel l by increasing predation mortality and consequent depensation in recruitment rates following stock collapse. Myers et al. (1995) found some evidence of depensation in 5 out of 128 marine fish populations for which time series of stock and recruitment data were available. Among them there are three clupeid fish stocks, the Spring spawning Icelandic herring, Pacific sardine and Georges Bank herring, which were severely overfished and remained commercially extinct for decades. The authors consider as possible causes of depensation multispecies effects such as nonlinear feeding responses of predators to changes in abundance of preys, as well as reduced reproductive success at low population densities (Allee effect). 106 Biomass / biomass baseline 3r Top-down control rel feeding tue Figure 4.8. Output predicted by Ecosim with selective predation mortality on juvenile sardine. Not all groups in the system are represented. The model includes a Delay-difference representation of changes in numbers and biomass of sardine, weakfish and triggerfish. The panels on the right represent the relationship between stock and recruitment for sardine obtained with the simulations on the left; after 4 years of heavy exploitation (2 times the baseline level) the sardine purse seine fleet is shut down to let the stock recover. Dotted lines represent the change in foraging time of pelagic feeding fish. Walters et al. (in press.a) warned about the possibility of strong depensatory effects, and multiple equilibria in community structure, when predators in the system are assigned higher limits to the time spend feeding (F m a x ) with decreasing food availability. That creates, according to the authors, results analog to the effect of including a Type II functional response o f predators to the availability of preys. Empirical and theoretical studies of predator-prey systems have tended to identify three types of functional responses (types I, II or III as in Hol l ing 1959) relating predation mortality rate and prey density. Type I response assumes that search by predators is random, and that the number of prey eaten by a predator would be directly proportional to prey density. Conversely, types II and III assume non-linear responses of predators to prey density as a result of behaviour processes, e.g. when predators are able to aggregate to areas where prey 107 numbers are high, as well as for the result of handling and satiation. In both cases predation rates may in effect increase with the decrease in prey density to cause depensation in production rates at low prey population size and multiple stability states. Field and laboratory studies have been able to demonstrate the existence of the latter two types of responses in both terrestrial (Holling, 1959; 1973) and aquatic systems (Peterman, 1977; Peterman and Gato, 1978; Peterman et a l , 1978). In these systems resilience seems to be conditioned by lower thresholds, e.g. in fish stock abundance, beyond which disturbance w i l l lead to completely different system stable states. Verifying this hypothesis in the field is however very difficult because it w i l l require estimating how predation mortality changes with the biomass of consumers and preys under uncontrolled and variable conditions determined by other environmental factors. Alternatively, the depensation hypothesis can be checked against the reconstructed time series of sardine spawning biomass and recruitment from V P A (Table 4.9; F ig . 4.9). Several processes can influence the productivity of an exploited fish population. Hilborn and Walters (1992) argued that sum of environmental and multispecies effects on a stock are likely to be most important during the fish early life stages, when larvae and juveniles are more vulnerable to sub-optimal environmental conditions, predation and competition for food. For convenience, in a population analysis the net effect of juvenile survival (i.e., recruitment) is usually represented by a stock-recruitment relationship. The focus on stock-recruitment relationships has also a practical reason, since it directly links a control variable (stock) and a rate of future stock production (recruitment). 108 T3 C3 \u00E2\u0080\u0094^N -4-\u00C2\u00BB 2 -9 0) ca N ,SP* M -a \" U (-1 3 o 13 T3 ca \u00C2\u00A33 <= ca \u00E2\u0080\u00A25 \"o \u00C2\u00A7 SP \u00E2\u0080\u00A2a \u00C2\u00AB IS 41 la ap o ca 2 03 . K \u00C2\u00AB * .s <; s Cu ca > s S -2 \"8 3 g 2 '3 O to f_| \" \u00E2\u0080\u0094 60,1-1 ca <\u00C2\u00AB -u o ca ^ CO \u00C2\u00A3 .2 2 CQ c . o OS \C u 'o \u00C2\u00A3 \"ft 0 I I 00 O ^_ CO OS 1 1 00 CO OS CO \"a- .\u00E2\u0080\u00941 'a-OS CO Os 00 r -OS t> CO OS 0 1 CO 1 OS co Os 00 t~- m CN in 00 00 in in in 00 00 00 ON OS in CN O in SO 0 CO 00 m OS O Ti- la- CO CO CN CN s o r o in 00 00 s o t ~ 00 in -a- OS 00 00 in in CN OS 0 in r o Os so Os r~ r o r o r o 0 ^J-1\u00E2\u0080\u00941 in so r - in OS r o 00 r~ in CO O in 1\u00E2\u0080\u0094\u00E2\u0080\u00A2 in Os Os t ~ 00 Os in Os OS 00 r - Os O O 00 0 in CO OS Os in CN r - 0 CN in \u00E2\u0080\u00A2 1\u00E2\u0080\u0094. ~\u00E2\u0080\u0094 SO 00 r~ 00 CN 00 Os CN SO ~ SO \u00C2\u00BBn t > CN 00 0 s o Os 00 Os Os CO CO CN OS \u00E2\u0080\u0094H OS 00 O O in s o SO SO in in CN r o CO O 00 t~-00 O CN CN Tt 1\u00E2\u0080\u00941 CN ~ ~ CN in 00 so so OS os r o r~ 0 00 CN C~ OS OS OS in so 00 Os m SO m 1\u00E2\u0080\u00941 CO Os in in 00 OS i ~ CO m 00 0 CN Os so CN in .\u00E2\u0080\u0094\u00E2\u0080\u00A2 r -CN \u00E2\u0080\u0094 CN \u00E2\u0080\u00A2\"3- \u00E2\u0080\u009E OO 00 0 -a- r o \u00E2\u0080\u00A2a- 00 00 CN 0O OS r o m CN OS SO CN SO OS r> SO 0 CO co 00 SO t ~ ^ \u00C2\u00AB s o CN CO CN SO SO r o r o SO CO in r o s o SO _ in SO 00 00 s o CN .\u00E2\u0080\u00941 00 CN CO s o 00 OS \u00E2\u0080\u0094 SO \u00E2\u0080\u00A2<* 0 OS s o OS co s o so OS CO CN CO 1\u00E2\u0080\u00941 SO \"a- CN CN 00 CN CO CN CN CO ' ' ' ' CO CN 0 r> so CN CO so 00 00 SO 00 OO 0 00 00 SO CO CN in -a- 0 m Os SO SO SO CO 00 CN CO r~ CN 0 SO 00 ,\u00E2\u0080\u00941 CN SO .\u00E2\u0080\u00941 00 SO in .\u00E2\u0080\u00941 SO CO r o CN CN f m Os CN CO \u00E2\u0080\u00A2a- m CO in ^ O OO Os O 0 in in r-~ r o vo OS CO CN r o 00 \u00E2\u0080\u00A2a- 1^ r -SO r~ O OS CO OS CN CN CN CN Os OS r - OS r o r o ~ 0 CN CO Os 0 t 0 CN r o 00 00 SO O t-- la- 0 m 00 OS OS Os CN co in 00 0 OS 00 .\u00E2\u0080\u00941 00 so CN SO CN r> 1\u00E2\u0080\u00941 CO CN r o OS r o -a-CN CN CN CN OS OS CN 00 CN in CN r o OS r-~ r o OS OS Os OS CN Os Os CN O O CN OS 1\u00E2\u0080\u00941 t> SO CN SO O 00 O SO \u00E2\u0080\u00A2\u00C2\u00ABt CN r o OS CN CN CN CN 00 SO in CN SO Os r~ 00 r-~ O OO C^ r - 00 CO 1^ r o r o s o OS Os 00 CN OS -^1 OS CO e- CN .\u00E2\u0080\u00941 0 CO SO CO .\u00E2\u0080\u0094< CN in 00 in 00 in in t> CN CN CO r o l ~ 0O _ OS \u00E2\u0080\u009E so 00 00 0 SO O CN m CN Tt 00 CN s o OS in 0 O SO CN s o SO 00 m 0 O in Os \u00E2\u0080\u00A2a- 00 CN 0 m CN Os CO in s o CO CO CO CO 0 m in CN in r o in 0 \u00E2\u0080\u0094< CN CO \"b 00 0 1). A l l recruitment variability is described by a log-normal error (v). Mode l parameterization was carried as follows. For both hypothesis, the maximum number of recruits was considered known from the historical experience, being set to 12*109, thus eliminating one unknown parameter (K). Parameter x was set to 1 to represent the hypothesis of no depensatory mortality (Hypothesis 1) and to a value of 2 to represent strong depensation i l l (Hypothesis 2). Mode l parameterization was completed by finding the maximum likelihood estimator of parameter a, i.e. the maximum recruitment rate at low spawning stock biomass. The kernel of the likelihood function with log-normal error was calculated according to Walters and Ludwig (1994). Final model parameterization is shown in table 4.10. Table 4.10. Parameters used in the representation of density-dependent hypothesis of stock-recruitment Parameters Hypothesis 1: no-depensation Hypothesis 2: depensation a 135 867 K 0.0888 0.0138 X 1 2 14 r 12 -10 -O 8 -w '5 i_ 6 -o * 0.003 -5 to J= o 0.002 -ro O 0.001 -0 -81 79 \u00E2\u0080\u00A2 82 \u00E2\u0080\u00A2 80 83 \u00E2\u0080\u00A2 78 \u00E2\u0080\u00A2 77 100000 200000 300000 400000 Biomass (tonnes) 500000 Figure 4.13. Relationship between catchability to the purse seine fleet and sardine stock biomass. Data points labeled with years. Figure 4.14. Changes in the catchability of sardine with stock size and the mean sea surface temperature in the Southeastern Brazilian Bight. Temperature was obtained from the C O A D S database and the average computed for the whole shelf area. 116 Table 4,12. Results of partial correlation analysis between catchability, stock size and sea surface temperature. Correlation Variables Catchability vs. Stock biomass Catchability vs. Temperature Temperature vs. Stock biomass Partial Correlation of Catchability and Stock biomass -0.887 -0.459 0.335 -0.877 In the case of the Brazilian sardine, the decrease in stock biomass observed from the mid-1970s to the early 1990s appeared to be followed by a substantial reduction in the distribution range of the species (Fig. 4.15), which tended to concentrate in smaller areas on the southern and central parts of the bight. This decrease in spatial range is also consistent with the changes in the regional distribution of landings (Fig. 4.16) which became highly concentrated in the south after the mid-1980s. However, it is still unclear whether the collapse in distribution range was due exclusively to the interaction between environmental and behavioural mechanisms, or a result of a depletion of geographically separated stocks. The hypothesis of two separated groups of sardine in the Southeastern Bight is supported by differences in size composition, spawning locations and time (Rossi-Wongtschowski, 1977), and protein biochemical types (Vazzoler and Phan, 1976), which all indicate the existence of one group in the north, between 22\u00C2\u00B0 and 25\u00C2\u00B0 S, that spawns during spring and summer, and a second group in the south (26\u00C2\u00B0 to 28\u00C2\u00B0 S) spawning mainly in spring. On the other hand, results of egg and larvae surveys indicate that spawning has taking place in different locations throughout the years and seems to be mostly determined by oceanographic conditions (Matsuura, 1988). In this sense, the observed concentration of sardine to the south would be a result of better oceanographic conditions in the area, and the depressed total biomass, which would force shoals to occupy a fairly restricted geographical range. The effect that the collapse in distribution range had on stock catchability is still unknown given the lack of information on fishing effort by each state fleet since the mid-1980s. Ultimately, the increase in stock catchability depends on how efficient are purse seiners in following the fish with the range collapse, that is, on how fishing effort is allocated. This question is addressed in the next sections by looking at the fleet characteristics and movement over the Bight. Fleet characteristics The purse seine fleet is basically divided among three states, R io de Janeiro (RJ), Sao Paulo (SP) and Santa Catarina (SC). 117 Rio de Janeiro In 1977 a total of 300 boats was fishing sardine in the Southeastern Brazil ian Bight. Wiefels & Jablonski (1979) concluded from sampling 2/3 of this total that ca. 52 % of the boats were fishing on waters off R J . From the sampled boats, 14.5 % had less than 15 m, 60.5 % between 15 and 22 m and 25 % between 22 and 27 m. Purse seiners targeted sardine and other pelagic species, specially mackerel Scomber japonicus. The importance of mackerel in the catches became more important after 1982, when the species made up to 70% of the total catches. From 1982 to 1985 the fleet was composed by an average of 167 boats constantly in activity and by 123 boats eventually in activity for less than 3 months per year. B y 1985 ca. 40% of this fleet had more than 20 tons (average size 18.6 m) and 60% had less than 20 tons (average size 12.3 m). Most of the sardine catches came from the \"constant\" boats and from those with tonnage larger than 20 tons. Both fleet components also fished on alternative species, although this activity was more important for the smaller boats usually restricted to nearshore areas (less than 50 m). A t that time there were at least 8 major landing ports along the coast. Variations in the oceanographic conditions and fish distribution determined the importance of each of the landing ports, as boats were highly mobile in the search for the most productive sites within the state waters ( S U D E P E , 1986). Larger boats (tonnage larger than 20 tons) were at least 3 times more efficient in finding and capturing fish shoals than smaller boats ( S U D E P E , 1986). 118 0.2 0.4 0.6 0.8 South 0 1964 1969 1974 1979 1984 1989 1994 Figure 4.16. Proportion of total sardine landings by fishing area. South refers to landings in Santa Catarina, central to Sao Paulo, and north to Rio de Janeiro (source I B A M A , 1995). In 1987 the number o f boats with tonnage superior to 20 tons decrease to 69 (Valentini and Cardoso, 1991). This decrease was balanced by an increase in the number o f smaller boats, who had sardine as one of the target species. Part of the larger boats that used to fish sardines were adapted to fish bonito, Katswonus pelamis. Valentini and Cardoso (1991) suggested that the number of purse seiners licensed on R J remained relatively high even after the decrease of the sardine catches after 1981, since sardine was not anymore the mainly elected species in the fishery, but one o f the alternative pelagic fishes. The first attempt to regulate the number of fishing licenses came in 1989 with the Portaria IBAMA 1347/89. A t that time there were a total of 105 licensed boats and about 180 illegal boats fishing for sardine off R J . Vessel numbers and characteristics did not change significantly until 1992, although a decrease in the number of active boats with the decrease in catches was observed. O f the 162 licensed boats, 131 were considered artisanal due to their small size and tonnage (< 20 tons), and the lack of technological devices. The 31 remaining boats were more equipped (sonar and power block), with tonnage between 20 and 81 tons and power between 110 and 150 H P . Even after the efforts to legalize all the purse seiners there are still boats fishing sardine without license. 120 Sao Paulo The number of purse seiners in Sao Paulo increased from 55 in 1973 to 120 boats in 1988 (Valentini and Cardoso, 1991). With the collapse of the fishery in the late 1980s and early 1990s, there was a decrease in the number of boats actively fishing. In 1989, only 77 of the 113 licensed vessels were active; 47 in 1990; 35 in 1991; and almost all fleet was inactive in 1992 ( I B A M A , 1991; 1992; 1993; 1994). Illegal boats were also observed in SP, making up to 34 boats in 1989 and only 6 in 1990. Overall, boat characteristics remained similar along the years, mean size of 19 meters, mean tonnage of ca. 55 tons and 260 H P power. However, some of the vessels acquired technological innovations such as sonar and power blocks after government incentives for fleet modernization during the 1980s. Santa Catarina Since the early 1970s the Santa Catarina's fleet was characteristically the most technologically equipped and with larger vessels among the three states. The number of boats increased from 54 to 105 between 1973 and 1988. Fleet mean characteristics were: size 21meters, tonnage 73 tons and 262 H P of power. The SC fleet significantly increased fishing power after 1985, following the government incentives, with the introduction of sonar and power-block. Technological improvements to locate and catch fish were particularly improved in 1988. With the collapse of the stock in the late 1980s the number of active boats progressively declined. Only 74 of the 107 licensed boats were actively fishing in 1991. A total of 12 boats still remained illegal in the fishery in 1991. Fleet movement and effort allocation Typically, fishing operation by purse seiners is very slow, involving sometimes up to 6 hours to complete a set. Catch volume and the long handling time determine that landings take place in ports close to the catching areas, thus creating a straight relationship between areas of sardine concentration and the landing volume in different ports along the coast (Valentini and Cardoso, 1991). Therefore, data on the origin of purse seiners at different landing ports along the coast can 121 provide information relevant for understanding fleet movement in the fishing ground. Figure 4.17 describes, for each state fleet, the proportion of licensed vessels that landed sardine in the different states between 1989 and 1991. The first conclusion from this analysis is that fishing boats distribute differently according to their origin; boats from R J maintain a fairly restricted fishing area, mostly concentrated in their own state waters; boats from SP appear to utilize ports from the other two states more regularly; and the SC fleet landed sardines either in SP or in SC. The second conclusion refers to changes in fleet distribution with time, where the two most mobile fleets (SP and SC) seemed to concentrate their activity (landings) to the southern part of the stock distribution during the period analyzed . In 1989 ca. 65% of the SP fleet were actively fishing in the waters of both SP and SC, while in 1991 this proportion was more than 80%. A s for the SC fleet, the southernmost state, the proportion of the fleet landing exclusively in the state waters increased from 30 to ca. 65% between 1989 and 1991. This information is consistent with the observed increase in the proportion of catches from the south (Fig. 4.16), and seems to corroborate to the hypothesis that purse seiners (at least those from SP and SC) followed sardines with the collapse in distribution range to the south in this time period. Results also indicate the possibility that the collapse in the distribution range limited access to the stock to roughly one third of the fleet, that from Rio de Janeiro, which is mostly composed by smaller and less technologically equipped boats that operate almost exclusively in the northern end of the species distribution area. 122 Fleet RJ SP SC RJ RJ/SP SP SP/SC SC R/S/S RJ \"\u00E2\u0080\u00A2\"SP SP SP/SC SC R/S/S RJ RJ/SP SP SP/SC SC R/S/S Landing Place U n d i n \u00C2\u00B0 P l a c e L>n*0 PLace Figure 4.17. Fleet distribution by landing place. Landing places were grouped in 6 categories: restricted to Rio de Janeiro (RJ); occurring in both Rio de Janeiro and Sao Paulo (RJ/SP); restricted to Sao Paulo (SP); occuring in both Sao Paulo and Santa Catarina (SP/SC); restricted to Santa Catarina (SC); occuring in the three states (R/S/S). Another likely consequence of an efficient process of search and effort allocation is that boats w i l l move to equalize catch per unit of effort across the stock distribution area according to an Ideal Free Distribution (Hilborn, 1985; Gi l l i s et al., 1993). Figure 4.18 shows the results of the regression of the proportion of catch in area i (c/C) on the proportion of the effort in area i (f/F) during 1990 and 1991. The ideal free distribution appears to be a good approximation to the dynamics of purse seiners off the coast of Santa Catarina, although slopes smaller than 1 were observed in the two years analyzed. That indicates higher than average return rates when proportional effort is high and smaller than average return rates when proportional effort i f low. Such small deviations from the IFD line can be explained by diverse factors, including increasing information exchange between boats with increasing effort, and by patterns of exploratory behaviour adopted by skippers, which by searching areas with low cpue for better fishing opportunities may cause a decrease in average return rates in areas with low proportional effort (Gill is et al., 1993). Overall, results confirm that purse seiners are highly efficient in following 123 shoal distributions, and that fleet dynamics may play an important role in creating the observed increase in catchability with decrease stock size and range. 1 0.8 0.6 0.4 0.2 1991 r2=0.95 slope=0.88 0.2 0.4 0.6 0.8 c/C Figure 4.18. Test for the equalization of cpue among fishing areas off Santa Catarina (see map on figure 3).c/C is the proportion of the total catch in area i , and f/F is the proportion of the total effort in area i . Regression slopes are statistically different from 1 (1990: df = 26; t = 3.69; P=0.001; 1991: df= 17; t=2.33; P=0.032). The analysis of stock catchability, range collapse and effort allocation has two general consequences for sardine stock assessment. First it indicates that cpue and other fishery dependent indexes of abundance are potentially biased by the behaviour of fish and by the characteristic non-random search of fishing vessels. Second, it questions the effectiveness of harvest control by effort limitation in preventing stock overfishing, specially with the overcapacity of the fleet. Besides being affected by the increasing catchability, tactics of effort control by fishing closures are particularly inefficient when fishers reallocate effort temporally (Fig. 4.19). For instance, up to 1990 effort was controlled by one fishing closure during the spawning season. With the prospect of stock collapse in 1991, a second fishing closure was established during the recruitment months. A s a consequence, effort (measured by the number of trips of the SC fleet) was simply reallocated to the other months of the year, total effort being almost the same between the two years. These results indicate that much insight can still be gained in controlling harvest by understanding fishers dynamics in allocating effort in space and time. 124 Figure 4.19. Monthly effort allocation by the Santa Catarina fleet before (1990) and after (1991) the establishment of a recruitment closure. 4.4. Summary This chapter reviews the status of sardine stock assessment and analyze the current ecological uncertainties in the management of the fishery. The ecosystem of the Southeastern Brazilian shelf is structured around the operation of physical-biological cycles of different speeds, the most documented ones being the seasonal upwelling cycle, and the decadal regimes determined by multi-year variability in the intensity of physical forces controlling biological production. The latter is particularly important for the management of fishery resources because it operates at the time scale commensurate with the life span of most marine fish populations. Fisheries in the Southeastern shelf have been targeting mostly low trophic level species and have since the late 1970s experienced an increase in the mean trophic level of landings. The scale of fisheries impact and the effect of environmental regimes are often associated to changes in the structure of the ecosystem and to a decrease in recruitment rates of the sardine population. The combined effect of fishery and natural processes make it difficult however to characterize the productivity of the population and to predict the results of rehabilitation measures for the sardine stock. Also , given the current regulatory mechanisms, the implementation of any measure to control harvest w i l l encounter difficulties created by the behavior of the fish and the dynamics of the fishing fleet, which seems to be responsible for changes in the stock catchability. 125 Chapter 5. Analysis of harvest decisions and information needs in the management of the Brazilian sardine. Comparing multi-species and single-species modeling approaches 5.1. Introduction This chapter aims to evaluate the short and long term predictions of the impacts of harvest strategies and controls for the Brazilian sardine, and to discuss the relative values of reducing current uncertainties on ecological processes. Results are used to recommend on the types of research that would most likely provide the type of information needed to improve the quality of decisions, and on the precautionary measures that should be adopted in face of the ecological uncertainties. One the major tasks of fisheries assessment is to estimate the tradeoffs of management choices. Also , as a scientific activity, fisheries research involves a process of learning about nature in which alternative ideas, or hypotheses, are contrasted against observations and experiences, so that a better understanding of the potential responses of systems being managed can be achieved through time. In these activities models are usually employed as tools to test our understanding of the mechanisms in the system, to provide the basis for both understanding of known patterns and prediction about situations not yet encountered, and to provide the means for evaluating the potential effects of various kinds of decisions (Hilborn and Mangel, 1997). Fisheries assessment can be based on two types of modeling approaches: single-species and ecosystem or multi-species models. Ecosystem models are in their infancy and have been little used to date. Single-species models are based solely on the analysis of population processes such as growth, mortality and recruitment, and has been the most widely applied approach to provide advice on the consequences of fishery regulatory tactics. B y definition, multi-species models consider not only the biological processes of the target population but also the ecological (e.g. predation, consumption, competition) and technological (e.g. by catch) interactions with other species in the ecosystem. Ecosystem impacts of fisheries has become an important concern in fisheries management and one of the most important issues in fisheries science. Therefore, a 126 central question for fisheries assessment today is on how to evaluate and communicate the impacts of alternative fishing policies on marine resources and ecosystems. Multi-species models have been proposed as tools for guiding the implementation o f ecosystem principles in fisheries management, but exactly how the approach could be used, and also what should be the role of single-species approaches in this new paradigm, are still unclear. Such comparisons have not been performed explicitly before. Following the framework outlined in chapter 1, this chapter analyzes how management decisions in the sardine fishery can be made under current ecological uncertainties, and how recommendations differ according to two modeling approaches: a single species and a multi-species model. Models are employed in this analysis to evaluate two types of decisions. The first concerns the evaluation of strategies for sardine stock recovery. The analysis aims to evaluate the short term predictions o f stock recovery according to different fishing strategies. The second type of decision relates to the choice of long term fishing strategies. I discuss the value of reducing uncertainties about ecological processes, the opportunities for improving understanding of these processes, and the precautionary measures that should be adopted in face of ecological uncertainties. 5.2. Methods Modeling approaches The evaluation of the performance of fishing strategies was carried with two modeling approaches: a single-species delay-differential model (Deriso, 1980; Fournier and Doonan, 1987) and a multi-species trophic model (Ecosim, Walters et al., 1997). Single-species approach Fishing strategies were evaluated with a Delay-Differential model used as an operating model in Monte Carlo simulations. The model predicts next year's biomass (B t + i ) and numbers (N t + 1 ) according to the equations (Hilborn and Walters, 1992): 127 Bt+x=st[aNt +pBt] + wkRt+l (1) Nt+l=stNt+Rt+l (2) where w k is the weight at recruitment age k (years), s t = A,(l-h,) is the total survival rate, X is the natural survival rate (A=e\"M), M is the instantaneous natural mortality rate (=0.95 year\"1), and h t is the exploitation rate. The exploitation rate can be calculated as the ratio catch/biomass in a given year, or as a function of fishing effort, i.e., h = l- e~qE , where E is effort and q the catchability parameter. Catchability was represented by an inverse function of stock abundance: q = q , B q 2 (3) where q l is a proportionality constant and q2 is the degree to which catchability increases with declining stock size. The delay-differential model also represents growth and recruitment processes. Growth in mean body weight at age is described by a Ford-Walford plot o f the type wa = a + pwa_i, where a and p are constants from the regression between weight w at consecutive age classes a = k, k+1, ...fully vulnerable to the fishery. Sardine parameter values for a (= 0.025) and p (= 0.896) were obtained by regressing data on weight at consecutive ages from Cergole (1994). Recruitment (Rt + 1) is included in the model as functions representing different stock-recruitment hypotheses (Table 5.1). Uncertainties are included in three processes/states: (1) the relationship between spawning stock and recruitment; (2) the relationship between stock biomass and catchability; and (3) the assessment o f stock biomass at sea. Uncertainties in the relationship between stock and recruitment were represented according to the hypotheses about recruitment changes outlined in chapter 4. For decision analysis, all hypotheses were assigned equal degree of belief, i.e., they are assumed to fit the available date equally well . The hypotheses and models used in this analysis are described in table 5.1. 128 Table 5.1. Hypothesis, models and parameters used to predict recruitment rates in the Delay-differential model (source chapter 4). Hypothesis Model Parameter values 1. Recruitment is a function of stock size 2. Recruitment is a function of stock size with depensation at low stock sizes. 3. Recruitment is a function of stock size and low frequency environmental regimes. aSSBx Recruits = \u00E2\u0080\u0094 ev 1 + -SSBX K aSSBx Recruits = \u00E2\u0080\u0094 ev 1 + SSB' K aSSB Recruits = ttt^t ev a = exp' M l K = -1 + -M , aSSB K M 2 ( e x p M l - l ) a= 135 K = 0.0888 x = 1 v = 0.4 a = 867 K = 0.0138 x = 2 v = 0.4 Mu = -3576 + 1 \u00E2\u0080\u00A2 Sin(-^-) M 2 = 0.152 v = 0.4 Uncertainties in the relationship between stock biomass and catchability were described by marginal posterior probability distributions for parameters q, and q 2 (equation 3) based on catchability data from chapter 4. Probabilities were computed using Bayesian analysis, D , , , , , L(datalql,q1)*p(ql,q2) Pyq\><72 < data) = \u00E2\u0080\u009E T / , \u00E2\u0080\u0094 ; -\u00E2\u0080\u0094\u00E2\u0080\u0094 ZL(data I q\,q2)*p(q\ ) where the likelihood of a given parameter combination L(data/q,,q 2) is calculated assuming that catchability values present a log-normal distribution (i.e. assume only positive values), and the prior probability distribution of parameters p(q l 5q 2) have an uniform distribution. The likelihood of a given parameter combination was calculated as ss - l / 2 ( \u00E2\u0080\u0094 ) L(data Iqi,q2) = e 0 where, SS = X(Inq - Inq) , SSq = X(Inq - Inq) , q is the catchability coefficient calculated t t from V P A and observed effort data (see chapter 4), q is the estimated catchability from q, and q 2 , and qis the estimated catchability at best fitting q, and q 2 . Figure 5.1 shows the estimated marginal posterior probability distribution of catchability parameters. 129 0.15 0.15 0.10 -s ro \"o \" ,\u00E2\u0080\u0094I\u00E2\u0080\u0094Tl\u00E2\u0080\u0094I\u00E2\u0080\u0094I \u00C2\u00A3 0.05 - \"~J-]\u00E2\u0080\u0094 o.oo I I M I M II II II I II M \u00E2\u0080\u0094 0.0001 0.0004 0.0007 0.001 0.0013 0.0016 0.0019 q i Figure 5.1. Marginal posterior probability distribution of catchability parameters q l and q2 (from equation 3) for the Brazilian sardine fishery based on data from 1977 to 1983 (see chapter 4). Finally, uncertainties in the estimation of stock biomass by direct methods (e.g., acoustic surveys, egg production, etc.) were introduced by including a normally distributed error around the true stock biomass (Frederick and Peterman, 1997), where B e s t = B t + (B t - C V -w) B e s t is the estimated stock biomass in year t, B t is the true biomass, C V is the coefficient of variation of the biomass estimation procedure (0 0.5), and w is a normally distributed variable with mean 0 and variance 1. Monte Carlo simulations were used to evaluate the expected outcomes of fishing strategies for sardine. Three criteria are used in the evaluation of strategies: average catches, catch variability, and the probability o f stock collapse. These criteria were selected because they represent three types of objectives commonly observed in fisheries management: - Maximize yield. In effect, the increase in fisheries catches represent more fish to the industry, more economic opportunities to the capture sector and, consequently, more job offers. - Maximize catch stability. Very often, the major interest with a management plan is to guarantee the stability or low variability of catches, and therefore maintain a constant supply of fish to the industry. - Minimize the chances o f fishery collapse. That is a fundamental objective to any fisheries management plan, considering the ecological and economic costs associated with the collapse of fisheries. ^ 0.10 -O ro o \u00C2\u00B0- 0.05 0.00 0.1 0.3 0.5 0.7 0.9 1.1 1.3 1.5 1.7 1.9 q2 130 Two types of fishing strategies are evaluated: - Catch control: the total catch allowed in a given year is defined in advance. Three types of catch control options are considered: constant allowable catch; total allowable catch defined by a constant exploitation rate; and total allowable catch defined by a minimum stock escapement. - Effort control: that represents the status quo policy of fishing control in the sardine fishery, involving the definition of a limit to the total effort allowed in a given year, and a minimum size of fish in the catches. In this specific case the effort unit used was the number of trips corrected by the mean tonnage of boats, as in chapter 4. Ideally the analysis of a purse seine fishery should use as effort unit any measure of fishing time such as fishing days, time searching, etc. Nonetheless, the characteristic short duration of purse seiners trips in the sardine fishery allows for a reasonably good approximation between the number of trips and the number of fishing days. Figure 5.2 represents the structure of the decision problem analyzed. The choice for one fishing strategy in the left hand side w i l l depend on the objectives considered and on the uncertainties on diverse processes and states (depicted as branches diverging from the circles). For instance, the choice for an strategy of catch control is made difficult by uncertainties on population characteristics, such as virgin biomass and productivity, as well as on the size of the stock in the sea. The choice for an allowed level of effort is on the other hand influenced by the stock productivity (i.e. recruitment) and the catchability to the purse seiners. A l l decisions are subjected to the inherent uncertainties about the future environmental regime. 131 Fishing Strategies Uncertain processes and states Figure 5.2. Conceptual structure of a decision analysis on the choice for fishing strategies for sardine. The squares on the left hand side represent decisions, the circles represent the main uncertainties on population processes and states. Expected outcomes of fishing strategies The expected outcome o f the different combination of fishing strategies and controls were obtained following the steps below (Figure 5.3). Starting conditions for simulations were set to be consistent with the observed status of the fishery in the late 1970s. 1) A number of alternative strategies in each strategy type (catch or effort control) was defined by test runs of the model and also by using historical data from the fishery. Specifically in the case of strategies of effort control, effort levels applied between 1977 and 1983 (Chapter 4) were used as reference points to define alternative fishing strategies. A n age at first capture of 1.5 years (approximate age of first maturity) was used as auxiliary measure to effort control. 2) Simulations were run for 10 years having as starting conditions the average spawning stock biomass between 1977 and 1983 from V P A (-250,000 tons), and a dome-shaped regime of environmental conditions as proposed by Bakun (1996) for the 1980s (see Chapter 4 ) . 3) The expected outcome of a given fishing strategy was calculated as the average of the 10 years runs obtained in 1000 Monte Carlo simulations, weighted by probabilities of parameters, models and variables used in the model. 4) The performance of each fishing strategy is defined by the expected average catch, catch variability, and the frequency of cases (simulations) in which the spawning stock was driven 132 below the historical lowest size, that is approximately 50,000 tons. This frequency was used as a measure of the probability of stock collapse. Predictions of stock recovery Simulations were carried also to examine the expected responses of the stock to strategies of stock rebuilding after the collapse. Stock biomass and exploitation rates for the. late 1980s were used as starting conditions for simulating biomass recovering trajectories using the same procedure outlined in Figure 5.3. Spawning stock per recruit Data on survival, maturity and reproductive output at age (Table 5.2) were used in the analysis of spawning per recruit 3.5 SPRA50% = ^(hfiRi) where S P R A 5 0 o / o is the number of spawners produced per recruit for a given combination of age at first capture ( A 5 0 % ) and exploitation rate (E=F/Z; [0.1]); /, is the survivorship to age i (/,=/\u00E2\u0080\u009Es(l-VjE), l0 is the survivorship to age 0 (=1), s = exp ( M ) , M is the instantaneous natural mortality rate; Vj is the age specific vulnerability , and R; is the age specific reproductive output. V ; is a function of the age at first capture and a concentration parameter (b=10), i.e. A50% +1 The reproductive output weights the relative reproductive contribution of an age class be accounting for differences in relative fecundity and frequency of spawning (Parrish et al., 1986). S P R is expected to decrease with E , and is normally expressed as the percentage of the SPR value calculated for and unfished stock (%SPR; Goodyear, 1993). 133 Table 5.2. Percentage of individuals mature and the reproductive output (A) at age for Sardinella brasiliensis (source Cergole, 1994). The reproductive output (A) was calculated using the equation proposed by Parrish et al. (1986) to E. mordax. The equation was adjusted so that the relationship A ; = 0.0000432W; 2 6 4 generates a factor A=l when the weight W = 45g, the approximate mean weight at age at first maturity Age 0 0.5 1 1.5 2 2.5 3 3.5 %Mature 0 0 0.5 0.75 1 1 1 1 A 0 0 0.83 1.00 1.95 3.21 5.19 10.58 Start A Choose fishing strategy type For each strategy 1 For each hy For each /pothesis loop over hypothesis 3 Carlo trial loop over M.C. trials Initialize stock abundance For years = 1 to 10 loop over years Estimate stock abundance (for strategies based on biomass estimations) t Harvest i Calculate remanining SSB (t) and recruitment (t) t Average outcome i.e., average catch catch variability prob. of collapse A Degree of belief on hypothesis Expected outcome for strategy Figure 5.3. Monte Carlo simulation procedure used in the evaluation of the outcomes of fishing strategies using a single-species approach. 134 Multi-species approach with Ecosim A trophic model o f the Southeastern shelf ecosystem (Fig. 4.4; chapter 4) is used in this section to evaluate fishing strategies for sardine when the population is placed in an ecosystem structured by trophic (i.e., predator-prey) relationships. The model has 22 trophic groups, and tracks changes in number and biomass of three groups (sardine, triggerfish and weakfish) that were split between juvenile and adult pools. Ecosim (Walter et al., 1997) was used to evaluate the predicted effects of short-term harvest strategies on the main harvested species in the ecosystem, and also to assess optimal fishing rates for the sardine stock. Mode l details are given in chapters 2 and 3. Uncertainties were accounted for the type of control of trophic relationships in the ecosystem, which has been shown to influence the magnitude and direction of changes in the ecosystem when subjected to fisheries (Chapter 2). Three hypotheses are considered to have the same degree of belief: a) bottom-up control, where the amount of prey available to predators is limited so that the mortality rate of a species in the ecosystem is largely independent of the abundance of predators. b) top-down control, where a larger part of prey biomass is vulnerable to predation, and mortality rate is largely dependent on the abundance of predators. c) \"wasp-waist\" control (Bakun, 1996; Cury et a l , in press), where the abundance of small pelagic fish control both their predators and prey. Following Cury et al., the interaction between small pelagic fish (sardine, anchovy, and other forage fish) and their prey (phyto and zooplankton) is assumed to be top-down controlled, while the interaction between small pelagics and their predators is bottom-up controlled. Top-down and bottom-up control were set in the model according to the scenarios a and b above. Trophic control hypotheses were combined with two contrasting assumptions about the maximum relative foraging time of apex predators in the system. The combination of hypotheses thus produced 6 different models used in the analysis of fishing policies. 135 To examine the short term effects of fishing strategies, dynamic simulations were run for 5 years according to fishing scenarios that halve, completely stop, and double the fishing rates of the four major fleets in the region, i.e., Purse seiners, Bottom trawlers, Shrimp trawlers, and Pole-and-line vessels. Scenarios 1, 2 and 3 represent fishing strategies solely directed to the Purse seine fleet, resulting in a 50% decrease, 100% decrease, and a 100% increase in F for sardine in 5 years, respectively. Strategies were also tested where all fleets (Purse Seiners, Bottom trawlers, Shrimp trawlers, and Pole-and-line vessels) are decreased by 50% (scenario 4); decreased by 100%, i.e., all fisheries closed (scenario 5); and increased by 100% (scenario 6) during a 5 years period. N o by-catch is included in the model, as the objective of the simulations was to investigate the likely effects of changes in fishing mortality rates for the target species. To assess the optimal fishing mortality for the sardine stock, simulations were run for 10 years and the average sardine yield compared among constant fishing rates. Valuing new information: the Expected Value of Perfect Information The quantitative decision analysis described above provides information useful for the computation of risks associated with harvest decisions, either in terms of the expected consequences of a given action or in terms of the odds of obtaining an outcome different from that prescribed as the most probable. That in turn allows evaluation of how well decisions can be made with present available information and uncertainties. Simulation results are thus used to evaluate the benefits of reducing uncertainties on the processes relevant for the decision problem. The objective of this analysis is to calculate the expected value o f a decision made with perfect information about each of the hypothesis is true. This value, termed the expected value of perfect information (EVPI) , represents the maximum, or upper bond, on what we should be wil l ing to pay for research that w i l l generate new information that reduces uncertainties. Two steps are necessary to calculate the E V P I (Walters, 1986; Morgan and Herion, 1990). First the expected value of a decision made with perfect information must be calculated. Imagining that we could forecast which state of nature exists, so that we could choose the optimal action for that state, then the value of a decision made with perfect information would simply be the forecasted result of that decision. However, at this point in time we are uncertain about which 136 state of nature exists, so the expected values of outcomes for each optimal action for each possible state of nature must be weighted by the probability o f occurrence of those states. Therefore the expected value with perfect information (EVWPI) is calculated as EVWPI = T.[p(i) \u00E2\u0080\u00A2 Max{i)\ i where, p(i) is the probability of state i occur, and Max(i) is the best outcome for state of nature i . To calculate the expected value of perfect information, then E V W P I needs to be subtracted from the maximum expected value from a decision analysis made under uncertainty, i.e. that takes uncertainty into account (ERJ). E V P I = E V W P I - E I U In this sense, E V P I indicates the total cost or value loss resulted from being uncertain. In the present analysis the expected value of perfect information is computed with both modeling approaches using the probabilities placed on each hypothesis or model, and the conjunct of decisions outlined above. For the single species approach, E V P I is computed for two types decisions: in the choice of an optimal exploitation rate; and in the choice of an optimal effort level. For the multi-species approach E V P I is computed for the choice of optimal fishing rates for sardine. The objective to be maximized in each case is the average catch in the time period specified. 5.3. Results and Discussion Single-species approach Simulations were run for 10 years, having as starting conditions the spawning stock biomass and environmental conditions hypothesized for the late 1970s. Results are shown in figures 5.4 to 5.7. Strategies of catching a constant proportion of the stock and allowing a constant escapement produce the highest yields among the strategies tested. Changes in stock catchability with stock size decrease considerably the performance of strategies of effort control. Similarly, constant catch strategies resulted in the lowest average yield among the strategies tested. The performance of constant escapement strategies are particularly sensitive to errors in biomass 137 estimates. The general effect is of decrease in average yield, increase in catch variability and in the chances of stock collapse with the increase in coefficient of variation of the estimation method (Figure 5.7). The adoption of more conservative escapement levels does not reduce the relatively high chances of stock collapse when the error in the estimation procedure is larger than 30%. Variations of this magnitude are commonly observed in acoustic estimation methods as a result of variations in the adjustment of ecosounders, changes in the acoustic properties of fishes, and the efficiency with which surveys cover the complete distribution area of the stock (MacLennan and Simmonds, 1992). Misund (1997), for instance, reported differences of ca. 100%) in the estimation of herring biomass obtained between surveys in the same year. Similarly, acoustic assessment of sardine biomass in the 1988 spawning season (Castello et al, 1991) produced a confidence interval for stock biomass between 38 and 77 thousand tons, corresponding to a coefficient of variation in the order of 30%o. Strategies of catching a constant proportion of the stock annually are less sensitive to errors in biomass estimates. The increase in the coefficient of variation of estimations has very little influence over the expected average yield, but increase the variability of catches and the chances of stock collapse (Figure 5.6). For a harvest rate of 0.5, the chances of bringing the stock to collapse increases from less than 1% to more than 50% when cv increases from 0 to 0.5. The expected yield under effort control peaks at ca. 350 effort units, with a probability of stock collapse of less than 20% (Figure 5.5). Figure 5.5 also represents the approximate fishing effort applied to the stock in the period from 1977 to 1983, which was very close to the predicted optimal effort level. Considering that since then the purse seine fleet has doubled in size and became more technologically equipped with sonar and power blocks, it is suggested that chances of stock collapse increased considerably during the following decade. 138 Expected Yield 0.2 0.1 to c o 0 0 0.1 0.2 0.3 Catch (tonnes. 106) Catch Variability 8 _ 6 x 3 4 > O 2 0 0 0.1 0.2 0.3 Catch (tonnes. 106) Probability of Collapse 1 0.8 ~ 0.6 \u00C2\u00B0- 0.4 0.2 0 0 0.1 0.2 0.3 Catch (tonnes. 106) Figure 5.4. Outcomes of fishing strategy that capture a constant catch of sardine as predicted by the single species model. 139 Expected Yield 0.25 i 0 200 400 600 800 1000 Effort Probability of Collapse Effort Figure 5.5. Outcomes of fishing strategies based on effort control as predicted by the single species model. The arrow indicates the approximate effort level in the period from 1977 to 1983. Effort index calculated as the product of number of trips and the mean tonnage of boats. 140 Expected Yield 0.4 i Probability of Collapse Harvest rate Figure 5.6. Outcomes of fishing strategies that harvest a constant proportion of the stock as predicted by the single species model. C V is the coefficient of variation of the biomass estimation procedure. 141 0 I i i i i I 0 0.1 0.2 0.3 0.4 0.5 Escapement (tons.106) Catch Variability 10 i 1 8 h 0 l i i i i l 0 0.1 0.2 0.3 0.4 0.5 Escapement (tons.106) Probability of Collapse Figure 5.7. Outcomes of fishing strategies based on a constant stock escapement as predicted by the single species model. C V is the coefficient of variation of the biomass estimation procedure. 142 Expected Value of Perfect Information In the present analysis the expected value of perfect information is computed for two types of decisions: in the choice for an optimal exploitation rate and for an optimal effort level. The objective is to find the fishing strategy (harvest rate or effort) that would maximize the yield from the fishery during 10 years after the collapse, i.e. after the early 1990's when the spawning stock biomass was about 70,000 tons. Uncertainties were included in three independent processes for each type of strategy (Table 5.3). Uncertainty on future environmental regimes represent a situation where decisions are made under complete ignorance on the environmental conditions and recruitment success in future years. Table 5.3. Decision table on the choice of harvest rate and effort level for the sardine fishery. Harvest strategy Uncertain processes/states Harvest rate h=0 to 1 Effort f=0to 1000 Stock production 3 hypotheses Stock production 3 hypotheses Estimated biomass Future env. reeime coefficient of variation of favorable or not estimations Stock catchability Future env. regime parameter uncertainty favorable or not Table 5.4 shows the calculated optimal fishing strategies, the correspondent average expected yield, and the expected value o f perfect information, which measures the expected gain to be obtained i f we could suddenly resolve all uncertainty about each o f the hypothesis is correct. Higher expected yields are predicted with constant harvest rate strategies, as has been previously suggested. Wi th constant harvest rates strategies the expected gain of reducing uncertainties is less than 5% of the expected yield obtained when decisions are made under uncertainty. Walters (1986) suggests three reasons why the value of learning, measured as a upper bound by E V P I , is often not as large as one would intuitively expect: first, because optimal policies for each of the various hypothesis considered need not to differ substantially from the policy that provides the maximum expected value; second, stock-recruitment hypothesis may predict nearly the same yield across a wide range of harvest policies; and finally, because the optimal harvest rate tend to be close to the optimal policy for those hypothesis that were assigned high initial probabilities. The latter reason is particularly unlikely in this case considering that the three stock production hypothesis were assigned equal probabilities. It seems that in this case uncertainties on 143 recruitment processes and in the estimation of stock biomass have very little influence on the choice for the optimal constant harvest rate strategy. The situation is quite different when effort control is employed. Uncertainties on stock catchability influence considerably the choice for the optimal effort level. The expected value of perfect information is in this case ca. 40% of what is expected to be obtained with a decision made under uncertainty. That indicates that research on the factors controlling stock catchability would cause a considerable improvement in the current management practice for the Brazilian sardine, more than that expected from oceanographic research aimed at increasing the ability to forecast future environmental regimes controlling recruitment. Table 5.4. Results of the analysis of the expected value of perfect information for harvest strategies for the Brazilian sardine after the collapse. Harvest strategy Best h or f Expected yield EVPI EVPI (tons.year\"1) (tons .year\"1) .(%) Harvest rate 0.4 208,015 7,800 3.75 Effort 150 85,467 34,886 40.81 Multi-species approach Results of simulations of the effect of fishing strategies on the biomass and catches of sardine are shown in tables 5.5 and 5.6. It is predicted that closing the purse seine fishery (scenario 2) w i l l lead to an average increase in sardine biomass ca. 6 times compared the baseline level of the late 1980s. Halving the current fishing rate (scenario 1) would lead to a biomass increase of ca. 2 times, and it is predicted to increase catch on average by 18% in five years (Table 5.6). Results vary however with the type of trophic control, being higher biomasses predicted to be recovered with top-down and wasp-waist control. In both cases releasing the fishing pressure would enable sardine to fully benefit from the available food by efficiently foraging on phyto and zooplankton. For a comparative analysis, results obtained with the single-species approach for the same short term policies are listed in table 5.7; for instance, stopping the purse seine fleet is expected to result in an increase in sardine biomass between 1.27 and 2.66 times the baseline biomass, depending on the environmental conditions prevailing during the 5 years. Table 5.5 and 5.6 also allow for a direct comparison of the effect of decisions in other fisheries (scenarios 4, 5 and 6) on the resulting sardine biomass and catches. For the rebuilding strategies 144 (4 and 5) results point to smaller biomass and catch than what is predicted under the purse seine only scenarios (1 and 2). On the other hand, doubling all fisheries has a smaller effect on sardine biomass than that predicted by increasing the purse seine fleet. These results illustrate the type of indirect effects expected when fish stocks are linked by predator-prey relationships, i.e., allowing the rebuilding of species at the top of the food web cause a reduction in the benefits expected from releasing fishing pressure on sardine, specially when the fisheries target species that directly affect sardine production, either by competing for the same food resources or by being a predator. The latter situation occurs with Bottom trawl and Pole-and-line fisheries, for harvesting adult weakfish and bonito which are predators of small forage fish in the model. However the most important sardine predator group in the ecosystem model (Other pelagic feeding fish) is not fished by any of the fleets considered in the simulations. Table 5.5. Relative change in sardine biomass after 5 year under different fishing scenarios. Values are relative to the Ecopath baseline level described in Chapter 3. F stands for the maximum relative feeding time of apex predators. Scenarios 1, 2 and 3 involve changes in the purse seine fleet only. Higher F Lower F Scenario Top-down Bottom-up Wasp-waist Top-down Bottom-up Wasp-waist Average 7 3~41 L51 2T16 3A1 L51 2A6 2.36 2 12.45 2.26 5.29 12.45 2.26 5.29 6.67 3 OJ 5 044 0.2_9_ __0A2 0.61 0.46_ 0.40_ 4 2.48 1.30 1.75 2.48 1.34 1.79 1.86 5 6.62 1.59 3.20 6.62 1.69 3.31 3.84 6 0.19 0.53 0.37 0.60 0.69 0.54 0.49 145 Table 5.6. Relative change in sardine catches after 5 years under different fishing scenarios. Values are relative to the Ecopath baseline level described in Chapter 3. F stands for the maximum relative feeding time of apex predators. Scenarios 1, 2 and 3 involve changes in the purse seine fleet only. Higher F Lower F Scenario Top-down Bottom-up Wasp-waist Top-down Bottom-up Wasp-waist Average 1 1/71 076 L08 TTl 0/76 L08 U 8 2 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 0.00 3_ 0.30 0.87_ 0.58 0.84_ 1_2_1 0.92_ 0/79 4 1.24 0.65 0.88 1.24 0.67 0.89 0.93 5 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 \u00E2\u0080\u0094 0.00 6 0.39 1.05 0.73 1.19 1.37 1.08 0.97 Table 5.7. Relative change in sardine biomass after 5 years under two strategies for stock recovery. Simulations Future, regime Stop fishing Halve fishing Bad 1.27 0.82 Good 2.66 2.20 Table 5.8 and 5.9 compare the predicted results of closing (scenario 4) and doubling (scenario 6) all fisheries on the biomass of the main harvested species in the ecosystem. Predicted changes in anchovy biomass are also shown. Closing all fisheries positively benefit all harvested groups with the exception of marine shrimps, anchovy and triggerfish. Conversely, doubling the current fishing rates negatively affect all groups in the system but marine shrimps and anchovy, which are predicted to have a net biomass increase in almost all trophic control hypotheses tested. Having small and fast growing groups such as marine shrimps benefiting from the overfishing of higher order consumers is not surprising given that reducing all predators in the system is expected to increase the abundance of preys, as has been the case with overfished tropical marine ecosystems (Pauly, 1979; Christensen, 1998). Also the higher turnover rates of marine shrimps allow them to withstand higher exploitation rates than that supported by slow growing fishes. The most affected groups are in both scenarios adult weakfish, rays and skates, and sardine. These species are either heavily fished (weakfish and sardine) or not being predated in the system (rays/skates and weakfish) so that a release in fishing w i l l lead to a prompt increase in biomass; or are slow growing species (rays and skates) so that an increase in fishing w i l l rapidly reduce their biomasses. Figure 5.8 illustrates these generic effects for a bottom-up control. 146 Table 5.8. Relative change in biomass of main harvested species in the Southeastern Brazilian Bight as predicted by Ecosim after 5 years with all fisheries closed. F stands for the maximum relative feeding time of apex predators. Shaded are groups that show a decrease in biomass. Groups Top down Higher F Bottom up Wasp-waist Top down Lower F Bottom up Wasp-waist Average Bonito 1.67 1.52 1.59 1.67 1.47 1.59 1.59 Adult Weakfish 14.45 4.53 4.57 14.44 4.50 4.54 7.84 Rays/Skates 5.01 4.14 4.13 5.01 4.14 4.13 4.43 Croaker 1.40 1.24 1.25 1.40 1.23 1.25 1.30 King Weakfish 1.27 1.48 1.50 1.27 1.41 1.46 1.40 Triggerfish 1.05 0.91 0.88 1.05 0.91 0.88 0.95 Marine shrimps 0.00 0.49 0.50 0.00 0.51 0.52 0.34 Anchovy 0.89 0.87 0.87 0.89 0.90 0.88 0.88 Sardine 6.62 1.59 3.20 6.62 1.69 3.31 3.84 Table 5.9. Relative change in biomass of main harvested species in the Southeastern Brazilian Bight as predicted by Ecosim after 5 years with all fishing fleets doubled. F stands for the maximum relative feeding time of apex predators. Shaded are groups that show an increase in biomass. Higher F Lower F Groups Top down Bottom up Wasp-waist Top down Bottom up Wasp-waist Average Bonito 0.63 0.65 0.65 0.64 0.66 0.66 0.65 Adult Weakfish 0.12 0.31 0.31 0.22 0.37 0.37 0.28 Rays/Skates 0.20 0.20 0.20 0.20 0.20 0.20 0.20 Croaker 0.69 0.78 0.78 0.73 0.80 0.80 0.76 King Weakfish 0.63 0.81 0.81 0.75 0.85 0.85 0.78 Triggerfish 0.27 1.02 1.02 0.33 1.02 1.02 0.78 Marine shrimps 0.37 1.39 1.38 1.26 1.51 1.51 1.24 Anchovy 1.07 1.14 1.15 1.36 1.22 1.26 1.20 Sardine 0.19 0.53 0.37 0.60 0.69 0.54 0.49 147 3x 1x LO CO o1/3x bo 0) CO 00 Stop fishing weakfish rays/skates \u00E2\u0080\u00A2triggerfish marine shrimps co E o in 3x 1x 1/3x Double fishing marine shrimps iggerfish bonito weakfish rays/skates 1 5 years Figure 5.8. Dynamic simulation of fishing scenarios where all fishing is closed (upper panel) and doubled (lower panel) in 5 years. Simulation ran under a bottom-up trophic control. Figure 5.9 compares the expected yield of sardine under constant fishing mortality rates. Results vary with the type of trophic control; higher yields are predicted with top-down control; top-down and wasp-waist control predict lower optimal fishing rates (F=0.3 year\"1) compared to bottom-up control (0.91.2 year\"1). Assuming lower maximum foraging time of apex predators result in the model predicting higher yields for fishing rates higher than the Ecopath baseline. This result has to do with the functional representation of consumption rates in Ecosim, where food consumption per time varies proportionally to the decrease in availability of resources and its is limited by a maximum foraging time. Decreasing the availability of preys (e.g. sardine) relative to the baseline level leads to an increase in the time apex predators spend feeding to balance the deficit in food intake; the end result being the increase in predation mortality when the stock is overfished. The higher the maximum foraging time the lower the yield obtained from the prey species. In one case, bottom-up control, the choice for the maximum foraging time directly influenced the predicted optimal fishing rates for sardine. 148 Ecopath baseline F Bottom-up Top-down \"Wasp-waist\" 0.5 1 t 1.5 F (year 1) Figure 5.9. Predicted average yield (10 years) of sardine under fishing mortality rates (F). Dashed line across graphs indicates the baseline F for sardine in the original Ecopath model (chapter 4). In the graphs the continuous line represents the predicted yield expected when the model allows for higher maximum relative foraging time of apex predators in the system. The dashed thin line represents results with lower maximum relative foraging time. Differences in production rates with each hypothesis do not imply however that much gain is expected to be obtained by reducing uncertainties on the operating type o f trophic control in the system; the expected value of perfect information (EVPI) indicates in this case a gain of only 149 7.36% over the optimal decision that takes uncertainty into account, which is to fish at a F of 0.3 year\"1. This value, which would also indicate an optimal harvest rate of approximately 0.3, is similar to that calculated with the single-species approach (h=0.4; Table 5.4) under a different set of constrains. The type of trophic control may however influence the transient responses of the ecosystem and consequently the output of a different set of fishing strategies. Predicted ecosystem response to overfishing sardine, for instance, raises concerns on the definition of minimum escapement levels and on the possible effects of pulse fishing strategies commonly adopted for small pelagics. Pulse fishing strategies, which basically rely on intensively fishing a stock to the point of overfishing and then let it recover before a new regime of intense fishing is initiated, have been applied to small pelagic stocks mainly as a result of the activity of Distant Water Fleets ( W W F , 1998). Some authors have in fact suggested that management of small pelagic stocks should be targeted to favor recoveries when favorable environmental conditions prevail, rather than try to prevent depletion which is normally accompanied by unfavorable environmental conditions (Beddington and May , 1977). The objective of promoting stock rebuilding is also intrinsic to escapement thresholds, defined by Quinn II et al (1990) as \"the population level below which the stock may be unable to rebuild its optimal level over an acceptable period of time\". Figure 5.10 exemplifies the type of response obtained in Ecosim when sardine is subjected to pulse fishing strategies. When the system is completely bottom-up controlled the stock is predicted to bounce according to the intensity of fishing, with minimal ecosystem-wide effects. When the system is top-down controlled, pulse fishing can however impose more drastic ecosystem effects and cause a delayed response of the sardine stock to the release in fishing pressure. 150 feiomass/original biomass 0 5 10 15 20 years Figure 5.10. Example of a pulse fishing strategy for sardine; upper panel refers to a top-down scenario, lower panel refers to a bottom-up scenario. A l l trophic groups are represented to show the scale of system wide effects of fishing under the two trophic control hypothesis. Conclusions from model simulations Policies o f effort control currently applied in the management o f the Brazilian sardine are particularly inadequate in avoiding stock collapse due to the combined effect of a large fleet capacity, changes in catchability with stock size, and the variability in production rates driven by environmental regimes. The prospect of reducing uncertainties in the output of harvest decisions, and in effect reduce the chances of stock collapse, for fisheries controlled by effort 151 limitation is still restricted by the proper understanding of stock catchability. That w i l l imply, in the case of the Brazilian sardine, that more effort should be put on evaluating the current fleet capacity, estimating the catchability of the stocks at contrasting stock sizes, and analyzing the population spatial dynamics (spatial range) and its interaction with the dynamics of effort allocation by purse seiners. In the latter context, two concurrent lines o f thought are currently developing. MacCal l ' s (1990) approach describes the spatial dynamics of schooling pelagic stocks according to a density dependent habitat selection model where a direct relationship between stock abundance and stock area is created by differences in habitat suitability (which is the result of environmental conditions combined with effects o f local competition). His hypothesis has some interesting corollaries useful for fisheries assessment, particularly that stock area could be a good indicator of the population state and thus could be used together with catch rates in the stock assessment. This approach has been successfully applied to the Pacific sardine, when the population was small and difficult to assess with conventional techniques (Barnes et al., 1992). A second approach (Pitcher, 1997) suggests that shoaling behaviour alone can cause range collapse in the absence of significant environmental gradients in space and time. If validated, this approach raises the prospect of obtaining cost-effective diagnostics of range collapse by monitoring behavioural parameters of shoaling fish. The consideration of stock spatial attributes into fisheries research w i l l require the gathering o f auxiliary information not only from surveys but also from the fishery activity such as time searching, density of schools, school size, inter-school distance, etc. The adoption of precautionary measures in the management of the fishery with effort control w i l l demand an effective reduction in the effort and fleet capacity at least to the level observed during the late 1970s. Also , strategies of effort control are more likely to succeed i f accompanied by auxiliary measures of control of the minimum age of recruitment to the fishery. Figure 5.11 shows the reproductive potential of the stock (measures as %SPR) for different combinations of age at first capture and the exploitation rate. When the first spawning age class (approximately 1.5 years) is effectively protected the total reproductive potential of a cohort is only slightly reduced with increasing fishing pressure. The combination of high natural mortality rate and early maturity makes the stock highly dependent on the first spawning age classes, which usually make the bulk of the reproductive capacity in many small pelagic stocks (Fig. 5.11c). In this sense protecting the first spawning age classes can help prevent early collapses of the stock, and encourage more rapid responses to favorable environmental conditions. Mace and Sissenwine 152 (1993) analysis of replacement thresholds indicated that to persist, i.e. for successive generations replace each other on average, small pelagic populations must maintain an average 40 to 60% of their unfished spawning per recruit (%SPR). This relatively high % S P R led the authors to infer low resilience of these stocks to fishing mortality, since a small reduction in SPR would compromise the future replacement of the stock. Clearly, exploitation must be lower in order to maintain the stock for lower ages at first capture (see 40 to 60% bands, F ig . 5.11a,b). But protecting the first spawning age class (1 to 1.5 years) provides the stock with high resilience, in the sense that a broader range of fishing mortality rates could be sustained without substantially diminishing the capacity of the population to react to favorable oceanographic events. For stocks with a dominant controlling influence by environmental regimes, that seems to be the most appropriate strategy that could be supported on biological grounds (Winters et al., 1985). Yet, such strategy is probably untenable under most conditions for pelagic species due to the difficulty in controlling the age at entry in the fishery. 153 0.9 0.8 0.7 %SPR 0.6 \u00E2\u0080\u00A2 80-100 \u00E2\u0080\u00A2 60-80 0.5 E \u00E2\u0080\u00A2 40-60 0.4 \u00E2\u0080\u00A2 20-40 0.3 \u00E2\u0080\u00A2 0-20 0.2 0.1 0 -0.9 -0.8 -0.7 %SPR -0.6 \u00E2\u0080\u00A2 80-100 \u00E2\u0080\u00A2 60-80 -0.5 E \u00E2\u0080\u00A2 40-60 -0.4 \u00E2\u0080\u00A2 20-40 -0.3 \u00E2\u0080\u00A2 0-20 -0.2 - 0.1 -0 50 40 30 20 10 0 \u00E2\u0080\u00A2 Accounting for A No age variation in A 0 20 40 60 80 %Total Spawning Potential 100 Figure 5.11. Percentage spawning per recruit (% SPR) for equilibrium exploitation rate (E) and age at first capture (A50o /o). A) taking into account the age specific reproductive output (A); B) i f reproductive output were constant with age; C) percentage contribution of sardine, Sardinella brasiliensis, first spawning age classes (ages 1 to 1.5) to the total reproductive potential of the stock. The y-axis represents the percentage of years from data compiled between 1979 to 1989 (Cergole, 1993). Better trade-offs between average catch, catch variability and the probability of collapse can be achieved with strategies o f catch control, such as the one obtained with a constant harvest rate 154 policy. Under this type of policy the expected yield of a best decision made under uncertainty on the productivity of the population is within a 10% difference from the expected result of a best decision made with complete knowledge about future environmental regime controlling recruitment or the type of control of trophic relationships. The predicted optimal harvest rates are very conservative compared to the ones usually applied to small pelagic stocks, but are consistent with the sustainable fishing rates defined by Patterson (1992). Optimal harvest rates between 0.3 and 0.4 are obtained according to the modeling approach used, and appear to be conditioned by two set of independent processes traditionally overlooked by fisheries assessment: in the single species approach conservative fishing rates result from the possibility of sub-optimal environmental conditions represented in the model by low frequency regimes in recruitment success. In the ecosystem model more conservative fishing rates are predicted for sardine when predation is a major mechanism controlling production at lower trophic levels. The response of the system to the interaction of both processes (trophic and environmental) still need to be evaluated. Constant harvest rate strategies are considered very robust strategies to cope with the inherent uncertainties created by climatic effects on marine fish populations (Walters and Parma, 1996). They are usually implemented by fishing control systems that rely on annual biomass estimates and on simple feedback rules that specify the proportion of the adult stock, or the total allowable catch, to be harvested each year. The success of catch control systems is however dependent on the accuracy of the stock assessment which often suffers from large uncertainties in parameters and variables (e.g. catch at age, relative index of abundance) used in the estimation procedure. The critical information for the success of catch control systems is therefore the frequency and accuracy of stock abundance estimates, which may call upon a combination of data from surveys (e.g. acoustic assessments of spawning biomass and recruitment), better monitoring of catch composition (age and size) used in virtual population analysis, tagging experiments, and may as well rely on the active participation of resource users in data collection (Walters and Pearse, 1996). Improvements in the understanding of biological and oceanographic processes controlling sardine production seems mostly needed i f the fishery is to be managed by escapement thresholds. The possibility of depensation at low spawning stock sizes complicates the definition of escapement policies for increasing the risk of shifts in stability domains. Environmental effects can have in this case either a positive or a negative impact by pushing the overfished stock in or out of a stability domain, or even by shifting the boundary o f stability domains. In 155 this situation a precautionary strategy that aims to diminish fishing and allow the stock to recover may not be very informative since the recovery of the stock might be interpreted both as a result of reduced fishing mortality or due to the occurrence of better environmental conditions; on the other hand, the failure of the stock to recover might be interpreted both as a result of depensation or due to a less favorable environmental regime. The trophic model provides in this case a tool to test and formulate different hypotheses about the possible causes of depensation and multiple ecosystem stability domains, such as the effect o f increasing predation mortality with decreasing stock size. The testing of these hypotheses in the field w i l l require data on recruitment, juvenile fish survival and abundance of potential predators; yet, in most cases the correct identification of mechanisms w i l l be problematic due to the confounding effect o f other environmental factors on juvenile survival. In the model of the Southeastern Brazilian shelf ecosystem the Pelagic feeding fish group is the most important sardine consumers. Pelagic feeding fish is, however, one of the less documented trophic components in the system. The group was originally included in the model to represent demersal fish species that actively feed on the pelagic system, such as weakfish and hake (Soares et al., 1993; Rocha et al., 1998), but may as well include a diversity of other demersal and pelagic species also targeted by fisheries and with important role at the top of the food web. A short list o f other pelagic feeding fish species include sharks, cutlass fish, Trichiurus lepturus, and pelagic predators such as Auxis spp. and Sarda sp. A better characterization of this group, which w i l l require data on biology and feeding habits, may lead to a more complete and accurate characterization of the effect o f fisheries on the southeastern shelf ecosystem. The evaluation of harvested strategies for the Brazilian sardine considered only the biological and ecological trade-offs involved between two types of widely applied fishing strategies, namely input (effort limitation) and output (catch limitation) controls. It was out of the scope of this work to discuss the likely socio-economic consequences of adopting one or other type of strategies and controls, neither was the objective of the analysis to examine all the possible combinations of strategies and controls for this fishery. Nonetheless, the results obtained here are expected to enrich this discussion which w i l l require, to be effective, the active participation of resource users. 156 On the choice of a modeling approach A central question for fisheries assessment today is on how to evaluate and communicate the consequences of alternative fishing policies to marine resources and ecosystems. For a quantitative analysis the question consequently concerns the choice for the modeling approach. Models have many possible purposes and uses, and no one model is right for the entire range of uses. In principle, the choice for which type of approach to be used depends on the type of products or outputs it is expected from the analysis, i.e., on what type of information is needed for deciding among harvest decisions. Costanza et al. (1993) suggests three criteria forjudging model performance: realism (simulating system behavior in a qualitatively realistic way); precision (simulating behavior in a quantitatively precise way); and generality (representing a broad range of system's behavior with the same model). No single model can do well on all three of these criteria, and the choice of which objective to pursue depends on the fundamental purposes of the model. In modeling complex systems tradeoffs must be made among realism, precision, and generality. For instance, when seeking for generality, models must give up some realism and/or precision. High precision (quantitative correspondence between data and model) w i l l often sacrifice realism and generality. When the goal is to develop realistic assessments of a system, generality and precision must be relaxed. High realism models are concerned with accurately representing the underlying processes in a specific system, rather than with precisely matching quantitative behavior or being generally applicable. Costanza et al. (1993) argued that in many types of system modeling, the desired outcome is to accurately determine the overall magnitude and direction of change, trading off realism for some moderate amount of generality and precision. Conventional fisheries stock assessment models strive for quantitative precision when describing time series of population abundance, calculating catch quotas, or predicting the consequences of policy options. Cochrane (1998), for instance, pointed to the need of an \"ecosystem operating model\" in management procedures for multi-species resources. A management procedure is defined as a set of rules which specify how a management recommendation is set and what data are used for this purpose (Butterworth et al., 1997). Rules are selected based on their anticipated 157 performance as estimated by simulation on an operating model of the resources and fishery. Successful implementation of formal management procedure for multi-species or ecosystem management w i l l require from the operating model (Cochrane, 1998): precision, so that the expected system response to a management strategy is quantitatively similar to that predicted by the model; realism in simulating adequately the fishery-resources interactions; the incorporation of all sources o f uncertainties to enable robust forecasts of ecosystem responses to a management strategy; and the output of meaningful performance criteria. A model with high precision would be able to describe perfectly wel l the quantitative changes in the variables of interest with changes in controls and strategies. Ideally, to the test the precision of the model it would be necessary to contrast predicted and observed values. Alternatively, we can compare the consistency of quantitative predictions of changes in sardine biomass obtained with the two modeling approaches in the same scenario. For instance, according to the single species model closing the purse seine fishery for five years is expected to cause an increase in sardine biomass between 1.27 and 2.66 times (Table 5.7). The same scenario in the ecosystem model produces an average increase in sardine biomass of 6.67 times the baseline level. A greater consistency between models is obtained when the ecosystem model is run under a bottom-up control of trophic relationships (sardine biomass 2.26 times higher than the baseline level; Table 5.5). Walters et al.(1997) also noted that Ecosim predictions approach that of a single-species model i f the trophic model is bottom-up controlled. In this sense, the precision of short term predictions may be quite similar between approaches, depending on the weight put on the trophic control hypothesis in the ecosystem model. Realism varies between approaches. The Ecopath/Ecosim model provides a more realistic description of the resource and ecosystem by representing not only the characteristics of the sardine stock but also the trophic interactions with other ecosystem components, and the possible interactions between fishing fleets. The ecosystem model however still lack a proper representation of important fishery-resource interactions such as the effect of changes in gear selectivity, which is often one of the most applied policy variables in fisheries management, and the effect of environmental factors on the availability and recruitment success of keystone species in the ecosystem. Experience with the use of the approach to explain the ecosystem changes observed in the Bering Sea suggested for instance that environmental factors, affecting recruitment or primary production, may be more important in determining the dynamics of the 158 ecosystem than predator-prey interactions alone (Trites et al., 1999). Lack of realism in representing important processes and uncertainties may in this sense apply for both modeling approaches, as models w i l l always involve some degree of simplification of the observed processes in nature. The critical question is therefore not which model is more realistic, but which level of simplification imposed when using a model is less adequate to the type of problem and policy variables in hand. When the objective of quantitative modeling is prediction, very simple models very often outperform more complex models. For instance, Ludwig and Walters (1985) showed that non-age-structured models produce better predictions of management actions than detailed age structured models. A s a rule, the increase in model details makes it more difficult to specify how the components functionally interact, and each additional model parameter becomes less well specified by the available historical data. Walters (1986) showed that uncertainty about a policy parameter is l ikely to be minimized, for a fixed data set, by basing its calculation on parameter estimates from a model of intermediate complexity. The increase in model complexity decreases the prediction error (increase the ability to fit historical data) but increases uncertainty on parameters used to calculate the policy of interest. In summary, the choice for model complexity depends basically on the purpose of the model and on the amount of available information, i.e., i f the model is too simple there is a risk of lacking realism, whereas i f the model is too complex there w i l l be not sufficient information in data to distinguish between the possible parameter values of the model. Also , the increase in model complexity should never compromise the transparency of arguments and conclusions of the model, and ideally, should be preceded by the analysis of its likely contribution to the final qualitative argument (Walters, 1986). Therefore, prediction of the dynamics of multi-species systems or ecosystems makes single-species models of little use, but detailed models involving all important species may not be the solution either, since the sensitivity of complex models to errors in parameters needed for their construction w i l l make them unreliable (May, 1984; Hilborn and Mangel, 1997). From a manager's point of view, the difference between the analysis o f scientific advice based on single or multi-species models may be irrelevant since the questions faced by managers concerning the quality of science and the political acceptance of regulations are exactly the same, regardless of the modeling approach (Brugge and Holden, 1991). In this respect, Brugge and Holden suggest situations where the use of multi-species models w i l l tend to go wrong. Among 159 the possible operational reasons for the failure are first, when recommendations based on multi-species models differ radically or are totally contradictory to those proposed in the immediate past with single species models. A s stated by Gulland (1991), new scientific advice is most likely to be used i f it implies changes in current management practice that are straightforward and preferably minor. Second, i f the use of multi-species models cause an increase in the complexity of both management and assessment. Complexity in the management system is likely to increase i f the conflicting interests are to discuss how the relative abundance of the multiple stocks should be changed. A s for assessment, data requirement w i l l most likely increase with multi-species models. Finally, multi-species models are inclined to fail i f the lack of data force extreme simplifications in overall system structure and process, thus compromising realism (an argument also constantly used against single-species models). The achievement of a comprehensive understanding that is useful for both realism and prediction w i l l perhaps call upon the integration of the different approaches. Not that the \"truth w i l l lie in the intersection of independent lies\" (Levins, 1966), but that the combination of approaches can provide complementary valuable results. Multi-species ecosystem models produce a more complete caricature of the system, explicitly recognizing its major components and processes, which enable us: i) to visualize the broad consequences of fishing policies (e.g. removing all the prey w i l l impact the predators) and hence provide guiding principles in which to ground ecosystem goals when defining conservative exploitation rates, catch quotas, etc.; and ii) to test and formulate hypotheses about the causes of the observed changes in marine fish populations, and about the functioning of marine ecosystems. The latter is particularly relevant in the scrutiny of research questions to which resources are to be allocated to improve the understanding of the processes affecting fish populations. The simpler structure and data requirement of single-species models, on the other hand, make them particularly useful for prediction of policy variables and monitoring purposes in the management of fisheries for single stocks, as is the case with most small pelagic fisheries. A n example of the use of a combination of approaches to fisheries management is found in the case of the Pacific herring off the west coast of Canada (Vandermeulen, 1998). From the ecosystem point of view, herring is an important source of food for higher trophic level species and act as an indicator of habitat loss due to their need for relatively pristine nearshore areas for spawning. The measure and monitoring of herring spawning biomass (undertaken with a single-species model) is thought to provide an indicator of overall marine ecosystem health and the effects of decisions regarding the protection of 160 nearshore habitats. In the case of the Brazilian sardine the trophic model provided the means to evaluate alternative hypothesis about the causes of species switches and changing stability domains that might have followed the stock collapse, besides indicating reasons other than environmental variability to the adoption of more conservative fishing rates for the species in the shelf ecosystem. 5.4. Summary This chapter assess the ecological risks of management decisions in the sardine fishery off southeastern Brazi l with currently available information and according to two modeling approaches: a single-species and a multi-species model. It is evaluated the short and long term predictions o f the impacts of harvest strategies and controls, and the relative values of reducing uncertainties on ecological processes. Better understanding of the processes controlling fishery-resource interactions are mostly needed i f the fishery is to continue being managed by effort control, or i f escapement thresholds have to be defined when adopting catch control strategies. A n alternative set of catch control strategies, particularly feedback rules for a constant harvest rate provide better trade-offs among the ecological indicators tested and appears robust to uncertainties on the prevailing ecological processes controlling production. Consistency between modeling approaches varies according to the weight put on the hypotheses represented in each model. A complementary role is suggested for the different modeling approaches in order to balance realism and prediction. Results are used to recommend on the type of research that would most likely provide the information needed to improve the quality of decisions, and on the precautionary measures that should be adopted in face of ecological uncertainties. 161 Chapter 6. Conclusions Marine fisheries have historically exploited the abundant forage fish resources as source of food and fishmeal. Small pelagics have increased in importance for world fisheries particularly during the last decades as a result of the depletion of important stocks of long-lived and high trophic level fish and the rapid increase in the demand for fishmeal for the feed industry. The exploitation pattern of capture fisheries has caused changes in marine ecosystems, evident as major shifts in the composition of species, which raises concerns on the limits of ecosystems carrying capacity to fisheries. Confronted with this reality, fisheries scientists have been faced with the need to evaluate and communicate the ecosystem consequences of fishing given an incomplete understanding of the complex dynamics of ecosystems. The objectives of this thesis were threefold: i) to evaluate the use of trophic models in the analysis of the ecosystem responses to fishing small pelagic forage fish; ii) to diagnose the fisheries in Brazi l to the fishing down marine food web phenomenon; and i i i) to provide an in-depth analysis of the sardine fishery off southeastern Brazi l to examine the hypotheses of ecosystem changes and stock collapse, and to evaluate how uncertainties about ecological processes influence the choice of harvest strategies and controls in this fishery. 6.1. Evaluation of ecosystem responses to fishing using dynamic trophic models. The comparative analysis of the ecosystem responses to fishing forage fish using trophic models indicated that small pelagics can play a central role in upwelling ecosystems, changes in their abundance can have considerable consequences to species at the top and the bottom of the food web. Also , as 'wasp-waist\" species in these ecosystems, small pelagic forage fish w i l l sustain much more conservative exploitation rates than what has been historically applied in the cases of stock collapse. Predicted results are generally in accordance with observed qualitative changes in upwelling ecosystems with the collapse of a forage species, which were normally followed by a decrease in the abundance of top predators and an increase in the abundance of a competing mid-trophic level species. Predictions of optimal fishing rates for the species are also generally in agreement with the suggested sustainable fishing rates for small pelagics, but are very sensitive to the type of trophic control (bottom-up or top-down) between predators and prey. 162 Model predictions are generally sensitive to the type of dominant trophic control assumptions and also appear influenced by the quality of the data used in the mass-balance assessments. A s a general rule, poor quality models w i l l perform badly in simulations run under top-down control, when models w i l l tend to self-simplify their structure through competition or predatory exclusion of some groups. The trophic model also allowed the testing of hypotheses about the functioning of ecosystems, their stability and resilience when impacted by fisheries. Comparing the recovery time of systems disturbed by fisheries with the structural, energetic and homeostatic characteristics of selected models it was possible to test hypotheses about the chief mechanisms controlling the stability of marine ecosystems. Results were in agreement with Odum's theory of ecosystem development, which suggests that nutrient recycling is one the main mechanisms developed by biological communities to increase control or homeostasis with the physical environment and achieve protection from its perturbations. Systems with a high percentage of the energy recycled are usually characterized by higher stability, or ability to recover from perturbations. In this analysis models of upwelling ecosystems were by far the most unstable systems, lacking well structured recycling routes to cope with perturbations imposed on the 'wasp-waist' forage species. A second type of stability, that related to the resilience of ecosystems or their ability to respond to perturbations without shifting to a different stability domain, was analyzed when exploring the effects of fishing sardine under different assumptions about the type of trophic control and the feeding time of apex predators in the system. Results indicate the possibility of depensation in sardine recruitment rates and consequent multiple stability domains in systems characterized with strong non-linear responses of predators to changes in prey densities. In these systems resilience appears to be conditioned by thresholds in the abundance of a forage species (sardine), beyond which disturbance w i l l lead to completely different system stable states. Another type of prediction explored with the trophic model relates to the long term effects of fishing down marine food webs on systems that have historically sustained catches of high trophic level species but that present the potential to sustain higher yields from unexploited forage species. Simulations showed that fishing down the food web for small pelagic planktivorous fishes, while at first increasing catches in intensively exploited regions, has the potential of actually decreasing yields by interrupting major energy pathways to already overexploited, high trophic level species. The diagnosis of this generic effect, described by a 163 backward bending curve between catches and trophic level of fisheries, corroborates global assessments of fisheries-induced changes in marine ecosystems where fishing down the food web is an observed phenomenon. The trophic model was also used to investigate the ecosystem effects of contrasting exploitation scenarios for the major fisheries resources in the southeastern Brazilian shelf ecosystem. Scenarios describing future options for managing fisheries in the region focused on the ecological effects of allowing the recovery of stocks or intensifying even more fishing by the main fleets. The predicted qualitative changes show that increasing exploitation w i l l lead to a decrease in biomass of high trophic level species and an increase in the importance of low trophic level species in the ecosystem. Conversely, allowing the recovery of stocks w i l l lead to major gains in abundance of commercially harvested species, but shrimps which are predicted to decrease in abundance. Predicted changes in sardine biomass were tested for consistency with the values obtained with a single-species model under the same scenarios. Biomass recovery values obtained with the trophic model converges to that predicted with the single-species model when the system is bottom-up controlled. However, simulations run under a bottom-up control produce very optimistic fishing mortality rates for sardine, usually above that predicted with the single-species model. Consistency between optimal fishing rates predicted with both models occurs when uncertainties on the type of trophic control are taken into account in the multi-species model. The models therefore justify the need to adopt more conservative fishing rates for the species by two interdependent reasons: in the single-species model conservative fishing rates result from the possibility of sub-optimal environmental condition controlling recruitment success in the future, while in the multi-species model conservative fishing rates are predicted for sardine when predation is considered as a major mechanism controlling production at lower trophic levels. A complementary role is suggested for the different modeling approaches in order to balance realism and prediction. Multi-species models produce a more complete caricature of the ecosystem which is. useful to examine the broad consequences of fishing policies and to test and formulate hypotheses about the causes of observed changes in marine ecosystems. On the other hand, the simpler structure and data needs of single-species models make them particularly useful for prediction of policy variables and monitoring purposes in the management of fisheries for single stocks, as is the case with many small pelagic fisheries. 164 The analysis of the ecosystem impacts of fishing small pelagics using dynamic trophic models (Ecosim) can be expanded in several ways. A question that still needs to be evaluated is how the impacts of fishing small pelagics compares with those resulting from exploiting other groups/species in the ecosystem. A systematic comparision o f the predicted ecosystem effects among exploited groups may provide a better characterization o f the ecological role of these species in marine food webs. Another question to be evaluated is how sensitive are the predicted optimal fishing mortality rates of small pelagics to parameter uncertainties other than the trophic control hypotheses tested in the thesis. Other candidate parameters for analysis are the productivity (P/B) of small pelagics, the rate of effective search and consumption rates of predators in the system. Improvements of the trophic model as a tool to test and formulate hypotheses about the ecosystem effects of fishing w i l l involve, in the specific case of the southeastern Brazilian Bight, a better representation of groups at the top of the food web, such apex fish predators, marine mammals and birds. Also , improvements in the quality and credibility of the model w i l l occur i f other scientists and local fisheries managers are fully involved in data input and evaluation of model structure. In general, model predictions are expected to be improved with prior information (hypotheses) on how trophic interaction strengths (v^) are distributed in the food web, and how environmental processes other than predation affects recruitment rates of individual trophic groups. That is expected to provide better insights on the dynamics of resilience of marine ecosystems when impacted by fisheries. 6.2. Fishing down marine food webs in Brazil In line with some of the trends in world fisheries, marine capture fisheries in Brazi l are in a state of crisis, with many of the traditionally harvested fish stocks either fully exploited or overexploited. Today, the prospect of increasing catches and recovering the status of fisheries activities rely on better management of the overfished stocks, fishing for offshore resources currently moderately exploited, and/or fishing down the food web for abundant short-lived, planktivorous fishes. Fishing down the food web is not yet an observed phenomenon in Brazi l ; fisheries in Braz i l had a relatively constant mean trophic level of the species landed from 1950 to the early 1980s, but show a recent increase in mean trophic level caused by the combined effect of the collapse of small and mid-size pelagic species (mostly sardine) and the increasing landings of large pelagic fishes with the development of offshore fisheries for tunas and sharks. 165 Characteristic of Brazi l is however the wide diversity of marine ecosystems and types of fishing strategies adopted regionally. In the southernmost region of the country, where the shelf ecosystem is under the influence of the subtropical convergence of Brazi l and Malvinas currents, fisheries have targeted mostly high trophic level stocks of demersal and pelagic species. In the upwelling ecosystem off the Southeastern Brazilian Bight sardine has been historically the main target of commercial fisheries, and sustained the largest fishery in the country. The collapse of this fishery in the late 1980s has apparently shifted the ecosystem into one dominated by an abundant population of anchovy, Engraulis anchoita, that is not commercially harvested. The extent to which the collapse of sardine stock and the switch to an anchovy dominated system was the result of fishing or environmental factors is still inconclusive. It was therefore the focus of this case study to examine hypotheses of ecosystem changes and to analyze how uncertainties on the ecological processes, as captured by different modeling approaches, influence the choice of harvest strategies and controls in this fishery. 6.3. The case of the sardine fishery off southeastern Brazil The analysis of hypotheses of ecosystem changes in response to fishing impacts relied on the identification of ecological processes controlling the dynamics of ecosystem succession and resilience. The ecosystem of the Southeastern Brazilian shelf is structured around the operation of physical-biological cycles of different speeds. Among the most documented cycles are the seasonal upwelling of the cold, nutrient-rich South Atlantic Central Water on the inner shelf, and the decadal regimes determined by multi-year variability in the intensity of physical forces controlling biological production. The latter is particularly important for the management of fisheries resources because it operates at time scales commensurate with the life span of most marine fish populations. A central question in the case study was therefore whether the collapse of the sardine stock and the apparent switches in species composition represent a natural and reversible change caused by oceanic regimes, or whether they reflect a loss of resilience of the ecosystem and a change of state caused by excessive exploitation. Literature review, model simulations and stock-recruitment data for the Brazilian sardine pointed at three generic hypotheses that could be used explain the changes in the structure of the ecosystem and the collapse of the sardine fishery. The overfishing hypothesis considers that fishing was the chief mechanism responsible for the collapse of the stock, so that once the intensity of fishing is 166 reduced the stock w i l l recover its original abundance. The depensation and ecosystem shift hypothesis considers that the stock declined in response to overfishing, but the recovery is hampered by depensation in recruitment rates and a change in state o f the ecosystem caused possibly by ecological and behavioral processes. Finally, the regime shift hypothesis considers that the stock collapsed in response to overfishing and recruitment failures caused by long-term low frequency environmental effects. This hypothesis accepts that radical fluctuations in abundance may be an intrinsic feature of small pelagics inhabiting upwelling ecosystems. If so the recovery o f the stock is less dependent on managing the fishery, but on favorable environmental conditions prevailing in the future. In summary, the combined effect o f both human and natural effects make it difficult to characterize the dynamics of the population and ecosystem, and to predict the results of rehabilitation measures for the stock. The combination of a third process, result of the interaction between the behavior of the fish and the dynamics of the fleet, further contributes to degrade the performance of current regulatory measures for stock rebuilding using effort control. On the other hand, the explicit recognition of uncertainties about the ecological processes controlling the dynamics of interactions between fisheries and ecosystem may allow for a sound choice of precautionary strategies for the fishery, and a better scrutiny of research programs in which to allocate resources to improve management. Policies of effort control currently applied in the management o f the Brazilian sardine are particularly inadequate in avoiding stock collapses due to the combined effect o f a large fleet capacity, changes in catchability with stock size, and the variability in production rates driven by environmental regimes. The prospect o f reducing uncertainties about the output of harvest decisions, and in effect reduce the chances of stock collapse, for fisheries managed with input control (e.g. effort limitation), is still restricted by the proper understanding of stock catchability. In the current situation the adoption of precautionary measures in the management of the sardine fishery w i l l imply an effective reduction in the effort and fleet capacity at least to the level observed during the late 1970s, and a more effective control of the minimum age of fish recruiting to the fishery, which has been shown to increase the resilience of the stock to overexploitation. A n alternative set of output control strategies (e.g. catch limit), particularly feedback rules for a constant harvest rate, provide better trade-offs among the ecological indicators tested and appears robust to uncertainties on the prevailing ecological processes controlling production. On the 167 other hand, the performance of another type of output control strategy based on minimal escapement levels deteriorates considerably in response to errors in the estimation of stock biomass. Also , the definition of escapement thresholds is complicated by the possibility of depensation in sardine recruitment rates at low spawning stock sizes and the risk of causing shifts in ecosystem stability domains. Consequently improvements in the understanding of biological and oceanographic processes controlling production and in the accuracy of biomass estimation methods seems mostly needed i f the fishery is to be managed by escapement thresholds. The implications o f this work to fisheries management were discussed in the individual chapters but w i l l be summarized here. A s 'wasp-waist' species in upwelling ecosystems, small pelagics should not be managed in isolation from other ecosystem components since changes in their abundance can have consequences throughout the food web. Their ecological role in marine food webs determines that small pelagics w i l l sustain much more conservative fishing rates than what has been historically suggested and in effect applied in the cases of fisheries collapse. The upwelling ecosystems inhabited by small pelagics have characteristics of unstable systems, lacking well developed biological mechanisms to dump the impact of fishing and environmental factors. The characteristics of unstable systems have served as a reason for the adoption of less restrictive fishing policies for the species, and helped to perpetuate management actions that continuously failed to allow the effective recovery of collapsed stocks. Conversely, seen from another perspective, the unstable characteristics of upwelling systems w i l l mean that variability imposed by nested physical-biological cycles is a key asset for the resilience of these ecosystems, and that management should in turn be prepared to cope with these dynamic conditions. In this respect fisheries that adopt more conservative exploitation rates for the species w i l l be more robust to the variability and ecological uncertainties inherent in these systems, and are more likely to be successful in the long run. 168 Notes 1. Part of Chapter 2 is published in Vasconcellos, M.; Mackinson, S.; Sloman, K. and D. Pauly. 1997. The stability of trophic models of marine ecosystems: a comparative analysis. Ecological Modeling 100: 125-134. 2. Chapter 3 is based on a manuscript in press in Fisheries Research: Vasconcellos, M. and M. Gasalla. Fisheries catches and the carrying capacity of marine ecosystems in southern Brazil. 169 Bibliography Aidar, E . ; Gaeta, S.A.; Gianesella-Galvao, S. M . F.; Kutner, M . B . B . and C. Teixeira. 1993. Ecossistema costeiro subtropical: nutrientes dissolvidos, fitoplancton e clorofila-a e suas relacoes com as condicoes oceanograficas da regiao de Ubatuba, SP. In Pires-Vanin, A . M . S. (Ed.) Estrutura e funcao do ecossistema de plataforma continental do Atlantico Sul brasileiro. Publicacao esp. Inst, oceanogr. S Paulo, Sao Paulo, n. 10, 245 p. Alheit, J. 1986. Eggs canibalism versus egg predation: Their significance in anchovies. ICES Council Meeting, 1986. 9 p. Alverson, D . L . ; Freeberg, M . H . ; Murawski, S. A . and J. G . Pope. 1994. A global assessment of fisheries bycatch and discards. F A O Fisheries Technical Paper 339, 233 p. Araujo, F . J. 1984. Habitos alimentares de tres especies de bagres marinhos (Ariidae) no estuario da Lagoa dos Patos, R S , Brasil . Atlantica (Rio Grande) 7:47-63. Armstrong, M . J. and P. A . Shelton. 1988. Bias in estimation o f stock-recruit function parameters caused by nonrandom environmental variability. Can. J. Fish. Aquat. Sci . 45: 554-557. Arreguin-Sanchez, F . ; Seijo, J. C. and Valero-Pacheco E . , 1993a. A n application of E C O P A T H II to the north continental shelf ecosystem of Yucatan, Mexico. In: V . Christensen and D . Pauly (Editors), Trophic Models of Aquatic Ecosystems, I C L A R M Conference Proceedings, 26: 269-278. Arreguin-Sanchez, F . ; Valero-Pacheco, E . and Chavez, E . A . , 1993b. A trophic box model of the coastal fish community of the Southwestern G u l f of Mexico . In: V . Christensen and D . Pauly (Editors), Trophic Models of Aquatic Ecosystems, I C L A R M Conference Proceedings, 26: 197-205. Baird, D . ; McGlade, J . M . and Ulanowicz, R. E . , 1991. The comparative ecology of six marine ecosystems. Phi l . Trans. R. Soc. Lond., 333:15-29. Bakun, A . 1992. Global greenhouse effects, multi-decadal wind trends, and potential impacts on coastal pelagic fish populations. ICES Mar. Sci . Symp. 195: 316-325. Bakun, A . 1996. Patterns in the Ocean. Ocean processes and marine population dynamics. California Sea Grant, 323 p. Bakun, A . 1998. The C E O S comparative analysis framework: Motivations and perceived opportunities, p. 7-26 In Durand. M . H . ; P. Cury; R. Mendelssohn; C. Roy; A . Bakun and D . Pauly (Eds.) Global versus local changes in upwelling systems. Editions de l'Orstom, Paris, 594 p. Bakun, A . and R. Parrish. 1990. Comparative studies of coastal pelagic fish reproductive habitat: The Brazilian sardine {Sardinella aurita). J. Cons. Int. Explor. Mer . 46: 269-283. Bakun, A . ; Beyer, J.; Pauly, D . ; Pope, J. G . and G . D . Sharp. 1982. Ocean sciences in relation to l iving resources. Can. J. Fish. Aquat. Sci . 39: 1059-1070. 170 Bakun A . ; Cury, P.; Durand, M . FL; Mendelsson, R.; Pauly, D . and C. Roy. 1998. The Climate and Eastern Ocean System Projects (CEOS), p. 1-7 In Durand. M . FL; P. Cury; R. Mendelssohn; C. Roy; A . Bakun and D . Pauly (Eds.) Global versus local changes in upwelling systems. Editions de l 'Orstom, Paris, 594 p. Barnes, J. T.; MacCa l l , A . D . ; Jacobson, L . D . and P. Wolf. 1992. Recent population trends and abundance estimates for the Pacific sardine (Sardinops sagax). C a l C O F I Rep. 33:60-74. Baumgartner, T. R., Soutar, A . and V . Ferreira-Bartrina. 1992. Reconstruction of the past history of Pacific sardine and northern anchovy populations over the past two millenia from sediments of the Santa Barbara Basin, California. C a l C O F I Rep. 33:24-40. Bax, N . J. 1998. The significance and prediction of predation in marine fisheries. ICES J. Mar. Sci . 55(6):997-1030. Beamish, R. J. 1995. (ed.) Climate change and northern fish populations. Canadian Special Publication in Fisheries and Aquatic Sciences 121, 739 p. Beddington, J. and R. M . May . 1977. Harvesting populations in a randomly fluctuating environment. Science 197:463-5. Begon, M . ; Harper, J. L . and Townsend, C. R., 1990. Ecology: Individuals, Populations and Communities. Blackwell Scientific Publications, London, 929 p. Berger, W . H . ; Smetacek, V . S. and G . Wefer. 1989. Ocean productivity and paleoproductivity -A n overview, pages 1 - 35 In W . H . Berger, V . S. Smetacek and G . Wefer (Eds.) Productivity of the Ocean: Present and Past. John Wiley & Sons, 471 p. Beverton, R . J .H . 1990. Small marine pelagic fish and the threat of fishing; are they endangered? J. Fish. B i o l . 37 (Suppl A ) : 5-16. Blalock, H . M . 1972. Social Statistics. M c G r a w - H i l l Company, New York . 583 p. Botsford, L . W. ; Castilla, J. C . & C. H . Peterson. 1997. The management of fisheries and marine ecosystems. Science 277: 509-514. Brandini, F . P. 1990. Primary production and phytoplankton photosynthetic characteristics in the Southeastern Brazilian coast. B o l m Inst, oceanogr., S. Paulo, 38(2): 147-159. Browder, J. A . , 1993. A pilot model of the G u l f o f Mexico continental shelf. In: V . Christensen and D . Pauly (Editors), Trophic Models of Aquatic Ecosystems, I C L A R M Conference Proceedings, 26: 279-284. Brugge, W.J . and M . J. Holden. 1991. Multispecies management: A manager's point of view. ICES Marine Science Symposia 193: 353-358. Butler, J. L . 1991. Mortality and recruitment of Pacific sardine, Sardinops sagax caerulea, larvae in the California Current. Canadian Journal of Fisheries and Aquatic Sciences, 48 (9): 1713-1723. 171 Butler, J. L . and D . Pickett. 1988. Age-specific vulnerability of Pacific sardine, Sardinops sagax, larvae to predation by northern anchovy, Engraulis mordax. Fishery Bulletin, 86(1): 163-167. Butterwoth, D . S.; Cochrane, K . L . and J. A . A . De Oliveira. 1997. Management procedures: a better way to manage fisheries? The South African experience. In Pikitch, E . K . ; Huppert, D . D . and Sissenwine, M . P. (Eds.) Global trends: Fisheries Management (Proceedings of the Symposium held at Seattle, Washington, 14-16 June 1994). A m . Fish. Soc. Symp. 20, 83-90. Caddy, J. F . and J. A . Gulland.1983. Historical patterns off ish stocks. Marine Policy 10/83: 267-278. Castello, J. P. 1992. Conservacao de recursos marinhos vivos no Brasil . In: Agostinho e Benedito-Cecilio (Eds). Situacao atual e perspectivas da ictiologia no Brasil . U E M -N U P E L I A - S B I ; p. 86-96. Castello, J.P. 1997. Pelagic teleosts, Pages 123 -128, In U . Seeliger, C. Odebrecht and J. P. Castello (Eds). Subtropical Converegence Environments. The coast and Sea in the Southwestern Atlantic. Springer, Berlin, 308 p. Castello, J. P. & M . Haimovici . 1991. Simposio da F U R G sobre pesquisa pesqueira: comentarios e recomendacoes. Atlantica 13(1): 5-10. Castello, J. P.; Habiaga, R. P.; Amaral, J. C . and I. D . L i m a Junior. 1991. Prospeccao hidroaciistica e avaliacao da biomassa de sardinha e anchoita na regiao sudeste do Brasil (outubro/novembro de 1988). B o l m Inst. Oceanogr., Sao Paulo, 8: 15:30. Castello, J. P.; Haimovici , M . ; Odebrecht, C. and C. M . Vooren. 1997. The continental shelf and slope, Pages 171-179 In U . Seeliger, C. Odebrecht and J. P. Castello (Eds). Subtropical Converegence Environments. The coast and Sea in the Southwestern Atlantic. Springer, Berlin, 308 p. Caveriviere, A . 1991. L'explosion demographique du baliste (Balistes carolinensis) en Afrique de l'Ouest et son evolution en relation avec les tendaces climatique. pp. 354'367 In P. Cury and C. Roy (Eds.). Pecheries Ouest-Africaines Variabilite, Instability et Changement. Editions de l ' O R S T O M , Paris, 524 p. Cergole, M . C. 1993. Avaliacao do estoque da sardinha-verdadeira, Sardinella brasiliensis, na costa sudeste do Brasil , periodo 1977-1990. PhD thesis. Universidade de Sao Paulo, 245p. Cergole, M . C. 1995. Stock assessment of the Brazilian sardine, Sardinella brasiliensis, of the southeastern coast of Brazi l . Sci. Mar. , 59(3-4):597-610. Chavez, E . A . ; Gardurio, M . and Arreguin-Sanchez, F. , 1993. Trophic dynamic structure of Celestun Lagoon, Southern G u l f of Mexico. In: V . Christensen and D . Pauly (Editors), Trophic Models of Aquatic Ecosystems, I C L A R M Conference Proceedings, 26: 186-192. Christensen, V . 1991. On E C O P A T H , Fishbyte, and fisheries management. Fishbyte, 9(2):62-66. 172 Christensen, V . , 1995a. Ecosystem maturity - towards quantification. Ecol . Modell ing, 77:3-32. Christensen, V . 1995b. A model of trophic interactions in the North Sea in 1981, the Year of the Stomach. Dana 11 (1): 1-28. Christensen, V . 1998. Fishery induced changes in a marine ecosystem: insights from models of the G u l f of Thailand. Journal of Fish Biology 53 (Suppl. A ) : 128-142. Christensen, V . and D . Pauly. 1992. E C O P A T H II - a software for balancing steady-state models and calculating network characteristics. Ecol . Model l ing 61: 169-185. Christensen, V . and D . Pauly. (Eds.) 1993. Trophic models o f aquatic ecosystems. I C L A R M Conference Proceedings No. 26. I C L A R M Manila , Philippines, 390 p. Cicin-Sain, B . and R. W . Knecht. 1998. Integrated Coastal Management and Ocean Management: Concepts and Practices. Island Press, Washington, D C , 518 p. Clark, C. & G . Munro. 1997. Why do fisheries collapse? H o w can protected marine reserves help? Fisheries Centre Report 5: 1-5. Cochrane, K . 1998. F A O ' s interest in and expectations of this workshop, p. 6-8. In D . Pauly. (Ed.) Use of Ecopath with Ecosim to Evaluate strategies for sustainable exploitation of multi-species resources: Proceedings of the workshop held at the Fisheries Centre University of British Columbia, March 25-27, 1998. Fisheries Centre Researh Reports 6(2), 49 p. Cochrane, K . L . ; Butterworth, D . S.; De Oliveira, J. A . A . and B . A . Roel . 1998. Management procedures in a fishery based on highly variable stocks and with conflicting objectives: experiences in the South African pelagic fishery. Reviews in Fish Biology and Fisheries 8, 177-214. Connell, J. H . 1978. Diversity in tropical rain forests and coral reefs. Science V o l . 199, 24:1302-1310. Costanza, R.; Wainger, L . ; Folke, C. and K . G . Mailer. 1993. Modeling complex ecological economic systems: towards and evolutionary dynamic understanding of people and nature, p. 148-163 In Samson, F. B . and Knopf, F . L . (Eds.) Ecosystem Management, Springer. Csirke, J. 1989. Changes in the catchability coefficient in the Peruvian anchoveta (Engraulis ringens ) fishery, p. 207-219 In Pauly, D . ; Muck, P.; Mendo, J. and Tsukayama, I. (Eds.) The Peruvian upwelling ecosystem: dynamics and interactions. I C L A R M conference proceedings, Manila , no. 18, 438p. Csirke, J. 1995. Fluctuations in abundance of small and mid-size pelagics. Scientia Marina 59(3-4): 481-490. Cury, P. and C. Roy. 1989. Optimal environmental window and pelagic fish recruitment success in upwelling areas. Can. J. Fish. Aquat. Sci. 46:670-680. 173 Cury, P.; Bakun, A . ; Crawford, J. M . ; Jarre-Teichman, A . ; Quinones, R.; Shannon, L . J. and H . M . Verheye. in press. Small pelagics in upwelling systems: patterns of interactions and structural changes in \"wasp-waist\" ecosystems. I C E S / S C O R Symposium on the Ecosystem Effects of Fishing, Montpellier, France, 15-19 March, 1999. Cushing, D . H . 1980. The decline of the herring stocks and the gadoid outburst. J. Cons. Int. Explor. Mer 39:70-81. DTncao, F . 1991. Pesca e biologia de Penaeuspaulensis na Lagoa dos Patos, R S . Atlantica (Rio Grande) 13(1):159-169. . Daan, N . 1980. A review of replacement of depleted stocks by other species and the mechanisms underlying such replacement. Rapp. P. -v. Reun. Cons. Int. Explor. Mer 177:405-421. DeAngelis, D . L . , 1980. Energy flow, nutrient cycling and ecosystem resilience. Ecology, 61:764-771. Deriso, R. B . 1980. Harvesting strategies and parameter estimation for an age-structured model. Can. J. Fish. Aquat. Sci . 37:268-282. Dias Neto, J. 1991a. A pesca da Piramutaba (Brachyplatystoma vaillanti, Val . ) na regiao norte do Brasil . Atlantica 13(1): 11-20. Dias Neto, J. 1991b. Pesca de camaroes no norte do Brasil . Atlantica 13(l):21-28. Diegues, A . C. S. 1995. Povos e Mares: Leituras em socio-antropologia maritma. N U P A U B , University of Sao Paulo, Brazi l . 260 p. Dorcey, A . H . J and K . J. Ha l l . 1983. Setting ecological research priorities for management: The art o f the impossible in the Fraser Estuary.Westwater Research Institute, University of British Columbia, 78 p. Durand, M . H . 1998. Fishmeal price behaviour: Global dynamics and short-term changes, p. 465-480 In Durand. M . H . ; P. Cury; R. Mendelssohn; C. Roy; A . Bakun and D . Pauly (Eds.) Global versus local changes in upwelling systems. Editions de l 'Orstom, Paris, 594 p. Durand. M . H . ; P. Cury; R. Mendelssohn; C . Roy; A . Bakun and D . Pauly (Eds.). 1998. Global versus local changes in upwelling systems. Editions de l 'Orstom, Paris, 594 p. F A O . 1997. Fisheries management. F A O Technical Guidelines for Responsible Fisheries, 4, Rome, 82 p. Finn, J. T. 1976. Measures of ecosystems structure and function derived from analysis of flows. J. Theor. B i o l . 56:363-380. Finn, J. T., 1980. F low analysis of models of the Hubbard Brook ecosystem. Ecology, 61(3):562-571. FishBase, 1998. FishBase 98 C D - R O M . I C L A R M , Mani l la . 174 Fogarty, M . J. and S. A . Murawski. 1998. Large-scale disturbance and the structure of marine systems: fisheries impacts on Georges Bank. Ecological Application, 8(1): 6-23. Fournier, D . A . and I. J. Doonan. 1987. A length-based stock assessment method utilizing a generalized delay-difference model. Can. J. Fish. Aquat. Sci . 44:422-437. Frederick, S. W . & R. M . Peterman. 1995. Choosing fisheries harvest policies: when does uncertainty matter? Can. J. Fish. Aqu . Sci. 52(2): 291-306. Froese, R. and D . Pauly (Ed). 1998. FishBase 98. Concepts, design and data sources. I C L A R M , Manila , Philippines. Garcia, S. M . and C. Newton. 1997. Current situation, trends, and prospects in world capture fisheries. Pages 3 - 27, In E . K . Pikitch; D . D . Huppert and M . P. Sissenwine (Eds). Global Trends: Fisheries Management. American Fisheries Society Symposium 20, Maryland. Gasalla, M . A . 1995. Organizacao trofica da ictiofauna do Saco de Mamangua, Parati, Estado do Rio de Janeiro. Instituto Oceanografico. University of Sao Paulo, Brazi l . M S c Thesis, 145p. Gasalla, M . A . & M . R. Oliveira. 1997. Papel trofico de clupeideos da costa sudeste do Brasil . XII Encontro Brasileiro de Ictiologia. Abstracts, Instituto Oceanografico, University of Sao Paulo, Sao Paulo. Gasalla, M . A . & Tomas, A . R . G . 1998. A n evaluation of the status o f Fisheries Data Collection and Stock Assessment Problems in Sao Paulo, Southeastern Brazi l . In: T.J. Quinn II, F . Funk, J. Heifetz, J .N. Ianelli, J.E. Powers, J.F. Schweigert, P.J. Sullivan, C.-I. Zhang (eds.), Fishery stock assessment models. Alaska Sea Grant College Program, University of Alaska Fairbanks, AK-SG-98-01 . Gi l l i s , D . M . ; Peterman, R. M . and A . V . Tyler. 1993. Movement dynamics in a fishery: Application of the Ideal Free Distribution of spatial allocation of effort. Can. J. Fish. Aquat. Sci . 50: 323-333. Glantz, M . H . 1983. Man , state and fisheries: an inquiry into some societal constraints that affect fisheries management, p. 871-918. In Sharp, G . D . and J. Csirke (Eds.) Proceedings of the expert consultation to examine changes in abundance and species composition of neretic fish resources. San Jose, Costa Rica, 18-29 Apr i l . F A O Fisheries Report 291 (3): 1224 p. Goitein, R. 1983. Sardinella brasiliensis (Steindachner, 1879): estudos sobre alimentacao nas regioes de Ubatuba (23\u00C2\u00B026 'S) , Santos (24\u00C2\u00B000 'S) e Cananeia (25\u00C2\u00B001 'S) . Naturalia, 8: 197-209. Goni, R. 1998. Ecosystem effects of marine fisheries: an overview. Ocean and Coastal Management 40, 37-64. Goodyear, C P . 1993. Spawning stock biomass per recruit in fisheries management: Foundation and current use. Canadian special publication of fisheries and aquatic sciences. 120: 67-81. 175 Grupo Executivo do Setor Pesqueiro (GESPE) . 1997. Piano basico para o ordenamento da pesca maritma nacional. Relatorio do Subgrupo de Trabalho para a Pesca Maritima. Junho de 1997. Ministerio da Ciencia e Tecnologia, Brasilia, Brazi l , 26 p. Gulland, J .A 1991. Under what conditions w i l l multispecies models lead to better fisheries management. ICES Marine Science Symposia 193: 348-352. Gunderson, L . H . ; Hol l ing, C. S. and S. S. Light. 1995. Barriers and bridges in the renewal of ecosystems and institutions. Columbia University Press, New York, 593 p. Haimovici , M . 1997. Cephalopods, Pages 146-150, In U . Seeliger, C. Odebrecht and J. P. Castello (Eds). Subtropical Converegence Environments. The coast and Sea in the Southwestern Atlantic. Springer, Berlin, 308 p. Haimovici , M . 1998. Present state and perspectives for the southern Braz i l shelf demersal fisheries. Fisheries Management and Ecology 5: 277-289. Haimovici , M . and L . C . Krug. 1992. Alimentacao e reproducao da enchova Pomatomus saltatrix no litoral sul do Brasil . Rev. Brasil . B i o l . (Rio de Janeiro) 52(3):503-513. Haimovici , M . ; Arruda, M . C. and R. L . Teixeira. 1989. Alimentacao da castanha Umbrina canosai no litoral sul de Brasil . Rev Brasil B i o l (Rio de Janeiro) 49(2):511-522. Haimovici , M . ; Castello, J. P. and C. M . Vooren. 1997. Fisheries, Pages 183-196 In, U . Seeliger, C. Odebrecht and J. P. Castello (Eds). Subtropical Converegence Environments. The coast and Sea in the Southwestern Atlantic. Springer, Berl in, 308 p. Haimovici , M . , Martins, A . S. and R. L . Teixeira. 1993. Distribuicion, alimentacion y observaciones sobre la la reproduccion de la merluza Merluccius hubbsi en el sur de Brasil . Frente Maritmo (Montevideo) 14: 81-86. Hansen, P. E . and J0rgensen, S. E . (Editors), 1990. Introduction to Environmental Management. Developments in Environmental Modell ing, 18, Elsevier, Amsterdam, 430 p. Hempel, G . 1971. Southwest Atlantic, In Gulland, J. A . (Ed) Fish resources of the oceans. Oxford, Fishing News Books, p. 146-152. Hilborn, R. 1985. Fleet dynamics and individual variation: Why some people catch more fish than others. Can. J. Fish. Aquat. Sci . 42:2-13. Hilborn, R. and Mangel, M . 1997. The ecological detective: confronting models with data. Princeton University Press, Princeton, New Jersey. 315p. Hilborn, R. and Walters, C.J . 1992. Quantitative fish stock assessment: choice, dynamics and uncertainty. Chapman and Ha l l Pub. Co. , New York. 570p. Holl ing, C. S. 1959. The components of predation as revealed by a study of small mammal predation of the European pine sawfly. Can. Ent. 91(5): 293-320. Holl ing, C. S., 1973. Resilience and stability of ecological systems. Ann . Rev. Ecol . Syst., 4:1-23. 176 Holl ing, C. S. 1986. Resilience of ecosystems; local surprise and global change, p. 292 - 317. In Clarck, W . C . and R. E . Munn (Eds.). Sustainable development o f the biosphere, Cambridge University Press, Cambridge, Mass. Holl ing, C. S. 1992. Cross-scale morphology, geometry and dynamics of ecosystems. Ecological Monographs 62(4): 447-502. Holl ing, C.S.; Schlindler, D . W. ; Walker, B . and Roughgarden, J. 1995. Biodiversity and the function of ecosystem. In: C. Perrings; K . Goran-Maler; C . Folke and C. S. Hol l ing (Editors). Biodiversity loss. Cambridge University Press, 332 p. Hunter, M . D . and P. W . Price. 1992. Playing chutes and ladders: heterogeneity and the relative roles o f bottom-up and top-down forces in natural communities. Ecology 73(3): 724-732. Hunter, J. R. and J. Alheit (Eds.). 1997. International G L O B E C Small Pelagic Fishes and Climate Cgange program. Implementation Plan. G L O B E C Report 11, 36 p. Hutchings, J. A . ; Walters, C. J. and R. L . Haedrich.1997. Is scientific inquiry incompatible with government information control? Can. J. Fish. Aquat. Sci . 54: 1198-1210. I B A M A . 1991. Relatorio da reuniao do grupo permanente de estudos sobre sardinha. C E P S U L -I B A M A , 22 a 26 de outubro 1990. Itajai, SC, 31p. I B A M A . 1992. Relatorio da reuniao do grupo permanente de estudos sobre sardinha. C E P S U L -I B A M A , 23 a 27 de setembro 1991. Itajai, S C , 41p. I B A M A . 1993a. Peixes demersais. Colecao Meio Ambiente, Series Estudos Pesca, Brasilia. Ministerio do Meio Ambiente 8, 94 p. I B A M A . 1993b. Relatorio da reuniao do grupo permanente de estudos sobre sardinha. C E P S U L -I B A M A , 19 a 23 de outubro 1992. Itajai, SC, 15p. I B A M A . 1994a. Lagosta, Carangueijo-uca e camarao do Nordeste. Relatorio das reunioes dos Grupos Permanentes de Estudos-GPE's. Brasilia. I B A M A Estudos de Pesca, 10; 190 p. I B A M A . 1994b. Sardinha, atuns e afins. Relatorio das reunioes dos Grupos Permanentes de Estudos-GPE's. Brasilia. I B A M A Estudos de Pesca, 11;201 p. I B A M A . 1994c. Relatorio da reuniao do Grupo Permanente de Estudos sobre camaroes das regioes sudeste e sul do Brasil . C E P S U L - I B A M A , 4 a 8 de outubro, 1993; Itajai, SC; 7 p. I B A M A . 1994d. Relatorio da reuniao do grupo permanente de estudos sobre sardinha. C E P S U L -I B A M A , 04 a 08 de outubro 1993. Itajai, SC, 19p. I B A M A , 1995. Situacao do estoque da sardinha no litoral Sudeste e Sul. Series Estudos de Pesca 17, 44p. I B A M A / I B G E . 1995. Estatistica da Pesca. C E P E N E , Tamandare,PE. 177 Instituto de Pesca, 1993. Forum de Debates: Administracao de recursos pesqueiros marinhos, problemas e perspectivas. Segunda Reuniao Anual do Instituto de Pesca, 13 de abril, 1993, Sao Paulo, Brazi l . Jablonski, S. and Y . Matsuura. 1985. Estimate of exploitation rate and population size of skipjack tuna of the southeastern coast of Brazi l . Bo lm. Inst. oceanogr.,Sao Paulo, 33(1): 29-38. Jarre-Teichman, A . and V . Christensen. 1998. Comparative modelling o f trophic flows in four Large Upwell ing Ecosystems: Global versus Local effects. Pages 425 - 443, In: Durand, M . PL; Cury, P.; Mendelssohn, R.; Roy, C ; Bakun, A . and D . Pauly (Editors) Global versus local changes in upwelling systems. O R S T O M Editions, Paris. Jarre, A . ; Muck, P. and Pauly, D . , 1991. Two approaches for modelling fish stocks interactions in the Peruvian upwelling ecosystem. ICES Mar. Sci. Symp., 193:178-184. Jasanoff, S. 1990. The fifith branch: Science advisers as policy makers. Harvard University Press, Cambridge, Mass. 302 p. Johannensson, K . A . 1975. Relatorio preliminar das observacoes acusticas quantitativas sobre tamanho e distribuicao dos recursos de peixes pelagicos ao largo da costa sul do Brasil . P D P - S U D E P E Serie Documentos Tecnicos, 10, 28 p. Juras, A . A . and N . Yamaguti. 1985. Food and feeding habits of the king weakfish (Macrodon ancylodon) (Bloch and Schneider 1801) caught in the southern coast of Brasi l (Lat 29( S -32( S). B o l m Inst Oceanogr (Sao Paulo) 33(2): 149-157. Kawasaki, T. 1983. Why do some pelagic fishes have wide fluctuactions in their numbers? Biological basis o f fluctuations from the viewpoint o f evolutionary ecology, pp. 1065-1080. In G . D . Sharp and J. Csirke (eds.). Reports of the Expert Consultation to Examine Changes in Abundance and Species Composition of Neritic Fish Resources. F A O Fish. Rep. 29(3): 1224p. Keeney, R. L . 1992. Value-focused thinking. Harvard Univ. Press, Cambridge, Mass. 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 (Suppl. a), 209-226. Lasker, R. 1975. Field criteria for survival of anchovy larvae: the relation between inshorechlorophyll maximum layers and successful first feeding. Fish. B u l l . U . S. 73: 453-462. Lasker, R. 1981. The role o f a stable ocean in larval fish survival and subsequent recruitment, pp. 80-87 In R. Lasker (ed.) Marine fish larvae: morphology, ecology and relation to fisheries. Univ . Washington Press., Seattle, 131 p. Levins, R. 1966. The strategy of model building in population biology. A m . Sci . 54:421-431. Light, S. S.; Gunderson,, L . H . and C. S. Holl ing. 1995. The Everglades: Evolution of management in a turbulent ecosystem, p. 103-168 In Gunderson, L . H . ; Holl ing, C. S. and 178 S. S. Light. 1995. Barriers and bridges in the renewal of ecosystems and institutions. Columbia University Press, New York, 593 p. Lima, I. D . and J. P. Castello. 1995. Distribution and abundance o f South-west Atlantic anchovy spawners (Engraulis anchoita) in relation to oceanographic processes in the southern Brazilian shelf. Fish. Oceanogr. 4(1): 1-16. Lluch-Belda, D . ; Crawford, R. J.; Kawasaki, T.; MacCa l l , A . D . ; Parrish, R. FL; Schwartzlose, R. A . and P .E . Smith. 1989. World-wide fluctuations of sardine and anchovy stocks: the regime problem. S. Afr. J. mar. Sci . 8: 195-205. Louw, G . G . ; V a n der Lingen, C. D . and M . J. Gibbons. 1998. Differential feeding by sardine Sardinops sagax and anchovy Engraulis capensis recruits in mixed shoals. South African Journal of Marine Science, 19: 227-232. Ludwig, D . and C. J. Walters. 1985. Are age-structured models appropriate for catch-effort data. Can. J. Fish. Aquat. Sci . 42 (6): 1066-1072. Ludwig, D . ; Hilborn, R. & C. Walters, 1993. Uncertainty, resource exploitation and conservation; lessons from history. Science 260, 17, 36. MacCa l l , A .D. I976 . Density dependence of catchability coeficient in the California Pacific Sardine, Sardinops sagax caerulea, purse seine fishery. C a l C O F I Rep. (18):136-148. MacCa l l , A . D.1990. Dynamic geography of marine fish populations. University of Washignton Press. 153 p. MacLennam, D . N . & E . J. Simmonds. 1992. Fisheries acoustics. London: Chapman and Hal l . 366 p. Mangel, M . et al. 1996. Principles for the conservation of w i ld l iving resources. Ecological Applications 6(2): 338-262. Mann, K . H . and J. R. N . Lazier. 1991. Dynamics of marine ecosystems. Biological-physical interactions in the oceans. Blackwell Scientific Publications, Boston, 466 p. Martins, A . S. 1992. Bioecologia do peixe-espada, Trichiurus lepturus Linnaeus, 1758, no sul do Brasil . M S c . Thesis, University of R io Grande, Brazi l . Martins, A . S. and M . Haimovici . 1997. Distribution, abundance and biological interactions of the cutlassfish Trichiurus lepturus in the southern Braz i l subtropical convergence ecosystem. Fisheries Research 30: 217-227. Matsuura, Y . 1988. Synospsis on the reproductive and early life history of the Brazilian sardine, Sardinella brasiliensis and related environmental conditions. IOC Workshop Report (65), Annex VIII, 8p. Matsuura, Y . 1995. Os ecossistemas brasileiros e os principals macro vetores de desenvolvimento. Subsidio ao Planejamento da Gestao Ambiental. Projeto Cenarios para o Planejamento da Gestao Ambiental ( M M A / P N M A ) , Brasilia, D F ; p. 39-104. 179 Matsuura, Y . ; Spach, H . L . and M . Katsuragawa. 1992. Comparison of spawning patterns of the Brazilian sardine (Sardinella brasiliensis) and anchoita (Engraulis anchoita) in Ubatuba region, southern Brazi l during 1985 through 1988. B o l m . Inst. Oceanogr., Sao Paulo, 40(1/2):101-115. May, R. M . , 1972. W i l l a large complex system be stable? Nature, 238:413-414. May, R. M . 1984. Exploitation of marine communities. Springer-Verlag, Berl in, 367 p. Mayr, E . 1997. This is biology. The science of the l iving world. Harvard University Press, Cambridge, Mass., 323 p. McCarthy, M . A . ; Ginzburg, L . R. and H . R. Akcakaya. 1995. Predator intereference across trophic chains. Ecology 76(4): 1310-1319. Mee, L . D . 1992. The Black Sea in crisis: A need for concerted international action. Ambio 21:278-286. Mel lo , R. M . ; Castello, J. P. and K . F. M . Freire. 1992. Associacion de especies pelagicas marinas en el sur de Brasil durante invierno y primavera. Frente Maritmo (Montevideo) 11: 63-69. Mendoza, J. J., 1993. A preliminary biomass budget for the Northeastern Venezuela shelf ecosystem. In: V . Christensen and D . Pauly (Editors), Trophic Models of Aquatic Ecosystems, I C L A R M Conference Proceedings, 26: 285-297. Misund, O. A . 1997. Underwater acoustics in marine fisheries and fisheries research. Reviews in Fish Biology and Fisheries 7 (1): 1-34. Morgan, M . G . & M . Henrion. 1990. Uncertainty: a guide to dealing with uncertainties in quantitative risk and policy analysis. Cambridge University Press, Cambridge. Mundy, P.R. 1985. Theory\u00E2\u0080\u0094a provisional definition of harvest control. Tech. Rep. Wash. Sea Grant, Fisheries dynamics: Harvest management and sampling, pp. 3-8. Myers, R. A . ; Barrowman, N . J.; Hutchings, J. A . and A . A . Rosenberg. 1995. Population dynamics of exploited fish stocks at low population levels. Science 269:1106-1108. National Research Council (NRC) , 1996. The Bering Sea ecosystem. National Academy Press, Washington, D . C. 207 p. Neiva, G . S. and S. J. C. Moura. 1977. Sumario sobre a exploracao de recursos marinhos do litoral brasileiro: Situacao atual e perspectivas. P D P - S U D E P E , Serie Doc. Ocasionais, Rio de Janeiro, 27: 1-44. O ' N e i l l , R. V . , 1976. Ecosystem persistence and heterotrophic regulation. Ecology, 57:1244-1253. Odebrecht, C. and V . M . T. Garcia. 1997. Phytoplankton, Pages 105-109, In In U . Seeliger, C. Odebrecht and J. P. Castello (Eds). Subtropical Converegence Environments. The coast and Sea in the Southwestern Atlantic. Springer, Berlin, 308 p. 180 Odum, E . P., 1969. The strategy of ecosystem development. Science, 164:262-270. Odum, E . P. 1997. Ecology. A Bridge between science and society. Sinauer, Massachusets. 330p. Odum, W . E . and Heald, E . J. 1975. The detritus-based food web o f an estuarine mangrove community, p. 265-286. In L . E . Cronin (Ed.) Estuarine research, V o l . 1. Academic Press, New York. Olivieri , R. A . , Cohen, A . and Chavez, F. P., 1993. A n ecosystem model of Monterey Bay, California. In: V . Christensen and D . Pauly (Editors), Trophic Models of Aquatic Ecosystems, I C L A R M Conference Proceedings, 26: 315-322. Palomares, M . L . D . and D . Pauly. 1989. A multiple regression model for predicting the food consumption of marine fish populations. Aust. J. Mar. Freshwat. res. 40:259-273. Palomares, M . L . ; Reyes-Marchant, P.; Lair, N . ; Zainure, M . ; Barnabe, G . and Lasserre, G . 1993. A trophic model of a Mediterranean Lagoon, Etang de Thau, France. In: V . Christensen and D . Pauly (Editors), Trophic Models of Aquatic Ecosystems, I C L A R M Conference Proceedings, 26: 224-229. Parma, A . M . 1990. Optimal harvesting of fish populations with non-stationary stock-recruitment relationships. Natural Resource Modeling 4(1): 39-76. Parma, A . M . and R. B . Deriso. 1990. Experimental harvesting o f cyclic stocks in the face of alternative recruitment hypotheses. Can. J. Fish. Aquat. Sci . 47(3): 595-610. Patterson, K . 1992. Fisheries for small pelagic species: A n empirical approach to management targets. Rev. Fish. B i o l . Fish. 2: 321-338. Paula e Silva, R.; Sousa, M . I. and Caramelo, A . M . ; 1993. The Maputo Bay ecosystem (Mozambique). In: V . Christensen and D . Pauly (Editors), Trophic Models of Aquatic Ecosystems, I C L A R M Conference Proceedings, 26: 214-223. Pauly, D . 1979. Theory and management of tropical multispecies stocks. A review with emphasis on the Southeast Asian demersal fisheries. I C L A R M Studies and Reviews 1, 35p. Pauly, D . 1980. On the interrelationships between natural mortality, growth parameters and mean environmental temperature in 175 fish stocks. J. Cons. Int. Explor. Mer. 39: 175-192. Pauly, D . and I. Tsukayama (Eds.). 1987. The Peruvian anchoveta and its upwelling ecosystem: three decades of change. I C L A R M Studies and Reviews 15, 351 p. Pauly, D . and V . Christensen 1993. Stratified models of large marine ecosystems: a general approach, and an application to the South China Sea. In: K . Sherman; L . M . Alexander and B . D . Gold (Editors), Stress, mitigation and sustainability of large marine ecosystems. A A A S Press, Washington, D C , p. 148-174. Pauly, D . and V . Christensen. 1995. Primary production required to sustain global fisheries. Nature 374:255-257. 181 Pauly, D . ; Christensen, V . and C. Walters, (in press). Ecopath as a tool for evaluating ecosystem impacts of fisheries. I C E S / S C O R Symposium on the Ecosystem Effects of Fishing, 15-19 March, 1999, Montpellier, France. Pauly, D . ; Christensen, V . ; Dalsgaard, J.; Froese, R. and F. Torres Jr. 1998. Fishing down marine food webs. Science 279:860 - 863. Pauly, D . ; Soriano-Bartz, M . L . and M . L . D . Palomares. 1993. Improved construction, parametrization and interpretation of steady-state ecosystem models, Pages 1-13, In V . Christensen and D . Pauly (eds) Trophic models of aquatic ecosystems. I C L A R M Conf. Proc. 26, 390 p.Pauly, D . and I. Tsukayama (Eds.). 1987. The Peruvian anchoveta and its upwelling ecosystem: Three decades of change. I C L A R M Studies and Reviews 15, 351p. Peterman, R. M . 1977. A simple mechanism that causes collapsing stability regions in exploited salmonid populations. J. Fish. Res. Board Can. 34(8): 1130-1142. Peterman, R. M . ; Clarck, W . C. and C. S. Holl ing. 1978. The dynamics of resilience: shifting stability domains in fish and insect systems, p. 321-341 In Anderson, R. M . ; Turner, B . D . and L . D . Taylor (Eds.) Population Dynamics. Blackwel l Scientific Publications, Melbourn. Peterman, R. M . and M . Gato. 1978. Estimation of functional responses of predation on juvenile slamon. J. Fish. Res. Board. Can. 35:797-808. Peterson, I. and J. S. Wroblewski. 1984. Mortality rate of fishes in the pelagic ecosystem. Can. J. Fish. Aquat. Sci . 41(7): 1117-1120. Pires-Vanin, A . M . S. (Ed.). 1993. Estrutura e funcao do ecossistema de plataforma continental do Atlantico Sul brasileiro. Publcao esp. Inst, oceanogr. S Paulo, Sao Paulo, 10, 245 p. Pires-Vanin, A . M . S. and Y . Matsuura. 1993. Estrutura e funcao do ecossistema de plataforma continental da regiao de Ubatuba, estado de Sao Paulo: uma introducao. p. 1-8 In A . M . S. Pires-Vanin (Ed.) Estrutura e funcao do ecossistema de plataforma continental do Atlantico Sul brasileiro. Publcao esp. Inst, oceanogr. S Paulo, Sao Paulo, 10:245 p. Pires-Vanin, A . M . S.; C. L . D . B . Rossi-Wongtschowski; E . Aidar; H . S. L . Mesquita; L . S. H . Soares; M . Katsuragawa and Y . Matsuura. 1993. Estrutura e funcao do ecossistema de plataforma continental da regiao de Ubatuba, Sao Paulo: sintese dos resultados, p. 217-231 In In A . M . S. Pires-Vanin (Ed.) Estrutura e funcao do ecossistema de plataforma continental do Atlantico Sul brasileiro. Publcao esp. Inst, oceanogr. S Paulo, Sao Paulo, 10, 245 p. Pitcher, T. J.1995. The impact of pelagic fish behaviour on fisheries. Sci . Mar . 59:295-306. Pitcher, T. and D . Pauly. 1998. Rebuilding ecosystems, not sustainability, is the proper goal of fisheries management. Pages 311-329. In: T. Pitcher, D . Pauly and P. Hart (Eds.) Reinventing Fisheries Management. Chapman & Hal l . London. Pitcher, T. J. and J. Parrish. 1993. The functions of shoaling behavior. P. 363-349. In T. J. Pitcher (ed.) The behaviour of teleost fishes, Chapman and Ha l l , London, 715 p. 182 Policanski, D . and J. Magnuson. 1998. Genetics, metapopulations, and ecosystem management of fisheries. Ecological Applications, 8 (1): 119-124. Polovina, J. J., 1984. Mode l of a reef ecosystem. I. The E C O P A T H model and its application to French Frigate Shoals. Coral Reefs, 3:1-11. Quinn II, T. J.; Fagen, R. and J. Zheng. 1990. Threshold management policies for exploited populations. Can. J. Fish. Aquat. Sci . 47:2016-2029. Radovich, J. 1982. The collapse of the California sardine fishery. C a l C O F I Rep. 23:56-78. Rees, W . E . 1996. Revisiting carrying capacity: Area-based indicators of sustainability. Population and Environment 17(3): 195-215. Reis, E . G . ; Vieira, P. C. e Duarte, W . S. 1994. Pesca artesanal de teleosteos no estuario da Lagoa dos Patos e costa do Rio Grande do Sul. Atlantica 16:69-86. Rice, J. 1995. Food web theory, marine food webs, and what climate changes may do to northern marine fish populations. In: R. J. Beamish (Editor), Climate Change and Northern Fish Populations. Canadian Special Publication of Fisheries and Aquatic Sciences, 121:561-568. Rijavec, L . and J. C. Amaral. 1977. Distribuicao e abundancia de peixes pelagicos na costa sul e sudeste do Brasil . P D P - S U D E P E Documentos Tecnicos 24, 55p. Rocha, G . R. A . ; Gasalla, M . A . ; Rossi-Wongtschowski, C . L . D . B . ; Soares, L . S. H . ; Pires-Vanin, A . M . S.; Muto, E . Y . ; Cergole, M . C ; Aidar, E . ; Mesquita, H . S. L . ; Gianesella-Galvao, S. M . F. ; Vega-Perez, L . A . and A . Jarre-Teichmann. 1998. Quantitative model of trophic interactions in the Ubatuba shelf system (Southeastern Brazil). Naga, The I C L A R M Quaterly, p. 25-32. Rossi-Wongtschowski, C . L . D . B . 1977. Estudo das variacoes da relacao peso total/comprimento total em funcao do ciclo reprodutivo e comportamento de Sardinella brasiliensis (Steindachner, 1879) da costa do Brasil entre 23\u00C2\u00B0S e 28\u00C2\u00B0S. Bo lm. Inst, oceanogr., S. Paulo, 26: 131-180. Rossi-Wongtschowski, C . L . D . B . ; Saccardo, S. A . and M . C. Cergole. 1996. Are fluctuations in Brazilian sardine catches related to global-scale climate changes? A n . Acad. bras. C i . 68 (Supl. 1): 239 - 250. Rowen, H . S. 1976. Pol icy analysis as heuristic aid: The design of means, ends, and institutions, p. 137-152 In L . H . Tribe; C. S. Schelling and J. Voss (Eds.) When values conflict: Essays on environmental analysis, discourse and decisions, Ballinger, Cambridge, Mass. Rudstam, L . G . ; G . Aneer and M . Hilden. 1994. Top-down control in the pelagic Baltic ecosystem. Dana, 10:105-129. Saccardo, S. A . 1980. Biologia e bionomia de Trachurus lathami (Nichols 1920) (Teleostei: Carangidae) na plataforma continental brasileira entre 23( S (RJ) e 30( S (RS). PhD Thesis, University of Sao Paulo, Brazi l . 183 Saccardo, S. A . 1983. Biologia y disponibilidada de sardina {Sardinella brasiliensis, Steindachner, 1879) en la costa sudeste del Brasil . F A O Fisheries Report 291, 2:449-464. Saccardo, S. A . & C. L . B . Rossi-Wongtschowski. 1991. Biologia e avaliacao do estoque da sardinha Sardinella brasiliensis: uma compilacao. Atlantica 13(l):29-44. Santander, H . J. Alheit; MacCa l l , A . D . and A . Alamo. 1983. Egg mortality of the Peruvian anchovy (Engraulis ringes) caused by cannibalism and predation by sardines (Sardinops sagax). F A O Fisheries Repors 291, vol . 3, p. 1011-1025. Scheffer, M . and R. J. de Boer. 1995. Implications of spatial heterogeneity for the paradox of \u00E2\u0080\u00A2enrichment. Ecology 76(7): 2270-2277. Schwingel, P. R. 1996. Influence of water masses on feeding behavior of anchovy off the Brazilian continental shelf (22\u00C2\u00B0S-34\u00C2\u00B0S). Abstract p. 2 Role of Forage Fishes in Marine Ecosystems, November 13-16, Anchorage, Alaska. Schwingel, P. R. and J. P. Castello. 1995. Alimentation de la anchoita, Engraulis anchoita, en el sur de Brasil . Frente Maritmo (Montevideo) 15: 67-86. Seeliger, C. Odebrecht and J. P. Castello. 1997. (Eds). Subtropical Converegence Environments. The coast and Sea in the Southwestern Atlantic. Springer, Berl in, 308 p. Silvestre, G . ; Selnanathan, S. and Salleh, A . H . M . , 1993. Preliminary Trophic Model of the Coastal Fisheries Resources of Brunei Darussalam, South China Sea. In: V . Christensen and D . Pauly (Editors), Trophic Models of Aquatic Ecosystems, I C L A R M Conference Proceedings, 26: 300-306. Soares, L . S. PL; M . A . Gasalla; M . A . T. Rios; M . V . Arrasa and C. L . D . B . Rossi-Wongtschowski. 1993. Grupos troficos de especies dominantes de peixes demersais da plataforma continental interna de Ubatuba, Brasil . p. 189-198 In A . M . S. Pires-Vanin (Ed.) Estrutura e funcao do ecossistema de plataforma continental do Atlantico Sul brasileiro. Publcao esp. Inst, oceanogr. S Paulo, Sao Paulo, 10:245 p. Soares, L . S. FL; Rossi-Wongtschowski, C. L . D . B . ; Alvares, L . M . C ; Muto, E . Y . and M . A . Gasalla. 1992. Grupos troficos de peixes demersais da plataforma continental interna de Ubatuba, Brasil . I. Chondrichthyes. B o l m Inst oceanogr., Sao Paulo, 40(l-2):79-85. Soutar, A . and J. D . Isaacs. 1974. Abundance of pelagic fish during the 19 t h and 20 t h century as recorded in anaerobic sediment off the Californias. Fish. B u l l . U . S. 72: 257-273. Sparre, P. 1991. Introduction to multispecies virtual population analysis. ICES Mar. Sci . Symp. 193,12-21. Steele, J. H . 1985. A comparison of terrestrial and marine ecological systems. Nature V o l . 313, 6001:355-358. Steele, J. H . 1996. Regime shifts in fisheries management. Fisheries reserach 25: 19-23. Steele, J. 1998. Regime shifts in marine ecosystems. Ecological Application, 8(1): 33-37. 184 Stergiou, K . I. 1998. Variability o f f i sh catches in different ecosystems, p. 359-370. In Durand. M . FL; P. Cury; R. Mendelssohn; C. Roy; A . Bakun and D . Pauly (Eds.) Global versus local changes in upwelling systems. Editions de l 'Orstom, Paris, 594 p. S U D E P E , 1986. A pesca da sardinha no estado do Rio de Janeiro. Producao e frota atuante. Documento Tecnico 769A, l i p . Teixeira, C. and J. Tundisi. 1967. Primary production and phytoplnakton in equatorial waters. Bulletin of Marine Science 17(4): 884-891. Tovar, H . ; Guillen, V . and M . E . Nakama. 1987. Monthly population size of three guano bird species off Peru, 1953 to 1982, p. 208-218 In Pauly, D . and Tsukayama, L . (Eds.) The Peruvian anchoveta and its upwelling ecosystem: Three decades of change. I C L A R M Studies & Reviews, no. 15. Trites, A . W. ; Livingston, P.; Vasconcellos, M . C ; Mackinson, S.; Springer, A . M . and D . Pauly. 1999. Ecosystem considerations and the limitations of ecosystem models in fisheries management: Insights from the Bering Sea. Proceedings of the Symposium on the Ecsosystem Effects of Fisheries, Anchorage, 1998. Turner, J. and S. Bencherifi. 1983. Replacement of sardine catches by anchovy catches in the Mediterranean pelagic fishery of Morocco. F A O Fisheries Reports, 305, pp. 63-72. Ulanowicz, R. E . , 1984. Community measures of marine food networks and their possible applications. In: M . J. R. Fasham (Editor), Flows of energy and materials in marine ecosystems, Plenum Press, New York, p. 23-47. Ulanowicz R. E . , 1986. Growth and development: Ecosystems Phenomenology. Springer-Verlag, New York. Ulanowicz, R. E . and Wulff, F. , 1991. Comparing ecosystem structures: The Chesapeake Bay and the Baltic Sea. In: J. Cole; G . Lovett and S. Findlay (Editors). Comparative analysis of ecosystems. Patterns, mechanisms and theories, Springer-Verlag, p. 140-166. Valdes-Szeinfeld, E . 1991. Cannibalism and intraguild predation in clupeoids. Marine Ecology Progress Series, 79(l-2):17-26. Valentini, H . & R. D . Cardoso. 1991. Analise da pesca da sardinha verdadeira, Sardinella brasiliensis, na costa sudeste-sul do Brasil . Atlantica 13(1): 45-54. Vandermeulen, H . 1998. The development of marine indicators for coastal zone management, Ocean & Coastal Management 39 (l-2):63-71. Vaske, T. Jr. 1992. Idade, crescimento e alimentacao de albacora de lage, Thunnus albacares (Bonnaterre, 1788) (Perciformes:Scombridae), explorada no sul do Brasil . Master Thesis, University of Rio Grande, Brazi l . Vazzoler, G . 1975. Distribuicao da fauna de peixes demersais e ecologia dos Sciaenedae da plataforma continental brasileira entre as latitudes 29\u00C2\u00B0 21 ' S (Torres) e 34\u00C2\u00B0 44' S (Chui). B o l m Inst Oceanogr (Sao Paulo) 24:85-169. 185 Vazzoler, A . E . A . M . and V . N . Phan. 1976. Electrophoretic patterns of eye-lens proteins of Sardinella brasiliensis (Steindachner, 1879) off Brazilian coast. Revue Trav. Inst. Pech. marit. 40 (3/4): 781-786. Vazzoler, A . E . A . M . , L . S . H . Soares & P . T . M . Cunningham (in press) Ictiofauna brasileira. Cap. 19. In: Lowe-MacConnell (traduction). Estudos ecologicos nas comunidades de peixes tropicais. Sao Paulo, E D U S P . Vega-Cendejas, M . E . ; Arreguin-Sanchez and Hernandez, M . , 1993. Trophic fluxes on the Campeche Bank, Mexico. In: V . Christensen and D . Pauly (Editors), Trophic Models of Aquatic Ecosystems, I C L A R M Conference Proceedings, 26: 206-213. Velarde, E . ; De la Soledad Torde Sillas, M . ; Vieyra, L . and R. Esouivel. 1994. Seabirds as indications of important fish populations in the G u l f of California. Ca lCOFI Rep. 35:137-143. Vieira, P. J. C. 1990. Biologia populacional de Cynoscion striatus (Pisces: Scianidae) no litoral sul do Brasil . Master Thesis, University of R io Grande, Brazi l . Vi le la , M . J. 1990. Idade, crescimento, alimentacao e avaliacao do estoque de bonito listado (Katsuwonus pelamis) (Scombridae:Thunnini), explotado na regiao sudeste-sul do Brasil. Master Thesis, University of Rio Grande, Brazi l . Vyalov Y . A . & V . V . Ovchinnikov. 1980. M a i n results o f the tuna, swordfish and sailfish studies in the Atlantic Research Institute for the Fisheries and Oceanography (Atlantniro) for the 20 year period (1957-1977). C o l . V o l . Sci . Pap., I C C A T , Madrid, v.9, no. 3, p. 662-674. Walsh, J. J. 1981. A carbon budget for overfishing off Peru. Nature 290:300-304. Walters, C. J. 1986. Adaptive management of renewable resources. Macmil lan Press, New York. Walters, C. J. 1987. Nonstationarity of production relationships in exploited populations. Can. J. Fish. Aquat. Sci . 44 (Suppl.2): 156-165. Walters, C . J. 1998. Improving links between ecosystem scientists and managers, p. 273-286 In Pace, M . L . and Groffman, P. M . (Eds.) Success, limitations, and frontiers in ecosystem science. Springer, Berlin. Walters C. J and J. S. Coll ie. 1988. Is research on environmental factors useful to fisheries management? Can. J. Fish. Aquat. Sci . 45:1848-1854. Walters C. J. and F. Juanes 1993. Recruitment limitation as a consequence of natural selection for use of restricted feeding habitats and predation risk taking by juvenile fish. Can. J. Fish. Aquat. Sci . 50:2058-2070. Walters, C J . and D.Ludwig. 1994. Calculation of Bayes posterior probability distribution for key population parameters. Can. J. Fish. Aquat. Sci. 51:713-722. Walters C. J. and J. J. Maguire. 1996. Lessons for stock assessment from the northern cod collapse. Rev. Fish. B i o l . Fish. 6:125-137. 186 Walters, C. J. and A . M . Parma. 1996. Fixed exploitation rate strategies for coping with effects of climate changes. Can. J. Fish. Aquat. Sci. 53: 148-158. Walters, C ; Christensen, V . and D . Pauly. 1997. Structuring dynamic models of exploited ecosystems from trophic mass-balance models of marine ecosystems. Reviews in Fish Biology and Fisheries 7(2): 139-172. Walters, C. J. and J. Korman. 1999. Linking recruitment to trophic factors: revisiting the Beverton\u00E2\u0080\u0094Holt recruitment model from a life history and multispecies perspective. Reviews in Fish Biology and Fisheries 9 (2): 187-202. Walters, C. J. and J. F . Kitchell . (in press). Cultivation/depensation effects on juvenile survival and recruitment: a serious flaw in the theory of fishing? Walters, C. J.; D . Pauly; Christensen, V . and J. F . Kitchell . (in press)a. Representing density dependent consequences of life history strategies in aquatic ecosystems: EcoSim II. Walters, C. J.; Pauly, D . and V . Christensen. (in press)b. Ecospace: prediction of mesoscale spatial patterns in trophic relationships of exploited ecosystems, with emphasis on the impacts of marine protected areas. ICES Journal of Marine Science. Walters, C. J. and A . E . Punt. 1994. Placing odds on sustainable catch using virtual population analysis and survey data. Can. J. Fish. Aquat. Sci . 51: 946-958. Wiefels, R. C. and S. Jablonski. 1979. Projeto Matriz da Pesca. R io de Janeiro, S U D E P E / P D P 276 p. W W F . 1998. The footprint of distant water fleets on world fisheries. Endangered Seas Campaing. 172 p. Zavalla-Camin, L . A . 1982. Distribuicao vertical e estacional de tunidos y otras especies pelagicas en el sudeste y sur del Brasil . C o l . Doc. Cient. (SIMP/81/2), I C C A T , Madrid, 17(2):439-443. Zavala-Camin, L . A . 1987. Ocorrencia de peixes, cefalopodos e crustaceos em estomagos de atuns e especies afins, capturadas com espinhel no Brasil (23 S-34S) 1972-1985. B . Inst. Pesca, Sao Paulo, v.14, no. unico, p.93-102. Zavala-Camin, L . A . and I. C. Lemos. 1997. Ocorrencia epipelagica de juvenis de Balistidae (Teleostei) e aspectos da biologia de Balistes capriscus no sudeste e sul do Brasil . Atlantica, R io Grande, 19: 183-195. 187 "@en . "Thesis/Dissertation"@en . "2000-05"@en . "10.14288/1.0074865"@en . "eng"@en . "Resource Management and Environmental Studies"@en . "Vancouver : University of British Columbia Library"@en . "University of British Columbia"@en . "For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use."@en . "Graduate"@en . "Ecosystem impacts of fishing forage fishes : an analysis of harvest strategies for the brazillian sardine"@en . "Text"@en . "http://hdl.handle.net/2429/10998"@en .