UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Strategic marine ecosystem restoration in northern British Columbia 2006

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

Item Metadata


ubc_2006-199244.pdf [ 33.74MB ]
JSON: 1.0074895.json
JSON-LD: 1.0074895+ld.json
RDF/XML (Pretty): 1.0074895.xml
RDF/JSON: 1.0074895+rdf.json
Turtle: 1.0074895+rdf-turtle.txt
N-Triples: 1.0074895+rdf-ntriples.txt

Full Text

S T R A T E G I C M A R I N E E C O S Y S T E M R E S T O R A T I O N I N N O R T H E R N B R I T I S H C O L U M B I A by C A M E R O N H . A I N S W O R T H B . Sc., The University of British Columbia, 1997 A THESIS S U B M I T T E D IN P A R T I A L F U L F I L 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) 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 M a y 2006 © Cameron H . Ainsworth, 2006 11 A B S T R A C T Innovative methodology is developed for Back to the Future (BTF) restoration policy analysis to aid long-term strategic planning of ecosystem-based restoration in marine ecosystems. Mass- balance and dynamic ecosystem simulation models (Ecopath with Ecosim: EwE) are developed to represent the marine system of northern British Columbia as it appeared in 1750, 1900, 1950 and 2000 A D . Time series statistics are assembled for biomass and catch, incorporating local ecological knowledge ( L E K ) from community interviews and new estimates of illegal, unreported and unregulated (IUU) fishery catch. The dynamic behaviour o f the historic models is fitted to agree with this time series information, when driven by historic catch rates and climate anomalies. Each historic period is evaluated in an optimal policy analysis for its potential to supply sustainable harvest benefits. Harvest benefits are quantified using socioeconomic and ecological indicators, including novel measures such as the Q-90 biodiversity statistic. Candidate goals for restoration are drafted based on these historic ecosystems. A new conceptual goal for ecosystem-based restoration is introduced, the optimal restorable biomass (ORB) that represents an optimized form of the historic ecosystems. It is structured to maximize sustainable harvest benefits, and to achieve a compromise between exploitation and the maintenance of historic abundance and biodiversity. Finally, restoration plans are drafted using a novel addition to Ecosim's policy search routine, the specific biomass objective function, which determines the pattern of fishing effort required to restore the depleted present-day ecosystem into one resembling a more productive O R B state. Cost-benefit analysis indicates that northern B C ecosystem restoration to an O R B state based on the 1950 ecosystem can deliver a rate o f economic return, in terms of increased fisheries yields, that is superior to bank interest. The effect of fleet structure is paramount; reducing bycatch w i l l greatly enhance the effectiveness of the fleet as a restoration tool. Restoration plans that sacrifice immediate fisheries profits tend to restore more biodiversity in a given amount of time, but a convex relationship between profit and biodiversity suggests there is an optimal rate of restoration. i i i T A B L E O F C O N T E N T S Abstract ii Table of Contents iii List of Tables v List of Figures vii List of Equations x Acknowledgements xi 1 Back to the Future 1 1.1 Introduction 1 1.2 Ecopath with Ecosim 7 1.3 Northern British Columbia 10 1.4 Structure of thesis 14 2 Harvest Policy Evaluation Techniques 19 2.1 Introduction 19 2.2 Economic index: Net present value (NPV) 19 2.3 Social utility index: Employment diversity 22 2.4 Ecological indices 23 2.5 Q-90 case study: NE Pacific ecosystems 26 3 Community Interviews 35 3.1 Introduction 35 3.2 Methods 36 3.3 Results : 42 3.4 Discussion 55 4 Estimating Illegal, Unreported and Unregulated Catch 59 4.1 Introduction 59 4.2 Methods 61 4.3 Results .' 71 4.4 Discussion 76 5 Modeling the Past and Present 82 5.1 Introduction...... , ; • 82 5.2 Fisheries 121 5.3 Ecosim parameterization 122 5.4 Assembling time series data 132 5.5 Analysis of fitted vulnerabilities 134 5.6 Validation of dynamic function 138 5.7 Discussion 142 iv 6 Evaluating Restoration Goals 146 6.1 Introduction 146 6.2 Methods 154 6.3 Results 165 6.4 Discussion 188 7 Achieving Restoration 197 7.1 Introduction 197 7.2 Methods 199 7.3 Results ; 215 7.4 Discussion : 244 8 Conclusions 250 8.1 Summary 250 8.2 An ecosystem approach to management : : 251 8.3 The developing role of ecosystem models 264 8.4 Policy recommendations 265 8.5 Concluding remarks 266 References 268 Appendices 298 Appendix 2.1 The Effect of Discounting on Fisheries 298 Appendix 2.2 Cost-benefit Analysis of Education 310 Appendix 3.1 LEK Trends of Relative Abundance 313 Appendix 4.1 BC Fisheries Timeline 314 Appendix 4.2 IUU Influences Table 322 Appendix 4.3 BC Reported Landings 330 Appendix 4.4 Average Species Weight 331 Appendix 5.1 Ecopath Parameters 332 Appendix 5.2 Ecosim Parameters 346 Appendix 5.3 Time Series 351 Appendix 5.4 Dynamic Fit to Data: 1950-2000 357 Appendix 5.5 Equilibrium Analysis of 2000 Model 361 Appendix 5.6 Comparison of Derived 2000 Model with Proper 2000 Model 363 Appendix 6.1 Policy Search Parameters 365 Appendix 6.2 Evaluation of ORB Ecosystems 367 Appendix 7.1 Input for Restoration Scenarios 372 Appendix 7.2 Candidate Restoration Trajectories 373 Appendix 8.1 Gwaii Haanas Spatial Investigations 376 Appendix 8.2 Ecospace Parameters 402 Appendix 9.1 References cited in the Appendices .' 407 V L I S T O F T A B L E S Table 1.1 Published materials appearing in this thesis 18 Table 2.1 Eight EwE models of the NE Pacific 28 Table 3.1 Percentage of interviewee comments that agree with stock assessment records 43 Table 3.2 Biomass estimates (tkm"2) used in Ecopath models compared to LEK trend 50 Table 3.3 Place names mentioned during interviews 53 Table 4.1 Incentive ratings 67 Table 4.2 Anchor point range 67 Table 4.3 Mean reported catch 67 Table 4.4 Absolute ranges of IUU catch rate for each incentive rating 72 Table 4.5 Monte Carlo input: IUU catch range 74 Table 4.6 Monte Carlo output: Mean IUU catch with 95% confidence intervals 74 Table 5.1 Data sources forNE Pacific environmental indices 128 Table 6.1 List of nine criteria for sustainable and responsible 'lost valley^ fisheries 153 Table 6.2 Lost valley fleet catch 156 Table 6.3 Lost valley fleet discards 157 Table 6.4 Rank order of ORB ecosystem performance in various evaluation fields 177 Table 6.5 Fishing rates of ORB solutions, analysis of response surface geometry and ecosystem stability 184 Table 7.1 Available settings for the specific biomass rebuilding objective function 203 Table 8.1 Criticisms of MSY and their applicability to the ORB concept 255 Table A2.1.1 CBA of education 312 Table A4.1.1 BC fisheries timeline 314 Table A4.2.1 IUU influences table 322 Table A4.3.1 BC reported landings 330 Table A4.4.1 Average species weight 331 Table A5.1.1 Species aggregation by functional group 332 Table A5.1.2 Basic parameters for all periods 336 Table A5.1.3 Diet composition 338 Table A5.1.4 Landings data for all time periods (tkm2) 342 Table A5.1.5 Discard data for 1950 and 2000 (tkm-2) 344 Table A5.1.6 Market prices ($jkg" ') for 2000 BC fleet 345 Table A5.2.1 Juvenile/adult stage transition parameters for all models 346 Table A5.2.2 Feeding parameters for 1950 347 Table A5.2.3 Trophic flow parameters for 1950 348 Table A5.3.1 Biomass time series data (t-km2): 1900-1950 351 Table A5.3.2 Biomass time series data (tkm"2): 1950-2000 352 v i Table A5.3.3 Catch time series data (tkm"2): 1900-1950 353 Table A5.3.4 Catch time series data (tkm2): 1950-2000 354 TableA5.3.5 Fishing mortality time series data (yr"'): 1900-1950 355 Table A5.3.6 Fishing mortality time series data (yr1): 1950-2000 356 Table A5.6.1 Comparison of derived 2000 model with proper 2000 model 363 Table A6.1.1 Market prices ($kg_l) for lost valley fleet 365 Table A6.1.2 Biomass/production (B/P) ratios by functional group 366 Table A6.2.1 Functional group biomass (tkm2) for selected ORB ecosystems 369 Table A6.2.2 Fisheries landings by gear type (tkm2) for selected ORB ecosystems 371 Table A7.1.1 Catch profile for maxdex fleet 372 Table A8.1.1 Ecospace habitat definitions 383 Table A8.1.2 Group behaviour guidelines used to standardize Ecospace functional groups 390 Table A8.2.1 Habitat occupancy 402 Table A8.2.2 Fishery activity by habitat 403 Table A8.2.3 Salmon straying rates 403 Table A8.2.4 Dispersal parameters 404 Table A8.2.5 Ecospace output region definitions 405 Table A9.2.1 IG discounting case study references cited in Appendix 2.1 407 Table A9.2.2 Cost-benefit analysis of education references cited in Appendix 2.2 408 Table A9.4.1 BC fisheries timeline references cited in Appendix 4.1 409 Table A9.4.2 Average species weight references cited in Appendix 4.4 414 Table A9.4.3 Illegal catch anchor point references cited in Appendix 4.2 415 Table A9.4.4 Discard anchor point references cited in Appendix 4.2 416 Table A9.4.5 Unreported catch anchor point references cited in Appendix 4.2 417 Table A9.5.3 Biomass, catch and effort time series data references cited in Appendix 5.3 418 Table A9.8.1 Spatial investigations for Gwaii Haanas references cited in Appendix 8.1-. 421 L I S T O F F I G U R E S Figure 1.1 Biodiversity and species abundance decline caused by fisheries 2 Figure 1.2 The Back to the Future approach to marine ecosystem restoration 4 Figure 1.3 Northern BC study area 12 Figure 2.1 Q-90 statistic definition 25 Figure 2.2 Dynamic ecosystem biodiversity (Q-90) of three example Ecopath with Ecosim simulations 29 Figure 2.3 Absolute Q-90 value at baseline (year zero) for eight northeastern Pacific Ecopath models 30 Figure 2.4 Change in Q-90 index after 30 years of fishing for eight EwE models of the NE Pacific 30 Figure 2.5 Q-90 sensitivity to changes in system biomass structure 31 Figure 2.6 Q-90 sensitivity to changes in ecosystem structure using three depletion filter thresholds 32 Figure 3.1 Fraction of comments that agree with DFO records by functional group 44 Figure 3.2 Interviewee agreement with stock assessment data by career length 45 Figure 3.3 LEK abundance trend versus stock assessment 48 Figure 3.4 Rank correlation of LEK abundance trend versus stock assessment 49 Figure 3.5 Correlation of LEK relative abundance trend and stock assessment with model ouputs 52 Figure 3.6 A map of the study area showing the number of LEK comments indicating species presence 54 Figure 4.1 A time series of numerical influence factors assigned semi-quantitative 'incentive' ratings 64 Figure 4.2 Salmon recreational catch estimates 66 Figure 4.3 Cumulative probability distribution of missing catch 70 Figure 4.4 Likely range of groundfish discards 70 Figure 4.5 Estimates of missing catch for salmon and groundfish fisheries 75 Figure 4.6 Total estimated extractions in BC salmon and groundfish fisheries 76 Figure 5.1 EwE's predicted climate anomalies versus their strongest correlating environmental indices ::' 128 Figure 5.2 Correlation of primary production and herring recruitment anomalies with environmental indices 129 Figure 5.3 Predicted and observed herring trend (1950-2000) under three conditions of climate forcing 130 Figure 5.4 Predicted and observed variance of group biomass trajectories (1950-2000) 131 Figure 5.5 Rank order of vulnerabilities in the fitted 1950 model versus predator and prey trophic level 135 Figure 5.6 Log vulnerabilities in fitted 1950 model versus predator and prey trophic level 135 Figure 5.7 Evaluation of short-cut methods used to parameterize Ecosim vulnerabilities 137 Figure 5.8 Group biomass predicted in 2050 by derived and proper 2000 models after fishing release 140 Figure 5.9 Biomass change predicted by the derived and proper 2000 models after fishing release 141 Figure 5.10 Direction of biomass change predicted by proper and derived 2000 models after fishing release 142 Figure 6.1 Optimal Restorable Biomass (ORB) concept 148 Figure 6.2 Cluster analysis of group biomass configurations for two example ORB ecosystems 166 Figure 6.3 Value equilibriums for ORB ecosystems based on various historical periods 167 Figure 6.4 ORB equilibrium catch value per gear type 170 Figure 6.5 ORB equilibrium value by group under various harvest objectives 171 Figure 6.6 Social utility provided by ORB ecosystems based on various historical periods 173 Figure 6.7 Biodiversity of ORB ecosystems based on various historical periods 175 Figure 6.8 Biodiversity of historic ecosystems under optimal fishing policies using lost valley fleet structure 176 Figure 6.9 Profit and biodiversity of ORB equilibriums based on 1750, 1900, 1950 and 2000 periods 179 Figure 6.10 Social utility provided by ORB ecosystems based on 1750, 1900, 1950 and 2000 periods 181 Figure 6.11 Response surface geometries 183 Figure 6.12 Biomass depletion risk of ORB solutions, considering Ecopath parameter uncertainty 186 Figure 6.13 The effects of data uncertainty on ORB equilibrium values determined by Monte Carlo 187 Figure 7.1 Controls added to Ecosim's policy search interface for SB algorithm 203 Figure 7.2 Three models describing marginal improvement in SB function as group biomass approaches goal.... 206 Figure 7.3 Constrained marginal improvement model (MIM) 208 Figure 7.4 Dynamic progress display form to monitor rebuilding success of the SB algorithm 211 Figure 7.5 Conceptual diagram showing cost-benefit analysis 214 Figure 7.6 Performance of SB algorithm towards achieving historic 1950 ecosystem structure 217 Figure 7.7 Commercial biomass increase under various restoration plans 218 Figure 7.8 Principle components analysis showing ecosystem configurations after restoration 219 Figure 7.9 Average improvement in functional group biomass towards target level after restoration 221 Figure 7.10 End-state group biomass after rebuilding relative to target 1950 goal biomass 222 Figure 7.11 End-state group biomass after rebuilding relative to target 1900 goal biomass 223 Figure 7.12 End-state profit and biodiversity of restoration plans targeting the historic 1950 ecosystem 225 Figure 7.13 Progress towards goal ecosystems 1950 and 1900 for all diagnostic optimizations 226 Figure 7.14 End-state ecosystem condition of nine restoration plans targeting the biodiversity ORB 228 Figure 7.15 End-state profit after 50 years of restoration versus sum of squares against goal ecosystem 229 Figure 7.16 Change in average system trophic level and biodiversity following restoration 231 Figure 7.17 Best reduction in sum of squares versus target system achieved by SB algorithm 232 Figure 7.18 End-state profit and biodiversity after restoration for all 50-year restoration plans tested 233 Figure 7.19 Worked example of a 30 year ecosystem restoration plan 234 Figure 7.20 Net present value of restoration plans achieving a minimum reduction in residuals versus goal 237 Figure 7.21 Equilibrium level profit and biodiversity achieved by restoration scenarios 241 Figure 7.22 Net present value of restoration scenarios 242 Figure 7.23 Internal rate of return (IRR) making restoration/harvest scenarios economically worthwhile 243 Figure A2.1.1 Stability analysis of dynamic ecosystem model 302 Figure A2.1.2 Real price of cod based on harvest from Atlantic Canada 304 Figure A2.1.3 Historic cod biomass trajectory estimated from VPA versus EwE optimal trajectories 305 Figure A2.1.4 Optimal end-state biomasses after 16 years of harvest under various discounting methods 306 Figure A2.1.5 Optimal end-state catches after 16 years of harvest under various discounting methods 307 Figure A2.1.6 Net present value of 40-year harvest profile based on real-world data and optimum solutions 308 Figure A2.1.7 Generational share of catch after 40 years for three harvest profiles 308 Figure A2.1.8 Sensitivity analysis showing the effect of discount rate on the optimal end-state biomass 309 Figure A2.2.1 Costs and benefits of education in BC discounted from a 1981 time perspective . 311 Figure A3.1.1 LEK trends of relative abundance 313 Figure A5.4.1 Biomass fit to data (tkm2) 357 Figure A5.4.2 Catch fit to data (tkm2) 360 Figure A5.5.1 Equilibrium analysis of 2000 model 361 Figure A6.2.1 Equilibrium harvest benefits from ORB ecosystems derived from 1750, 1900, 1950 and 2000 367 Figure A7.2.1 Restoration scenarios using the BC fishing fleet 373 Figure A7.2.2 Restoration scenarios using the lost valley fishing fleet 374 Figure A7.2.3 Restoration scenarios using the maxdex fishing fleet 375 Figure A8.1.1 Ecospace habitats 382 Figure A8.1.2 Bathymetry 382 Figure A8.1.3 Tidal speed 382 Figure A8.1.4 Primary production forcing pattern used in Ecospace 385 Figure A8.1.5 Modeled current circulation 387 Figure A8.1.6 Ecospace output regions used to summarize results by area 391 Figure A8.1.7 Catch by output region 393 Figure A8.1.8 Regional effects of NMCA area closures on landings 394 Figure A8.1.9 Equilibrium trophic level of catch in regions adjacent to MPA 395 Figure A8.1.10 Group biomass change within MPA resulting from area closures.: 397 Figure A8.1.11 Equilibrium biodiversity in MPA and adjacent regions following fishery closure 398 Figure A8.1.12 Equilibrium state changes within the MPA under zero to twelve month area closures 399 Figure A8.2.1 Value of catch per gear type 406 X L I S T O F E Q U A T I O N S Equation 1.1 Ecopath production equation 8 Equation 1.2 Ecopath consumption equation 8 Equation 1.3 Ecosim biomass dynamics 9 Equation 2.1 Conventional discounting model 20 Equation 2.2 Discount factor 20 Equation 2.3 Intergenerational discounting model 21 Equation 2.4 Intergenerational discount factor 21 Equation 2.5 Shannon entropy function 22 Equation 2.6 Shannon-Weaver biodiversity model 22 Equation 2.7 Q-90 statistic definition 25 Equation 2.8 Q-90 10th percentile 26 Equation 2.9 Q-90 90th percentile 26 Equation 4.1 Likely error range used for IUU Monte Carlo analysis 68 Equation 4.2 Probability density function of triangular IUU catch error distribution 68 Equation 6.1 Policy search routine objective function....! 159 Equation 7.1 SB algorithm summation term 200 Equation 7.2 Proximity to goal index (9) used by SB algorithm 201 Equation 7.3 Proximity to goal index (9) modified for biomass unit of improvement 202 Equation 7.4 Proximity to goal index (9) modified for combined unit of improvement 202 Equation 7.5 Linear marginal improvement valuation model 204 Equation 7.6 Quadratic marginal improvement valuation model 204 Equation 7.7 Gamma marginal improvement valuation model : 205 Equation 7.8 Biomass term substitution for functional groups already close to target in SB algorithm 207 Equation 7.9 Fast-track modification to SB algorithm summation term 210 Equation A2.1 Cost-abundance relationship of fishing 303 XI A C K N O W L E D G E M E N T S I extend deep gratitude to my supervisor, Tony Pitcher, for his advice and assistance, and the many opportunities he gave me. This project could not have been done without the support of the Fisheries Centre. I thank Daniel Pauly, V i l l y Christensen, Rashid Sumaila, Carl Walters, Nigel Haggan, Sheila Heymans, Sylvie Guenette and many other friends and colleagues who supported my work. Many thanks go to Alan Sinclair and my research committee. I especially want to express my most sincere gratitude to Les Lavkulich, who went beyond his obligation to help me. Interviews were conducted under the guidelines and approval o f the U B C Ethical Review Committee. On behalf of myself, Coasts Under Stress and the U B C Fisheries Centre, I would very much like to thank all our interviewees for lending their time and expertise to this project. I also thank the following organizations for project funding: the University o f British Columbia Graduate Fellowship; Natural Sciences and Engineering Research Council ; Coasts Under Stress, World Wildlife Fund Canada and the B C Ministry of Water, Land and A i r Protection. I offer thanks and love to Er in Foulkes for her support, her patience and encouragement. To my dear parents, Herb and V i Ainsworth, who did everything to help me, I dedicate this report. 1 1 B A C K T O T H E F U T U R E The significant problems we face cannot be solved at the same level of thinking we were at when we created them. Albert Einstein Qu. Dukas and Hoffman (1979) 1.1 Introduction For thousands of years, humans have been exploiting the seas for food. Paleoecological and archaeological evidence records the significant impacts that we have caused (Jackson et al. 2001). Fishing is thought to have become important to humans during the Upper Paleolithic period, 10 to 30 thousand years ago (Bar-Yosef, 2004), although fish may have contributed to 'our diet much earlier than that (Yellen et ai, 1995; Fiore et al, 2004). From the earliest harpoons, nets and bone hooks, each advancement made in capturing fish must have opened up new habitats and new species to exploitation. But it was not until the development of industrial fisheries, less than 200 years ago, that the major depletion of marine systems began (e.g., Myers and Worm, 2003; Pauly et al, 2005). With the advent of sail, steam and diesel powered boats, areas became accessible that were once out of reach. The end of the Second Wor ld War saw the modernization of fleets, including the addition of at-sea freezers, radar navigation, acoustic fish finders and other conveniences that increase catching power (Pauly et al, 2002). The trend continues today with satellite navigation systems and communication networks that make fishing easier, safer and more efficient than ever before. Unfortunately, a step up in technology has proven to be a step down in the biodiversity and abundance of marine ecosystems (Pitcher and Pauly, 1998) (Fig. 1.1). The effect is cumulative. Globally, fisheries are in crisis (Pauly et ai, 1998; Myers and Worm, 2003). Many factors can potentially contribute to the decline of fish stocks and the failure of fisheries. Climate is known to influence productivity of fish populations (e.g. Beamish et al, 1995; 2 Polovina, 2005), and changes in climate may be related to long-duration environmental cycles that are poorly understood (Finney et al, 2002). Other culprits like coastal development, land- based pollution and marine industries are also identified. In some cases, scientific error may contribute to fishery declines (e.g., Hutchings, 1996). However, it is overfishing that many scientists now believe has been the primary driver of fisheries collapse world-wide. Pleistocene Recorded Present Near history day future TimeC^> Figure 1 .1 Biodiversity and species abundance decline caused by fisheries. The stepped downward line represents the serial depletion of marine ecosystems. Each fishing innovation, from simple harpoons to factory trawlers, opens up new species and habitats to exploitation. Horizontal arrows show sustainable use, which could have been achieved, in theory, at any level of ecosystem abundance. The three-way arrow shows policy options currently open to us. Modified from Pitcher and Pauly (1998). Fishing overcapacity is viewed by some as the single greatest threat to sustainable fisheries (Mace, 1997; Greboval and Munro, 1999; Ward et al, 2001). Ludwig (1993) suggested that overcapitalization in the fishing fleet is driven by a dangerous bioeconomic ratcheting effect, where good fishing years encourage over-investment and bad fishing years demand government subsides to keep the industry afloat. Compounding the problem, investors in the fishing industry may also expect a rate o f return that is comparable to other types o f enterprises, but cannot be supported sustainably by the natural growth rate of fish populations (Clark, 1973). Therefore, 3 overfishing is driven by complex social, economic and political factors. A n y lasting solution w i l l require cooperation across disciplines, and the commitment of many stakeholders groups. To form this alliance we w i l l need tools that can weigh the interests of all resource users, we w i l l need to improve our understanding of human impacts on marine systems, and we w i l l need to agree on a proper goal for fisheries management. Although sustainability is usually pursued as an explicit objective in regulated fisheries, repeated failures indicate that it is rarely achieved in practice (Ludwig et al, 1993; Botsford et al., 1997). When environmental conditions are favorable, sustainable fisheries may be achieved without careful restraints on human activities. But when climate turns against the interests of people, which may be increasingly of our own causing, our management systems need to operate according to strict precautionary principles. Sustainability is now too low of standard to aim for; we realize this when we look to the past as a reference point and understand the enormous benefits that a healthy ecosystem is capable of providing. A new perspective on fisheries management Many traditional target species have declined to only a fraction of their abundance prior to the industrialization of fisheries (Christensen et al., 2003; Worm and Myers, 2003; Reid et al, 2005; Rosenberg et al, 2005; Ward and Myers, 2005). The public, and scientists as wel l , are generally unaware of the magnitude of the historic decline. It is perhaps because of Pauly's (1995) shifting baseline syndrome. He suggested that one's concept or perception of ecosystem abundance is based on a mental benchmark set at the beginning of the career. A s the ecosystem is slowly degraded, each generation accepts a lower standard as the rule. This can apply to fisheries scientists as well as the general public. Considering the poor state of the oceans, it has been argued that the proper goal for fisheries management should not be to sustain current fish populations, but rather to restore them to historic levels (Pitcher et al. 1998; Pitcher and Pauly, 1998; Pitcher, 2001). The Back to the Future (BTF) approach to restorative marine ecology offers a new perspective on what management objectives should be (see Pitcher, 2001a, 2004, 2005; Pitcher et al, 1999, 4 2004, 2005; Ainsworth and Pitcher, 2005b). Under the BTF approach, an initial objective for any ecosystem-based restoration initiative should be to establish long-term goals for restoration. Candidate goals should be quantitatively evaluated for their potential to provide benefits to stakeholders and maintain ecological health. Using ecosystem models, BTF simulates fishing of historic ecosystems to determine their long- term sustainable production potential. From this we can estimate what resource value has been lost due to human influences, and what a restored ecosystem might be worth to society. Fig. 1.2. shows a schematic illustration of the BTF concept. The symbols in Fig. 1.2 document many new and unconventional sources of information that must be relied upon to create whole ecosystem models of the past. Although there will be some aspects of historical ecosystems that are unknowable, multidisciplinary data on fish stocks and the environment can be used to form a picture of what the ecosystem looked like before heavy exploitation. A N C I E N T P A S T P A S T P R E S E N T A L T E R N A T I V E F U T U R E S Figure 1.2 The Back to the Future approach to marine ecosystem restoration. Triangles represent trophic pyramids; height is directly related to biomass and internal connectance. Internal boxes show biomasses of representative species through time, with closed circles indicating extirpations. Ecosystems of the past contained longer trophic chains than they do now, greater biodiversity and predator biomass. The BTF approach advocates setting restoration goals based on historic ecosystems (right). Ecosystem models are constructed to evaluate various periods using historical documents (paper sheet symbol), data archives (tall data table symbol), archaeological data (trowel), the traditional environmental knowledge of indigenous Peoples (open balloons) and local environmental knowledge (solid balloons). Reproduced from Pitcher et al. (2004). 5 Historic ecosystems may hold special resonance with stakeholders as restoration goals if people can appreciate the long-term impacts that fisheries have had (Pitcher, 2000; Pitcher and Haggan, 2003). There may also be a scientific rationale for selecting restoration goals based on historic ecosystems. Because they existed, their relative species compositions may represent workable ecosystem goals, more so than an arbitrary design. If we can allow for environmental changes that have occurred since their time, then historic ecosystems can serve as an analogue for the future. The study of historic ecosystems can inform us as to what level of abundance and productivity can be expected from a natural system, given any constraints that regional oceanographic conditions impose. Pitcher et al (2004) imagined a bright future for marine fisheries, where the ecosystem is restored to something resembling a historic condition. They likened the reconstituted ecosystem to a lost valley1, an untouched area as discovered in Sir Arthur Conan Doyle's "The Lost World". This lost valley offers humans a second chance to responsibly use the marine ecosystem. BTF asks the following questions: what might this lost valley look like, how might we sustainably harvest it, and what would be the costs and benefits of rebuilding to this goal? To answer these questions, a new methodology has been developed that makes use of the ecosystem simulator, Ecopath with Ecosim (EwE: Christensen and Pauly, 1992; Walters et al, 1997; Christensen and Walters, 2004a). A quantitative goal for ecosystem based approaches Quantitative techniques are often called upon to help set safe removal rates. Numerical targets and reference points have been established to guide fisheries management and allow the responsible use of living marine resources. Historically, a widely used paradigm has been the maintenance of maximum sustainable yield (MSY) from fisheries. For a given stock size, it is the theoretical amount of catch that can be taken each year, under average environmental conditions, without influencing the abundance of the stock. The "puritanical philosophy" identified by Larkin (1977), to take only surplus stock production and forever maintain MSY 1 The term lost valley was suggested by Prof. Daniel Pauly (Pitcher et ai, 2004). 6 once promised to solve all fisheries issues. N o w people question whether M S Y has ever been achieved in practice and whether it is achievable in theory (Larkin, 1977; Sissenwine, 1978; Punt and Smith, 2001). Amendments have been proposed to address the well-known inadequacies of M S Y ; for example, optimum sustainable yield ( O S Y : Roedel, 1975), maximum economic yield ( M E Y ) and Fn.i (see Hilborn and Walters, 1992). However, some question whether proper fisheries management is at all possible through a reductionist approach (Ludwig et al. 1993), which is the traditional mechanism of single species science. More and more, scientists are turning towards ecosystem based approaches in the hopes that a holistic view of ecosystem functioning w i l l provide a better foundation for fisheries management. Ecosystem based management ( E B M ) could benefit from a new objective reference point; one that considers the health and productivity of the ecosystem as a whole. Such a standard could do for E B M what indices like M S Y , O S Y , M E Y and Fo.i did for single species management - provide a quantitative policy goal that can potentially set the benchmark for sustainable use. This volume presents a new conceptual target for ecosystem based approaches. It is the optimal restorable biomass (ORB) , an equilibrium biomass configuration for the ecosystem that maximizes sustainable harvest benefits, and is designed to meet specific criteria for ecosystem health. O R B is calculated based on historic ecosystems. It is the species biomass vector, defining the relative abundance of each ecosystem component, that would naturally result after the long-term responsible use of historic ecosystems. Sidestepping the serial depletion of stocks witnessed in reality, it takes into account the activities of fisheries and determines the best compromise between maintaining historic abundance and diversity, while still providing for the needs of humans. Mace (2001) pointed out that even i f we could establish suitable goals for whole-ecosystem restoration, it is doubtful whether we would have the capability to manipulate the ecosystem into the desired state. The work presented in this volume offers a first step towards developing an integrated approach to management that can accomplish just that. Tools and techniques developed here for use with E w E models provide a strategic aid to help draft restoration plans 7 that would use selective fishing as a tool to manipulate the marine ecosystem, and ultimately restore it to some former level of abundance and productivity. 1.2 Ecopath with Ecosim E w E provides a fresh tool to explore the complex interactions of marine organisms. To enable multi-sector fishery policy analysis, the competing effects of fisheries must be considered, as well as trophic interactions throughout the food web. Single species models, versatile and informative, are completely indispensable to whole ecosystem work, as they form the basis of our understanding for key ecosystem components. Nevertheless, they are limited in scope. Even traditional multi-species models can isolate and examine only a small number of interactions, and strict data requirements limit these analyses to well understood ecosystem components. Although ecosystem models offer no panacea, they can provide a new perspective on population dynamics and help us understand unintuitive processes. They can complement well-established analysis methods and provide an integrated overview of ecosystem functioning and the impact o f fisheries. The mass-balance approach, in particular, makes it possible to construct a virtual ecosystem without the need for exhaustive supporting science. Invented by Polovina (1984) and advanced by Christensen and Pauly (1992, 1993), Walters et al. (1997, 1998) and Christensen and Walters (2004a) among others, E w E is a mass-balance trophic simulator that acts as a thermodynamic accounting system. Summarizing all ecosystem components into a small number of functional groups (i.e., species aggregated by trophic similarity), the box model describes the flux of matter and energy in and out of each group, and can represent human influence through removals and other ways'. There are now dozens of published articles that use E w E to describe ecosystems, qualify data, test hypotheses and demonstrate other applications (see review in Christensen and Walters, in press). E w E has been used in actual fisheries management, but to a limited extent. Reviews and criticisms of the E w E approach are provided by Fulton et al. (2003), Christensen and Walters (2004a), and Plaganyi and Butterworth (2004). 8 Ecopath The static model Ecopath (Polovina, 1984; Christensen and Pauly, 1992) implicitly represents all biotic components of the ecosystem. The model operates under two main assumptions. The first assumption is that biological production within a functional group equals the sum of mortality caused by fisheries and predators, net migration, biomass accumulation and other unexplained mortality. Eq. 1.1 expresses this relationship: B, • {P/B), = Y, + £ Bj • (Q/B)j • DC, + E, + BA, + B, {p/B), • (l - EE,) Equation 1.1 7=1 Where B , and B , are biomasses of prey (/) and predator (/"),• respectively; P/B, is the production/biomass ratio; Y , is the total fishery catch rate of group (/); Q/B, is the consumption/biomass ratio; DC,y is the fraction of prey (/) in the average diet o f predator (/'); E, is the net migration rate (emigration - immigration); and B A , is the biomass accumulation rate for group (i). E E , is the ecotrophic efficiency; the fraction of group mortality explained in the model; The second assumption is that consumption within a group equals the sum of production, respiration and unassimilated food, as in eq. 1.2. B-(Q/B) = B-(P/B)+(l-GS)-Q-(\-TM)-P + B(Q/B)-GS Equation 1.2 Where GS is the proportion of food unassimilated; and T M is the trophic mode expressing the degree of heterotrophy; 0 and 1 represent autotrophs and heterotrophs, respectively. Intermediate values represent facultative consumers. Ecopath uses a set of algorithms (Mackay, 1981) to simultaneously solve n linear equations of the form in eq. 1.1, where n is the number of functional groups. Under the assumption of mass- balance, Ecopath can estimate missing parameters. This allows modelers to select their inputs. 9 Ecopath uses the constraint of mass-balance to infer qualities of unsure ecosystem components based on our knowledge of well-understood groups. It places piecemeal information on a framework that allows us to analyze the compatibility of data, and it offers heuristic value by providing scientists a forum to summarize what is known about the ecosystem and to identify gaps in knowledge. Ecosim Ecosim (Walters et al, 1997) adds temporal dynamics to turn the mass-balance model into a simulation. It describes biomass flux between groups through coupled differential equations derived from the first Ecopath master equation. The set of differential equations is solved using the Adams-Bashford integration method by default. Biomass dynamics are described by eq. 1.3. dB " i \ " i \ ^ - g ^ f \ B p B ) - Y J f \ B l , B ^ l l - ( M l 4 - F ^ e y B i Equation 1.3 Where dB/dt represents biomass growth rate of group (i) during the interval dt; gi represents the net growth efficiency (production/consumption ratio); It is the immigration rate; Mt and F , are natural and fishing mortality rates of group (i), respectively; e, is emigration rate; and f(Bj,B,) is a function used to predict consumption rates of predator (J) on prey (/) according to the assumptions of foraging arena theory (Walters and Juanes 1993; Walters and Korman, 1999; Walters and Martell , 2004). The principle innovation in Ecosim considers risk-dependant growth by attributing a specific vulnerability term for each predator-prey interaction. The vulnerability parameter is directly related to the carrying capacity of the system, and it describes the maximum increase in the rate of predation mortality on a given prey. A high value represents a top-down (Lotka-Volterra) interaction, a low value represents a bottom-up (donor-driven) interaction, and an intermediate value indicates mixed trophic control. Variable speed splitting enables Ecosim to simulate the 10 trophic dynamics of both slow and fast growing groups (e.g., whales/plankton), while juvenile/adult split pools allow us to represent life histories and model ontogenetic dynamics. A new multi-stanza routine in Ecopath (Christensen and Walters, 2004a) back-calculates juvenile cohorts based on the adult pool biomass and on life history parameters. The multi-stanza routine has replaced former the split-pool method; however, it was not available at the time of this work. A s such, recruitment to juvenile stanzas in this model are determined by Ecosim using a Deriso- Schnute delay difference model (Walters et al, 2000). Ecospace Ecospace (Walters et al 1998) models the feeding interactions of functional groups in a spatially explicit way. A simple grid represents the study area, and it is divided into a number of habitat types. Each functional group is allocated to its appropriate habitat(s), where it must find enough food to eat, grow and reproduce - while providing energy to its predators and to fisheries. Each cell hosts its own Ecosim simulation and cells are linked through symmetrical biomass flux in four directions; the rate of transfer is affected by habitat quality. Optimal and sub-optimal habitat can be distinguished using various parameters such as the availability of food, vulnerability to predation and immigration/emigration rate. B y delimiting an area as a protected zone, and by defining which gear types are allowed to fish there and when, we can explore the effects of marine protected areas ( M P A s ) and test hypotheses regarding ecological function and the effect of fisheries. Many authors have used Ecospace in this capacity (e.g., Walters et al, 1998; Beattie, 2001; Pitcher and Buchary, 2002a/b; Pitcher et al, 2001; Salomon et al, 2002; Sayer et al, 2005). 1.3 Northern British Columbia Whenever viable fisheries are lost, communities and cultures that have traditionally relied on the sea can be impacted in deep and lasting ways. This is especially true when social and cultural values are tied closely to the sea. That is the case in northern British Columbia (BC). Fishery failures, such as the herring collapse o f early 1960s, the Northern abalone collapse of the 1980s, 11 and the present decline o f the salmon fisheries displaces workers, disrupts communities and sabotages a sustainable source of revenue. This volume evaluates restoration scenarios for northern B C that would return the ecosystem to historic conditions of biodiversity and abundance. For this, I create ecosystem models o f northern B C at various points in history: 1750, 1900, 1950 and 2000 A D . The models are described in Chapter 5. The 1750 model represents the marine ecosystem prior to contact by Europeans. It contains the most abundant array of marine fish and animals, although it does not represent an unexploited system since indigenous coastal human populations are thought to have relied on the sea to a great extent (Haggan et al, in press; Turner et al, in press). A model o f 1900 represents the ecosystem as it appeared prior to the industrialization of fisheries, and before the advent of major advances in fishing technology such as steam trawlers. The 1950 model demonstrates what the ecosystem looked like during the heyday o f the Pacific salmon fisheries, and before most major depletions of commercial fish populations. Finally, the present-day model, 2000, provides a contemporary representation of the ecosystem. It is from this vantage point that restoration plans are drafted. Physical area 12 This study models the marine environment of northern B C , from the northern tip of Vancouver Island to the southern tip of the Alaskan panhandle, including the waters of Dixon Entrance (DE), Hecate Strait (HS) and Queen Charlotte Sound (QCS) (Fig. 1.3). It covers the shelf and continental slope, about 70,000 k m z o f ocean, using the same delineation as in Beattie (2001), including Department of Fisheries and Oceans (DFO) statistical areas 1-10. Oceanography of the region was described by Crean (1967), Thomson (1981), Ware and McFarlane (1989) and Crawford (1997). Dixon Entrance G r a h a m Is M o r e s b y Is ; si hfecate Strait '/ «? 4 i Queen Charlotte Sound c 280 Kilometers The area roughly corresponds to the „. „ . , „ t , ° J r Figure 1.3 Northern BC study area. The study area eastern region of the Coastal includes the shelf waters of Queen Charlotte Sound, Hecate Downwelling Domain identified by Ware S t r a i t a n d D i x o n E n t r a n c e ( D F O statistical areas l-io). and McFarlane (1989). Water movement is influenced by the counterclockwise flow of the Alaska gyre, which creates a northeastern flowing Alaskan current year round. The Alaskan current enters Q C S and extends northward along the coast into H S . In the south of the study region there is a transitional zone, where the clockwise flowing California Current diverges from the Alaska current and flows south. Coastal convergence occurs mainly on the west coast of Haida Gwai i and along the mainland shoreline of Q C S and HS . The shelf area is relatively shallow, more than two thirds of the total area is less than 200 m in depth. Three major gullies transect the continental shelf from east to west. Crossing H S and 13 terminating south of Moresby Island (S. Haida Gwaii) is the Moresby Trough. Q C S is divided twice, by Mitchel l ' s Trough in the north and Goose Island Trough in the south. The mainland coastline is rugged, with many islands and inlets. Biological system The waters o f northern B C host a diverse marine biota. With the greatest human populations concentrated in the south of the province, the marine ecosystem of northern B C remains relatively intact compared to the Strait o f Georgia and Southern B C . The complex coastline provides a range of habitats including rock, sand and mud flats, with various degrees of wave exposure. With its large expanse open to the Pacific Ocean, Q C S offers an 'oceanic' habitat which is subject to oceanographic intrusions. H S and D E provide a more shallow and protected zone. Deep troughs and productive banks in Q C S support large populations of rockfish, flatfish and demersal fish species. The coastal corridor is migrated annually by five salmon species, each an important commercial asset. Important nesting areas for seabirds, like cormorants (Phalacrocoracidae), gulls (Laridae) and auklets (Alcidae), are located along the coastal islands and on the mainland. Large kelp beds covering much of the coast provide habitat for juvenile fish, and support a large population of benthic invertebrates. Echinoderms like urchins, sea stars and sea cucumbers are common. Also occurring in the tidal and subtidal zones are massive beds of bivalves and barnacles. Seals and seal lions occur throughout northern B C . There are five species of pinnepeds: two Phocidae (true seals) and three Otariidae (eared seals). Cetacean species like killer whale (Orcinus orca), minke whales (Balaenoptera acutorostrata) and dolphins can occur throughout the year, and there are seasonal populations of migratory gray (Eschrichtius robustus) and humpback whales (Megaptera novaeangliae). Four hexactinellid sponge reefs in central Q C S and HS are noted for their uniqueness and conservation utility (Conway, 1999; Sloan et al. 2001; Ardron, 2005). 14 Fisheries Commercial fisheries in northern B C are conducted by seine boats, gillnetters, trawlers (or draggers), trailers, demersal traps, hook and line, scuba diving and other gear types. Commercial capture fisheries yielded a value of $359 mil l ion in 2004 (DFO, 2004d), contributing a meagre 0.1% to the provincial gross domestic product. B y comparison, recreational fisheries and their supporting industries contributed an estimated $675 mil l ion in the same period, while aquaculture, mainly for Atlantic salmon (Salmo solar), contributed another $287 mill ion. Pacific salmon constitutes the most valuable component of the commercial catch. Salmon species include sockeye (Oncorhynchus nerka), pink (O. gorbuscha), chum (O. keta), chinook (O. tshawytscha) and coho (O. kisutch). The large majority of salmon captures is achieved by the seine net fishery, followed by gillnets and trailers. The halibut (Hippoglossus stenolepis) fishery is second in importance after the salmon species. It mainly uses longline gear and trolling methods. Herring roe purse seine fisheries and shrimp trawl and trap fisheries follow. Fisheries for rockfish, sablefish (traps), crabs, lingcod and other invertebrates also contribute to the coastal economy. 1.4 Structure of thesis Chapter 1 summarizes the Back to the Future approach to restorative marine ecology. It describes the E w E ecosystem modeling software and provides background on the study area of northern B C . A new conceptual and quantitative target for ecosystem restoration is introduced: optimal restorable biomass (ORB) . Chapter 2 introduces quantitative indices used throughout this volume to evaluate harvest benefits in economic, social and ecological terms. Case studies are provided to demonstrate the use of these indices within the E w E framework and their application to restoration ecology. Economic valuation indices include net present value (NPV) , calculated using conventional and intergenerational discounting approaches. A case study examines the Newfoundland cod collapse, and demonstrates that intergenerational valuation of fisheries resources advocates better maintenance of fish stocks than conventional valuation. A n employment diversity index is 15 developed to help quantify social benefits of fishing, and a new ecological index is introduced to describe species biodiversity, the Q-90 biodiversity statistic. A case study compares biodiversity impacts of fishing policies using the Q-90 index across eight E w E models of the N E Pacific, and demonstrates that the index is invariant to model structure. In Chapter 3, I describe the B T F community interviews conducted in northern B C , and explain the methodology used to turn the subjective comments of interviewees into a relative abundance trend for E w E functional groups. These trends help set the dynamics for data-poor functional groups in the northern B C models. The perceived changes in biomass are compared with stock assessment information and with preliminary model outputs as a diagnostic tool used to identify problem dynamics. Chapter 4 quantifies illegal, unreported and unregulated ( IUU) catch in B C for salmon and groundfish fleets using a new subjective methodology. It is part of a larger effort to establish reliable estimates of extractions, which can be used to tune the dynamic models. A timeline of B C fisheries is compiled that includes regulatory, technological and political changes likely to have affected the quantity o f unreported catch. From this, a semi-quantitative Monte Carlo procedure provides estimates of I U U catch for each 5 year period between 1950 and 2000 based on qualified anchor points (i.e., real-world estimates of misreporting from the literature and expert opinion). Chapter 5 explains the northern B C models in detail, including basic parameterization and all fitting procedures used to improve dynamic predictions. Climate factors are addressed that may have influenced observed ecosystem dynamics, and some generalizations are drawn concerning predator-prey trophic vulnerabilities: Ecosim's chief dynamic parameters. A novel procedure is introduced whereby the dynamics of ancient E w E models are tuned based on the fitted dynamics of more recent models. This assumes stationarity in density-dependent foraging tactics. It is demonstrated that this method improves predictions by the 1900 northern B C model over other common parameterization methods. In Chapter 6, I demonstrate O R B as a new ecosystem-based goal for restoration. Various O R B restoration targets are determined from historic ecosystems. O R B equilibriums are structured to 16 maximize socioeconomic or ecological benefits in varying degrees, and a trade-off spectrum of available benefits is presented for each historical period. This analysis demonstrates what wealth we have sacrificed over the last 250 years through our unsustainable fishing practices, and it also demonstrates what restoration could be worth to stakeholders in monetary and non-monetary terms. New techniques are used to relate the geometry of the optimization response surface to various policy considerations. Uncertainties surrounding historic model parameter estimates are also considered in the ORB solutions through use of a Monte Carlo routine. Chapter 7 describes a new procedure integrated into Ecosim that can be used to determine optimal restoration plans to transform the current ecosystem into a desired configuration. A new objective function called specific biomass is created for EwE's policy search routine, and possible restoration policies are evaluated that would turn the present-day depleted system into one resembling a more productive ORB state. Plans are tested that provide various degrees of continued harvest benefits during the restoration period. A cost-benefit analysis tests the economic feasibility of ecosystem restoration. A conservative approach to restoration is demonstrated to provide a better economic return than bank interest. Chapter 8 offers conclusions on the strengths and weaknesses of this restoration approach, and suggests new avenues of research that could take this integrated methodology from theory into practice. A comparison is made between ORB as an ecosystem management target and Maximum Sustainable Yield (MSY), an analogous single species index. Criticisms of the BTF approach are addressed, and comment is made on the usefulness of EwE as a policy aid for restoration ecology. Finally, policy recommendations are provided based on the general conclusions of this study. The appendices provide results and supporting information for each chapter. Appendix 2 includes a cost-benefit analysis of education, as an existing example of a multigenerational enterprise, that can be used to set the intergenerational discount rate for valuation of fisheries resources. Appendix 3 presents qualitative trends of relative abundance for EwE functional groups based on L E K information. Appendix 4 provides supporting materials for the IUU analysis, including a timeline of BC fisheries, a table summarizing influences in the rate of misreporting, as well as reported landings and species weights used to estimate the IUU trend. 17 Appendix 5 presents parameters used in the Ecopath and Ecosim models o f northern B C for all time periods. Time series data for biomass and catch are presented; other information includes model outputs such as dynamic biomass and catch, and an equilibrium analysis o f the present- day (2000) model. Appendix 5 also compares the present-day 2000 model with the one derived from the 1950 model (following a 50 year simulation). Appendix 6 first presents the parameters used in the policy optimization routine in Chapter 6, and then presents the results of the optimizations, listing harvest benefits of O R B ecosystems measured using various indices of harvest utility. Appendix 7 provides supporting information used to parameterize the policy search in Chapter 7, and biomass trajectories are presented for restoration plans that vary the speed of restoration and the level of sustained harvest benefits. Appendix 8 provides a spatial analysis of the consequences of marine protected area ( M P A ) zonation in northern B C . Various harvest strategies are analyzed for the National Marine Conservation Area ( N M C A ) surrounding Moresby Island in southern Haida Gwai i . Appendix 9 lists references cited in the appendices. The published materials appearing in this thesis are presented in Table 1.1. Table 1.1 Published materials appearing in this thesis. Articles in review or in preparation are available from this author (contact: c. ainsworth@fisheries. ubc. ca) Thesis section Subject Reference Journal or publisher Description Chapter 1 Back to the Future policy approach. Pitcher et al. (2004) American Fisheries Society Symposium * Conference procedings Chapter 2 Application of Q-90 biodiversity statistic to EwE models of NE Pacific. Ainsworth and Pitcher (inpress) Ecological Indicators * Primary literature As above. Ainsworth and Pitcher (2004b) Fisheries Centre Research Reports Grey literature Economic valuation technqiues. Ainsworth and Sumaila (2004a) Fisheries Centre Research Reports Grey literature Employment diversity index. Ainsworth and Sumaila (2004b) Fisheries Centre Research Reports Grey literature Chapter 3 Use oflocal ecological knowledge in ecosystem models. Ainsworth and Pitcher (2005a) Alaska Sea Grant * Conference procedings Interview methodology. Ainsworth (2004) Fisheries Centre Research Reports Grey literature Chapter 4 Estimation of IUU catch in BC Ainsworth and Pitcher (2005c) Fisheries Research * Primary literature As above. Ainsworth and Pitcher (2005d) State of the Environment Reporting (MWLAP) Internal government report Chapter 5 Preliminary Northern BC models for 1750, 1900, 1950 and 2000 AD. Ainsworth et al. (2002) Fisheries Centre Research Reports Grey literature As above. Alcock et al. (in prep.) McGill-Queens University Press * Book Analysis of predator-prey vulnerabilities for northern BC models Ainsworth and Pitcher (2004a) Fisheries Centre Research Reports Grey literature Chapter 6 Evaluation of restoration goals based on ORB concept. Ainsworth and Pitcher (2005b) Alaska Sea Grant * Conference procedings As above. Pitcher et al. (2005) NATO Science Series IV: Earth and Env. Sciences. Book Introduction of ORB concept. Baker et al. (inprep.) McGill-Queens University Press i Book Trade-off analysis of ORB benefits. Pitcher and Ainsworth (in review) Procedings of the 4th World Fisheries Congress * Conference procedings Policy search methods. Ainsworth et al. (2004) Fisheries Centre Research Reports Grey literature Chapter 7 Restoration policy optimization; introduction of specific biomass algorithm. Ainsworth and Pitcher (in review) Procedings of the 4th World Fisheries Congress * Conference procedings Cost benefit analysis of ecosystem restoration to various ORB states. Ainsworth and Pitcher (in review) ICES Annual Science Conference Proceedings (2005) * Conference procedings Demonstration of ecosystem-based population viability analysis. Pitcher et al. (in review) ICES Annual Science Conference Proceedings (2005) * Conference procedings Appendix 2 Intergenerational discounting case study: Newfoundland Northern cod collapse. Ainsworth and Sumaila, (2005) Canadian Journal of Fisheries and Aquatic Sciences * Primary literature As above. Ainsworth and Sumaila (2003) Fisheries Centre Research Reports Grey literature Appendix 8 Spatial analysis of Gwaii Haanas NMCA zonation options. Ainsworth (2004) (Available from author) Workshop proceedings * Peer reviewed contribution 2 H A R V E S T P O L I C Y E V A L U A T I O N T E C H N I Q U E S 19 The prudent heir takes careful inventory of his legacies and gives a faithful accounting to those whom he owes an obligation of trust. John F. Kennedy State of the Union Address, 1961 2.1 Introduction To evaluate economic, social and ecological benefits of harvest policies for the B T F procedure, I have adapted standard evaluation techniques and developed new ones for use with E w E models. The indices described in this chapter include an economic index, net present value ( N P V ) calculated under conventional and intergenerational discounting equations (Ainsworth and Sumaila, 2004a; Sumaila, 2004; Sumaila and Walters, 2005), a social utility index based on employment diversity (Ainsworth and Sumaila, 2004b) and an ecological index used to represent biodiversity (Q-90 statistic) (Ainsworth and Pitcher, 2004b; in press). Chapters 6 and 7 use these indices along with standard E w E outputs to compare candidate ecosystem goals for restoration, and evaluate the success of fishing plans to achieve those goals. 2.2 Economic index: Net present value (NPV) Although benefits of marine ecosystem restoration may be measured in ecological and social terms, economic considerations w i l l l ikely weigh heavily in determining the practicability of any long-term restoration agenda. The N P V term is used to summarize the economic success of harvest plans because it condenses the flow of future benefits into a single expression, while introducing a time component through discounting that reflects the preference of an investor for immediate benefit and delayed payment. The conventional discounting N P V term weights immediate harvest benefits heavily, but the present value of benefits to be received far in the future is reduced exponentially with time. Under the intergenerational discounting approach 20 (Sumaila, 2004; Sumaila and Walters, 2005), future benefits are discounted less, and the welfare of future generations is considered explicitly in the present value term. Conventional discounting Under the conventional model discounting, the flow of fishery benefits is summarized in the N P V term using the expression in eq. 2.1. T NPV = YJ{d'xNBt) Equation 2.1 /=o Where N B is net benefit accruing in year t; d is the discount factor defined in eq. 2.2, d — T; r r Equation 2.2 (i + s) Where 5 is the discount rate. Intergenerational discounting The intergenerational discounting equation considers a continuous interlacing of generations, where the discounting of future benefit is countered each year by the addition of 1/G stakeholders (G is generation time). These new entrants bring with them a renewed perspective on future earnings, partially resetting the discounting clock. The equation requires a standard annual discount factor (d) and a discount factor to evaluate benefits destined for future generations (<ifg). N P V is represented as in eq. 2.3. 21 1-A' 1-A if 8*5, fg NPV=\ Equation 2.3 I-r (1+S)'X + G) otherwise G is assumed to be 20 years, the average age at which a Canadian woman has her first child, and The conventional approach to discounting w i l l favour fishing policies that provide immediate benefits to stakeholders, while the intergenerational approach wi l l assign a higher N P V to policies that spread out benefits over several decades. The need for a new resource valuation method In cost-benefit analysis ( C B A ) , standard discounting is often unable to sanction long-term environmental policies that fulfill the frequently stated mandate to provide for the needs of future generations (e.g., D F O , 2001; E C , 2002). Scaling down the value of future benefits exponentially through time ensures that immediate costs w i l l outweigh far-off benefits at any practicable level of discounting, so that only myopic policies can result (Clark, 1973; Sumaila, 2001; 2004). In valuing the stream of benefits from a fisheries resource, use of conventional discounting may lead to early profit taking at the expense of sustained productive potential. Evidence of this type of 'front-loading' of fisheries benefits is clear in the harvest record of Northern cod (Gadus morhua) in the years before the 1992 collapse. Appendix 2.1 presents a case study on the Newfoundland cod fishery that suggests conventional valuation of fishery resources may have contributed to the decline and collapse of the Atlantic cod fishery. The case study also demonstrates that intergenerational valuation of fisheries resources could make long-term conservation an affordable prospect. The discount rates I use to A = Equation 2.4 22 evaluate fishery benefits are based on a C B A of education (Appendix 2.2). Schooling o f children serves as an example of an existing multi generational investment. B y applying the apparent discount rate that people use to value the education o f their children, I implicitly account for a variety of non-monetary benefits which could also apply to resource conservation. 2.3 Social utility index: Employment diversity Ainsworth and Sumaila (2004b) used an employment diversity index to evaluate harvest plans after the methodology of Attaran (1986). Based on the Shannon's entropy function (Shannon and Weaver, 1949), this measure describes the diversity of employment across fishing sectors. The entropy function is defined as in eq. 2.5: n D(EX ,E2,...EN) = -^T Et log2 E- Equation 2.5 /=1 Where n is the number of (possible) fishing sectors active in the ecosystem, and E is the proportion of total employment that is located in the ith fishing sector. . The measure is normalized across sectors with respect to their maximum possible diversity so that D ( E i , E2, . . . E n ) ranges from 0 to 1, as in eq. 2.6. D(E],E2...En)= - X / v l o g 2 £ , /MAX(D(EvE2,..En)) Equation2.6 V 1=1 )i When D - 0, this indicates that all fishing activity is concentrated in a single sector; D = 1 indicates the maximum possible employment diversity with all sectors contributing equally to employment (all Ej are equal). 23 Application to Ecosim A V B algorithm uses this descriptor to assess the annual employment diversity of the dynamic 50-year harvest schedule for each optimal policy suggested by the E W E policy search routine. Beginning with Ecosim's output C S V file, total value per gear type is calculated as the sum of all functional group landings, multiplied by gear-specific prices (Chapter 5; Appendix Table A5.1.6). Total value per gear type is converted to relative number of jobs using an estimated "jobs per catch value". It is considered equal for all fleets, so employment is proportional to landed value. Employment per sector ( E i ) is then calculated as a fraction of total employment. 2.4 Ecological indices Although the commercial value of fishing a restored ecosystem may offset the costs of rebuilding (Chapter 7), any practical restoration agenda wi l l need to include ecological criteria for ecosystem improvement. A range of ecological indicators is useful for forecasting non-monetary benefits in fishing scenarios, and many have been developed or adapted for use with E w E models. It can be difficult to define appropriate indices to summarize ecosystem model outputs (Fulton et al, 2003), but considering the generic nature of E w E , its wide availability and comparatively simple implementation, there is a need to develop standardized outputs that can help users interpret ecosystem effects of experimental harvest scenarios. Ecosystem modelers have begun to realize that functional group aggregation styles and other nuances of model structure can have significant impacts on the dynamic predictions (Fulton et al, 2003; Pinnegar et al, 2005). Output indices therefore need to be robust and deliver consistent results despite subjective variations in model structure. Ecosystem modelers must often make judgments on the applicability of imperfect data, but there are also fundamentally subjective components in E w E . 1) Functional groups of species are aggregated depending on the objectives o f modeling, fishery and policy targets and availability of data. 2) When time-series data are unavailable for fitting, flow parameters may be set according to rules of convention. 3) The model diet matrix is usually based on incomplete and 24 imprecise data, and arbitrary manipulation of the matrix may be required to achieve mass- balance. Some attempts have been made to standardize the E w E model construction process (e.g., automatic mass-balance: Kavanagh et al, 2004; semi-automated data retrieval from Fishbase: Froese and Pauly, 2005). Existing EwE outputs Ecological indicators automated in E w E include the Finn cycling index (Finn, 1976), indices relating to emergy and primary production required (Odum, 1988; Christensen and Pauly, 1993), trophic flow indices (Ulanowicz, 1986), resource niche overlap (based on Pianka, 1973), system omnivory index (Pauly et al, 1993), fishing-in-balance index (Pauly et al, 2000), mixed trophic impacts (Ulanowicz and Puccia, 1990), among other system state and trophodynamic indictors. Indicators developed for the B T F approach include an ecosystem resiliency index based on information theory (Heymans, 2004), a fuzzy logic algorithm to estimate local extinction risk based on fish life history parameters (Cheung and Pitcher, 2004; Cheung et al 2005) and a biodiversity statistic, Q-90, which is described here (also see Ainsworth and Pitcher, in press). Q-90 biodiversity statistic The Q-90 biodiversity statistic is a variant on Kempton's Q index (Kempton and Taylor, 1976) that has been adapted for use with E w E , where taxonomic species are grouped into aggregate biomass pools of functionally similar organisms. When used in conjunction with other indicators, the Q-90 index offers a useful method to evaluate consequences of alternative fishing plans, track the effect o f climate fluctuations and changes on biodiversity, estimate the non- consumptive value of ecosystems, and generally inform the ecosystem-based approach to marine science. Although ecological indicators of all varieties are of scientific interest, biodiversity holds special appeal to the public and is often addressed directly by policy - even though the appropriate scientific definition may not be made explicit (Harper and Hawksworth, 1994; Hamilton, 2005). In this section, I refer to biodiversity as organismal diversity at the level of species functional groups. 25 Definition Kempton's Q index describes the slope of the cumulative species abundance curve (Fig. 2.1). A s applied here, each functional group in the E w E model represents one 'species', and the biomasses of these groups, sorted into bins, serves as a proxy for the number of individuals in that species. Kempton and Taylor (1976) suggested using the inter-quartile slope o f the species abundance curve in order to avoid problems arising from the inclusion of tails, which, in field sampling, may be long and include a high number of low- abundance species. In applying this methodology to Ecosim, tails are less of a problem since modelers do not log log R 2 log-abundance Figure 2.1 Q-90 statistic definition. S is number of functional groups in reference model; R| and R2 are lower and upper 10-percentiles of the species abundance distribution. Modified from Kempton and Taylor (1976). normally represent a large number of low abundance functional groups. I therefore used the slope between the upper and lower 10-percentiles rather than quartiles. The Q-90 statistic is defined as in eq. 2.7. g90 = 0.85/[log(i22 Equation 2.7 Where S is the total number of functional groups in the model; R i and R 2 are the representative biomass values of the 10 t h and 90 t h percentiles in the cumulative abundance distribution. 26 The 10 and 90 percentiles are determined by eq. 2.8 and eq. 2.9, respectively, R Equation 2.8 R2-\ R2 Y,nR < 0 . 9 - S < £ « R Equation 2.9 Where n R is the total number of functional groups with abundance R. Magurran (1988) describes the qualities of Kempton's index that make it well suited to this application. Kempton's index is not dependent on the assumption of a particular species abundance model, which makes it generically applicable to a wide variety of ecosystem types. It is not biased by very abundant or very rare species, and this can be advantageous i f there are highly aggregated functional groups, as is sometimes the case with data-poor models. It expresses both species richness and evenness, which allows it to discriminate ecosystem effects among harvest plans (since exploitative fishing strategies can result in depletions or extirpations), while also capturing changes in the ecosystem that occur outside of harvested functional groups. In field studies, Kempton's index is robust against changes in sample size i f very small samples are avoided, but this is not critical with E w E models since the entire ecosystem is represented explicitly or implicitly. The following case study evaluates the effect of fisheries on ecosystem biodiversity, and demonstrates that the Q-90 statistic delivers consistent results regardless of model structure. 2.5 Q-90 case study: N E Pacific ecosystems I use the Q-90 statistic to evaluate biodiversity after 25 years of fishing for eight ecosystem models of the northeastern Pacific under a variety of fishing plans. I test the ability of Q-90 to respond to fishing influence on the ecosystem, and compare predictions made using simple and complex ecosystem models. B y choosing similar shelf ecosystems, biodiversity predictions 27 should be comparable across models. A n y real differences in biodiversity among the ecosystems should be minimized so that we can examine the effects of model structure on the index. Applying index to Ecosim output Using a Visual Basic algorithm, a user-defined number of bins is established that represents the complete range of functional group biomasses. The biomass of each functional group is then sorted into its appropriate bin as a count; this serves as a proxy for the number of individuals in that group. Bins may be linear or logarithmic. The Q-90 index is the slope of the cumulative species abundance curve is determined between the 10- and 90-percentiles; the Q-90 value may be plotted for each year in the simulation. At present, E w E does not permit absolute extinctions; it returns a low but non-zero biomass value for critically depleted groups. Therefore, every fishing scenario at its conclusion w i l l contain the same number of functional groups as the base model. To increase the sensitivity of the index to group depletions, a filter is passed over group biomasses for each year o f the simulation. If the biomass of a given functional group falls below a reference value, that group is omitted from the~Q-90 calculation, reducing the overall biodiversity score. In previous applications of this index, the depletion filter threshold has been set as an arbitrary 60% of the unfished biomass (Bo) and pristine biomasses represented in models of ancient ecosystems have been used as a proxy for Bo (e.g., Ainsworth and Pitcher, 2005b; also see Chapter 6). Setting a high threshold makes the index more sensitive to group depletions; the Q- 90 value therefore drops off quickly as fishing plans tend towards heavy exploitation and the index provides greater discrimination between conservative and exploitative fishing plans. The filter threshold may be reduced when evaluating severely depleted ecosystems; alternatively, one may set the threshold at a fraction of the baseline biomass. 28 Methods Using eight E w E models of present-day ecosystems in the northeastern Pacific (Table 2.1), I compare the effects of three simple fishing policies on biodiversity: a reduction in fishing mortality to one-half the model baseline (0.5 F), baseline fishing mortality (1 F) and a five-times increase in fishing mortality (5 F). Baseline represents an estimate of current real-world fishing mortality. In lieu of biomass estimates for unfished populations (i.e., that correspond to the species group aggregation style used by the original modelers), the depletion filter is set here as a proportion of baseline group population size for all simulations. If groups fall below 80% of their initial biomass, they are removed from the Q-90 calculation. Table 2.1 Eight EwE models of the NE Pacific. Abbreviation Model area # of groups Reference WCVI West Coast Vancouver Is. 15 Pauly et al. (1996) SNBC Northern BC - small model 26 Ainsworth, C. {unpublished manuscript)* SOG Strait of Georgia 27 Dalsgaarde/a/. (1998) ALU Aleutians 40 Heymans (2005) HEC Hecate Strait 50 Beattie(2001) ' • PWS Prince William Sound 51 Okey and Pauly (1999); Okey and Wright (2004) LNBC Northern BC - large model 53 Ainsworth et al. (2002) NCC Northern California Current 65 Field (2004) 1 Contact: c.ainsworth@fisheries.ubc.ca Results Long-term fishing simulations show a relationship between biodiversity maintenance and the overall level of fishing mortality applied. F ig . 2.2 shows Q-90 biodiversity predictions from the Ecosim model of the Northern California Current (Field 2004). A s we increase extractions from the ecosystem, biodiversity is sacrificed. Under the exploitative fishing policy described by the (5 F) scenario, there is an initial 50% drop in ecosystem biodiversity. 29 10 8 o a> O G Q O O O O O O O O O O O 0 o<~, o o o o o o o o 0 0 0 o o o 5 0 0 0 0 o o o o o o o O o o o o o o o o o o o o o •••••••• 10 20 Years 30 140 E 120 ro 100 0.5 F 1 F 5 F Fishing mortality Figure 2.2 Dynamic ecosystem biodiversity (Q-90) of three example Ecopath with Ecosim simulations. A.) White circles show reduced fishing mortality from model baseline (0.5 F); grey circles show baseline fishing mortality (1 F); black circles show increased fishing mortality (5 F). B.) Bar graph shows total catch for these policies. Model of Northern California Current ecosystem (Field, 2004). Fig. 2.3 compares baseline biodiversity among Ecopath models of the N E Pacific constructed by various authors using independent group aggregation criteria. The absolute value of the Q-90 statistic increases in direct relation with the total number o f functional groups. The scatter around the trendline represents differences in functional group aggregation style and real ecological differences, although I have tried to minimize this factor by using models of similar ecosystems. Fig. 2.4 suggests that the relative change in the Q-90 statistic is not dependant on model complexity. However, model complexity itself can affect dynamic function i f functional groups are over- or under- aggregated and a key ecological interaction is misrepresented (Fulton et al, 2003). In that case, the Q-90 index w i l l report the errant model behaviour. Because of this, we may expect a small degree of variation around the trend line in F ig 2.3, owing to inherent behavioural differences between models o f varying complexity. However, Q-90 measurements for complex models (containing many functional groups and interactions) should be resistant to the compounded data uncertainty (see Hakanson, 1995) i f errors surrounding the slope line in Fig 2.3 tend to cancel. 10 0 0 20 40 60 80 Number of functional groups Figure 2.3 Absolute Q-90 value at baseline (year zero) for eight northeastern Pacific Ecopath models. The simplest Ecopath model represents the ecosystem using only 15 functional groups, while the most complex model uses 65 functional groups. 20 -60 J WCVI SNBC S O G ALU HEC PWS LNBC NCC # groups 15 26 27 40 50 51 53 65 Figure 2.4 Change in Q-90 index after 30 years of fishing for eight EwE models of the NE Pacific. From left to right, models increase in number of functional groups. White bars show reduced fishing mortality from model baseline (0.5 F); grey bars show baseline fishing mortality (1 F); black bars show increased fishing mortality (5 F). West Coast Vancouver Is. (WCVI); small-Northern British Columbia (SNBC); Straight of Georgia (SOG); Aleutian Islands (ALU); Hecate Strait (HEC); Prince William Sound (PWS); large-Northern British Columbia (LNBC); Northern California Current (NCC). 31 Fig . 2.4 also shows the effects of fishing on the biodiversity of the ecosystem. At five times model baseline fishing mortality, every E w E model predicts a drop in biodiversity over 30 years. Except for S N B C , L N B C and P W S , which were designed to be steady state under baseline fishing mortality, all models predict a biodiversity decline under baseline fishing mortality. Several models suggest that even halving the exploitation rate w i l l not prevent biodiversity from declining over the long-term. However, the fishing scenarios tested are simplistic because all assume a constant level of fishing mortality without regard to changing stock size and the fishing rates tested also assume a uniform change in fishing mortality across all gear sectors. Index resolution The Q-90 statistic tends to change in a step-wise fashion with dynamic biomass predictions. Models containing many functional groups allow the index to resolve more precise changes in species composition, but models containing fewer functional groups tend to produce coarse changes in the biodiversity index over time reflecting only large-scale changes in species composition. Resolving power of the index is therefore reduced in models containing fewer functional 20 o cn • O c (/) CD cn c ro B cn 15 10 0 0 1 5 - 4 0 5 0 - 6 5 # functional groups Figure 2.5 Q-90 sensitivity to changes in system biomass structure. Q-90 sensitivity is compared in small (< 40 functional groups) and large models (> 50 groups). Y-axis shows mean number of step-wise changes in Q-90 value (i.e., resolving power) for a standard set of harvest simulations (30 year simulations at 0, 0.5, 1 and 5 times baseline fishing mortality). Closed circles show logarithmic bins, open circles show linear bins (error bars; 1 SD). groups (Fig. 2.5). A one tailed Student's t test indicates that resolving power is significantly less for small models (< 40 functional groups) than large models (> 50 groups) (p < 0.05). Linear 32 bins provide better resolution for small models than logarithmic bins (t test; p < 0.01), but logarithmic bins produce less variable results overall (F test; p < 0.05). Fig . 2.6 shows application o f the depletion filter at 30%, 50% and 80% of baseline functional group biomass. A high filter threshold causes functional groups to fall out of the Q-90 calculation, and increases the sensitivity of the index to ecosystem changes. Under high depletion filter thresholds, linear bins may be slightly better at resolving biodiversity changes than logarithmic bins (t test; p = 0.104). Discussion This application of Kempton and Taylor's (1976) index to ecosystem models considers both evenness and richness in the biodiversity score. Although most ecological studies determine biodiversity based on, occurrence and abundance of taxonomic species ('speciosity'), the number of functional groups in a E w E model is fixed and species-level population changes are not considered in the dynamics unless those species are explicitly represented. The method introduced here therefore provides an approximation to the original Kempton index, which was developed for field studies. o CD i O to CD c 03 0) CO CO 20 15 10 0 0 0 30 50 80 Depletion filter threshold (%) Figure 2.6 Q-90 sensitivity to changes in ecosystem structure using three depletion filter thresholds. Thresholds are set at 30%, 50% and 80% of baseline functional group biomass. Y-axis shows mean number of step-wise changes in Q-90 value (i.e., resolving power) for a standard set of harvest simulations (30-year simulations at 0, 0.5, 1 and 5 times baseline fishing mortality). Closed circles show logarithmic bins, open circles show linear bins (error bars; 1 SD). 33 Evenness can be represented in the ecosystem models, with biomass serving as a proxy for the number of individuals in each functional group. Under some circumstances this proxy could produce a bias; for example, i f the average weight of animals changes suddenly as a result of fishing, as new technologies are introduced or in response to market influences. When comparing ecosystem models of different time periods, evolutionary changes in response to fishing could also cause a bias. Calculating richness is less straightforward. Since the number of model functional groups is fixed, the depletion filter is used to drop groups from the calculation and the total number o f functional groups active in the calculation is therefore analogous to species richness. B y setting a high depletion filter threshold we increase the contribution of species richness to the overall biodiversity score, but without the filter the index solely represents evenness. Eliminating groups from the computation with the filter increases the sensitivity of the index to depletion events or effects and reduces the overall Q-90 value. However, as functional groups are removed, the remaining biodiversity calculation is based on fewer groups and the ability of the index to recognize small changes in biodiversity is compromised. I suggest using a high threshold to increase sensitivity of the index for models containing many functional groups, which can stand to loose a few from the calculation, or to exaggerate small differences in ecosystem biodiversity when comparing similar models or fishing plans. The algorithm could be adapted to work with any static or dynamic multispecies or ecosystem model that represents species biomass in aggregated functional groups; see Fulton et al. (2003), Hollowed et al. (2000) and Whipple et al. (2000) for reviews of multispecies and ecosystem models. Model dynamics do not need to be based on trophic flows, but the biomass of functional groups must be accessible. Models which are primarily oceanographic or biogeochemical likely could not benefit; nor could E w E models that use nutrients as the currency of group exchange instead of biomass (e.g., Watkinson, 2001). The Kempton Q index is now automated in E w E V5.1 and is available as a dynamic output for simulations (Christensen and Walters, 2004b). However, the integrated version is not exactly as described here. It considers only high trophic level functional groups (> T L 3), it uses the inter- quartile slope of the cumulative abundance curve rather than 90-percentiles. It can also 34 accommodate only linear species biomass bins, and as it does not employ a depletion filter it mainly serves as an indicator of biodiversity evenness. Contribution to ecosystem studies The use of ecological indicators is recognized as a critical component o f E B M (e.g., F A O , 2003; Cury et al, 2005; Garcia and Cochrane, 2005), although firm ecological theory is needed to relate changes in ecological indices to proper remedial management actions (Hall and Mainprize, 2004). A s the relatively new field of ecosystem modeling continues to advance, facilitated by an increase in inexpensive computing power and the current drive towards- ecosystem-based approaches in marine systems (Link, 2002), standardized indices w i l l make ecosystem models tools that are more effective toward understanding fisheries and climate effects on marine communities. Not only can ecosystem models be used to evaluate potential repercussions o f fishing on non- target organisms, broad indicators which describe the state of the natural environment may hold special resonance with the general public (Rogers and Greenway, 2005); and public buy-in is critical since fishery stakeholders become a far more encompassing group once the entire marine ecosystem is factored in to management decisions. The next chapter w i l l summarize work done with communities in northern B C . Community members helped evaluate candidate restoration goals and suggested fisheries that could be used to harvest a restored ecosystem. Through interviews, they provided local ecological knowledge to supplement scientific information and help satisfy the vast data requirements of the ecosystem models used in B T F research. 3 5 3 C O M M U N I T Y I N T E R V I E W S All our knowledge has its origins in our perceptions. Leonardo de V i n c i Qu. E. MacCurdy(1954) 3.1 Introduction In modeling whole marine ecosystems, data deficiencies become especially apparent among species that hold no commercial appeal. Stock assessment records exist for only a small minority o f species so modelers must borrow parameters from other ecosystems, or rely on guesswork. Although E W E grants modelers some reprieve by automatically estimating biomasses of data-poor groups based on the assumption of mass-balance, there is a clear need to reduce uncertainty in our estimates by incorporating supplemental information, particularly for historic ecosystems. Local ecological knowledge ( L E K ) held by fishing community members is one such resource. L E K can be used to fine-tune static Ecopath models, to confirm dynamic Ecosim function, or to inform us how the ecosystem might have been structured decades ago - before time-series data began for most species. L E K therefore holds obvious application for B T F , which seeks to quantify ecosystem changes over time. The key step in adapting L E K to our modeling needs comes in producing a quantitative data series from qualitative accounts. This section describes how that was done for the northern B C models, and how the L E K trends are used to improve dynamics in the northern B C models. I also compare L E K trends with stock assessment in the hope that fishers' perceptions can help establish criteria by which we can assess the quality of scientific data - by challenging it with an independent authority and identifying where fishers' perceptions depart from the scientific understanding. Interview methods used in this chapter are published in Ainsworth (2004); results are in Ainsworth and Pitcher (2005a). 36 3.2 Methods Interviews Under approval of the University of British Columbia Ethical Review Committee, workers from the Fisheries Centre interviewed forty-eight community members from the Prince Rupert region and Haida Gwai i , B C in two community workshops in 2002 (Pitcher et al, 2002b; Pitcher, 2004). The processed anonymous data is searchable online at [www.fisheries.ubc.ca/projects/btf/]. Interviewees represented a broad cross-section of commercial, recreational and aboriginal fishers as well as processors and others who are familiar with the marine system in Hecate Strait, Dixon Entrance and Queen Charlotte Sound. A s the aim was to improve the northern B C models, participants were not selected randomly; snowball sampling was used to find the most knowledgeable contributors as recommended by partners and participants. One hundred and eighteen flashcards of marine mammals, birds, fish and invertebrates were shown to each interviewee. L E K information recorded included species population changes, fisheries interactions and spatial information - such as animal aggregations and seasonal movements. These data along with career and demographic information were processed to ensure anonymity, and entered into the B T F Historical and Interview Database (Erfan,. in press): Creating a time-series of relative abundance Respondents were asked whether the abundance of marine creatures had increased, remained the same or decreased during their careers. This method assumes that respondents made implicit allowance in their answers for any changes in catchability arising from new methods or fishing technology. To create a numerical trend, an interviewee's comment of increase, stable or decrease is assigned the numerical value of +1, 0 or -1, respectively. Every year that the respondent fished receives one numerical 'vote' for that organism. Summing votes from all 2 Dr. Tony Pitcher (P.L), Dr. Ussif Rashid Sumaila, Dr. Sheila Heymans, Dr. Melanie Powers, Nigel Haggan, Russ Jones (Haida Fisheries Council), Eny Buchary, Cameron Ainsworth, Pablo Trujillo, Louisa Wood, Richard Stanford, Erin Foulkes and Aftab Erfan. 37 respondents, the annual total is assumed to indicate the average perception for that year. A value greater than zero therefore indicates that the fishers perceived an increase in biomass during that period, while a value less than zero indicates a perceived decline. The resulting time-series provides an index of the rate of change for each organism, which is converted into a running total to serve as a proxy for relative abundance. Data trends for organisms are compiled into Ecopath functional groups (see Chapter 5 for group descriptions). Some functional groups include multiple species, so I assume that the abundance trend of the group closely follows the species mentioned most often by respondents. For example, only eight comments out of 59 concerning the functional group Odontocetae mentioned the Northern right whale dolphin, while 36 comments were made for orca. The abundance trend of Odontocetae therefore more closely reflects the trend for orca; it is a weighted average of the relative number of comments. Ideally, one would weigh the contribution of each species to the overall functional group abundance trend using some independent estimate of relative abundance. However, in the base Ecopath model, important and commercial species (i.e., species for which independent abundance data are available), are typically assigned their own dedicated functional group. 38 Weighting by expertise The interview data captures a diverse sample of local knowledge. Many fisheries (and industries) were represented at the interviews, and each sector is expected to carry its own special expertise in species of particular importance to the specialization. I therefore applied weighting to the votes offered by each participant according to their expertise. 'Expert' opinions were taken to be worth twice as much towards calculating the average (i.e., +2 and - 2 for increasing and decreasing votes), 'Novice ' opinions were taken to be worth half as much (i.e., +0.5 and -0.5). The following criteria are used to define 'expert' and 'novice' comments: 1. Fishers are considered expert in their target functional groups; 2. Group interviews are novice in all functional groups; 3. First Nation group interviews remain unchanged in First Nation specialties; 4. Non-fishers are novice in all functional groups; 5. Recreational fishers are novice on all functional groups except their specialty, in which they are expert; 6. Interviewee #20 was judged expert in all functional groups; 7. Interviewee #21 was judged expert in all rockfish functional groups. Group interviews operated on consensus; their vote is reduced in importance to limit the effect of influence between respondents in the analysis. However, one exception is made. Since the majority o f First Nations respondents participated in group interviews, I do not want to reduce the impact of their comments on the L E K abundance trends. Comments made during First Nations group interviews therefore remain 'unchanged' in importance regarding the abundance of traditionally harvested species. I assume that non-fishers and recreational fishers spend less time at sea than commercial harvesters, so their contribution to the overall trend is weighted half as much. In addition to years of fishing experience, interviewees 20 and 21 had formal ecological training. . ' 39 Alternatively, a weighting scheme based on years of experience could be used, although some degree of ranking by gear specialization should still be included. Information from experienced fishers does actually influence the LEK abundance trend more than information from less experienced fishers under the current methodology, since their comments apply to more years in the analysis. Qualitative agreement of L E K versus stock assessment To determine how often comments agreed with stock assessment records, I compare the qualitative change in abundance offered by each interviewee with time series biomass data assembled from stock assessment. For the period in which a respondent fished, an Excel macro consults time series stock assessment records assembled from various DFO publications3. The algorithm determines whether the abundance of the subject functional group had increased, stayed the same or decreased in the stock assessment record. It compares this result against the suggested population change provided by the interviewee to determine agreement. This procedure is conducted for functional groups that have continuous stock assessment information. Comments are used from only the respondents whose career spanned a period covered by stock assessment data. For each comment made concerning a particular functional group, the span of the interviewee's career at sea is divided into two halves. The average abundance of that functional group in the first and second halves of the fisher's career is determined from stock assessment records. If the average abundance was greater in the second half than in the first, the functional group is said to have increased. If the fisher had indicated an increase in abundance, then their comment is considered 'true' (indicating only agreement between datasets). Similarly, if the stock had declined according to stock assessment, then comments that indicate a decrease in abundance are considered 'true'. Vasconcellos, M. (2001. Unpublished manuscript. Marine Resources Service. Food and Agriculture Organization, Viale delle Terme d