ISSN 1198-6727 Centre Research Reports

2002 Volume 10 Number 2

The Use of Ecosystem Models to Investigate Multispecies Management Strategies for

Capture Fisheries

Fisheries Centre, University of British Columbia, Canada

ISSN 1198-6727

The Use of Ecosystem Models to Investigate Multispecies Management Strategies for Capture Fisheries

Fisheries Centre Research Reports 2002 Volume 10 Number 2

edited by

Tony Pitcher and Kevern Cochrane

156 pages © published by

The Fisheries Centre, University of British Columbia

2204 Main Mall Vancouver, B.C., Canada 2002

ISSN 1198-6727

THE USE OF ECOSYSTEM MODELS TO INVESTIGATE MULTISPECIES MANAGEMENT STRATEGIES FOR CAPTURE FISHERIES

Edited by

Tony Pitcher and Kevern Cochrane 2002 Fisheries Centre Research Reports 10(2),156pp

CONTENTS Page Directors Foreword Ecosim and the Land of Cockayne ...... 3

Executive Summary ...... 4

Introduction

The Use of Ecosystem Models to Investigate Ecosystem-Based Management Strategies for Capture Fisheries: Introduction Kevern Cochrane ...... 5 Searching for Optimum Strategies for Development, Recovery and Sustainability Carl J. Walters, V. Christensen and ...... 11

Contributed Papers

Simulating Strategies in the Strait of Georgia Ecosystem using and Ecosim Steven J. D. Martell, Alasdair I. Beattie, Carl J. Walters, Tarun Nayar and Robyn Briese ...... 16 The Use of Ecosystem-based Modelling to Investigate Multi-species Management Strategies for Capture Fisheries in the Bali Strait, Indonesia Eny Anggraini Buchary, Jackie Alder, Subhat Nurhakim and Tonny Wagey...... 24 The Eastern Bering Sea Kerim Aydin ...... 33 A Preliminary North-East Atlantic Marine : the Faroe Islands and ICES Area Vb Dirk Zeller and Katia Freire ...... 39 Policy Simulation of Fisheries in the Hong Kong Marine Ecosystem Wai-Lung Cheung, Reg Watson and Tony Pitcher...... 46 Exploration of Management and Conservation Strategies for the Multispecies Fisheries of Lake Malawi using an Ecosystem Modelling Approach Edward Nsiku ...... 54 The Use of Ecosim to Investigate Multispecies Harvesting Strategies for Capture Fisheries of the Newfoundland-Labrador Shelf Marcelo Vasconcellos, Johanna Heymans and Alida Bundy ...... 68 Simulating Management Options for the North Sea in the 1880s Steven Mackinson...... 73 Ecosim Case Study: Port Phillip Bay, Australia Beth Fulton and Tony Smith ...... 83

Page 2, Using Ecosim for Fisheries Management

Simulating Extreme Fishing Polices in Prince William Sound, Alaska: a preliminary evaluation of an ecosystem-based policy analysis tool Thomas A. Okey ...... 94 Exploring Alternative Management Policies: A Preliminary Ecological Approach for the San Matias Gulf Fishery, Argentina Villarino María Fernanda, Mario L. Lasta and Marcelo Pájaro...... 109 Exploring Multispecies Harvesting Strategies on the Eastern Scotian Shelf with Ecosim Alida Bundy and Sylvie Guénette ...... 112 The Use of Ecosystem Models to Investigate Multispecies Management Strategies for Capture Fisheries: Report on Southern Benguela Simulations Lynne Shannon ...... 118 Impact of Harvesting Strategies on Fisheries and Community Structure on the Continental Shelf of the Campeche Sound, Southern Gulf of Mexico Francisco Arreguin-Sanchez ...... 127 Evaluating Harvesting Strategies for Fisheries of the Ecosystem of the Central Gulf of California Francisco Arreguín-Sánchez and Luis E. Calderón-Aguilera ...... 135 Testing Responses of a Tropical Shelf Ecosystem to Fisheries Management Strategies: a Small-scale Fishery from the Colombian Caribbean Sea Luis Orlando Duarte and Camilo B. García...... 142 Exploratory analysis of possible management strategies in Lake Victoria fisheries (Kenyan sec- tor) using the recent Ecosim software Jacques Moreau and Maria Concepcion Villanueva ...... 150

Participants in the Workshop...... 155

This report is co-published as an FAO Fisheries Report

A Research Report from The UBC Fisheries Centre and the Food and Agriculture Organisation of the United Nations

156 pages © Fisheries Centre, University of British Columbia, 2002

FAO/Fisheries Centre Workshop, Page 3

Director’s Foreword

Ecosim in the Land of Cockayne

There is a famous Breughal engraving, ‘Big Fish Eat Little Fish’, that illus- trates a marine food web by showing how the stomach of each size of fish contains fish of the smaller size class below. I use it to dramatise trophic re- lationships and, many years ago, it made a neat cover for my textbook (Pitcher and Hart 1982). One student later said that the book cover taught him all he knew. Another Breughal painting is reproduced opposite in hopes it will be equally instructive.

Like the Breughal engraving, Ecopath models take all trophic relationships The Land of Cockayne, by Pieter Breughal the Elder, 1567. As well as within an aquatic ecosystem into ac- illustrating a Fool’s Paradise (note the egg on legs and the walking count, and the approach will be famil- roast pig), the painting also shows how some of the workshop par- iar to many readers of Fisheries Cen- ticipants felt after a week of simulations. tre Research Reports, being the sub- Oil on panel, 52 x 78 cm, Alte Pinakothek, Munich. ject of 5 previous issues from 1996 (see below). Ecosim, the dynamic version of Eco- also became the ‘Big Rock Candy Mountain’ of US path (Walters et al. 1997, 2000) has recently been hobo lore. endowed with routines that not only simulate the consequences of changes in fisheries for all ele- What’s the moral here? Well - it is all too easy to ments in the ecosystem, but can also search for be seduced by these elegant simulations. The fishing rates that will maximize ecological, social ‘Land of Cockayne’ is intended to provide a plan- or economic goals. The first report of this impor- gent warning that we should continually check tant advance made by Carl Walters is presented simulation results with the real world. And not be here. tempted to boast that Ecosim optima place us in a far better world. It is a reality check. This new facility has interested FAO because it could lead to a way of managing multispecies The Fisheries Centre Research Reports series fisheries that takes into consideration all ele- publishes results of research work carried out, or ments of the ecosystem, not just those that are workshops held, at the UBC Fisheries Centre. The subjected to fishing. Hence, the Fisheries Centre series focusses on multidisciplinary problems in and FAO decided to hold a workshop to explore fisheries management, and aims to provide a syn- the potential of the new software, and the papers optic overview of the foundations, themes and in this report are the result. prospects of current research.

Back to that Breughal. The ‘Land of Cockayne’ is a Fisheries Centre Research Reports are distributed derisive medieval comment on those who imagine to appropriate workshop participants or project themselves in a paradise where food and luxuries partners, and are recorded in the Aquatic Sci- are so easily obtainable that life is comprised of ences and Fisheries Abstracts. A full list appears little more than a happy indolence. Or boast that on the Fisheries Centre's Web site, they are in such a place. As Breugal’s painting www.fisheries.ubc.ca. Copies are available on re- shows (above), Cockayne is evidently such a fool’s quest for a modest cost-recovery charge. paradise. The term ‘Cockayne’ probably derives from ‘cake’ (OED), and there are similar terms in Tony J. Pitcher medieval French. It gave rise to ‘Cockneys’, in- Professor of Fisheries habitants of London noted for boasting of the mi- Director, UBC Fisheries Centre raculous nature of their city, and a similar idea

Page 4, Using Ecosim for Fisheries Management

Literature cited Executive Summary

Pitcher, T.J. and P.J.B. Hart (1982) Fisheries . Chapman and Hall, London. 414 pp This report comprises the edited proceedings of Walters, C., Christensen, V., and Pauly, D. (1997) Struc- workshop held at the Fisheries Centre, University turing dynamic models of exploited ecosystems of British Columbia in July 2000, jointly spon- from trophic mass-balance assessments. Reviews sored by FAO, Rome and thr government of Ja- in Fish Biology and Fisheries 7(2): 139-172. pan. Walters, C., Pauly, D., Christensen, V., and Kitchell, J. F. (2000) Representing density dependent conse- quences of life history strategies in aquatic ecosys- This is the first published account of new Ecosim tems: EcoSim II. Ecosystems 3(1): 70-83. policy search software that aims to find fishing Mass-Balance Models of North-eastern Pacific Ecosys- rates that maximize objective functions for eco- tems. Fisheries Centre Research Reports 1996, nomic, ecological, employment or mixed goals. Vol.4 (1), 131 pp. Two papers set out the numerical basis for the Use of Ecopath with Ecosim to Evaluate Strategies for software and the procedure that was adopted for Sustainable Exploitation of Multi-Species Re- the workshop case studies. sources. Fisheries Centre Research Reports 1998, Vol.6 (2), 49 pp The report contains 18 case study papers explor- A Trophic Mass-Balance Model of Alaska's Prince Wil- liam Sound Ecosystem for the Post-Spill Period ing the use of the ecosim policy software. 1994-1996. Fisheries Centre Research Reports Papers examine fisheries and their ecosystems in 1999, Vol.6 (4), 143 pp the Strait of Georgia (BC), the Bali Strait (Indone- Ecosystem Change and the Decline of Marine Mam- sia), the North Sea, The Faroe Islands, Hong mals in the Eastern Bering Sea. Fisheries Centre Kong, Lake Malawi, Newfoundland, Port Philip Research Reports 1999, Vol.7 (1), 106 pp. Bay (Tasmania), Prince William Sound (Alaska), A Trophic Mass-Balance Model of Alaska's Prince Wil- the San Matias Gulf (Argentina), the Scotian Shelf liam Sound Ecosystem for the Post-Spill Period (Canada), Southern Benguela (South Africa), 1994-1996, 2nd Edition. Fisheries Centre Research Campeche Sound (Mexico), Gulf of California Reports 1999, Vol.7 (4), 137 pp. (Mexico), the Caribbean (Columbia).

FAO/Fisheries Centre Workshop, Page 11

Searching for optimum fishing measures or ‘objective functions’ for manage- strategies for fishery development, ment. recovery and sustainability These approaches can be used in combination, e.g. by doing a formal optimization search then ‘reshaping’ the fishing rate estimates from this Carl J. Walters, V. Christensen search in order to meet other objectives besides and Daniel Pauly those recognized during the search process. Fisheries Centre, UBC The first of these approaches is what has up to Abstract now been the standard simulation form in Eco- sim, and does not require further description Policy may be defined as an approach towards reaching here, (see Walters et al., 1997; Christensen et al., a broadly defined goal. In fisheries, policies are often 2000; Walters et al., 2000). The newly added implemented via total allowable catches, TACs, that are formal optimization involves three steps: recalculated annually, and through regulations that af- fect fleet and deployment. The task of fisheries scien- 1. Define blocks of fleet/year groupings to be in- tists should be to advise both on policy formulation and on its implementation. However, so far ecosystem- cluded in the search procedure; based policy explorations have rarely been conducted. 2. Define objective function weights for the four This can, however, now be addressed by the recent de- optimization objectives: velopment of a policy exploration routine for the Eco- (i) net economic value (total landed value of catch path with Ecosim approach and software. The paper minus total operating cost to take this landed gives an overview of the background for the policy value); search, and how it has been implemented. A brief over- (ii) employment (a social indicator, assumed pro- view of a new routine for examining uncertainty in the portional to gross landed value of catch for each management process is also included. fleet with a different jobs/landed value ratio for each fleet); (iii) mandated rebuilding of target species (obtained by setting a threshold biomass for the relevant species relative to their biomass in Ecopath); On Policy Exploration using (iv) ecological ‘stability’ (measured by assign-ing a Ecopath with Ecosim (EwE) weighting factor to each group based on their longevity, and optimizing for the weighted sum). A central aim of fisheries management is to regu- 3. Invoke the search procedure by clicking the late fishing mortality rates over time so as to search button. achieve economic, social, legal and ecological sus- tainability objectives. An important dynamic When a search has been completed, the resulting modeling and assessment objective is to provide ‘optimum’ fishing rates by year/fleet block are insight about how high these mortality rates transferred to the Ecosim ‘Temporal Simulation’, should be, and how they should be varied over where the optimized fishing rates will have re- time (at least during development or recovery placed the baseline (or previously sketched) rela- from past ). We cannot expect models tive efforts by fleet/gear type. to provide very precise estimates of optimum fishing mortality rates, but we should at least be Methodology able to define reasonable and prudent ranges for the rates. The impacts of alternative time patterns Invoking the search option causes Ecosim to use a of fishing mortalities can be explored using two nonlinear optimization procedure known as the different approaches in Ecosim: Davidson-Fletcher-Powell (DFP) method to itera- tively improve an objective function by changing 1. Fishing rates can be ‘sketched’ over time in relative fishing rates, where each year/fleet block the Ecosim simulation interface, and simulated defines one parameter to be varied by the proce- results (catches, economic performance indica- dure, (e.g., setting four color code blocks means a tors, biomass changes) examined for each sketch. 4-parameter nonlinear search). DFP runs the This is using Ecosim in a ‘gaming’ mode, where Ecosim model repeatedly while varying these pa- the aim is to encourage rapid exploration of op- rameters. tions. The parameter variation scheme used by DFP is 2. Formal optimization methods can be used to known as a ‘conjugate-gradient’ method, which search for time patterns of fishing rates (actually, involves testing alternative parameter values so as relative fishing efforts by fishing fleet/gear types), to locally approximate the objective function as a which would maximize particular performance

Page 12, Using Ecosim for Fisheries Management quadratic function of the parameter values, and maximizes the biomass of long-lived organisms, using this approximation to make parameter up- as defined by the inverse of their produc- date steps. It is one of the more efficient algo- tion/biomass ratios. The optimization of ecosys- rithms for complex and highly nonlinear optimi- tem ‘health’ optimization often implies phasing zation problems like the one of finding a best fish- out of all fisheries except those targeting species ing pattern over time for a nonlinear dynamic with low weighting factors. model. The search procedure results in what control sys- Nonlinear optimization methods like DFP can be tems analysts call an ‘open loop policy’, i.e. a pre- tricky to use, and can give grossly misleading re- scription for what to do at different future times sults. In particular the method can ‘hang up on a without reference to what the system actually local maximum’, and can give extreme answers ends up doing along the way to those times. It due to an inappropriate objective function. To would obviously be crazy to just apply an open check for false converge to local maxima, an op- loop policy blindly over time, each year commit- tion to use random starting F’s should be used in ting a fishery to fishing rates calculated at some addition to forcing additional iterations using the past time from only the data available as of that option to redo the analysis based on the current time. F’s. To test for sensitivity of the results to objec- tive function parameters, searches for a variety of In practice, actual management needs to be im- values of the objective function weights and pa- plemented using ‘feedback policies’ where harvest rameters should be accessed. goals are adjusted over time as new information becomes available and in response to unpredicted The objective function can be thought of as a ecological changes due to environmental factors. ‘multi-criterion objective’, represented as a But this need for feedback in application does not weighted sum of four criterion components or in- mean that open loop policy calculations are use- dicators: economic, social, legal, and ecological. less: rather, we see the open loop calculations as Assigning alternative weights to these compo- being done regularly over time as new informa- nents is a way to see how they conflict or tradeoff tion becomes available, to keep providing a gen- with one another in terms of policy choice. For eral blueprint (or directional guidance) for where example: the system can/should be heading. Also, we can often gain valuable insight about the functional (a) placing a high weight on the net economic value form of better feedback policies, (how to relate component (total fishing profits) typically causes harvest rates to changes in abundance as these the optimization to favor lower fleet sizes and se- changes occur) by examining how the open loop vere simplification of the simulated ecosystem to fishing rates vary with changes in abundance, es- maximize production of only those species that are pecially when the open loop calculations are done most profitable to harvest; with Ecosim ‘time forcing’ to represent possible (b) placing a high weight on the employment (social) changes in environmental conditions and produc- indicator typically results in favoring larger fleet tivity in the future. For an example of this ap- sizes, and again often severe ecological simplifica- proach to design of policies for dealing with de- tion in order to maximize production for the fleet cadal-scale variation in ocean productivity for that employs the most people. single species management, see Walters and Parma (1996). External pressure, (e.g. in form of legal decisions) may force policy makers to concentrate on pre- Maximizing Risk-averse Log Utility for serving or rebuilding the population of a given Economic and Existence Values species in a given area. In Ecosim, this corre- sponds to setting a threshold biomass (relative to One option in the search procedure for optimum the biomass in Ecopath) and identifying the fleet fishing patterns is to search for relative fleet sizes structure that will ensure this objective. The im- that would maximize a utility function of the form plication of this policy tends to be case-specific, w1⋅log(NPV) + w2⋅S⋅log(B) - w3⋅V, where the wi’s and to depend on the trophic role of the group are utility weights chosen by the user, and the whose biomass is to be rebuilt. utility components NPV, S⋅log(B), and V are de- fined as: The ecosystem criterion component is inspired by the work of E.P. Odum (1971) in terms of ‘matur- (1) NPV is net present economic value of harvests, ity,’ wherein mature ecosystems are dominated by calculated as discounted sum over all fleets and large, long-lived organisms. This is implemented times of catches times prices minus costs of fish- in Ecosim by identifying the fleet structure that ing, i.e., the discounted total profit from fishing the ecosystem. FAO/Fisheries Centre Workshop, Page 13

(2) S⋅log(B) is an existence value index for all compo- ing policy’, in which fishing is deployed so as to nents of the ecosystem over time. It is calculated severely simplify the ecosystem to maximize pro- as the discounted sum over times and biomass duction of one or a few species that appear at pre- pools of user-entered structure weights times logs sent to be the most valuable (price, potential total of biomasses, scaled to per-time and per-pool by catch). This may even involve deploying some dividing the sum by the number of simulation years and number of living biomass pools. fleets just to remove predators and competitors for the most valued species, just like deploying (3) V is a variance measure for the prediction of pesticides and herbicides to remove “pests” in ag- log(NPV) + S⋅log(B). It is assumed to be propor- ricultural systems. But simplifying an ecosystem tional to how severely the ecosystem is disturbed in such ways can make the behavior of the system away from the Ecopath base state, where distur- deeply unpredictable, by creating opportunities bance is measured at each time in the simulation for ecological response (population growth) by a by the multidimensional distance of the ecosystem variety of species that are rare in the “normal” biomass state from the Ecopath base state. This ecosystem, and hence are not well researched or term is negative, implying that increased uncer- understood in terms of their potential impacts on tainty about the predictions for more severe dis- valued species should they become abundant. turbances causes a decrease in the mean of log(NPV). The term represents both aversion to management portfolio choices that have high vari- Simplifying an ecosystem is hence much like in- ance in predicted returns, and the observation that vesting in high-risk, high-return stock market op- the mean of the log of a random variable (NPV⋅PB) tions; such investments may make you rich, but is approximately equal to the log of the mean of they may also bankrupt you. Most people are that variable minus ½ the variance of the variable. risk-averse as investors, and seek to “spread risk” Large w3-values can be used to represent both by investing in “balanced portfolios” with lower high uncertainty about predictions that involve expected returns on investment but much lower large deviations of biomass from the Ecopath base probabilities of severe loss. state, and strong risk aversion to policy choices

that have high uncertainty. The prediction variance measure V is not meant This utility function combines several basic con- to represent all components of variation or uncer- cepts of utility. First, the log scaling of value tainty about future biomasses and fishery values. components represents the notion of “diminish- V goes to zero for policies that hold or maintain ing returns”, that adding some amount to any the ecosystem at the Ecopath base state Bo for value measure is less important when the value every biomass, for all simulation times. It is ob- measure is already large than it is when the value viously not correct to suggest that we would ex- measure is small. Second, the log scaling also pect no variance in future biomasses (and hence represents the notion of “balance”, that no value in the harvest components of NPV as well) if such component should be ignored entirely (unless it is a policy were implemented. Imagine running a very large number of simulations of future bio- assigned a zero wi); the overall utility measure approaches minus infinity if either net economic mass changes under such a policy, while varying performance (NPV) or if any biomass component all possible uncertain quantities such as the Eco- path base biomasses and biomass accumulation of the ecosystem (any biomass B in S⋅log(B)) ap- i rates, productivities, Ecosim vulnerability pa- proaches zero. Third, it represents the notion that rameters, environmental forcing inputs repre- our predictions about the future of both economic senting oceanographic productivity regimes, fu- performance and biodiversity (biomasses) be- ture demand and price patterns, and changing come progressively more uncertain for policies vulnerabilities to fishing due to biophysical and that result in more extreme departures from the technological factors. Even for the baseline policy Ecopath base state about which we presume to where Ecosim predicts stable (‘flat line trajec- have at least some knowledge. tory’) expected or mean biomasses over time,

these simulations would likely reveal high vari- In the terminology of portfolio selection theory in ances and complex covariance patterns for most economics, fishing policies result in a portfolio of biomasses over time, i.e. we would see wide prob- value components with “expected total returns on ability distributions of possible future biomass investment” equal to NPV + S⋅B. But policies that states for the ecosystem. We should not be arro- have higher expected total returns are most often gant enough to suggest that we can describe all also ones that would push the ecosystem into the uncertainties well enough to accurately calcu- more extreme states, and hence represent portfo- late the variances of such distributions. But note lio choices with higher variance in total returns. that much of that variance in predictions of future

biomasses, (and hence variance in the value com- For example, maximizing the deterministic pre- ponents) would be due to sources of uncertainty diction of NPV in Ecosim often involves a ‘farm-

Page 14, Using Ecosim for Fisheries Management and variability that are the same no matter what and ecological stability criteria. In Ecosim, the the policy choice, i.e., would cause about the same relative fleet sizes are used to calculate relative amount of variance in predictions for any future fishing mortality rates by each fleet type, assum- harvest policy that we might simulate. ing the mix of fishing rates over biomass groups remains constant for each fleet type, (i.e. reducing When comparing policy choices using an optimi- a fleet type by some percentage results in the zation objective function, there is no point in in- same percentage decrease in the fishing rates that cluding extra constant terms that do not change it causes on all the groups that it catches). The with the policy variables, (e.g., a base variance Vo fisheries and ecosystem may be simulated with in predictions that does not change with fishing Ecosim using the solution found in the policy rate policy and just represents uncertainty about search interface: this is termed an ‘open loop’ any prediction that Ecosim might make). Hence simulation. the V distance measure is meant to represent only extra variance or uncertainty in predictions for Note that when density-dependent catchability ef- policy scenarios that would likely drive biomasses fects are included in the simulations, reductions far from the Ecopath mean state. in biomass for a group may result in fishing rate remaining high despite reductions in total effort Note that Ecosim does not deliberately advocate by any/all fleets that harvest it. Despite this ca- or promote any particular risk-averse portfolio veat, the basic philosophy in the fishing policy approach to public investment in ecosystem har- search interface is that future management will be vest and existence values. Rather, it provides the based on control of relative fishing efforts by fleet logarithmic utility function option so that users type, rather than on multispecies quota systems. who do have highly risk-averse attitudes about It is in any case not yet clear that there is any way ecosystem values can identify policy options that to implement multispecies quotas safely, without would better meet their objectives. Users should either using some arbitrary conservative rule like always construct a series of policy scenarios with closing the fleet when it reaches the quota for the varying utility weights w1, w2, and w3 on the log first (weakest) species taken, or alternatively al- utility components, to see how placing different lowing wasteful discarding of species once their emphases on these components would alter the quotas are reached. predicted best policy choice. If future multispecies management is indeed im- Use of these functions for policy exploration will plemented by regulation of fleet fishing efforts so generally involve some balance between these ob- as to track time-varying fishing mortality rate tar- jectives. Indeed, identifying the weighting factors gets as closely as possible, then a key practical is- to be given to each of these objectives may be the sue is how to monitor changes in gear efficiency most valuable aspect of this Ecosim routine. (catchability coefficients) so as to set effort limits each year that account for such changes in effi- Thus, to assist the user in achieving this, the ciency. Such monitoring is particularly important starting values of the objective functions have for fisheries that can show strong density- each been standardized relative to their base val- dependence in catchability, such that a unit of ues (from Ecopath), making them roughly com- fishing effort takes a much higher proportion of parable. The first two of these measures tend to some stocks (exerts a higher fishing mortality rate pull towards increasing fishing effort, while the per unit of effort) when stock size(s) is/are small. two others tend to pull towards reducing effort. Care should be taken to consider this balance There are at least two possible ways to monitor when giving relative weightings to the objectives. the changes in catchability (gear efficiency) dis- Also note that the optimizations should be per- cussed above. Both are based on monitoring fish- formed with a range of weighting factors for each ing mortality rates Ft over time, and using the re- objective function, rather than with single values, lationship qt=Ft/ft, where qt is fishing rate per which may miss a well-balanced solution (see unit effort and ft is effort. Cochrane, this volume). The first approach is to do traditional biomass Open-loop Policy Simulations stock assessments each year, and to estimate Ft as Ft=Ct/Bt, where Ct is total catch and Bt is esti- The fishing policy search interface of EwE de- mated vulnerable stock biomass. The second ap- scribed above estimates time series of relative proach is to directly monitor the fishing mortality fleet sizes that maximize a multi-criterion objec- rate, estimating probabilities of harvest using tive function that includes net economic value, methods such as annual tagging experiments and social employment value, mandated rebuilding, within-year estimates of relative decrease in fish