Current approaches to the estimation and implementation of biological reference points are often criticized for failing to incorporate environmental factors. The implicit increase in realism of population models that incorporate such factors is assumed to imply improved management. In most cases, however, this assumption remains untested and unquantified. Theoretical considerations and simulation studies are used to explore whether, and in which circumstances, the inclusion of environmental factors in the estimation of reference points could imply benefits to management.

Mark V. Bravington.   Measurement errors in stock-recruit data and reference points (poster)

Stock-recruit data are usually modeled as if stock were known exactly, but in practice there can be substantial measurement errors, especially on recent estimates. Ignoring measurement error can cause systematic and risk-prone bias in reference-point calculations. The problem is worst for stocks that are currently at historical low levels and thus are of greatest concern. This paper presents a new method for handling moderate measurement error, applies it to some examples of stock-recruit data, and discusses its applicability to stock-recruit data in general.

A given level of fishing effort is ultimately sustainable if it will not drive a stock to extinction. The key determinant is the slope at origin of the stock-recruit curve, and various proxies for it are commonly used as biological reference points. This paper starts by reviewing ways of making inferences about this parameter, culminating in the CONCR and DIMPOS methods, which resolve many problems with existing approaches. Sustainability is arguably the only purely biological basis for a reference point; other "biological" reference points involve implicit trade-offs with yield. However, management plans designed around sustainability reference points can lead to economically ludicrous levels of over- or underexploitation. Implications and possible solutions are explored.

The ultimate goal of an ecosystem approach to fisheries management is sustainability of fishery yields over a broad range of species. Because these systems are inherently complex, effective ecosystem management will entail a broad-based, interdisciplinary approach. Sustainability goals need to be translated into quantifiable terms, i.e., metrics. Biotic metrics range from single-species to whole-system attributes. Abiotic metrics describe environmental conditions. Human metrics (e.g., fishery capitalization, profitability recreational fishing opportunities, pollution, and regulatory compliance) are essential for grounding policy discourse in the face of differing value systems. Directionality, sensitivity, generality, feasibility, and uncertainty are important criteria when both ecosystem and human metrics are selected. We present several metrics for the Georges Bank ecosystem, ranging from single-species to whole-system measures, and discuss criteria for their usefulness and implementation. Ideally, a full suite of metrics should be developed to account for diverse system properties, e.g., spawner abundance, fishery capitalization, habitat quality, biodiversity, by-catch, primary production, or performance of regulations. These metrics can serve as the basis for decision criteria and reference points for the management process. We submit that a suite of metrics can help to quantify the inherent trade-offs that occur ecosystem management much more clearly than can a single-species focus. Although tangible goals, multiple metrics, and adaptive management will be needed to sustain marine ecosystems, it is clear that the science of ecosystem management is still developing.

Underlying this presentation is a focus on the challenges posed by structural uncertainty, a basic lack of understanding of the nature of fishery systems. Such uncertainties are widespread in fisheries, often produce surprises (dramatic, unanticipated changes to the system), and thus have major consequences for management, but they are not particularly amenable to standard analytical approaches. This presentation argues, therefore, that (a) adjusting the burden of proof is necessary not just analytically but also at the policy level, and (b) a precautionary approach to structural uncertainty requires structural adjustment to fishery management itself. Two major questions are examined. First, how should the burden of proof operate in the face of structural uncertainty? Second, how can structural changes, through robust management, lead to more inherently precautionary fisheries? These questions are explored by drawing on insights from theoretical models and on practical experience in Atlantic Canada’s groundfish fisheries, notably with the Fisheries Resource Conservation Council. In that context, technical challenges, such as deriving precautionary reference points, combine with fundamental human challenges involving such issues as whose data are to be used in the analysis and how to deal with the inertia that limits introduction of new management approaches. The discussion leads finally to a set of policy recommendations for future fishery management.

Many if not most marine fisheries today suffer from overcapacity of fishing fleets, in the sense that current capacity exceeds, often by a wide margin, the capacity that would be required to harvest a sustained yield. Such overcapacity makes the control of catches difficult and engenders continued demands for excessive quotas. This paper presents an analysis of the economics of overcapacity and shows that overcapacity is an inevitable consequence of management strategies that attempt to control annual catches without using individual quotas. It also shows that attempting to reduce overcapacity by means of vessel buy-backs may well exacerbate, rather than alleviate, the problem, inasmuch as vessel owners will see buy-backs as a form of subsidy.

As sport fisheries increase in coastal marine waters, it is expected that management agencies will lose direct control over fishing effort, total harvest, and fishery-based information for stock assessment. Effort control will be difficult because most sport fisheries remain open to unlimited public use without direct license or effort limitation. Without effort control, management tactics such as bag limits aimed at controlling total harvest typically fail because they regulate individual anglers. We developed an open-access model of sport fisheries in order to assess the effects of management actions on total sport fishing effort and exploitation. The basic model predictions are (i) linear effort response to fish abundance, (ii) fish abundance limit below which effort is not attracted, and (iii) asymptotic fishing mortality. Our predictions were consistent with observations in both freshwater and marine systems. Comparisons among fisheries indicated that the slope of proportionality is determined by factors such as access cost (lower slope) and development of facilities (greater slope). However, where our model predicted that restrictive bag limits would increase the effort response slope, we observed a lower slope. An alternative formulation of the model suggested that harvest restrictions might increase the apparent access cost or decrease the attractiveness of a fishery. We conclude that although bag limits are usually ineffective for their intended purpose, they probably act as an indirect means of effort control by limiting participation of consumptive anglers. We suggest that adaptive management of sport fisheries should include systematic bag limit experiments using angler effort as a key response variable.

Most quantitative fisheries management schemes involve some estimation of parameters of a population model as well as population states from a time series of data. Whatever the management approach, better methods for differentiating between models, estimating parameters, and quantifying uncertainty can produce better results. I consider the problem of incorporating both process noise (e.g. environmental variability) and observation error (e.g. random variation in catch-effort relationships) simultaneously in the estimation procedure. It has long been known that failure to incorporate noises properly can cause bias in parameter estimates and hence in functions of parameters such as MSY. I implemented a method that allows for incorporation of both process noise and observation error in model-fitting that not been explored for population models (Kitigawa, 1987). It works by numerically manipulating entire probability distributions of population states and observations. This method provides a nearly exact numerical extension of the linear, Gaussian, Kalman filter to nonlinear, nonGaussian cases. I present simulation results showing improvements in parameter estimates obtained by this method as compared to least squares methods. The relationship between this method and Markov chain Monte Carlo methods that have recently been used for fisheries models will be discussed.

Kitigawa, G. 1987. Non-Gaussian state-space modeling of nonstationary time series (with discussion). Journal of the American Statistical Association 82:1032-1063.

Management recommendations are often based on reference points derived from age-structured population models. Age-structured data are vital to getting reliable information on recruitment strength and mortality rates. To link these data to the model fishery, scientists commonly use a likelihood model. The multinomial function has been the default likelihood for most of the age-structured assessment models, but recently two alternative models have been proposed, namely Fournier's Robust Likelihood Function for Proportions (Fournier et al., 1989) and a Log-normal Error Model (Schnute and Richards, 1995).

Given the importance of these likelihood components to stock assessment models and the lack of comparative studies, a simulation analysis was performed to assess statistically the performance of model outputs derived from each of these three likelihood functions. A simulation model was implemented in a Monte Carlo framework to assess for bias and precision in management parameters (B_c, B₀, B_c/B₀, B_msy and F_msy). Precision was assessed by 95% confidence bounds estimators (derived from likelihood profiles) and summarized across different runs by the success rate (success of containing the true value) and length and shape of the interval. Life-history parameters were taken from groundfish stocks (Clark, 1993).

Clark, W. G. 1993. Groundfish exploitation rates based on life history parameters. Can. J. Fish. Aquat. Sci. 48:734-750.

Fournier, D. A., J. R. Sibert, J. Majokowski, and J. Hampton. 1989. MULTIFAN a likelihood-based age composition method for estimating growth parameters and age composition from multiple length frequency data sets illustrated using southern bluefin tuna (Thunnus maccoyii). Can. J. Fish. Aquat. Sci. 47:301-317.

Schnute, J. T., and J. R. Richards. 1995. The influence of error on population estimates from catch-age models. Can. J. Fish. Aquat. Sci. 52:2063-2077.

Sustainability issues in marine fisheries have recently grown to include concerns about the sustainability of the ecosystems that support fish. However, much of the detailed information about food web structure needed to manage fisheries from an ecosystem basis is not available. General patterns across different types of fisheries need to be identified to provide insight about which types of fisheries are associated with large ecological changes. To this end, we evaluated the trophic impact of pelagic fisheries aimed at pelagic fish that span a gradient of life-history strategies, ranging from fast-growing, short-lived fish (skipjack tuna, mahi mahi), to intermediate species (albacore tuna, bigeye tuna, yellowfin tuna), and slow-growing and long-lived species (blue marlin, swordfish, blue shark). The trophic impact of fishing was measured as the reduction in predation rate associated with fishery harvest, determined from species-specific bioenergetics and age-structured population models. The trophic impact of fishery exploitation was highly variable across this life-history gradient; greatest impacts were found for fisheries aimed at intermediate-life-history fishes. For example, the trophic impact of yellowfin tuna harvest was five times greater than that associated with blue shark harvest. This pattern emerged because trophic impact is determined both by the population’s susceptibility to fishing mortality and by its relative predation rate. Both of these properties vary along a life-history gradient, but in opposite directions; long-lived, slow-growing species are susceptible to fishing mortality but have low predation rates, whereas short-lived, fast-growing species have high predation rates but are robust to fishing mortality. We suggest a "rule of thumb" that the largest ecological changes are likely to be associated with fisheries aimed at species with intermediate life-history characteristics.

A precautionary approach to fisheries management has been clearly stated as a desirable goal in various laws and international agreements. Achievement of this goal, however, will require a change in the way fisheries-management decisions are made. Science and policy have complementary but interrelated roles to play. Although there are many important and challenging scientific problems, there are also policy decisions that must be made explicit, without which a precautionary approach is likely to fail. Rules to set target and limit reference points (the standards of proof) and the acceptable probabilities that the limits will be exceeded (the level of proof) are key policy decisions that must be made. The way these standards and levels apply to other species in the ecosystem, affected by the fishery either directly as by-catch or indirectly through habitat modification, must also be articulated. An effective precautionary management procedure should have: (1) transparent and precise decision rules, (2) explicit recognition of uncertainty, (3) fishing targets that are inversely related to uncertainty, (4) conservative defaults in the absence of data, and (5) testing of proposed management procedures to assure they will meet policy goals. Both successful and unsuccessful application of these ideas to marine mammal and invertebrate by-catch in U.S. domestic fisheries and the international purse-seine tuna fishery are illustrated.

Seamounts are island habitats of enhanced productivity located in the open ocean, dependent on larval drift and attractive to fish communities. The few unfished seamounts that have been surveyed exhibit a stunning biodiversity and abundance of both long-lived resident demersal and transient pelagic fishes. Not surprisingly, over the past 40 years, seamounts have become the object of increasing fishing pressure as coastal resources are depleted. Typically, fisheries have mined the unexploited biomass very fast and have maintained high cpue by moving on to fish new seamounts, a form of serial depletion. Many of these depletions are largely unreported, as they have occurred in international waters. The data are therefore often irretrievable, leaving fisheries biologists with no good reference points. How could we find reference points? We propose to use an ecosystem model based on present knowledge, together with the past state of seamount ecosystems inferred from the constraints of mass-balance and ecosystem simulations and validated from many sources. Data would come from seamounts from other regions, partial surveys and commercial data cpue, qualitative information on the benthic complexity recorded by photographs. Preliminary reference points for seamount fisheries can therefore be set and proposed harvest levels tested with the model under a "burden of proof" approach. Subsequent experience will allow the models and reference points to be fine-tuned and improved.

Scientific advice for management of New Zealand hoki fisheries is currently presented as risk analyses (probabilities of exceeding specified reference points) resulting from two competing assessment models, each fit to two contradictory abundance indices. Because scientists do not agree on the most appropriate model structure and data to use, managers are faced with the quandary of which of the alternate scenarios to pay attention to. Reference points are an incomplete specification of a management procedure; a true evaluation of the risk of a management plan should specify all aspects, including data collection and analysis. Fully specified management procedures can be tested through simulation to ensure they are robust to key uncertainties, providing a natural way to deal with uncertainty while ensuring adherence to the precautionary approach. In this paper we present results from a simulation study to evaluate the performance of alternate management procedures for New Zealand hoki fisheries. Each management procedure specifies the data that will be collected and the algorithm for translating the data observations into management actions (e.g. TAC's). The operating model encompasses the structural form of the two current hoki stock assessment models, with parameter uncertainty based on the Bayesian posterior estimates from the assessment model fits. Alternate hypotheses considered in the operating model include the form of the S-R relationship and a nonlinear relationship between an abundance index and stock size. Results suggest that decision rules based on estimates of current surplus production are more robust to uncertainties than those based directly on abundance index trends. Although decision rules that increase average catch do so at the expense of increasing stock risk and catch variability measures, economic performance (net present value) is maximized at intermediate levels of stock risk and catch variability.

Despite their well-documented shortcomings, commercial catch per unit effort (CPUE) data play a significant role in the assessment and management of many exploited fish populations. In an examination of the relationship between CPUE and abundance in exploited fish populations, a meta-analysis was conducted incorporating data from fisheries worldwide. A methodology for quantifying density-dependent catchability will be introduced, and the implications for both stock assessments that use CPUE data and management targets based solely on CPUE data will be discussed.

As more pressure is placed on fished populations around the world, fishery managers are often asked to place increasingly severe restrictions on the exploitation of a given stock. In the United States, managers are required by the Magnuson-Stevens act to base their decisions on the "best available data." The most commonly available data in the southeastern United States are landings data from commercial fisheries, although some length-frequency information is available and sometimes enough fish are sampled to provide age-length keys. Recreational data are available for some species. Fishery-independent data, however, are available for many species from this region. Samples from commercial and recreational fisheries are often biased because of the selectivity of the commercial or recreational gear--fishermen wish to maximize their harvest in the face of a bag limit, size limits, or other restrictions. Fishery-independent data are usually collected without regard to size and may be more representative of a population. Comparisons were conducted between various population parameters from data sources for five species of reef-associated fish: red porgy, Pagrus pagrus; grey triggerfish, Balistes capriscus; vermilion snapper, Rhomboplites aurorubens; white grunt, Haemulon plumieri; and scamp, Mycteroperca phenax. Our data show that estimates of yield per recruit, F_0.1, and F_max can vary tremendously depending on the data source used and that estimates based solely on fishery-dependent data are often higher than those derived from fishery-independent data.

Today, fishing is the major source of mortality in most commercially exploited fish stocks. Fishing can induce different selective pressures. These can be caused directly by fishing, i.e., result from elevated mortality (which is often highly selective). Selection may also be induced by ecosystem-level responses to fisheries as exploitation affects food availability and predation risk in both target and nontarget species. Moreover, fishing gear may alter the physical environment through mechanical disturbances. Responses to selection can be observed at two levels. First, at the community level, some species may suffer more from direct and indirect effects of harvesting than others; some species may even increase in abundance. Responses of species to exploitation are correlated with their life history: especially, species with late maturation at large size and with a low population growth rate tend to suffer from more pronounced declines than early-maturing species with rapid growth. Second, the phenotypic composition within species may also change. If phenotypic variability has a genetic basis, then fisheries-induced selection can result in evolutionary change. Traits potentially evolving include life-history traits, as well as behavioral traits (e.g., gear avoidance behavior) and morphological traits.

As an example I present data on life-history changes in the northeast arctic cod, showing a strong trend toward earlier maturation at smaller sizes. These changes are concordant with changes expected in selection responses to increased fishing effort at the feeder fishery. However, exploitation has also resulted in changes in food availability, which could have induced similar life-history changes. An analysis focusing on changes in the reaction norms for age and size at maturation helps to disentangle life-history changes caused by changes in growth (i.e., phenotypic plasticity) from potentially genetic changes caused by fishing mortality. The presented data analysis indicates that the documented life-history change in the northeast arctic cod has a significant genetic component.

A mixed Monte Carlo-Bootstrap procedure was used to incorporate assessment model uncertainty for decisions regarding the status of the Delaware Bay blue crab population. Incorporating this uncertainty is important because fishing mortality rates and the biological reference points to which they are compared are both estimated with error. The procedure generates probability density functions (PDFs) for both the current fishing mortality rate (Ft) and the replacement fishing mortality rate (FREP) as an overfishing definition. A probabilistic framework was then developed that specifies the probability that current fishing mortality rates are equal to or greater than the overfishing definition [Pr(Ft > FREP)], and its associated statistical confidence (i.e. 90%). Probability profiles were generated by integrating the area under the Ft PDF for different confidence levels (e.g., 80%) which can be thought of as the a-probability. Probability density functions for Ft and FREP, show that recent Fs (80% CI from 0.6 to 1.2) are generally below the FREP overfishing definition (80% CI from 0.9 to 1.6) but overlap significantly. At the 80% decision confidence level, the Pr(Ft > FREP) is 0.03. Thus, with high confidence (80%), we can state that blue crab is not currently being overfished.

The United States and many other countries have developed a set of standard reference points that can be used to determine allowable harvests. Here, I explore some of the problems that have arisen in this practice, including (1) uncertainties in current stock biomass as applied in reference point formula, (2) uncertainty in virgin stock biomass, (3) the inappropriateness of reference points derived for some species when applied to others, and (4) how use of reference points has led to an environment where stock-assessment scientists rarely evaluate alternative management policies. I suggest that alternative data-based rather than model-based approaches to setting quotas may be preferable. I also consider the distinction between overfishing and nonsustainability and point out that many stocks may be sustainably overfished. I consider the true meaning of the precautionary approach and how many people seem to have forgotten that the purpose of a fishery is to produce social and economic benefits to society and that those benefits are what need protection. Finally I suggest that the key to successful fisheries management is not better science, better reference points, or more precautionary approaches but rather implementing systems of marine governance that provide incentives for individual fishermen, scientists and managers to make decisions in their own interest that contribute towards societal goals.

Traditionally, fish stocks have been treated as homogeneous assemblages when, in reality, they are composed of individually varying stock components: populations or subpopulations. Ignoring the variability among stock components and prescribing harvest levels based on averaged parameter estimates may have substantial impacts for components that are below average in some sense. We describe how different spawner-recruitment relationships give indistinguishable statistical fits to the data but result in quite different parameter estimates and thus differing harvest prescriptions.

Here, we present the results of extinction models for Coho salmon (Onchorhynchus kisutch) on the west coast of North America. For each different spawner-recruitment relationship, we use models that incorporate both the variability in the parameter estimates for individual populations (streams) and the variability among different populations. We discuss management strategies that maximize long-term economic yield while minimizing local extinctions.

The franciscana dolphin, Pontoporia blainvillei, is endemic to the western South Atlantic Ocean. Its distribution is restricted to waters up to 30 m of depth, making it vulnerable to human actions. In southern Brazil the coastal artisanal gillnet fisheries have increased substantially since the early 1980s, and the incidental mortality of franciscana caused by these fisheries has been a source for concern. Even though the strandings of franciscanas killed in these fisheries have been documented for almost 20 years, the impact of these captures has remained unknown. For 1994 the number of franciscanas incidentally killed by the artisanal gillnet fleet based in Rio Grande (Brazil) and population size were estimated. Also, information on fertility and longevity have been used recently to estimate the population’s intrinsic growth rate. I integrate all available information into a population model in order to quantify the impact of incidental kills. The statistical analysis is conducted within a Bayesian framework to keep coherence when biological knowledge is weighted with observational data. My posterior analysis indicates a 98% probability that the population is currently decreasing. I analyse the impact of the incidental kill in the future by defining the population at "quasicollapse" when it reaches 10% of its current size, and calculating the probability of quasi-collapse within 30 years under alternative scenarios. Although numbers change slighly according to reasonable changes in model and prior specifications, results indicate that current levels of incidental kills cannot be sustained, calling for protective measures.

MSY is usually considered a biological reference point, for example equal to rK/4 in the logistic production model. In some cases there is a trend of MSY, which may be due to changes in fishers' practices. For the Atlantic yellowfin tuna fishery, MSY estimates were multiplied by two from 1970 to 1985, a possible reason being the increase of fishing area. Therefore MSY may be a process, combining biological and exploitation characteristics. We present an analysis of "catch, nominal effort, area fished" data for the Atlantic skipjack fishery using the production model:

dB_t/d_t = r (1-a_t) B_t (1 - B_t/K) - q f_t (B_t - a_t K)

where a_t = 1 - S_t/s. S_t is the size of area where catches are made. S_t results from fishers’ decisions. The parameter s may correspond to the whole area of the stock, but also to biological characteristics of the stock.

MSY, F_MSY, and f_MSY are therefore functions of the S_t time series. Accuracy of those process estimates should be considered together with the question of their corresponding models. For a given threshold on biomass level we can obtain different threshold values on mortality and on nominal effort. For example, if the biomass must remain greater than K/2 and if S_t is always lower than s/2, there is no need of any threshold on nominal effort. Conversely, if S_t remains equal to s, a value of f = r/(2q) leads to a long-term mean biomass value of K/2, and thresholds on nominal effort must be considered.

Considering such a model makes necessary the analysis of the covariation of nominal effort and fished area. We discuss this analysis.

The so-called "precautionary principle" seems to have appeared first as a legal requirement. The interpretation of this principle is awkward because it is nowhere clearly defined. The goal is to avoid overexploitation of resources, especially under conditions of uncertainty. Here I present a quantitative version. The idea is to require insurance to cover the possible collapse below a certain point of the harvested resource. The calculation of the risk is borne by the underwriter, and the premium must be paid by the exploiter.

This procedure has several desirable properties:

1. Because the risk of collapse depends upon our state of knowledge, the premium decreases as our knowledge increases.
2. The burden of evaluation of the risk of collapse is placed on the underwriter, who will have an interest in not underestimating that risk.
3. The cost of the premium is deducted from the discounted present value of the resource. Hence many risky enterprises would never be undertaken if the premium were required.
4. The risk depends strongly upon the exploitation strategy proposed. Hence the more risky procedures are discouraged.
5. The strategy that maximizes the net discounted present value depends upon the penalty. If the penalty is sufficiently large, the resulting optimal strategy may be far more conservative than otherwise.

The main disadvantage of this idea is the political difficulty of adopting it. All of the usual arguments against restraint of exploitation will undoubtedly be advanced. It also goes against the prevailing practice, which is to have the general public bear all substantial risks.

Recent FAO guidelines for the implementation of a precautionary approach to fishery management include the following: (1) Evaluate the potential consequences of alternative fishery management actions based on prespecified biological reference points. (2) Design management actions that account for uncertainties in stock assessment procedures. (3) Apply risk assessment methods to help formulate management advice. These recommendations have been adopted in many different fisheries, both established and developing, using a wide variety of stock-assessment methods that apply simulation modeling and a variety of different methods to account for uncertainty in model parameters. Several recent works, however, point to the need to account also for uncertainty in the structural formulation of stock-assessment models, e.g., over the stock-recruit relationship, assumptions about catchability, or stock structure. The reason is partly that model-based advice can be much more sensitive to structural uncertainty than parameter uncertainty. Although it may be desirable to account for both parameter and structural uncertainties, the fisheries literature has devoted attention mostly to the former; relatively little has been devoted to developing empirical methods to account for structural uncertainties. Instead, structural uncertainties have typically been accounted for by rerunning the stock assessment with structurally different models. Managers have then been presented with a variety of results in which the "best" action depends upon which structural alternative is assumed to be true. Because little formal scientific guidance is given about how to weight these alternatives, empirical considerations about how well alternative assumptions fit with existing evidence come to be ignored in the decision-making process. For example, when precautionary considerations have held sway, advice has often been based on the most pessimistic model formulated. Rather than being based on a formalized assessment of alternative assumptions for their credibility against scientific evidence, the decision is based instead on the imaginations and argumentative capacities of the scientists present. We argue for a more formalized empirical basis for dealing with structural uncertainties in providing precautionary management advice. We review statistical and decision theoretic methodologies to provide empirically based weightings (probabilities) for stock assessment results from structurally different models. The practical significance of these is illustrated with the fishery for Namibian orange roughy which began in 1994 and has been managed on a rigorously precautionary basis. Despite this caution, an apparent collapse in abundance has recently occurred. The example illustrates that the application of the statistical decision theory to account for both structural and parameter uncertainty, rather than only the latter, would empirically favor models that resulted in even lower total allowable catches and would pose lower risks of damaging the future productivity of the stocks.

Low-frequency or interdecadal climate variability presents a difficult challenge for fishery science and fishery management. A constant harvest-rate policy may work well for long-lived fishes. Exploitation of short-lived fishes benefits from harvest policies that link harvest rate to environmental conditions, but a delayed response can be desirable. Thus, rapid identification of regime shifts is not necessarily needed for fishery management, but early identification could provide planners with valuable advance notice of long-term increases or decreases in expected fish harvests. Planning horizons for fishery management, especially for rebuilding of depleted stocks, may have to be much longer than has been typical of past fishery management, and in some cases may require a century or more. Low-frequency climate variability may result in prolonged adverse periods of several decades in which little rebuilding occurs even in the total cessation of fishing. Adverse species interactions may further prolong rebuilding and can also greatly reduce sustainable yields. Optimal biomasses of large predators may be above half the unfished biomass if their removal releases smaller competitors. A critical weakness in fishery management is the shortness of historical time series of climate and fishery data, which make it extremely difficult to provide the information needed for management to address the problems of climate variability.

Striped mullet (Mugil cephalus) is an important component of the food web dynamics of coastal waters and estuaries throughout the southeastern United States. The spawning stocks of striped mullet have been under intense fishing pressure since the development of the roe export market in the late 1970’s in Florida and other regions along the Gulf of Mexico. Results of stock assessments showed that spawning potential ratio (SPR) values for striped mullet in Florida ranged from 18 to 22% during 1989-1992. In 1993, the Florida Marine Fisheries Commission established a target SPR level of 35% for striped mullet and adopted several management regulations aimed at reducing fishing pressure on the spawning stocks. With the implementation of the net-ban in July 1995, which extended the regulations adopted in 1993, it was expected that (1) the commercial landings of mullet would be substantially reduced, (2) fishing effort and cathability would be reduced, and (3) the target SPR would be reached earlier than was projected on the basis of the 1993 regulations. An age-structured population model, using fishery-dependent and fishery-independent data, was used in 1998 to evaluate the impact of regulations and the efficacy of the target SPR for management of the mullet fishery in Florida. The 1998 fishing mortality rate was estimated at 0.55 per year, corresponding to an SPR value of 32%. The 1998 F was slightly higher than the target level (F_0.35 = 0.45) necessary to achieve the 35% SPR level. The model forecast shows that, if the recovery continues, the target SPR is achievable within two to three years.

Ricker defined MSY in this way (italics added):

MAXIMUM SUSTAINABLE YIELD (MSY OR YS): The largest average catch or yield that can continuously be taken from a stock under existing environmental conditions. (For species with fluctuating recruitment, the maximum might be obtained by taking fewer fish in some years than in others).

The usefulness (or lack thereof) of the concept of MSY hinges of how we understand the words that are marked in italics. I will explore this understanding by using a model of squid fisheries in Central California. The main prey of squid are krill, and the abundance of krill is determined by environmental conditions, and oceanic upwelling can be used as signal for krill abundance. In the simplest case, the evironment exists in two states that correspond to high and low krill (and thus squid) abundance. I will use a sequence of submodels, with different assumptions about our knowledge of the system (upwelling-krill-squid-fisherman behavior), to illustrate how a sophisticated understanding of the definition of MSY can be used in management of fishing intensity. One concept that will emerge from this discussion is that MSY should function as much as a constraint as a target.

Many fisheries regulatory systems for achieving harvest-rate goals are based on estimating stock sizes and setting a total allowable catch based on an optimal harvest rate. These systems are vulnerable to errors in estimating current stock size. A second approach is to regulate fishing effort; exploitation is limited by a set number of fishing days. This approach assumes that fishing mortality is proportional to fishing effort and often fails because of changes in catchability over time. A third approach, one that is independent of estimating stock size, is to monitor exploitation rates as the fishery proceeds and to calculate catchability coefficients as a function of fishing mortality rate and fishing effort. Before changes in catchability can be tracked over time, a direct estimate of exploitation rate is required. Options for directly monitoring exploitation rates include tagging programs, area-swept estimates, and monitoring of changes in total mortality from mean body size and/or age composition data. We develop "closed-loop" simulation models for evaluating management systems that rely on annual estimates of stock size and for management systems that monitor exploitation rates using data from surveys and tagging programs. Simulation studies suggest that intensive tagging programs, or extensive tagging combined with annual survey data, are a more cost-effective and risk-averse method of fisheries management. Survey information on stock parameters for developing fisheries, or fisheries that lack time-series data, are greatly improved when combined with tagging data.

Spotted seatrout, Cynoscion nebulosus, is managed with an objective of maintaining a spawning potential ratio (SPR) of 35%. The 1999 stock assessment showed that, overall, the SPR values have increased from an average of 15% in the 1986-1989 period to 21% in the 1990-1995 period to 26% since 1996 in response to increased regulation. The regional median SPR values for spotted seatrout from the period 1986-1998 were 15-17%, and the 1999 SPR values ranged from 20% to 37%. If managers were to distinguish between threshold and targets, the objective of 35% SPR, which has been called both a target and a threshold, is more properly a target and the threshold would be 15-17%. However, SPRs only track fishing mortality rates and cannot identify issues with recruitment. For example, although SPR values have generally been increasing in the Northwest, the spawning biomass in that region has been decreasing since 1989. Spawning potential ratios by themselves do not capture the condition of the stock, and it is necessary incorporate a measure that tracks spawning biomass. This is the same conclusion reached by a 1995 study sponsored by the Gulf of Mexico Fishery Management Council. Selection of appropriate biomass benchmarks for spotted seatrout will be difficult because the species has undergone extensive habitat loss, has been managed under a variety of size limits, has a short time series of suitable assessment data, and is primarily a recreational species.

Life histories of fish populations presumably reflect optimization of the age-specific reproductive effort and survival through natural selection. The addition of fishing mortality may radically change the optimal life history, depending on its magnitude and age/size-specific pattern. Such evolutionary changes in life history may have detrimental effects on stock dynamics by reducing yield and increasing the probability of stock collapse. We developed a model that integrates demographics and quantitative genetics to evaluate these possibilities. Our model allows the age-specific allocation of surplus energy to growth or to reproduction to evolve in response to a change in mortality scheduling. We include two sources of stochasticity--recruitment to age-0 and annual surplus energy--resulting in variable growth, reproductive output, and survival. Using this model, we track the change in optimal life history, yield, and probability of a population collapse (75% decrease in biomass) in response to changes in harvest regime. Specific harvest regimes are characterized by the fishing mortality and the minimum size subject to harvesting. Model predictions agree with empirical evidence: i.e. except under special circumstances, the evolutionary response to harvesting resulted in earlier ages at maturity and lower population biomass. By explicitly recognizing that stock production is a function of the size-specific allocation of energy to growth or to reproduction, and assuming that the allocation pattern will evolve in response to harvesting, our model allows evaluation of the long-term consequences of various harvest regimes for stock dynamics.

Most data sets used to estimate biological reference points, e.g. spawner recruit data or data for production models, contain enough information to estimate less than one and a half parameters. This information content is simply not good enough to estimate biological reference points reliably. I will review how meta-analytic approaches can be used to improve such estimates greatly.

A dominant trend in the implementation of the precautionary approach has been to develop conservative harvest rules based on target and threshold reference points for biomass and fishing mortality. These are intended to prevent overfishing by advocating more conservative fishing targets and by triggering corrective management actions when thresholds are approached or exceeded. For this procedure to work, stock assessments must be reliable and consistent from year to year, and reference points must be well determined. Neither of these conditions is commonly met in reality. On the west coast of North America, for example, where the data used in assessments are often insufficient and information gaps are filled by model assumptions, the uncertainty in the assessments tends to be large and dominated by structural uncertainty. Competing models often result in very different biomass levels and even different biomass trends, which complicate the implementation of generic harvest rules. In addition, changes in model structure between successive assessments can bring about substantial changes in estimates, which may unnecessarily disrupt management if point estimates are carried into catch quota recommendations. A second main problem is that generic reference points are often related to population parameters such as unfished biomass and spawning biomass per recruit and to MSY-related parameters, whose estimation, meaning, and usefulness are elusive in the face of decadal changes in productivity and regime shifts. In these situations, simple management procedures that respond more directly to monitoring signals may be as or more effective at maintaining stocks at reasonably productive levels than are more elaborate rules based on uncertain and evolving stock assessments. I discuss these problems using Pacific halibut as an example.

Biological invasions are increasingly recognized to be an important proximate cause of biodiversity loss. They have also disrupted key ecological functions in a wide range of systems. In marine systems this has led to the near collapse of many fisheries, but they have not yet received much attention from economists. This paper considers a generic model of invasions and invasion control. It shows how economic parameters drive both the threshold of the spread of invasive species, and the dynamics of the interactions between invasive and "native" species.

Large uncertainties that prevail in fisheries create widely recognized biological, economic, and social risks. These risks led to the development of FAO's Precautionary Approach and in certain cases to reversing the burden of proof. However, management agencies still need to implement these approaches more widely, and harvesters must comply more fully with the resulting regulations to reduce implementation error. Perhaps the long-term goal of achieving biologically and economically productive fish stocks could be reached in a less adversarial and more cooperative way by management agencies and industry working together to establish ecocertification criteria for management and harvesting practices in specific fisheries. Consumers' preferences for ecolabeled products would help align industry's goals with those of management agencies. The threat of losing certification would create incentives for appropriate levels of precaution in agencies' regulations and increased compliance with them by industry. Ecolabeling initiatives will be aided by recent developments in "sustainability indicators" and scientifically rigorous target, limit, and threshold reference points against which those indicators can be compared. Particular types of ecolabeling (e.g. voluntary certification combined with consumer education programs) can potentially create appropriate incentives for agencies and industry while avoiding negative reactions by GATT/WTO to the resulting changes in trade.

Is adaptive management fundamentally incompatible with the precautionary approach? This question is examined in terms of the controversy surrounding experimental fishing for southern bluefin tuna (SBT).

SBT is a highly depleted stock with a rebuilding target and time frame defined by the responsible management body (CCSBT). All recent stock assessments have found that the stock is depleted, but large differences exist in estimates of recovery probabilities under current catches. In 1996, CCSBT adopted a set of principles and a process for considering adaptive management involving experimental fishing, which are reviewed for their consistency with the precautionary approach. Substantial efforts to develop a program in line with these principles did not succeed, in part because of the lack of a decision-making framework. In 1998 and 1999, Japan conducted unilateral experimental fishing, arguing that the additional substantial catches could reduce uncertainty in stock assessments and thus were justified. This situation led to international legal proceedings under UNCLOS, in which preliminary measures were issued preventing further unilateral experimental fishing. This decision has been cited as a manifestation of industry’s "worst fear with the implementation of the precautionary approach." Results from recent stock assessments and evaluations of possible management strategies indicate that the Japanese experimental fishing, even if successful, was unlikely to resolve the disparity in estimates of the recovery probabilities or substantially improve the performance of these management strategies. In this context, it is the absence of a management framework, rather then a fundamental problem with adaptive management, that challenges the compatibility of these experimental fishing catches and the precautionary approach.

We developed a Bayesian approach to resolving the problem of accounting for measurement errors in the estimation of the parameters and the Biological Reference Points (BRPs: MSY, stock generating MSY, harvest rate at MSY) from stock and recruitment (SR) data. The key feature is the application of the Rao-Blackwell formula in a four-step procedure: (1) measurement errors are quantified by calculation of the posterior probability distributions of the annual stock and recruitment values conditioned on the field observations; (2) SR data sets are generated from those distributions; (3) for each data set, the parameters of the Ricker model and the associated BRP posterior distributions are generated, accounting for recruitment process variability; and (4) posterior distributions integrated over SR data uncertainty are estimated by means of the Rao-Blackwell formula; i.e. the distributions from step 3 are averaged over the data sets generated in step 2. This procedure is easy to implement and provides smooth and precise estimates of posterior distributions with minimal additional computational effort. Because a Bayesian framework is used throughout the process, the resultant posterior distributions are ultimately conditioned on the field observations. We illustrate the procedure using Atlantic salmon data from a river in France, where measurement errors of stock and recruitment estimates are inferred from mark-recapture observations according to a Bayesian hierarchical model. This case study demonstrates that (1) measurement errors tend to increase parameter and BRPs uncertainty and shift the distributions, resulting in more conservative management advice, and (2) the disruptive effect of measurement errors is minor when there is a high contrast in the stock variable.

In many fisheries management agencies, current stock state is compared to a set of standard reference points to provide advice on allowable catch. One problem is that both current stock indices and reference points are estimated on the basis of models and data that bear high uncertainty. Consequently, current indices and reference points are not independent, nor are their uncertainties. An alternative scheme would be to use sets of directly measurable parameters as current stock indices; they would be known with a measurable sampling error. Models would be needed to set the corresponding reference points, but the propagated errors and possible model misspecification would be limited to the latter. Clearly, survey indices of population abundance could be used in such a way. They may be combined with reproductive traits, because there is increasing evidence that fishing affects life-history traits of fish. Fish compensate for fishing mortality by faster growth, earlier maturity, and increased fecundity. Reproductive traits are also critical in determining the population growth rate; hence they may be used as indices of population viability. The question is, which reference points are appropriate for, e.g., age at maturity? Reference points based on a simple life-history model are proposed. Optimal age at maturity that maximizes lifetime reproductive output under the current fishing mortality, or that allows a given (minimum) lifetime reproductive output, is a possible reference point. The question is whether they should be targets or thresholds. North Sea cod is used as an example.

In this paper I briefly review the policy and technical background for the development and implementation of a precautionary approach to resource management. Barriers to the application of the precautionary approach are described on the basis of domestic and international experience over the last five years. International regional fishery organizations have made some progress in implementation, but negotiations on actual management measures have still not really made a transition to a precautionary framework in most cases, partly because most fisheries management is focusing on ending overfishing and starting rebuilding. Although precaution is needed for rebuilding programs, the initial conservation issue concerns seem to override needed changes in the approach to management all too often. Domestically the situation is better but still not full implementation of precautionary management by any means. Finally, I describe some possible means of making progress, in both arenas, toward more explicit application of the tenets of the precautionary approach to fisheries management.

Management targets (such as maximum sustainable yield), although theoretically tractable, have been criticized for their failure to protect harvested stocks that are subject to uncertain natural variation in their dynamics. These limitations arise, in part, because the fisheries of interest are often difficult to sample (hence, stock size or other state variables are poorly known), and they are often ill-defined with respect to their dynamic properties. For example, we seldom know the qualitative form of density dependence that governs different demographic rates and rarely, if ever, know the parameter values that define these relationships. The fisheries of historic concern to managers are those harvested for food. An important emerging fishery in many parts of the world is the marine ornamental fishery--fishes and invertebrates harvested and sold in the aquarium trade. Few management strategies have been devised for these fisheries, yet they may represent the most tractable systems: e.g., the benthic life stages are relatively site attached and easily sampled, and many species are amenable to experimental study. Here, we investigate, through stochastic population models, the efficacy of strategies designed to manage a marine ornamental fishery. We use field data for a model system to define functions and associated parameters that govern the dynamics of the unharvested population, and we simulate the responses to harvesting and different management strategies. The data are taken from extensive field studies of the tropical damselfish Dascyllus trimaculatus. If management targets are going to work anywhere, it may be in systems like this. If targets work in these systems, but not more traditional fisheries, then the "requiem of MSY" may be premature--there is not a single management approach for all systems. Instead we may require a more diverse arsenal, from which we match the general strategy to the nature of the system being managed.

Maximum Sustainable Yield (MSY) control rules, established under the U.S. Magnuson-Stevens Fishery Conservation and Management Act for precautionary management of fisheries stocks, define two reference points. The maximum fishing mortality threshold (MFMT) and minimum stock size threshold (MSST) are used to determine overfishing rates and stock status with respect to overfished levels. These limits also guide the development of rebuilding plans intended to increase stock biomass to MSY-producing levels. Under constant fishing mortality control rule, the MFMT is equated to the MSY level of fishing mortality (F_MSY). Various biological reference points have been suggested as surrogates for F_MSY under situations with limited data and little analysis. Among them, natural mortality rate (M) is most popular and has been used in Bering Sea and Aleutian Islands crab fisheries management. This paper addresses the shortcomings of equating F_MSY to M under classical surplus production (Schaefer and Fox) and spawner-recruit (Beverton & Holt and Ricker) settings and suggests appropriate formulae and methods for determining surrogates for F_MSY in terms of M. It also addresses difficulties in using other popular biological reference points for F_MSY, such as fishing mortality associated with recruitment levels that would exactly replace their parent stocks 50% of the time (F_REP), with a simulation study.

Using control optimization theory, we developed a precautionary and ecosystem-based framework for fisheries management. This framework can be formulated with one or more variables and can be modified to produce quota-setting systems ranging continuously from constant catch to constant escapement. By modifying target and limit points of the framework, one can provide appropriate precaution inversely proportional to the state of knowledge about the stock and protect the ecosystem roles of key species, particularly those in lower trophic levels. We will present the framework, demonstrate its utility, and discuss how it can be particularly useful in data-poor situations. We will also demonstrate how, in the case of virtually no information, the framework can be applied through no-take marine reserves. Finally, we will discuss the degree to which the lessons of our study have been or are being applied in U.S. fisheries management.

In 1995, the U.N. Fisheries and Agriculture Organization (FAO) adopted the precautionary approach for fisheries management, and the U.S. passed the 1996 Sustainable Fisheries Act, which led to the 1998 National Standard Guidelines recommending that fisheries management councils adopt a precautionary approach. In agreement with solicited scientific guidance, the implementation of the precautionary approach by FAO and the U.S. has followed the path of risk assessment, with the stipulation that fisheries management be risk averse. Equating the precautionary approach with risk assessment, however, has diminished its scope and potential for protection and has preserved the guiding question, "How much can we get away with when exploiting natural ecosystems?"--a question that directly opposes the precautionary principle and has led to the worldwide severe decline of fish populations. The difference in fisheries now is that managers acknowledge that we cannot get away with as much fishing as we once thought, so they aim to massage the risk-analysis process accordingly.

Instead, it is here proposed that risk assessment more appropriately be relegated to a subsidiary role, useful in specific, data-rich circumstances, and be placed in the context of a larger alternatives assessment more appropriate for implementing the precautionary approach and taking into account all its scientific, social, and ethical aspects. This shift changes the guiding question to, "What action(s) will allow us to have the smallest possible impact upon the natural ecosystem while enabling us to sustain ourselves and future generations?" This poster contrasts the two approaches.

Michael D. Tringali.   Thresholds for genetic change in exploited and/or supplemented populations (poster, YI)

Two genetic challenges to the maintainence of population viability--inbreeding depression and the accumulation of mildly deleterious mutations (i.e., mutational meltdown)--have become prevalent in discussions of fishery resource management and conservation. Fishery harvest potentially increases these genetic risks indirectly by reducing overall population abundances; supplemental stocking may do so directly by artificially magnifying the reproductive contributions from a subset of the population. Ultimately, exploitation and stocking may cause lowered effective population numbers and, possibly, increased rates of genetic change. Generally accepted guidelines for "acceptable" values of effective population number are on the order of 10² to preclude inbreeding depression and 10³ to preclude mutational meltdown. These numbers are often portrayed as threshold values, below which the population is at risk but beyond which the population behaves genetically as if it were infinitely large. In this presentation, I review the genetic theory that forms the bases of these recommended values, explore sources and directions of bias in their estimation, and address some common misconceptions relating to their application in resource management.

Much of the machinery of modern fish stock assessment is aimed at providing better estimates of current stock size, but the adaptive and feedback policies needed for sustainable long-term management are specified in terms of fishing mortality-rate goals in relation to relative stock size, and estimates of optimum fishing mortality rates have dropped sharply since 1990. Should we elect to redirect assessment methods and data gathering more directly toward estimation of current and optimum fishing mortality rates, we could probably bypass many of the costs and pitfalls that have plagued traditional assessment approaches. Further, we could concentrate more on providing advice about how to implement safe feedback policies, rather than just contributing to debates about how much is out there to be harvested.

All fisheries management models incorporate simplifying assumptions about important ecological and oceanographic mechanisms that are fundamentally uncertain or stochastic. But exploited fisheries are also subject to equally important uncertainties associated with fisherman behavior. Fishermen make a broad range of decisions from long-term entry/exit decisions to daily or even hourly decisions about where and how to fish. These decisions are complex, because they are influenced by factors including regulations, technology, weather, and expectations about prices, costs, and abundance. They are important determinants of the character of exploited fisheries, however, because they ultimately determine the spatial and intertemporal pattern of fishing mortality in a fishery. Although biologists have taken a stab at incorporating fisherman behavior in management models, much of the work is ad hoc. At the same time, there is a rich tradition of both conceptual and empirical behavioral modeling in economics. This paper discusses some of this modeling, with an aim toward demonstrating the potential usefulness of some economics-based behavioral modeling. We utilize data typically collected for biological management and estimate some models of participation and spatial choice. These models are then used to forecast the implications of various management measures that might be applied to the red sea urchin fishery in California.

A directed commercial fishery on horse (gaper) clams, Tresus capax and T. nuttallii (Bivalvia: Mactridae) is being developed in British Columbia, Canada. We used analysis of spawning stock biomass per recruit (SPR) to evaluate the effect of different numbers of rotation years and alternative exploitation rates on the stock reproduction potential. Conventionally, SPR modelling is applied to annual fisheries and fitted in a deterministic manner. We adapted the analysis for a multiyear rotation fishery. Our analysis also incorporated variations in the estimates of natural mortality rate and growth, converting uncertainties inherent in scientific knowledge into estimates of biological risk. The information produced permits fishery managers to choose exploitation rates and number of rotation years on the basis of the degree of risk managers are willing to take. Because of poor understanding of the resilience of the horse clam stocks to fishing, highly precautionary target and limit reference points are recommended for each possible fishing scheme until further information on stock dynamics is obtained.