Accounting for Risk and Uncertainty in the Management of Renewable Marine Resources

Andre E. Punt

School of Aquatic and Fishery Sciences, University of Washington, Seattle, WA 98195-5020

Fisheries management has been argued to be in crisis. Two examples of the perceived failure of fisheries management are the collapse of the northern cod fishery, and the depleted state of the groundfish resources off the US west and east coasts. Fisheries science has been implicated in this perceived failure as some of the assessments of the status and productivity of resources have been shown to be overly optimistic. Improved scientific management advice for fisheries must therefore better account for the uncertainty associated with the system being managed and the consequent risk attached to alternative management actions. The uncertainties routinely encountered when managing fisheries can be categorized into five classes: process uncertainty, observation uncertainty, model uncertainty, estimation uncertainty, and implementation uncertainty. The last of these relates to the ability to enforce management decisions adequately and is such largely outside of the scientific arena. However, implementation uncertainty is often the major uncertainty faced by decision makers. The remaining uncertainties can be divided into those that can be resolved relatively straightforwardly through scientific endeavor (e.g. more frequent resource surveys, experimental manipulation) and those (e.g. the size of future recruitment) for which this is probably not the case. In the past, scientific management advice was based on providing "best estimates" of the current status of, and prospects for, the fishery. There is, however, now increasing focus on attempts to reflect uncertainty and hence risk when providing scientific management advice to decision makers. Some of the current avenues of investigation relate to the use of Bayesian methods as the statistical basis for analyses, quantification of model structure uncertainty, and the use of management procedures. Bayesian methods provide the basis for quantifying the probability (risk) of various undesirable outcomes. However, the major reason for the use of these methods relates to the ability to formally incorporate "prior" information from studies of other species on the quantities that most strongly influence management advice (e.g. the extent of depensation) but for which data are usually lacking. Model uncertainty relates to the sensitivity of predicted outcomes to the assumptions underlying the analyses. Although it remains difficult to weight different models, exploration of model uncertainty allows the decision makers to be at least aware of the impact of basing management advice on the incorrect model. Management procedures are sets of rules that are tested by means of Monte Carlo simulation in terms of their ability to satisfy the (quantified) management objectives. Their adoption and use forces long-term thinking and detailed consideration of uncertainty. They are now being used relatively successfully in several fisheries jurisdictions (e.g. South Africa, Namibia, New Zealand, Australia, and the International Whaling Commission).