Hammond, T. R., and Trenkel, V. M. 2005. Censored-catch data in fisheries stock assessment. e ICES Journal of Marine Science, 62: 1118e1130.Landings statistics can be lower than true catches because many fish are discarded or landed illegally. Since many discards do not survive, treating landings as true catches can lead to biased stock assessments. This paper proposes treating catch as censored by bounding it below by the landings, L, and above by cL (for scalar c O 1). We demonstrate the approach with a simulation study, using a Schaefer surplus production model. Parameters were estimated in a Bayesian framework with BUGS software using two sets of priors. Both the traditional true-catch method and a survey-and-effort method (which was landings free) performed worse on average than the censored approach, as measured by the Bayes risk associated with estimates of maximum sustainable yield (MSY) and of an index of depletion (X). Recursive partitioning (regression trees) was used to associate simulation parameters to best-performing methods, showing that higher commercial fish catchability favoured the censored method at estimating X. In conclusion, censored methods provide a means of dealing with discarding and misreporting that can outperform some traditional alternatives.

Schaefer-model stock assessments can be imprecise when they are fitted to catch rate data (CPUE) because a large, unproductive stock can often explain CPUE trends as well as a small, productive one. However, consideration of life-history characteristics can improve parameter estimates by constraining maximum productivity. Therefore, we applied the methods of McAllister et al. (2001) to Northeast Atlantic spurdog (Squalus acanthias) by using demographic techniques to convert prior distributions for age-specific fecundity and natural mortality (the latter based on published estimates from tagging studies) to prior distributions for the intrinsic rate of population growth (r). The priors for r generated in this manner were then used in a Bayesian, Schaefer-model assessment of spurdog, fitted to bottom trawl survey CPUE data. Results suggest the stock is depleted to about 5% of virgin biomass.

Bayesian methods have a number of advantages that make them especially useful in the provision of fisheries management advice: they permit formal decision analysis, and they facilitate the incorporation of model uncertainty. The latter may be particularly useful in the management of contentious fisheries, where different nations and interest groups may suggest alternative assessment models and managementeach likely to imply different findings, even when using the same data. Such situations might be approached in a number of different ways. For example, one might attempt to choose a best model from all those available and to base decisions on it alone. Alternatively, one might make decisions that lead to acceptable outcomes under all envisaged models; or one could reach decisions that are good on average (where average is taken over the set of all competing models and is weighted by a measure of how well each model coheres with available information). This last approach is advocated in this paper, and a Bayesian technique for achieving it is presented and discussed. The main points of the paper are illustrated with a hypothetical application of the technique to the rebuilding of the biomass of haddock by a selective culling of seals.

This paper applies BASCET, a Bayesian Spatial Composition Estimation Tool for clusters of acoustically identified schools, to Bering Sea acoustic survey data collected during 1994. As the method employs prior information from an acoustic expert, procedures for eliciting such information are suggested and pitfalls of the process are indicated. Techniques for model checking using the posterior predictive distribution are employed, as is a multi-chain method for evaluating the convergence of the Markov-Chain Monte Carlo algorithm used in BASCET. Unlike methods based on neural networks, BASCET is able to provide confidence regions for its estimates of school cluster composition. In addition, it can indicate which school cluster attributes were most influential in determining a given estimate, a useful tool for model checking that is here demonstrated on a randomly selected cluster. Estimated abundance ratios of juvenile to adult pollock (Theragra chalcogramma) were compared, in two regions, to the values used by expert technicians. Ratios differed from expert values by less than 0.03 in both regions. The encouraging results reported here suggest that the BASCET method, originally tested on simulated data, may be usefully applied to real surveys.