A Bayesian approach to choosing a design for surveying fishery resources: Application to the eastern Bering Sea trawl survey

Decision analytic approaches (DAAs) are useful for designing surveys of fishery resources because they enable evaluation of the trade-offs of each alternative and account for uncertainty. However, such problems may require models with many uncertain parameters. We present a DAA that uses a Bayesian posterior probability density function (pdf) of multiple uncertain population model parameters. The pdf is estimated in a companion paper (this issue). Draws are taken from this pdf to evaluate the management consequences of alternative survey designs. We illustrate the approach using yellowfin sole (Limanda aspera) in the eastern Bering Sea. The example illustrates the bias-variance trade-off of treating trawl survey abundance estimates as relative versus absolute biomass indices. Treating the indices as absolute resulted in low variances but large biases in allowable biological catch (i.e., coefficients of variation (CVs) ≤19% and biases >50%) in all survey designs considered. Treating the indices as relative resulted in decreased absolute biases (i.e., to ≤16%) but increased variance especially for a survey once every 3 years (e.g., the CV increased from 17 to 36%).