Quantifying the effects of sample size and species distribution on the precision and accuracy of abundance estimates from bottom-trawl surveys in the Gulf of Alaska

We assessed the effect of survey effort reduction on the accuracy and precision of estimates of abundance for 4 commercially or ecologically important species with differing distributions observed in a bottom-trawl survey conducted in the Gulf of Alaska. Simulations from a spatiotemporal generalized linear mixed model based on historical observations of catch densities were used to evaluate the statistical robustness, measured in terms of coefficient of variation, relative bias, and relative root mean square error, of the abundance estimates and their variances. These metrics were used to compare estimates between the traditional design-based estimator and the alternative estimator, based on a vector autoregressive spatiotemporal model, at 4 different sampling densities, representing 2 historical and 2 theoretical sampling effort levels on either side of the historical range. The recent reduction in the density of survey sampling from 820 to 550 stations had only a modest effect on the performance metrics for both estimators for arrowtooth flounder (Atheresthes stomias), Pacific cod (Gadus macrocephalus), and Pacific ocean perch (Sebastes alutus). However, the effect on the abundance estimates for sablefish (Anoplopoma fimbria) was substantial. We attribute this difference in results to the wider depth range utilized by sablefish, which preferentially occupy the relatively under-sampled deep strata (>500 m), and to the truncated survey area at the reduced sampling levels where the deepest strata (>700 m) have been eliminated.