Development of Bayesian production models for assessing the North Pacific swordfish population

Abstract

Bayesian surplus production models were developed for assessing the North Pacific swordfish population under alternative scenarios: a single-stock scenario and a two-stock scenario with subareas that represented the western central and eastern Pacific Ocean regions. Biomass production was modeled with a three-parameter production model that allowed production to vary from the symmetric Schaefer curve using an estimated shape parameter. Lognormal prior distributions for intrinsic growth rate and carrying capacity were assumed. Goodness-of-fit diagnostics were developed for comparing the fits of alternative model configurations based on the root-mean squared error of catch per unit effort (CPUE) fits and the standardized CPUE residuals. Production model fits for 1952–2006 indicated that the Japanese longline CPUE numbers were influential under each stock scenario because these scenarios were the longest time series of relative abundance indices. Model results also indicated that assumptions about the prior means for intrinsic growth rate and carrying capacity may be important based on the model configuration.