Model averaging and probable status of North Pacific striped marlin, Tetrapturus audax

Abstract

We show how model averaging can be applied to estimate the probable status of a fishery resource under assessment scenario uncertainty. This approach is applied to North Pacific striped marlin ( Tetrapturus audax ), an apex predator that may be vulnerable to recruitment overfishing in pelagic longline fisheries targeting tunas. In the current striped marlin assessment, two assessment scenarios were used to account for different hypotheses about the steepness of the stock–recruitment dynamics. Estimates of spawning stock and recruitment from these scenarios are used post hoc to fit age-structured production models that represent alternative hypotheses about the degree of compensation in stock–recruitment dynamics and the degree of serial correlation of environmental forcing. Model-averaged estimates of target spawning biomass to produce maximum sustainable yield (SMSY) and the associated limit fishing mortality (FMSY) characterize relative stock status (S/SMSY and F/FMSY) under each scenario. Scenario-weighted averages of relative status determine probable stock status, with weightings reflecting the credibility of each scenario. Estimates of the variance of probable status account for both model selection and assessment scenario uncertainty in risk analyses. Using model averaging to estimate probable stock status from multiple assessment scenarios is analogous to using ensemble averages from multiple predictive models to make weather forecasts.