A Bayesian hierarchical approach to integrating historical and in-season genetic data for real-time assessment of a mixed stock fishery

Abstract

With advances in molecular genetics, it is becoming increasingly feasible to conduct genetic stock identification (GSI) to inform management actions that occur within a fishing season. While applications of in-season GSI are becoming widespread, such programs seldom integrate data from previous years, underutilizing the full breadth of information available for real-time inference. In this study, we developed a Bayesian hierarchical model that integrates historical and in-season GSI data to estimate temporal changes in the composition of a mixed stock of sockeye salmon (Oncorhynchus nerka) returning to Alaska’s Chignik watershed across the fishing season. Simulations showed that even after accounting for time constraints of transporting and analyzing genetic samples, a hierarchical approach can rapidly achieve accurate in-season stock allocation, outperforming alternative methods that rely solely on historical or in-season data by themselves. As the distribution and phenology of fish populations becomes more variable and difficult to predict under climate change, in-season management tools will likely be increasingly relied upon to protect biocomplexity while maximizing harvest opportunity in mixed stock fisheries.