Application of Bayesian Hierarchical Models to the Development of Precautionary Biological Reference Points for American Eel, Anguilla Rostrata, in Canadian Marine and Freshwater Systems: Constant or Time-Varying Kernel Model Parameter Selections?

Recent precipitous collapses of American Eel fisheries in Canadian freshwater systems have raised considerable concerns about the sustainability of population production and its susceptibility to anthropogenic activities. Numerous models have been developed to account for the observed dramatic fluctuations, but no quantitative reference points have been put in place to regulate total allowable harvests. In this study, we incorporated fishery-dependent and fishery-independent information collected since the mid-1950s in analyses of the southern Gulf of St. Lawrence (SGSL) as a representative marine area and the upper St. Lawrence River and Lake Ontario (USLR-LO) as a typical freshwater system. The overall objectives of this study were to establish biological reference points and thresholds for harvest control rules by applying Bayesian state-space statistics and generalized surplus production models. We composed four probability-based scenarios in the contexts of constant or time-varying kernel model parameterizations to assess model performance and multi-model inference. The best model in the SGSL was the constant K and r model (KCRC), while the time-varying K and r model (KVRV) was the best in the USLR-LO system. Combined with the modeled outputs, a set of adaptive management strategies are discussed in relation to individual population trajectories in the two habitats.