A general model for reconstructing salmon runs

A general model is developed for salmon run reconstruction based on catch, escapement, and age composition data. The model is based on groups'' of salmon, each of which share the same characteristics but can differ from other groups in run timing, abundance, gear selectivity, and migration routes. The model is highly flexible so that it can be adapted to a variety of fisheries and can compare the fits of alternative hypotheses to available data. The model is applied to three sockeye salmon (Oncorhynchus nerka) districts in Bristol Bay, Alaska, USA, to show the effect of allowing age classes to arrive at different times and the impact of including process errors to mimic day-to-day arrival variability. The model predicts that in 2005, Wood River salmon comprised only 54% of the catch in the Nushagak fishing district (but 71% of the escapement), although these predictions are contradicted by genetic data for 2006-2008 showing high harvest rates of Wood River and Nushagak River fish but only light harvest rates of Igushik River fish. The genetics highlight the importance of including stock-specific availability parameters in future versions of the model to account for differences in harvest rates among stocks caught in the same fishing district.