Cooperative Fish and Wildlife Research Units Program: Oregon
Education, Research and Technical Assistance for Managing Our Natural Resources

Management of mixed-stock recreational fisheries requires balancing fishery access and conservation of vulnerable stocks. Although accurate, timely estimates of stock-specific harvest are crucial in achieving competing objectives, limited sample sizes of stock assignments (e.g., physical tag recoveries) can limit the utility of estimates. Using empirically-based simulations of a high-effort Chinook salmon fishery in the Columbia River, we assessed the performance of competing model frameworks, sources of prior information, extents of data aggregation, and monitoring actions using limited tag-based stock assignments. We sought to improve accuracy for point estimates of cumulative harvest and harvest trajectories over time. Using simulated datasets, Bayesian likelihood-based models performed similarly to existing models and provided novel estimates of uncertainty. Incorporation of prior information most benefited stocks with fewer tag returns and produced the most accurate harvest trajectories with limited data aggregation. Among management actions yielding similar sample sizes, enhancing low tagging rates resulted in the greatest estimate precision. Analysis results highlight new assessment practices for the fishery, and our approach to model assessment may be applied to similar mixed-stock fisheries.