Bayesian information updating procedures for Pacific salmon run size indicators: evaluation in the presence and absence of auxiliary migration timing information

Abstract

Preseason forecasts of Pacific salmon run size are notoriously uncertain and are thus often updated using various abundance indices collected during the run. However, interpretation of these in-season indices is confounded by uncertainty in migration timing. We assessed the performance of two Bayesian information-updating procedures for Kuskokwim River Chinook salmon (Oncorhynchus tshawytscha), one that uses auxiliary run timing information and one that does not, and compared the performance with methods that did not involve updating. We found that in-season Bayesian updating provided more accurate run size estimates during the time when harvest decisions needed to be made, but that the incorporation of run timing forecasts had little utility in terms of providing more accurate run size estimates. The latter finding is conditional on the performance of the run timing forecast model we used; a more accurate timing forecast model might yield a different conclusion. The Bayesian approach we developed provided a probabilistic expression of run size beliefs, which could be useful in a transparent risk-assessment framework for setting and altering harvest targets during the season.