Test fishery indices for sockeye salmon (Oncorhynchus nerka) as affected by age composition and environmental variables

Abstract

In this paper, we explore sources of variability in test fishing indices for sockeye salmon (Oncorhynchus nerka) and illustrate potential solutions. For our analysis, we use the Port Moller test fishery in Bristol Bay, Alaska. The traditional model used to forecast run size using Port Moller test fishery data is based on a simple linear regression of total returns to the fishing districts against the cumulative test fishery catch-per-unit-effort (CPUE) index on a given date. We show that much of the inaccuracy of recent forecasts based on test fishery indices is due to the variable age composition of sockeye runs and the unequal vulnerability of respective age classes to the test fishing gear. We also illustrate solutions to the problems presented by variability in migration route and variability in return timing. In warmer years, we found that there was an increase in effective vulnerability to the test fishing gear and that the strength of this relationship increased linearly with the average body length of age classes. Retrospective analysis shows that correcting for age composition, migration route, and timing provides an in-season index of abundance with an average error of 6.7 million fish or 21%, compared with 33% for previous models.