An in-season management system for sockeye salmon returns to Lynn Canal, southeast Alaska

Abstract

An in-season management system was developed for four stocks of sockeye salmon (Oncorhynchus nerka) from Chilkoot and Chilkat Lakes in Southeast Alaska, using a total run (catch + escapement) database built from analysis of 132,000 scale samples. Separate management objectives were defined for early and late stocks from each lake. A Ricker model, with nonlinear least squares and a bootstrap procedure, was used to estimate optimal escapement levels, which were lower than previous estimates, and precision bounds. Comparison of several in-season forecasting models based on run timing showed that a model combining a preseason forecast with average proportion forecasts was most accurate. Average proportion forecasts were improved by timing shifts identified by inflection points. Linear regression was used to forecast Chilkat Lake escapement in-season. Accuracy of the best forecasting models for each analysis was ≤ 25% by the first quartile of abundance.