In-season forecasting of North Coast coho salmon marine survival: a decision-analytic method and retrospective analysis

, ,

Abstract

This paper develops an estimation algorithm that is intended to provide early warnings of poor marine survival conditions for Canadian North Coast coho stocks. The modelling approach attempts to partition weekly variation in observed coded-wire tagged (CWT) coho catches in boundary troll fisheries into components that depend upon total CWT coho smolts released, troll fishing effort, and marine survival rates. Uncertainty in marine survival rate forecasts is addressed via a Bayesian decision analytic framework, which accounts for overfishing risk. Statistical approaches, forecast accuracy, and forecast biases were tested using simulated data. Retrospective marine survival forecasts compared favorably with actual marine survival estimates obtained from post-season catch and escapement estimates and with previous assessments that attempted to forecast marine survival. The algorithm provided accurate warnings of poor marine survival conditions in all years for which such warnings were required. One false warning was issued from 19 possible cases, or a 5% Type I error rate. The marine survival forecasting procedure provides early warnings of poor marine survival up to 6 weeks in advance of the Canadian coho fishery opening. Therefore, it allows sufficient time for Canadian coho fishery managers to react to adverse marine survival conditions.