The impact of management objectives and performance measures on model selection for Fraser River sockeye salmon.

Fisheries management entails decision making in the face of complex, uncertain systems, often resulting in decisions with imperfect outcomes that fail to achieve management objectives. For Fraser River sockeye salmon (Oncorhynchus nerka), fishery managers use models to better anticipate the magnitude of in-river loss of adults migrating upstream to spawn, thereby improving the chance of avoiding conservation concerns or losing fisheries revenue. Ecosystem-based models (Management Adjustment (MA) models), which predict in-river loss from forecasts of river environmental conditions, provide management advice on appropriate harvest adjustments (in terms of reduced catch) to increase the probability of achieving spawning escapement targets. The performance of a suite of MA model structures and predictor variables was assessed using a retrospective analysis and a range of asymmetric loss functions. Rank order of best models depended on the performance measures chosen, the relative importance of specific management objectives, and the degree of asymmetry in loss functions.