The paper puts forward a model of the Atlantic salmon fishery in the Baltic Sea that integrates the salient biological and economic characteristics of migratory fish stocks. Designed to be compatible with the framework used for actual stock assessments, the model accounts for agestructured population dynamics, the seasonal harvest and competing harvesting by commercial and recreational fishermen. It is calibrated using data and parameter estimates for the Simojoki River stock. The socially optimal policy for maximizing discounted net benefits from the fishery within an uncertain environment is determined using a dynamic programming approach and numerical solution method. Our results indicate that substantial economic benefits could be realized under optimal management without compromising stock sustainability.