TO INCLUDE TIME-VARYING PARAMETERS AND SPATIAL STRUCTURE

A good understanding of the Ricker stock-recruitment relationship is important for estimating productivity of Pacific salmon (Oncorhynchus spp.) stocks and managing them. However, the best-fit Ricker model for a given data set typically leaves large residual variation. Identification of environmental effects to account for some of that variation is hampered by short data series, measurement error, and natural variation in survival rates on short and long time scales. We extended the Ricker stock-recruitment model in two ways. First, we used a multi-stock version of the model to incorporate environmental variables. Estimating model parameters across a large number of stocks simultaneously should improve the precision of parameter estimates and power to detect environmental effects. Second, in a separate analysis, we formulated the Ricker model in the context of a Kalman filter, which allowed the Ricker a parameter to 21 vary over time to reflect possible climate-driven changes. Such non-stationarit