Comparison of parameter estimation methods for detecting climate-induced changes in productivity of Pacific salmon (Oncorhynchusspp.)

Abstract

Pacific salmon (Oncorhynchus spp.) populations can experience persistent changes in productivity, possibly due to climatic shifts. Management agencies need to rapidly and reliably detect such changes to avoid costly suboptimal harvests or depletion of stocks. However, given the inherent variability of salmon populations, it is difficult to detect changes quickly, let alone forecast them. We therefore compared three methods of annually updating estimates of stock-recruitment parameters: standard linear regression, Walters' bias-corrected regression, and a Kalman filter. We used Monte Carlo simulations that hypothesized a wide range of future climate-induced changes in the Ricker a parameter of a salmon stock. We then used each parameter estimation method on the simulated stock and recruitment data and set escapement targets and harvest goals accordingly. In these situations with a time-varying true Ricker a parameter, Kalman filter estimation resulted in greater mean cumulative catch than was produced by the standard linear regression approach, Walters' bias correction method, or a fixed harvest rate policy. This benefit of the Kalman filter resulted from its better ability to track changing parameter values, thereby producing escapements closer to the optimal escapement each year. However, errors in implementing desired management actions can significantly reduce benefits from all parameter estimation techniques.