An Improved Bayesian State-Space Model for Dealing With Multiple Sources of Uncertainty in Stock-Recruitment Data

Obtaining reliable parameter estimates for stock-recruitment (SR) relationships is essential for fisheries management. To this end, we developed a state-space model (SSM) for Pacific salmon that accounts for measurement errors in spawner abundance and catch, variation in recruitment, and time-series bias. We also developed two estimation methods for the SSM: Bayesian estimation via Markov chain Monte Carlo and an extended Kalman filter. Using extensive simulations, we provided clearly evidence that state-space modeling is the preferred approach for modeling stock-recruitment data that usually contain multiple sources of uncertainty. The SSM is robust to the effects of errors in measuring spawner abundance and catch, stock productivity and harvest rates. It reliably estimated optimal spawning abundance and harvest rate over a wide range of conditions. In contrast, a simpler SSM that ignored error in measuring catch and a traditional SR model that only considered recruitment variation perform poorly in various conditions. Our findings have the potential to improve the management of Pacific salmon and to promote the adoption of state-space models in ecological applications.