An age-structured state-space stock–recruit model for Pacific salmon (Oncorhynchusspp.)

Abstract

We describe an age-structured state-space model for stock–recruit analysis of Pacific salmon data. The model allows for incorporation of process variation in stock productivity, recruitment, and maturation schedules, as well as observation error in run abundance, harvest, and age composition. Explicit consideration of age structure allows for realistic depiction of system dynamics and sample design, more complete use of recent data, and forecasts that consider sibling relationships. A Bayesian framework is adopted, implemented with Markov chain Monte Carlo methods, which provides an enhanced ability to incorporate auxiliary information, convenient and rigorous consideration of measurement error and missing data, and a more complete assessment of uncertainty. We fit the model to annual upstream weir counts, commercial and recreational harvest estimates, and age composition data from Chinook salmon (Oncorhynchus tshawytscha) in Karluk River, Alaska. For the case study, the model is configured with a Ricker stock–recruit relationship, autoregressive lag-1 productivity, and Dirichlet age-at-maturity. Details of alternate configurations are also described. We introduce the optimal yield probability profile as an objective tool for informing the selection of escapement goals based on yield considerations and describe alternative versions useful for addressing other management questions.