Estimating salmon stock-recruitment relationships from catch and escapement data

This paper presents an analysis of stock-recruitment data that takes account of natural variation in stock productivity (process error) and inaccurate escapement counts (measurement error). We formulate the model using dynamic state variables and take advantage of related techniques for parameter estimation, such as an extended Kalman filter. Our recruitment function depends explicitly on parameters relevant to management and includes various cases of historical interest. We adopt Bayesian methods for assessing uncertainty and use Markov chain Monte Carlo (MCMC) techniques to obtain posterior samples. A worked example, based on simulated data, illustrates geometric relationships among model choices, estimated recruitment curves, and data interpretations.