How variable is recruitment for exploited marine fishes? A hierarchical model for testing life history theory

Recruitment often varies substantially in fish populations, and residual variability may have serial autocorrelation due to environmental effects even after accounting for a stock-recruitment relationship. However, the likely magnitude of variability and autocorrelation in recruitment has yet to be formally estimated. We therefore developed a hierarchical model for recruitment variability and autocorrelation and applied it to data for 154 fish populations. Results were similar when using either the Ricker or Beverton-Holt stock-recruitment model, and showed that autocorrelated recruitment has a marginal standard deviation of 0.74 (SD = 0.35) and a mean autocorrelation of 0.43 (SD = 0.28) when predicting for an unobserved taxonomic order. Estimates differed somewhat among taxonomic orders and stocks, and also supported a hypothesized positive relationship between age at maturity and autocorrelation in recruitment. Our results can be used as a Bayesian prior for recruitment variability in models for data-poor stocks and to distinguish recruitment from other process errors in models for data-rich stocks. Estimates can also be used in the design of future simulation models and management strategy evaluations and in theoretical research regarding life history variation.