Time Variation in Natural Mortality, Gear Selectivity, and Maturity in Sitka Sound Herring

Integrated age-structured assessment models often estimate parameters such as natural mortality, gear selectivity, and maturity as constant over time or age. If parameters are really time-dependent, then significant bias can occur in results from age-structured assessment models with constant parameters. The goal of this study is to determine whether these models can be improved by using time dependent parameters. An age-structured assessment model of Pacific herring (Clupea pallasi) in Sitka Sound, Alaska provided sufficient information to evaluate time-dependent parameterizations. The types of time variation included regime shifts, relationships with covariates related to weight-at-age, age composition, and climate, and random walks. Models that estimated natural mortality and maturity with respect to regime shifts resulted in the lowest Akaike information criterion values. Model results indicated a significant change in natural mortality for Sitka Sound herring after the 1998 regime shift and a decrease in maturity after the 1989 and 1998 regime shifts. Mortality estimates were comparable to estimates from a model with a random walk for natural mortality. Gear selectivity did not appear to change over time in any models.