Identification and performance of environmentally-driven recruitment relationships in state space assessment models

Abstract

Environmentally-driven recruitment relationships are important for understanding fisheries responses to climate change; however they are difficult to estimate due in part to large variability in the recruitment process. State space models provide a promising way forward in allowing the characterization of multiple sources of stochastic variability. Here we conducted a large simulation-estimation study using environmentally-driven recruitment relationships to evaluate the effects of operating model characteristics on state space assessment model performance. We generally find low parameter and assessment bias across operating and estimating model combinations; however, some assessment and parameter bias is present under conditions of high recruitment variability and low spawning stock biomass contrast. Model identifiability for the correct functional form of the stock recruit relationship and the environmental relationship was generally poor, and projections were insensitive to assumed values of the environmental driver. We recommend the use of random effects on recruitment in state space assessments and caution against explicit stock-recruitment relationships. We encourage future work on environmental nonstationarity, which will be of increasing importance as exploited fish stocks experience accelerating rates of environmental change.