Improving assessment of Pandalus stocks using a seasonal, size-structured assessment model with environmental variables. Part II: Model evaluation and simulation

Integrated, size-structured stock assessment models are now being used widely for assessment and management of hard-to-age species. However, few studies have attempted to evaluate their performance. A seasonal, size-structured assessment model with environmental covariates has been developed for hermaphroditic Pandalidae. We conducted simulations to evaluate its sensitivity to model configuration and performance with various misspecifications. Ignoring the seasonal fishing pattern (half-year closure) led to risk-prone assessment results of overestimating spawning stock biomass (SSB) and recruitment (R) and underestimating fishing mortality (F). Failure to incorporate environmental signals when the recruitment dynamics was environmentally driven led to bias in recent estimates of SSB, R, and F in the simulation. Ignoring annual variability in growth resulted in large estimation bias. Failing to account for time-varying natural mortality (M) led to strong biases; however, misspecifying size-specific M produced even stronger estimation bias. This result may depend on the variation of M among size classes. Under no model misspecifications, an unbiased estimate of M could be obtained by taking advantage of the seasonal fishery closure. Annual growth parameters were also estimable, but the large number of parameters with annual growth made it difficult for the model to converge.