Effects of model misspecification and data availability on spatially-explicit assessment model performance Fisheries Research

Simulation studies are useful for determining the implications of available data and mis-specified model structure on the accuracy of model estimates of abundance and fishing mortality. Spatially-explicit multi-stock, age-structured stock assessment models have not been evaluated to understand how the accuracy of model estimates is affected by different assumptions of stock structure, age composition, and movement, compared to current methods used to inform management decisions. Our objective was to evaluate the accuracy of spatially-explicit multi-stock and spatially-implicit single stock assessment models for striped bass under alternative scenarios of data availability and quality. We conducted a simulation study to evaluate the accuracy of spatial assessment models under alternative scenarios of data quality and appropriateness of assumptions including investigating the effect of ageing error on model accuracy. A range of estimation models with alternative assumptions about spatial dynamics, stock composition, and ageing error were fitted to the datasets. Spatially-explicit estimates were approximately unbiased in estimating abundance and fishing mortality when they closely matched the assumptions of the data generating model and accounted for ageing error. Data to inform stock composition or informative priors on occupancy probabilities were necessary for the spatially-explicit models, but estimates were sensitive to correct specification of the occupancy probability priors. Models that ignored potential ageing error in datasets resulted in biased and inaccurate estimates of abundance and fishing mortality. All models provided inaccurate estimates of reference points, although estimates from spatially-explicit models were the least biased. Spatially-explicit stock assessments can potentially improve accuracy of estimates when they match spatial dynamics of fish populations and when ageing error was corrected. Grant no. NA21OAR4170244