Multispecies Stock Assessment for Georges Bank: Model Development, Performance Testing, and Multimodel Inference

Accounting for species interactions in both stock assessment modeling and fisheries management is of increasing interest in the Northeast US. However, multispecies assessment models in development require testing to evaluate their capabilities. We review progress on the ongoing Georges Bank multispecies assessment project, which incorporates multispecies production models, multispecies delay difference models, and empirical nonlinear time series forecast models as assessment models within a multi-model inference framework. The ability of each model to estimate the current status and historical trend of 10 Georges Bank species was evaluated using simulated datasets derived from a multispecies size-structured operating model. The operating model incorporates multiple functional forms for growth and recruitment, three effort-driven fishing fleets with different selectivities, and environmental effects on growth. Hundreds of simulated survey biomass and catch time series were generated with stochastic recruitment and varying catchabilities and levels of observation error. The multispecies assessment models were then fit to this simulated data, and assessment model estimates of biomass and catch trends were compared with "true" operating model values for each time series.  Estimates from the multispecies assessment models were also examined for retrospective bias.  This process both improves the multispecies models and informs managers of their strengths and weaknesses.