When are estimates of spawning stock biomass for small pelagic fishes improved by taking spatial structure into account?

Consideration should be taken of including spatial structure in the models on which stock assessments are based. Abstract A simulation-estimation approach is used to evaluate the efficacy of stock assessment methods that incorporate various levels of spatial complexity. The evaluated methods estimate historical and future biomass for a situation that roughly mimics Pacific herring Clupea pallasii at Haida Gwaii, British Columbia, Canada. The baseline operating model theorizes ten areas arranged such that there is post-recruitment dispersal among all areas. Simulated data (catches, catch age-composition, estimates of spawning stock biomass and its associated age structure) generated for each area are analyzed using estimation methods that range in complexity from ignoring spatial structure to explicitly modelling ten areas. Estimation methods that matched the operating model in terms of spatial structure performed best for hindcast performance and short-term forecasting, i.e., adding spatial structure to assessments improved estimation performance. Even with similar time trajectories among sub-stocks, accounting for spatial structure when conducting the assessment leads to improved estimates of spawning stock biomass. In contrast, assuming spatial variation in productivity when conducting assessments did not appreciably improve estimation performance, even when productivity actually varied spatially. Estimates of forecast biomass and of spawning stock biomass relative to the unfished level were poorer than estimates of biomass for years with data, i.e., hindcasts. Overall, the results of this study further support efforts to base stock assessments for small pelagic fishes on spatially-structured population dynamics models when there is a reasonable likelihood of identifying the sub-stocks that should form the basis for the assessment