Abstract
Accounting for patterns of movement and mixing in stock assessment of highly migratory species is important for managing these fisheries sustainably. Despite the growing wealth of information characterizing mixing between the western and eastern populations of Atlantic bluefin tuna (Thunnus thynnus), the stock assessment methods on which their management is based do not explicitly account for this critical aspect of stock structure. Simulation was used to test the performance of a virtual population analysis estimation model for estimating Atlantic bluefin tuna population abundance using pseudodata from an operating model incorporating movement and mixing. Model misspecification caused estimation models to frequently produce biased estimates of recruitment and spawning stock biomass. Western recruitment was significantly overestimated (~200% positive bias) but eastern recruitment was underestimated (~30% negative bias). Similarly, spawning stock biomass was underestimated for the eastern population (~70% negative bias) but overestimated for the western population (~100% positive bias). These biases appear to result from the model's inability to capture a net subsidy of the eastern population into western stock areas and fisheries. Estimation models applied to alternative operating model scenarios that modeled different potential recruitment trajectories and maturity assumptions behaved similarly. Models were better able to predict the size of mixed-population stocks than populations, suggesting that model predictions may be more effective for informing short-term trends in the resources available to fisheries than for implementing management decisions required for conservation of populations. The results suggest that stock mixing should be more explicitly considered in stock assessment of Atlantic bluefin tuna, and underscore the importance of testing stock assessment models and communicating their biases and limitations to managers.