The complexities of stock assignment using otolith stable isotope chemistry in Atlantic Bluefin Tuna in the Gulf of Maine, a stock-mixing hot spot

Objective The Atlantic Bluefin Tuna Thunnus thynnus is a highly migratory, internationally managed species whose stock structure and dynamic life history are challenging to capture in scientific advice for management. The large measurement and process uncertainty generated by Atlantic Bluefin Tuna stock identification using stable isotope chemistry, the central method of individual stock identification for the stock assessment, suggests that a more thorough examination of the methodology is needed. The Gulf of Maine, United States, is a key commercial fishing area where substantial mixing occurs between two Atlantic Bluefin Tuna stocks. The goal of this study was to determine the influence of alternative stable isotope baselines (adults and yearlings) and different random forest input parameters and decision thresholds on the stock assignment of Atlantic Bluefin Tuna caught in the Gulf of Maine. Methods Using different stable isotope baselines, random forest modeling was used to explore the impact of sample baseline assumptions, training–test set proportions, and probability thresholds on stock identification in the Gulf of Maine. Baseline-to-baseline classification was conducted to examine the alignment between presupposed and predicted identities. Results Baseline-to-baseline classification revealed that when samples from one baseline were treated as unknown and reclassified using a random forest model trained on another baseline, the assigned stock identities differed from their original classifications. Additionally, baseline choice and model specifications altered classification performance metrics, thus affecting the perceived mixing proportions and age structure in the Gulf of Maine. Conclusions Critical examination of baselines and establishment of robust approaches for addressing new baselines are necessary because their inclusion has the potential to influence stock perceptions in areas of significant fishing pressure and stock mixing.