Integrating diverse marine predator data for robust species distribution models in a dynamic ocean ICES Journal of Marine Science

Species distribution models (SDMs) are an important tool for marine conservation and management, yet guidance on leveraging diverse data to build robust models is limited. We evaluated whether an integrated SDM (iSDM) framework outperforms traditional data pooling or ensemble approaches when synthesizing multiple data types. We trained traditional SDMs and iSDMs using three data types for the blue shark (Prionace glauca) in the North Atlantic: fishery-dependent marker tags, observer records, and fishery-independent electronic tags. We compared pooled and ensembled SDMs, built with boosted regression trees, to an iSDM explicitly designed to address data-specific biases while leveraging each dataset’s strengths. While all approaches produced robust models, performance varied among data types, with fishery-dependent data consistently yielding more accurate than fishery-independent data. Differences in performance stemmed from models’ abilities to capture spatiotemporal dynamics in training data. iSDMs accounting for seasonal variability yielded the most accurate estimates but were computationally intensive, emphasizing the need to align model purpose with integration methods. Our findings reveal key trade-offs in data integration methods, particularly in balancing predictive accuracy and feasibility. As diverse data sources grow, leveraging robust approaches will be vital for improving conservation and management strategies and understanding dynamic species distributions in a changing ocean. Grant no. NA21OAR4170247