RESEARCH ON HIGH ACCURACY DETECTION OF RED TIDE HYPERSPECRRAL BASED ON DEEP LEARNING CNN

Mapping the 3D distribution of oceanic nitrate is challenging. We developed a continual learning-based multilayer perceptron, integrating prior knowledge from numerical models and Biogeochemical Argo validation to reconstruct the pan-European 3D nitrate field from 2010 to 2023 (0–2000 m depth, monthly, 0.25° horizontal resolution) using sea surface environmental features. This dataset helps bridge observational gaps and enhances understanding of the ocean's interior environment.