Advanced Deep Learning Approach for Accurate Upwelling Detection Along Morocco’s Atlantic Coast Using SST Imagery

International audience The study of coastal upwelling through the analysisof sea surface temperature (SST) satellite imagery has beena valuable approach because of its efficiency and practicality.Building on inception and residual structures, we introduceIncepResup-Net, a novel deep learning model for identifyingupwelling regions along Morocco’s Atlantic coast. This modeleffectively addresses limitations in recent methods targeting thesame upwelling system and outperforms them by more accuratelydetecting true upwelling areas, thereby minimizing false positives.Applied to SST data spanning from 2000 to 2022, IncepResup-Net demonstrates superior performance over traditional andcontemporary deep learning models, marked by its precisesegmentation capabilities and robustness in real-world detectionscenarios. Our findings highlight the model’s effectiveness inleveraging SST imagery for upwelling detection, establishinga new benchmark in the application of deep learning withingeoscience and remote sensing fields.