Spatial delineation of the compound flood transition zone using deep learning

• Introduced a deep learning framework to quantify coastal vs. hydrologic flood drivers. • Applied SHAP to spatially delineate the compound flood transition zone in Galveston Bay, TX. • The resulting transition zone delineation is consistent with physical principles. • Framework reduces subjectivity in defining transition zones, aiding flood risk assessment. Coastal and hydrologic floods are distinct yet interconnected phenomena, driven by oceanic and terrestrial processes, respectively. Their interaction—known as compound flooding—occurs when storm surge, heavy precipitation, and river flow coincide, significantly amplifying flood impacts in coastal riverine regions. These interactions give rise to a transition zone, where coastal and hydrologic flood processes converge, resulting in complex, prolonged inundation that is challenging to predict using traditional hydrodynamic models. Accurately delineating this zone is essential for improving flood risk assessment and mitigation strategies. In this study, we employ deep learning to quantify the relative contributions of terrestrial hydrologic and coastal flood drivers, enabling spatial delineation of the transition zone within Galveston Bay in Texas. This data-driven approach addresses the limitations of conventional models and supports more effective flood-resilience planning for vulnerable coastal communities. Our results reveal spatial patterns of flood driver dominance, with storm tide influencing coastal zones and river flow playing a greater role inland. The use of SHapley Additive exPlanations (SHAP) enables the delineation of a transition zone where no single driver dominates, underscoring the importance of compound flood modeling in such areas. This framework offers a scalable and interpretable solution for identifying high-risk zones, enhancing the precision of flood risk assessments, and informing targeted mitigation efforts in coastal regions.