Integrated Flood Hazard and Economic Risk Assessment Using Satellite Data and Hybrid Modeling

Floods represent one of the most destructive natural hazards, inflicting substantial environmental degradation, economic losses, and human casualties worldwide. This study presents an integrated framework combining deep learning, remote sensing and hydraulic modelling to estimate flood depth, hazard and economic losses in the Dez River Basin (southwestern Iran). A U-Net convolutional network was trained on Sentinel-2 and Landsat 8/9 imagery to delineate inundation extents for six major flood events (2016–2023). Flood depths were estimated using the Floodwater Depth Estimation Tool (FwDET) applied to the U-Net inundation maps and 10-m TanDEM-X DEM. Flow velocities were obtained from HEC-RAS simulations driven by observed discharge records. Depth and velocity maps were combined with a debris-flow coefficient to compute pixel-wise Hazard Risk (HR) indices and classify areas into four risk levels. Economic damages were estimated by overlaying HR maps with land-use data and applying depth-dependent monetary coefficients. The U-Net model achieved mean IoU values of 70–71.3% and accuracy above 95%, while the integrated approach produced spatially consistent HR maps highlighting urban concentrations of high risk. Results show that urban losses were substantially greater than agricultural losses for the examined events. The workflow demonstrates an operationally feasible, data-efficient pathway to deliver flood hazard products and monetary loss estimates that can support emergency response and planning. The methodology is adaptable to other basins and can be extended using higher-resolution imagery or additional in-situ measurements to improve accuracy and operational readiness.