Validation of flood inundation modelling using multi-source SAR imagery

The increasing frequency of extreme precipitation events due to climate change pose significant challenges to flood risk management. While various hydraulic models are widely used for flood monitoring, their validation in data-scarce regions remains a critical obstacle. Satellite remote sensing, particularly synthetic aperture radar (SAR) imagery, provides a viable solution by capturing flood dynamic observation independent of weather conditions. In this study, we validate two-dimension (2D) hydraulic simulations using flood extents derived from multi-source SAR data. The FloodMap-HydroInundation2D model was applied to simulate a flood event in Nanyang, China (2024). An automated threshold-based framework was developed to extract flood extents from both GF-3 and Sentinel-1 SAR imagery. The results indicate that the threshold-based segmentation method can extract water extents rapidly and effectively, with overall accuracy above 85% and Kappa coefficients exceeding 0.6. The model simulations showed strong agreement with the SAR-derived flood maps, reaching a hit rate of 88.8%, a low bias of 0.058, and a critical success index of 72.7%. A sensitivity analysis revealed that hydraulic conductivity had a greater influence on simulation results than Manning's roughness coefficient. Our study contributes a robust and scalable framework for flood model validation in data-scarce regions, thereby supporting flood disaster mitigation and emergency response.