Abstract:
The accuracy of flood hazard mapping in dam-break modeling depends strongly on the quality of the Digital Elevation Model (DEM). This study evaluates six open-access global DEMs (SRTM, ALOS PALSAR, AW3D30, COPERNICUS, FABDEM, and ANADEM) in dam-break simulations for two Brazilian dams (Negreiros and Copiti), using HEC-RAS 2D and high-resolution LiDAR as reference. Four aspects were analyzed: (1) elevation accuracy; (2) flood wave propagation, including peak discharge and time-to-peak; (3) inundation extent, with over- and underestimation indices; and (4) hydrodynamic risk. Results show that AW3D30 consistently produced flood propagation, extent, and risk estimates that are closest to those from LiDAR. In contrast, SRTM and PALSAR systematically underestimated inundation areas, whereas FABDEM and ANADEM tended to overestimate, resulting in conservative risk scenarios. COPERNICUS and FABDEM achieved the highest altimetric accuracy but only moderate hydraulic performance. Across case studies, dam size, geomorphology, and distance downstream significantly affected DEM reliability, with mismatches becoming more pronounced the further downstream. The findings highlight the influence of DEM choice on flood modeling and guide dam-break risk assessment in semi-arid and data-scarce regions, where high-resolution DEMs are often unavailable.
