A comparative assessment of flood mapping methods for urban risk management in data-poor environments

Flood mapping is essential to urban resilience, but it often relies on traditional hydrological models, which are poorly suited to data-scarce contexts due to their complexity and input requirements. This study, conducted in Antananarivo, Madagascar, evaluates alternative approaches combining remote sensing (Pleiades and Sentinel-1), simplified hydrological modeling (Fast Flood Simulation—FFS), and multicriteria analysis (MCA), with validation from field observations. The results show contrasting performances: FFS (10 cm threshold) detected 45% of flooded zones in the urban center, while Sentinel-1 identified only 3% due to poor performance in dense built-up areas. Pleiades was also affected by structural artifacts, despite its high resolution. MCA identified 40% of the city as being highly vulnerable. In agricultural zones, FFS captured over 80% of flooded areas, compared to 56% for MCA and less than 10% for Sentinel-1. In residential neighborhoods, flooded building detection ranged from under 1% (Sentinel-1) to 27% (FFS) and 37% (MCA). Overall, FFS emerged as the most suitable tool, providing flood depth estimates for operational decision-making, but its reliability depends heavily on input data quality and calibration processes. Satellite imagery offers complementary post-crisis validation but remains limited in dense urban environments. MCA supports strategic planning but cannot model event-specific dynamics. Combined within a GIS platform or interactive dashboard, these tools provide complementary capabilities across risk management phases—prevention (FFS, remote sensing, MCA), crisis response (FFS), and reconstruction (FFS, remote sensing)—offering decision-support solutions adapted to data-poor environments.