Synthesizing flood exposure maps for comprehensive risk assessment

Flooding poses a major risk to rapidly growing cities like Addis Ababa, Ethiopia, where urban expansion, increasing impervious surfaces, and climate change continually intensify flood hazards. This study evaluates urban flood exposure by systematically comparing flood maps derived from fundamentally distinct methodologies. The primary objectives are to (i) delineate empirical flood exposure by integrating a multi-temporal Sentinel-1 SAR inundation map (2017–2023) with datasets on population, buildings, and road networks; and (ii) assess spatial agreement among exposure maps produced via hydrodynamic modeling, machine learning-based susceptibility analysis, and SAR mapping. The SAR-derived map was validated using a synthesized dataset of institutional records and citizen-science observations. Results reveal that 57 km² of buildings and 18 km of road networks are highly exposed to recurrent flooding. Furthermore, a critical spatial comparison using Jaccard (0.014–0.072) and Dice (0.027–0.143) coefficients, Moran’s I, and topographic correlation analysis revealed profound disagreement in flood risk extent and spatial distribution. These extreme discrepancies mathematically confirm the models capture distinct physical mechanisms rather than acting as redundant proxies. Consequently, relying on a single mapping framework inevitably biases urban flood risk estimation. This study provides baseline evidence advocating for a multi-criteria approach to support robust urban disaster risk management.