Integrating sentinel-1 SAR and AHP for flood susceptibility and agricultural impact assessment in the Awash River Basin, Ethiopia

Abstract Ethiopia’s economy is highly dependent on agriculture, yet recurring natural hazards continue to expose the country to significant vulnerability. This challenge is particularly severe in the Awash River Basin, where seasonal rainfall frequently triggers widespread flooding, threatening rural livelihoods and food security. Accurate flood monitoring is therefore essential for effective flood risk managment, disaster preparedness and mitigation. Although remote sensing images are widely used for large-scale flood detection, persistent wet-season cloud cover limits optical imagery. Sentinel-1 Synthetic Aperture Radar (SAR), processed in Google Earth Engine (GEE), provides a suitable alternative. However, in the complex terrain of the Awash Basin, conventional threshold-based SAR flood mapping produces substantial errors because terrain effects can be misinterpreted as flood signals. For the 2025 flood event examined in this study, the application of Height Above Nearest Drainage (HAND), Morphology, and SRTM DEM-derived slope masks reduced flood overestimation by an estimated 21.6%, eliminating approximately 199.44 km² of potentially misclassified areas from the initial SAR-derived flood extent and resulting in an estimated inundation area of 923.36 km². SAR-based flood mapping also identified a raw flooded cropland area of 84.62 km², which translates to an error-adjusted cropland exposure of 65.69 ± 7.98 km² with 95% confidence interval (CI). The Analytical Hierarchy Process (AHP) model identifies that the high-risk class covers 51.86% of the basin, while the very high-risk class accounts for 0.61%. To improve flood preparedness, this study integrates diagnostic SAR-based flood observations with a prognostic AHP flood susceptibility model. Spatial analysis shows strong agreement between the two approaches, with 89.25% of the mapped flood extent occurring within areas classified as “High” and “Very High” susceptibility, whereas only 0.25% occurred in low-risk zones, indicating agreement for this event rather than predictive proof. This suggests that the AHP model captures the spatial distribution of flood susceptibility for the studied event. The integrated framework achieves an overall accuracy of 81.5% ± 3.8% (Cohen’s Kappa = 0.630 ± 0.076) with 95% CI. Overall, the findings demonstrate the potential of combining terrain-corrected SAR flood mapping with AHP-based susceptibility assessment for flood risk analysis and climate-resilient planning in data-scarce regions.