Flood Susceptibility Mapping Using Machine Learning: Spatial Risk Assessment and Implications for the Green Riyadh Project

Flash flooding presents a critical challenge for sustainable development in arid cities such as Riyadh, where rapid urbanization and complex geomorphology exacerbate hydrological risks. This study integrates high-resolution remote sensing, geospatial analysis, and machine learning (Light Gradient Boosting Machine, LGBM) to produce a 30-meter flood susceptibility map for Riyadh, complemented by spatial analysis of vegetation distribution using NDVI clustering. The results indicate that the overall flood risk in the studied sub-catchment is low, but highly concentrated in low-lying drainage corridors, especially along the Wadi Hanifa watershed. Notably, spatial overlap between green infrastructure clusters and flood-prone zones is observed, suggesting that green spaces are often situated in hydrologically sensitive areas. A detailed assessment of three major urban parks under construction as part of the Green Riyadh Project—King Salman Park, Wadi Hanifa Park, and Al-Uruba Park—reveals substantial variability in flood risk, primarily governed by local topography and hydrological connectivity. These findings highlight the necessity of integrating targeted green infrastructure planning with flood risk management and underscore the importance of adaptive, data-driven approaches for future urban resilience in arid environments.