20-6 Towards Spatially Explicit Flood Risk Assessment: Integrating Machine Learning-Based Hazard Mapping with Multi-Dimensional Exposure Assessment in the Western Himalayas

Abstract:

Floods are among the most destructive hydrometeorological hazards, accounting for 35-40% of all weather-related disasters globally. Between 1990-2022, floods affected more than 3.2 billion people, 218,000 fatalities, and resulted in US$ 1.3 trillion losses globally. Despite these impacts, studies continue to emphasize susceptibility mapping, while the risk assessment of socio-economic assets remains comparatively limited, particularly in Himalayan basins. Therefore, this study uses a four-dimensional multivariate flood exposure framework to quantify the flood risk across 35 administrative blocks of the Beas River Sub-Basin (BRSB) in Western Himalaya. The hazard component developed through sequentially optimized XGBoost model, whereas the flood exposure index was developed using the CRiteria Importance Through Intercriteria Correlation (CRITIC) method. The robustness of the derived exposure index was further evaluated using the joint cumulative distribution function and Entropy Weight Method. Subsequently, the normalized flood hazard and exposure indices were integrated to develop a composite Flood Risk Index (FRI). The study shows that 2061 km² (16.5%) area of BRSB is highly prone to floods, and approx. 1.20 million people, 75.9 km² of built-up area, 553.5 km² of agricultural land, and 2,921 km of road network are exposed to flood within the BRSB. The FRI identified Panchrukhi, Nagrota Bagwan, Nurpur, Dharamshala, Mandi Sadar, and Kangra blocks is highly prone to flood and should be prioritized for flood mitigation, emergency preparedness, and resilient infrastructure planning. The proposed framework provides a systematic approach for integrating flood hazard and exposure information, thereby, supporting evidence-based flood risk management and spatial planning in mountainous river basins of the Himalayas.

© Copyright 2026 The Geological Society of America (GSA), all rights reserved.