Multi-Scenario Flood Susceptibility Projections in the Eastern Hindu Kush: Integrating Machine Learning, CMIP6, and Land Use Change

Floods are natural hazards characterized by the temporary overflow of water onto normally dry land, often causing significant damage to communities, infrastructure, and ecosystems. While most studies focus on historical or current flood risk, there is a critical gap in assessing future flood risk under changing climatic and land use conditions. This study addresses this gap by developing an integrated multi-scenario flood susceptibility framework for the eastern Hindu Kush region of Pakistan. The framework integrates both static and dynamic flood susceptibility conditioning factors to assess baseline and future flood susceptibility. Static topographic and hydrologic variables (e.g., elevation, slope, aspect, plan and profile curvature, geology, distance to roads, distance to rivers, topographic wetness index (TWI), stream power index (SPI), normalized difference vegetation index (NDVI), and historical rainfall are used to train the model for baseline conditions (1992–2020). To assess future flood scenarios, we incorporate rainfall projections from CMIP6 climate models under three Shared Socioeconomic Pathways (SSP1-2.6, SSP2-4.5, and SSP5-8.5) and simulate future land use patterns using the CA-Markov model for three time periods: 2040, 2070, and 2100. This allows us to evaluate how both climatic and anthropogenic changes influence future flood risk. Results indicate a significant increase in rainfall coupled with expansion of built-up and agricultural areas, leading to an escalation in flood risk across the eastern Hindu Kush region. In the baseline period, approximately 3% of the total area was classified as very highly prone to flooding. By the end of the century (2100), very high flood-prone areas are projected to increase by 14.91% (SSP1-2.6), 31.17% (SSP2-4.5), and 30.25% (SSP5-8.5) compared to baseline. Mid-century projections (2070s) show increases in very highly susceptible areas by 23.85% (SSP1-2.6), 17.34% (SSP2-4.5), and 12.19% (SSP5-8.5). In the near future (2040s), flood susceptibility remains largely similar to the baseline, although a slight decrease (1.63%) in high-risk zones is observed under SSP2-4.5. Among the conditioning factors, distance to rivers, rainfall, and slope emerged as the most influential predictors of flood susceptibility, based on the Mean Decrease Gini (MDG) index. The proposed framework offers practical insights for policymakers, urban planners, and disaster risk managers, enabling the formulation of forward-looking strategies to manage flood hazards and enhance climate resilience in vulnerable mountainous regions.