cUrban–rural flood risk inequality in Pakistan: Coupling social vulnerability and flood susceptibility

Study region Pakistan, South Asia Study focus Against the backdrop of global climate change, rapid urbanization, and persistent urban–rural development disparities, Pakistan has experienced a sustained increase in flood frequency and the magnitude of associated losses. However, the spatial distributions of the country’s Social Vulnerability Index (SVI) and Flood Susceptibility Index (FSI), along with their coupled interactions, remain insufficiently characterized, thereby constraining the design of targeted disaster prevention and mitigation strategies. To this end, we developed an urban–rural flood risk assessment framework (URSV-FS). First, we constructed the urban–rural SVI using principal component analysis (PCA) and estimated the FSI using an Extreme Gradient Boosting (XGBoost) model. We then integrated SVI and FSI to derive flood risk and examined the spatial coupling between the two indices. Finally, nighttime light (NTL) remote-sensing data capturing the 2022 extreme flood event were used to validate the risk results and assess urban–rural flood resilience. New hydrological insights for the region The results indicate that: (1) Rural SVI was significantly higher than urban SVI. The proportion of rural districts with very high SVI was 25.4%, and the spatial variability was closely related to socioeconomic development level. (2) The XGBoost model performed well, with an AUC of 0.948 and an accuracy of 0.877. Soil type (SHAP = 0.833), NDVI (SHAP = 0.662), and slope (SHAP = 0.634) were the most influential predictors of FSI. The model classified 24.4% of the study area as high FSI, primarily along river corridors on the Indus Plain and in highly urbanized areas. (3) Flood risk was widespread in rural areas: 13.9% of rural districts were classified as very high risk, and 12.09% of the districts simultaneously exhibited high SVI and high FSI; both proportions were significantly higher than in urban areas. (4) During the flood disaster period, the decline in NTL in rural areas was 2.4 times larger than that in urban areas. However, after the disaster, the recovery of NTL in urban was 2.7 times larger than that in rural, suggesting a weaker recovery capacity in rural areas.