Explainable AI Analysis of Future Domestic Water Shortages in Local Water Supply Systems: A Case Study of the Seomjin River Basin

This study applied an integrated K-WEAP and XGBoost-SHAP framework to identify the characteristics of domestic water shortages and their key explanatory factors under climate change conditions in the Seomjin River water system. Three representative drought scenarios with high drought vulnerability were selected from CMIP6 climate projections, and domestic water shortage rates were simulated for 11 administrative districts using K-WEAP. XGBoost surrogate models developed for four districts with sufficient shortage samples achieved coefficients of determination exceeding 0.91 on the test dataset. SHAP analysis identified regional and local water supplies, reservoir supply, and direct river intake as the primary explanatory factors, although their relative importance varied across districts. In general, decreases in major water supply variables were associated with increases in predicted shortage rates, suggesting that the dominant drivers of domestic water shortages differ depending on the composition of local water sources and the structure of the water supply system.