Investigating LST evolution and heatwave patterns using machine learning in the Beijing-Tianjin-Hebei Urban agglomeration

Rapid urbanization and persistent climate warming are jointly intensifying summer thermal risks in large urban agglomerations, with increasingly pronounced coupling between daytime heat exposure and insufficient nighttime cooling. Using MODIS land surface temperature (LST) data, heatwave metrics, Moran's I and Getis-Ord Gi∗ statistics, and XGBoost-SHAP interpretation, we examined daytime and nighttime thermal evolution across the Beijing-Tianjin-Hebei urban agglomeration (BTHUA) from 2000 to 2024. Daytime LST showed weak net warming with strong interannual fluctuations, whereas nighttime LST increased steadily by 2.3°C. Pixels exposed to compound day-night heatwaves expanded from 6.2% to 18.6%, with persistent hotspots in the southern urban belt and coldspots in northern mountains. SHAP results indicate that elevation exerted the strongest cooling association, while vegetation and albedo mainly shaped daytime LST and nighttime lights and built-up land shaped nighttime LST. These findings may support time-specific and regionally differentiated heat-risk governance.