Spatially heterogeneous thermal responses to urban form and greening: A multi-model analysis of seasonal and diurnal dynamics in beijing

Urban morphology and vegetation jointly regulate land surface temperature (LST) across seasonal and diurnal cycles, with implications for mitigating urban heat island (UHI) effects. Focusing on Beijing, we quantified the effects of two- and three-dimensional (2D/3D) urban form parameters and vegetation on LST using an integrated framework combining ordinary least squares (OLS), geographically weighted regression (GWR), extreme gradient boosting (XGBoost), and SHapley Additive exPlanations (SHAP). Results revealed pronounced nonlinear and spatially heterogeneous urban morphology-LST relationships and identified 250-300 m as the optimal explanatory scale. Compared with two-dimensional parameters, three-dimensional urban morphology exerted stronger control on LST, especially at night, whereas daytime cooling was dominated by the normalized difference vegetation index (NDVI). SHAP interaction analysis further showed that vegetation cooling and building-induced warming were jointly conditioned by urban form and ventilation across seasonal and diurnal contexts. These findings provide a structural basis for differentiated, scale-sensitive heat-mitigation strategies in high-density cities.