Comparative analysis of urban heat Island dynamics in China's coastal megaregions: an evolution-mechanism-heterogeneity framework

Climate change and rapid urbanization are intensifying the urban heat island (UHI) effect, with serious implications for public health, energy systems, and ecological environments. However, comparative analyses across multiple national-level urban agglomerations remain limited, and existing studies often fail to integrate natural and anthropogenic factors with spatial heterogeneity in UHI mechanisms. This study employed a novel “Evolution-Mechanism-Heterogeneity” framework to systematically analyze UHI dynamics in China's three major coastal agglomerations: Beijing-Tianjin-Hebei (BTH), Yangtze River Delta (YRD), and Pearl River Delta (PRD). Remote sensing data (2013−2023), spatial statistical methods, machine learning algorithms (XGBoost, Random Forest), Shapley Additive Explanations (SHAP), GeoShapley decomposition, and Multi-Scale Geographically Weighted Regression (MGWR) were employed. Results reveal distinct UHI patterns: BTH exhibits a “gradient mosaic” shaped mainly by vegetation and elevation, with the combined contribution of the Normalized Difference Vegetation Index (NDVI) and Elevation (DEM) exceeding 45%; YRD displays “networked” characteristics associated with urban form, with Shape Index (SI) contributing about 25%; and PRD forms a “core agglomeration” linked to urbanization, where SI, the Normalized Difference Bare Soil Index (NDBSI), and Nighttime Light (NTL) together contribute about 50%. Nonlinear threshold effects and significant multi-factor interactions were identified across all regions. MGWR analysis demonstrated substantial spatial heterogeneity in factor influences, with natural factors localized in BTH but urbanization factors regionally consistent in YRD. This framework innovatively bridges macro-evolution to micro-heterogeneity, overcoming limitations of traditional single-scale analysis and providing a scalable paradigm for differentiated UHI mitigation strategies globally.