Thermal environment disparities across 11 cities in the greater bay area: Nonlinear relationships and urban greening-based solutions

Urbanization intensifies urban thermal environment (UTE) degradation. While urban morphology affects UTE, its nonlinear, cross-city contributions under shared climatic conditions remain understudied. We analyze 11 cities in China's Greater Bay Area (GBA) using LightGBM and SHapley Additive exPlanations (SHAP) to reveal nonlinear relationships between urban form and daytime/nighttime summer land surface temperatures. Results show NDVI is the strongest cooling factor, with peak effectiveness at 0.6–0.8, beyond which benefits diminish. Water cools during the day but loses efficacy at night, except in Hong Kong and Guangzhou, revealing intra-regional heterogeneity. Landscape Shape Index (LSI) increases daytime warming but saturates at high values—Foshan and Dongguan shift to cooling at LSI > 1.2, while Macao shows amplified warming. Building heights >20 m exacerbate nighttime heat, especially at 20–40 m. High floor area ratio (FAR) generally worsens UTE, yet Hong Kong and Zhuhai exhibit anomalous daytime cooling. We propose a novel classification method based on the nonlinear response patterns of urban morphologies to UTE, enabling data-driven city clustering. Using this approach, cities are grouped into four UTE-behavior clusters: (1) balanced forms (Dongguan, Foshan, Zhongshan), (2) extensive green spaces (Guangzhou, Huizhou, Zhaoqing), (3) complex environments (Hong Kong, Macao, Zhuhai), and (4) diverse green spaces (Shenzhen, Jiangmen). This typology captures morphological-thermal synergies. By isolating morphological drivers in a climatically homogeneous region, our study reveals critical thresholds and divergent thermal responses, underscoring the need for city-specific mitigation. The framework advances nonlinear UTE analysis and supports climate-resilient urban design in rapidly urbanizing regions.