本研究以福州市主城区为例，基于POI与路网数据识别功能区，结合ECOSTRESS夜间地表温度（LST）及多维城市形态指标，构建XGBoost-SHAP模型分析夜间LST空间分异与非线性驱动机制。结果表明：（1）夜间LST功能区异质性显著，居住用地是核心热区，而工矿用地表现为显著低温区。（2）社会经济因子（夜间灯光强度NTL、人口密度POP）与三维形态因子（平均建筑高度BH）是夜间升温 的主导因素，影响力高于二维形态因子。（3）关键因子对夜间LST的影响存在显著阈值：NTL在值达到约40后边际效应趋于饱和；POP和BH的关键阈值分别约为150人/公顷和20米。本研究为精细化热环境调控提供科学依据。

Integrating ECOSTRESS and Urban Functional Zones identified via POI and road network data, this study uses an XGBoost-SHAP framework to analyze spatial differentiation and non-linear drivers of nighttime Land Surface Temperature (LST) in Fuzhou’s main urban area. Results indicate: (1) Residential zones are heat sources whereas industrial zones are cold spots. (2) Socio-economic factors of Nighttime Light Intensity(NTL) and Population Density (POP), plus the 3D morphology factor Average Building Height (BH), dominate warming over 2D factors. (3) NTL saturates near 40; POP and BH thresholds are approximately 150 people/ha and 20 m. Findings support zone-specific regulation.