Evolution of Urban Spatial Structure and Its Impact on the Thermal Environment Based on Local Climate Zones (LCZ): A Case Study of Qingdao
Yanjun Wang, Jiaxin Li, Kaipeng Huo, Jinqiao Ren, Yi Bian, Yan Gao, Mingshuo PanUrban spatial structure represents a fundamental determinant of urban thermal dynamics; however, the nonlinear responses of different Local Climate Zones (LCZs) and their spatial extents to thermal environments remain insufficiently understood. This study focuses on the central urban area of Qingdao, China, and establishes an integrated analytical framework for urban thermal environments by coupling remote sensing-based classification, WRF–SLUCM numerical simulation, and interpretable machine learning approaches. Based on 10 m Sentinel-2A imagery acquired in 2016, 2020, and 2025, a 17-class LCZ classification scheme was developed using the Google Earth Engine (GEE) platform and an improved Random Forest algorithm. The derived LCZ maps were subsequently incorporated into the WRF 4.7–SLUCM model with a four-level nested configuration to simulate 24 h near-surface meteorological conditions at a finest spatial resolution of 0.333 km, from which the Wet-Bulb Globe Temperature (WBGT) was calculated. Furthermore, the LightGBM model combined with SHAP interpretation was employed to quantitatively examine the nonlinear statistical response between LCZ area and WBGT and to identify the associated critical scales. The overall classification accuracy of the LCZ maps for the three study years ranged from 74.4% to 75.0%, with Kappa coefficients ranging from 0.7333 to 0.7446. The SHAP analysis identified four response patterns: negative-to-positive, positive-to-negative, consistently negative, and consistently positive. Different LCZ types exhibited distinct area thresholds, and some types showed a reversal in their thermal-environment response after exceeding a critical scale. These results indicate that the relationship between LCZ and WBGT is not simply linear but represents a scale-dependent nonlinear statistical association jointly shaped by land-cover type and patch size. This study provides methodological support for the quantitative and fine-scale analysis of urban thermal environments in coastal hilly cities.