Abstract:

As a key indicator for measuring the urban thermal environment, the spatiotemporal heterogeneity of land surface temperature (LST) and its driving mechanisms have become major research directions. Traditional linear models exhibit limitations in analyzing the nonlinear dynamic characteristics of thermal environment systems, whereas the LightGBM model combined with the interpretable Shapley Additive exPlanations (SHAP) algorithm provides a novel approach for uncovering complex driving mechanisms. Addressing the shortcomings of existing research regarding the quantification of “source-sink” scale landscape effects and the analysis of seasonal dynamic mechanisms, this study innovatively constructs a two-dimensional “season-source-sink” analytical framework based on Local Climate Zones. Taking the main urban area of Jinan City as the study area and integrating multi-source remote sensing and geospatial data, this study investigated the coupled influence mechanisms of urban spatial form, natural environmental factors, and human activities on LST. It quantitatively analyzed the contributions of nine types of driving factors to LST within "source-sink" landscapes across the four seasons. The findings revealed that natural environmental factors play a dominant role in regulating the urban thermal environment, with the Digital Elevation Model, Normalized Difference Vegetation Index, and Modified Normalized Difference Water Index identified as key regulatory factors. Although the impact of urban spatial form on LST is less pronounced than that of natural environmental factors, elements such as the Floor Area Ratio, Sky View Factor, and Building Coverage Ratio exert significant effects. The influence of human activities on LST was relatively minor; however, point data and Road Density exhibited certain impacts at the local scale. These findings provide strategic recommendations for differentiated planning in different seasons and source-sink areas, offering a scientific basis for managing urban thermal environments in the future. It is essential to fully utilize natural environmental resources and rationally plan building layouts to optimize the urban thermal environment and enhance the urban ecological livability of cities.

Key words: urban heat island effect, LightGBM model, shapley additive exPlanations (SHAP), “source-sink” scale, season, local climate zone (LCZ)

摘要：

地表温度（LST）作为衡量城市热环境的关键指标，其时空分异特征与驱动机制已成为当前研究的前沿方向。传统线性模型在解析热环境系统的非线性动力学特征时存在局限性，而LightGBM模型结合Shapley加性解释（SHAP）的可解释性算法为揭示复杂驱动机制提供了新方法。该研究针对现有研究中“源－汇”尺度景观效应量化与季节动态机制解析的不足，创新性地构建了以局地气候区为依托的“季节－源汇”二维分析框架。以济南市主城区为研究区，融合多源遥感数据与地理空间数据，深入探究了城市空间形态、自然环境要素及人类活动对LST的耦合影响机制，量化分析了9类驱动因子在四季“源－汇”景观中对LST的贡献度，发现自然环境因素在城市热环境调控中占据主导地位，数字高程、归一化植被指数和改进归一化水体指数是关键调控因子。城市空间形态对LST的影响虽小于自然环境因素，但建筑容积率、天空开阔度和建筑覆盖率等因素仍具有显著作用。人类活动对LST的影响相对较小，但兴趣点数据和道路密度在局部区域仍存在一定的影响。这些发现为不同季节和源汇区域的差异化规划提供了战略性建议，为城市热环境管理提供了科学依据。应充分利用自然环境资源，合理规划建筑布局，以优化城市热环境，提升城市生态宜居性。

关键词: 城市热岛效应, LightGBM模型, Shapley加性解释（SHAP）, “源－汇”尺度, 季节, 局地气候区（LCZ）