Analysis of the spatiotemporal evolution of the surface thermal environment in high-density urban areas based on multi-source data, and its influencing factors: a case study of Haidian District, Beijing

Abstract

As global urbanization advances, surface thermal environment issues in high-density urban areas are intensifying, constraining sustainable development and human livability. Existing studies often focus on single indicators or localized areas, lacking a comprehensive analysis of the spatiotemporal evolution and multi-source driving factors. Taking Haidian District, Beijing, as a case study, this research retrieves land surface temperature (LST) using multi-temporal remote sensing data from 2015, 2020, and 2025. We employed spatial autocorrelation to reveal the distribution and trends of the urban heat island (UHI) effect. Furthermore, an XGBoost model was utilized to quantitatively analyze the spatial effects and interaction mechanisms of a comprehensive framework encompassing natural (e.g., green spaces, roads), social (e.g., building density), and economic (e.g., population density) dimensions. The findings indicate: (1) From 2015 to 2025, the LST in Haidian District rose by an average of 1.22°C annually, significantly exacerbating the UHI effect. This is particularly pronounced in high-density areas and transportation hubs, with temperature differences exceeding 6°C. (2) Night-time light (NL), building density (BD), and the normalized difference water index (NDWI) emerged as the dominant economic, social, and natural drivers, respectively. (3) The driving mechanisms exhibit strong scale dependency. At a 500m grid scale, population density (PD) and NL most significantly impact LST, whereas at a 1000m macro-scale, the dominant factors shift to NL and NDBI. This study deepens the understanding of dynamic thermal patterns, providing a scientific basis for spatial form regulation and refined planning in high-density cities.