Spatial-temporal evolution of land surface temperature and influencing factors: Case of the Pearl River Delta urban agglomeration in China

This is the authors' abstract. We don't add key points for this paper.

Global warming and rapid urbanization aggravate the deterioration of urban thermal environments, especially in urban agglomerations. Assessing the spatial-temporal trends in land surface temperature (LST) within urban agglomerations is essential for optimizing and mitigating warming. This study examines spatial-temporal trends in LST in the Pearl River Delta (PRD) urban agglomeration from 2002 to 2022 and identifies their dominant drivers. By integrating machine learning techniques—specifically SHAP and XGBoost—this research not only improves the accuracy of LST prediction but also quantifies the contribution and relative importance of each influencing factor, thereby addressing existing gaps in future LST projections and in understanding the dominant control factors. Key findings reveal: (1) The 20-year average LST in the PRD shows an increasing trend, with a higher rate at nighttime than during the daytime. (2) Spatially, 33.66% and 43.49% of the area showed increasing LST trends for daytime and nighttime, respectively. Predictive models indicate that 0.82% of the area (daytime) and 0.23% of the area (nighttime) are projected to experience significant increases. (3) Land cover, climatic, socio-economic, and topographic factors significantly correlate with LST, and the dominant factors have heterogeneity in different cities. Collectively, this study supports the development of an urban thermal environment evolution model for the PRD. The multi-source (regional and inter-city) analysis offers actionable insights to inform climate-adaptive urban planning and long-term thermal environment management.