XGBoost-Based Analysis of the Relationship Between Urban 2-D/3-D Morphology and Seasonal Gradient Land Surface Temperature

Abstract

The escalation of greenhouse gas emissions has led to a continuous rise in land surface temperature (LST). Studies have highlighted the substantial influence of urban morphology on LST; however, the impact of different dimensional indicators and their gradient effects remain unexplored. Selecting the urban area of Shenyang as a case, we chose various indicators representing different dimensions. By employing XGBoost for regression analysis, we aimed to explore the effects of urban 2D and 3D morphology on seasonal LST and its gradient effect. The following results were obtained: (1) The spatial pattern of LST in spring and winter in Shenyang was higher in the suburbs than in the center. (2) The correlation patterns of the indicators in spring and winter were similar, except for the proportion of woodland and grass (PWG), digital elevation model (DEM), and sky view factor (SVF), which exhibited opposing trends in summer and autumn. (3) Vegetation and construction had the highest influence on LST in the 2D index, followed by building forms and natural landscapes in the 3D urban morphology. (4) The influence of each indicator varied significantly across different gradients. Among all the indicators, the landscape index, social development, building forms, and skyscape had the highest impacts on urban areas. Vegetation and built-up areas had a greater influence on suburban areas. The findings of this study can assist in adjusting urban morphology and provide valuable recommendations for targeted improvements in thermal environments, thereby contributing to urban sustainable development.