Analysis of the impact of street physical morphology on thermal environment based on GBDT-SHAP machine learning model：A Case Study of the Five Central Districts of Chengdu

Under the influence of rapid urbanization and extreme climate, the urban heat island effect is prominent, becoming a common problem faced by cities. Although studies have focused on the relationship between urban physical morphology and thermal environment, there is still a lack of in-depth understanding of the dynamic mechanism of action between various factors of physical morphology and thermal environment on the minimum unit analysis scale, and the degree of contribution of different factors in different time periods is also controversial. Therefore, it is of great practical significance to systematically explore the dynamic influence mechanism of urban physical forms on the thermal environment. This paper takes the five central districts of Chengdu as the research area, combines the surface temperature inversion data in 2015, 2020 and 2025, and uses the individual unit surfaces divided by the transportation road network as the minimum analysis unit, and selects 7 influencing factor indicators from the dimension of urban physical morphology. Using multi-source remote sensing data and geospatial analysis technology, the dynamic influence mechanism of urban physical forms on the thermal environment is systematically explored. The results show that: (1) The average surface temperatures of each block in the three periods were 35.03℃, 41.59℃ and 33.80℃ respectively. The high-temperature zone shows a trend of expansion to the periphery, and will show the characteristics of center-high in the southeast and low in the northwest in 2025 (2) The proportion of impermeable surfaces (PIS) and water area (PWD) in the study area do not change significantly. The average building height (ABH) tends to be high in the center and increasing in the outer edges; the building density (BD) shows a distribution pattern of dense in the center and sparse in the outer edges; the sky visibility (SVF) shows a trend of low in the center and increasing in the edges. Road density (RD) is showing a continuous upward trend. (3) OLS reveals the overall overall correlation characteristics of the two, with the positive contribution of BD and the negative impact of ABH. GWR reveals the spatial heterogeneity of each factor. The mean of the regression coefficient of the GWR factor in each year is basically the same as that of the OLS, but the effects of each factor vary in different regions and times. GBDT captures nonlinear relationships. In 2020, urban physical morphology has a stronger impact on temperature, and the degree of impact of urban physical morphology dimensions is from high to low: building morphology, surface covering and roads. This study provides accurate data support and decision-making basis for optimizing the layout of urban physical morphology to alleviate the heat island effect, and has important practical value for promoting the sustainable development of cities.