Quantifying the Effects of Urban Form on Land Surface Temperature in Subtropical High-Density Urban Areas Using Machine Learning.

It is widely acknowledged that urban form significantly affects urban thermal environment, which is a key element to adapt and mitigate extreme high temperature weather in high-density urban areas. However, few studies have discussed the impact of physical urban form features on the land surface temperature (LST) from a perspective of comprehensive urban spatial structures. This study used the ordinary least-squares regression (OLS) and random forest regression (RF) to distinguish the relative contributions of urban form metrics on LST at three observation scales. Results of this study indicate that more than 90% of the LST variations were explained by selected urban form metrics using RF. Effects of the magnitude and direction of urban form metrics on LST varied with the changes of seasons and observation scales. Overall, building morphology and urban ecological infrastructure had dominant effects on LST variations in high-density urban centers. Urban green space and water bodies demonstrated stronger cooling effects, especially in summer. Building density (BD) exhibited significant positive effects on LST, whereas the floor area ratio (FAR) showed a negative influence on LST. The results can be applied to investigate and implement urban thermal environment mitigation planning for city managers and planners.