Seasonal effects of urban morphology on land surface temperature in a three-dimensional perspective: A case study in Hangzhou, China

Architecture influences the design of urban landscapes and hence plays an important role in the regulation of urban thermal environment. Studies on the correlation between multi-dimensional building morphology and land surface temperature (LST) have been conducted in many cities around the world. However, general conclusions on their relationship have not been drawn, owing to the differences in climatic, social, and economic conditions between cities. Moreover, the multivariate and nonlinearity of the three-dimensional (3D) indicators also require more robust and specific statistical models. In this study, the random forest (RF) regression model was used to quantitatively analyse the seasonal differences in the responses of nine multi-dimensional building indicators to LST. The results show that the building coverage ratio (BCR), high building ratio (HBR), and architecture height standard deviation (AHSD) always had the most substantial influence. Across seasons, the average relative contribution ratios of BCR, HBR, and AHSD were 13.23–33.50%, 11.37–20.06%, and 9.02–24.12%, respectively. In general, 2D indicators tended to have a higher impact in warmer seasons while 3D indicators showed more impact on LST in cooler seasons. According to the marginal effect analysis, the BCR, cubic index (CI), and aggregation index (AI) were the main heating indicators, whereas AHSD and HBR were the main cooling indicators. Overall, our findings contribute to the understanding of the urban heat island phenomenon, which remains an important issue in the context of sustainable landscape and urban planning.