关键词: 城市形态, 地表温度, 尺度效应, 季节变化, 随机森林模型

Abstract: With the intensified urbanization, urban morphology has a significant impact on land surface temperature (LST). Although the influence of urban morphology on the thermal environment has been widely recognized, further investigation is required in terms of multi-scale grid-based analyses and seasonal comparisons. Taking the area within Beijing's Fourth Ring Road as the study area, we employed a random forest model to quantify the inpact of urban morphology factors on LST. We investigated the seasonal effects of different urban morphology factors on LST across multiple grid scales, and examined the spatial correlation between urban morphology factors and the spatial distribution of LST using bivariate spatial autocorrelation analysis. The results showed that within the grid scale range of 180-360 m, the goodness of fit between urban morphology factors and LST generally exhibited an increasing-decreasing trend with increasing grid size. Among these, the optimal model performance was achieved at the 300 m grid scale, suggesting that this scale could be applied in urban planning to mitigate the urban heat island effect. The effects of urban morphology on LST exhibited seasonal differences. The influence of building height was the strongest in spring, with a relative importance value of 3.36. The effect of the density of built-up land peaked in summer and autumn, with relative importance values of 4.21 and 4.39, respectively. The influence of urban vegetation cover was more pronounced in winter, with a relative importance value of 2.15. Among the urban morphology factors, normalized difference built-up index was positively correlated with LST, while normalized difference vegetation index, building height, building volume, and sky view factor were negatively correlated with LST. All urban morphology factors and LST exhibited a certain degree of local spatial correlation. By taking urban morphology as the analytical entry point, this study would advance the understanding of the multi-scale and seasonal variation patterns of LST, reveal its spatial heterogeneity, and provide scientific support for climate-adaptive urban planning and the development of differentiated thermal environment regulation strategies.

Key words: urban morphology, land surface temperature, scale effect, seasonal variation, random forest model