Revealing the impact of urban spatial morphology on land surface temperature in plain and plateau cities using explainable machine learning

Abstract

Rapid urbanization has intensified urban heat island (UHI) effects, highlighting the need to understand UHI drivers to improve local thermal environments. While previous research has shown Urban spatial morphology significantly influences land surface temperature (LST), the mechanisms and characteristics of this impact across different geographic conditions remain unclear. Based on this, we selected the main urban areas of Chengdu and Lhasa as examples, using machine learning models and Shapley additive explanation (SHAP) method to reveal the linear and nonlinear relationships between Urban spatial morphology and LST from a morphological perspective. The results show that: (1) The built environment has the most significant impact on LST in plain cities, while the morphology of green space more strongly regulates LST in plateau cities. (2) Building height and density of core both reflect a role in reducing LST in plateau cities. (3) The interaction mechanisms of building density and building height features show the same trend in both plain and plateau cities. However, density of branch between 0.1 and 0.2 reduces LST in plain cities, while densities below 0.1 are more effective in reducing LST in plateau cities. Our results can provide refined and differentiated references for urban planners dedicated to mitigating UHI.