Identifying urban morphological archetypes using k-means to analyze spatial heterogeneity of the urban thermal environment

Complex urban spatial morphology continues to shape the urban thermal environment. Although prior studies have highlighted the influence of urban morphological parameters on thermal conditions, few have translated city-specific characteristics into morphology-based design recommendations for heat mitigation. This study introduces a data-driven method using the k-means algorithm to identify urban morphological archetypes based on spatial morphology parameters. By integrating Random Forest with interpretable machine learning, we examined how morphology affects land surface temperature (LST) within each archetype and derived differentiated implications for thermal-environment optimization. The results identified 11 morphological archetypes and revealed significant LST differences among them. Compared with OLS and XGBoost, Random Forest achieved better performance in most archetypes. NDVI showed the strongest overall association with LST, while interpretable analysis further revealed nonlinear responses and threshold effects of key morphology parameters. These findings provide a basis for archetype-specific and spatially adaptive design recommendations for urban heat mitigation.