A 3D perspective for understanding the mechanisms of urban heat island and urban morphology using multi-modal geospatial data and interpretable machine learning

The urban heat island (UHI) effect influenced by 3D urban morphology exacerbates urban thermal environments and presents significant challenges to sustainable urban development. While previous studies have emphasized the impact of urban morphology indicators on UHI, fine-scale variations and the intricate relationships between these factors remain underexplored. This study employs LiDAR and geotagged data to obtain nine morphological indicators using the deep learning based semantic segmentation methods. An explainable machine learning framework, specifically an ensemble learning model based on Shapley Additive exPlanations (SHAP), is applied to assess the impact of these indicators and their complex interactions on the thermal environment. Using Austin, Texas as a case study, we present a 3D perspective on the morphology-UHI relationship. The results reveal that urban indicators have more significant impact on UHI, with the sky view factor and impervious surface ratio contributing the most. The influence of urban morphological features on UHI exhibits spatial heterogeneity and boundary effects. For example, building volume initially exacerbates UHI, but once it exceeds a certain threshold, it starts to mitigate the heat island effect. Additionally, the interaction between small buildings and dense road networks intensifies UHI, whereas high-rise buildings can alleviate the effects of extensive urbanization on UHI. These findings offer valuable insights into the driving mechanisms of 2D and 3D urban morphology on UHI and provide guidance for optimizing urban design to reduce the urban heat island effect.