Heterogeneous dominant factors of heat resilience across diverse urban patterns

Under urban stock renewal, the diverse characteristics of cities pose challenges for building climate-resilient environments. This urgently requires systematic investigation to identify dominant factors affecting both high heat (HH) and heat resilience (HR) across different urban patterns. This study integrates 2D land cover data and 3D street-view derived features to classify urban forms in Zhanjiang, China, into distinct patterns. Predictive models for HH and HR were developed using the XGBoost algorithm and interpreted via SHAP. The results reveal that the dominant factors of HR differ across diverse patterns: impervious surfaces in historical and development areas, water bodies in waterfront areas, and building view factor in residential areas. We further examined the synergistic threshold effects on reducing HH and enhancing HR. Based on these HH-HR threshold effects and the distinct characteristics of various areas, we proposed tailored urban control strategies. The study provides empirical evidence for differentiated heat adaptive planning, advancing urban climate resilience research from generalized frameworks to spatially targeted strategies and presenting methodologies and strategies for developing nations to foster HR urban environments.