LAUSR

ABSTRACT This study focused on urban high-temperature expansion (HTE) and its spatiotemporal patterns, risks, and underlying driving mechanisms, a critical yet underexplored aspect of the urban climate that differs from traditional urban heat alone. We factored vitality factors in 22 drivers derived from geospatial big data using 3S technology and then conducted a fine-grained interpretable model by integrating machine learning with the Shapley additive explanation (SHAP) method, offering novel insights into the exploration of HTE. This is the first in-depth exploration of urban HTE. The results showed that: (1) the severe urban heat island effect occurred in summer, and the HTE patterns were highly consistent with policy-led development trends. (2) XGB_LR integrating with the SHAP method was used to explore the underlying driving mechanisms of HTE. Among the top five contributors, the first two are morphological factors (NDVI and BuilD) that remain dominant as expected, followed by the three vitality factors (BusRouK, TaxE, and Uuse) instead of FAR and IS. (3) A fine-grained perspective of how the interactions between factors affect the HTE risk was presented. These findings and risk maps provide a significant reference for urban planning and high-temperature risk mitigation.