The rapid pace of urbanization has intensified the urban heat environment, posing significant challenges to sustainable urban development. This study takes the Yangtze River Delta (YRD) urban agglomeration as its research area and utilizes MODIS summer land surface temperature (LST) remote sensing data with a spatial resolution of 1 km from 2000 to 2022. It proposes a multi-modal AI-driven integrated framework that combines Getis-Ord G spatial clustering analysis, Isolation Forest anomaly detection, KMeans-DBSCAN hybrid clustering, and the Percentile threshold method through weighted fusion, and evaluates the integrated results using hotspot coverage, temperature contrast, and spatial consistency. The study found that the area of heat spots in the YRD region increased significantly from less than 1% in 2000 to approximately 12% in 2022, with a phased surge after 2010, forming a continuous heat island belt spanning core cities such as Shanghai, Hangzhou, and Suzhou. The integrated model achieved over a 20% improvement in hotspot identification performance compared to single methods. Based on XGBoost-SHAP analysis results, a 1% increase in per capita GDP and impervious surface area (ISA) expansion reduces heat island area by 1.02% and 0.87%, respectively. Conversely, a 1 unit increase in annual average LST and nighttime light index increases heat island area by 0.94% and 0.88%, respectively. This study emphasizes the mitigation of regional urban heat island effects through differentiated spatial strategies and green infrastructure development, which may provide important scientific basis for urban climate adaptation planning and sustainable development.
Guoyin Liu (Thu,) studied this question.