Rapid urbanization and climate change intensified the spatiotemporal heterogeneity of urban thermal environments (UTEs). Conventional UTE studies primarily focus on physical and land cover indicators, neglecting the underlying socioeconomic fabrics that describe human–environment interactions and drive thermal heterogeneity. This study develops a CNN (Convolutional Neural Networks) intelligent framework for socioeconomic indicator simulation, using for accurate UTE modeling. Socioeconomic indicators are intelligently derived from land use attributes contained in publicly available urban planning maps (UPMs), overcoming the long-standing data scarcity in UTE studies. These indicators were represented as probabilistic maps, offering spatially explicit and interpretable socioeconomic representations. The framework was first applied in Zhengzhou, Henan Province, China, and was further validated in an adjacent city, Kaifeng. Results indicate that the method achieves 0.61 to 0.69 test values across different urban contexts and timings, demonstrating robust and stable generalization. Feature importance analyses further proved that these socioeconomic indicators contribute 48.3% to 54.6% to the UTE modeling across the four seasons, implying the high importance of the socioeconomic indicators on UTE modeling. Ultimately, block-level UTE responses were derived to identify high-risk areas, with quantitative land use optimization strategies established. The framework translates UTE risk diagnostics into concrete spatial governance strategies, shifting from qualitative guidance toward data-driven risk mitigation. It provides urban planners with a quantitative tool to assess block-level thermal responses to land use patterns and to optimize urban form for improved thermal comfort and climate resilience in metropolitan areas.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号