Cluster-based multiscale attribution and spatial mechanism optimization of urban heat and cold islands in Beijing

Abstract

Amid accelerating urbanization and climate change, it has become increasingly urgent to understand the spatial mechanisms governing heat and cold island dynamics in megacities. This study establishes an integrated, cluster-based typological framework by combining multi-source data with interpretable machine learning, statistical modeling, and spatial network analysis, aiming to disentangle the drivers and spatial organization of urban thermal environments in Beijing. Our analysis demonstrates that: (1) Intense built-up density and insufficient ecological buffering are primary contributors to urban heat intensification. (2) Regulatory mechanisms differ between heat and cold islands. In heat island zones—especially where space is constrained and blue–green coverage is limited—vegetation functions as the dominant and resilient cooling agent, whereas in cold island formation, large, contiguous water bodies and vegetated buffers play a crucial role by facilitating ventilation and delivering broad, sustained cooling effects. (3) Urban thermal resilience or vulnerability is closely tied to both spatial heterogeneity and connectivity of thermal patterns. Well-connected cold island backbones enable broader and more sustainable citywide cooling, whereas fragmented or isolated patches offer limited mitigation. Conversely, “enclosure-core” spatial configurations—such as the Urban Commercial–Business Area surrounding the Metropolitan Core Area—exacerbate heat entrapment by limiting ventilation and intensifying internal thermal buildup. These findings advance understanding of how urban form, landscape structure, and functional zoning jointly influence heat risk, and provide an operational framework to inform adaptive, differentiated strategies for thermal mitigation and sustainable urban planning.