Inferred vertical coupling of urban heat islands across morphology zones: conditional effects and planning implications

Understanding the three-dimensional structure of the urban heat island (UHI) is essential for climate-adaptive planning. Beyond surface and canopy-layer warming, UHIs involve vertically integrated heat accumulation; however, most studies assess temperature differences at a single height. This study infers vertical UHI coupling from the joint yet differentiated responses of surface (SUHII), canopy-layer (CUHII), and vertically integrated boundary-layer (VUHII) indicators, without directly quantifying coupling or simulating energy exchange. It further examines morphology-dependent relationships using morphology-zone classification and double machine learning (DML)-based statistical attribution. The results show that SUHII and CUHII exhibit continuous urban-rural gradients, whereas VUHII is more localized and fragmented. At night, mean VUHII reaches 963.5 K·m in dense central districts, compared with 206.6 K·m in peripheral plains, indicating that conventional surface- or canopy-layer metrics alone may not fully represent vertically integrated heat accumulation within the lower atmosphere. Thermal intensity generally follows high-rise compact (HC) > low-rise compact (LC) > low-rise sparse (LS) > low-rise open (LO). Nighttime VUHII trends are 59.1, 18.8, and 13.1 K·m·a⁻¹ in HC, LC, and LS zones, respectively, and winter nighttime VUHII exceeds 2500 K·m in HC zones. Transition-zone analysis indicates that VUHII changes lag behind those of SUHII and CUHII during urban-form evolution: during LS to HC transitions, VUHII remains 200–400 K·m, far below the 1500–2000 K·m observed in stable HC zones. After adjustment for observed covariates, DML-based statistical attribution indicates that land cover and surface properties have the strongest estimated conditional associations with SUHII and CUHII. In contrast, VUHII is more strongly associated with three-dimensional urban structure, atmospheric conditions, and human activity. At night, building-height heterogeneity, impervious-surface connectivity, and population density show positive conditional effects on VUHII, whereas vegetation shows a negative conditional effect. These findings provide evidence-informed planning priorities for morphology-aware mitigation of heat accumulation.