How do urban morphological blocks shape spatial patterns of land surface temperature over different seasons? A multifactorial driving analysis of Beijing, China

Globally, rapid urbanization accompanied by a great deal of anthropogenic heat emission, further led to severe urban heat island (UHI) effects, posing a significant risk to human and environmental health. Understanding the combined driving effects of intra-urban temperature variation is vital to UHI mitigation. Although previous studies have extensively revealed the essential contribution of two-dimensional (2D) and three-dimensional (3D) architectural morphology to the land surface temperature (LST), their combined effects and seasonality on shaping the LST is still inconclusive, especially at finer scales. Therefore, this study aimed at broadening insight for thermal environment optimization from a block perspective. For the first time, the urban morphological blocks (UMBs) were introduced as the basic units to characterize the spatial-temporal variability of LST under various architectural morphological conditions. Multifactorial LST drivers, comprising architectural morphology, land cover, land use, and functions, were rigorously examined by the geographical detector model. The results showed that the highest LST generally occurred in the UMBs with the lowest building height and highest building density. The contribution of building aggregation to hot spots was most significant in the low-rise UMBs, while the cooling effects of the shadows were more pronounced in middle-density and high-density UMBs. In terms of different seasons, architectural morphological characteristics significantly contributed to LST variations under cold temperature, and the factors characterizing impervious surface and greening were always the two most powerful drivers in warm seasons. More factors contributed to the spatial patterns of LST variation in winter, suggesting a more complex LST driving mechanism under lower temperatures. Hierarchically, architectural morphologies, surface biophysical parameters, and urban land cover largely affected LST; however, the functional property factors were comparatively weak driving forces. The interaction detector model identified bilinear-enhanced and nonlinear-enhanced interactions from pairs of factors affecting LST over different seasons.