Prioritizing environmental determinants of urban heat islands: A machine learning study for major cities in China

The exacerbated thermal environment in cities, with the urban heat island (UHI) effect as a prominent example, has been the source of many adverse urban environmental issues, including the increase of health risks, degradation of air quality and ecosystem services, and reduced resiliency of engineering infrastructure. Last decades have witnessed tremendous efforts and resources being invested to find sustainable solutions for urban heat mitigation, whereas the relative contributions of different UHI attributes and their patterns of spatio-temporal variability remain obscure. In this study, we employed the random forest (RF) method to quantify the relative importance of four categories of urban surface characteristics that regulate the surface UHI, namely the urban greenery fraction, land surface albedo, urban morphology, and level of human activities. We selected seventeen major cities from six megaregions in China as our study areas, with the RF training and test sets obtained from multi-sourced remote sensing and observational data products. It is found that the urban greenery coverage manifests as the most important environmental determinants of UHI, followed by surface albedo. The results are informative for urban planners, policymakers, and engineering practitioners to design and implement sustainable strategies for urban heat mitigation.