The nonlinear effects of multi-scale built environments on CO2 emissions from commuting

With survey data from 18 neighborhoods in Guangzhou, China, this study employs a gradient boosting decision tree (GBDT) method to measure the contributions of the built environment (BE) at the place of residence and the workplace to the carbon dioxide (CO2) emissions from commuting at three spatial scales and explore the nonlinear associations between them. Results show that the systematic impact of built environments on CO2 emissions from commuting is about four times that of residents’ socio-demographics. The BE has a scale effect on CO2 emissions and each scale makes an unequable contribution. In general, the BE at the workplace plays a more important role than that at the place of residence. Some BE factors, like residential density, land-use mix, and bus stop density, have a complex and nonlinear effect on CO2 emissions, which needs to be fully and critically considered in land use and transport planning and community design.