Which matters to commuting-related CO2 emissions? Parking, population suburbanization, or employment decentralization?

Understanding the influences of built environment characteristics at residence and workplace on commuting-related CO2 emission (CCE) is important for achieving net zero carbon emissions from the urban planning perspective. Using 2019 travel survey data and parking survey data in Xi'an, China, this study applies gradient boosting decision trees (GBDT) to reveal the associations between the built environment and CCE. The results show that distance from workplace to CBD is the most important predictor, contributing 54% of the predictive power, and distance from home to CBD contributes 15%. The findings highlight the detrimental effect of population and job decentralization on CCE. Parking spaces at workplace and residence are more important to CCE than parking costs at the respective locations. However, their contributions to predicting CCE are limited once other built environment characteristics are controlled for. Therefore, omitting parking measures may not materially bias the relationships between the characteristics and CCE.