Impacts of built environment on commuting mode choice considering spatial autocorrelation

Considering the similarity in residents ′commuting behavior living in nearby zones，several multilevel Bayesian models are employed to examine the impacts of built environment at traffic analysis zone（TAZ）levels on commuting mode choice. Bayesian models can capture the spatial autocorrelation of commuting behavior by incorporating adjacency matrixes to represent the spatial relationships between TAZs. The results show that the spatial autocorrelation significantly exists in residents′ commuting behavior. In addition，the model incorporating a centroid distance adjacent matrix has the best performance among the compared models. After controlling for socioeconomic characteristics at the individual levels，built environment characteristics are important factors of car commuting. Specifically，land use mix，public transit station density and intersection density have negative impacts on commuting by car. These results suggest that increasing the number of public transit stations，promoting more balanced land use and optimizing road network designs are important for car commuting reductions. The findings suggest that the optimization of built environment is important for encouraging low-carbon travel. © 2023 Editorial Board of Jilin University. All rights reserved.