Search

This study adopts a gradient boosting decision tree (GBDT) model to contrast the relative importance of improving regional accessibility versus local land use for reducing driving distance, as well as their nonlinear and threshold effects. The nonlinear analysis allows us to scrutinize the effective impact scopes of urban and suburban centers, which can provide planning guidelines for polycentric development, while threshold effects can help planners determine normative ranges of land use settings for planning pedestrian-scale neighborhoods (e.g., the 15-minute life-circle plan in China). Based on a 2017 travel diary survey in Beijing, empirical findings reveal the synergistic effects of the proximities to urban and suburban centers on driving distance, demonstrating the importance of polycentric development for reduced driving. In addition, many local land use variables have distinctive threshold effects, including local access to daily facilities and density indicators. For driving reduction, effective densities of health, civil, and retail/service facilities are around 15, 10, and 200 establishments (points of interest) per km2, respectively. These findings demonstrate the importance of nonlinear and threshold analysis and suggest incorporating polycentric development and neighborhood life-circle planning would lower vehicle use in Beijing. • Adopting a GBDT model to analyze nonlinear and threshold effects of regional accessibility and local land use on driving. • Estimating the synergistic effective impact scopes of urban and suburban centers for planning polycentric development. • Using threshold effects to determine normative ranges of land use settings for planning neighborhood life circle. • Incorporating polycentric development and neighborhood life-circle planning to lower vehicle use in Beijing. [ABSTRACT FROM AUTHOR]