Abstract:

With an increasing proportion of middle-income groups in the population structure, quantitative analysis of the nonlinear impact of community built environment on vehicle miles traveled (VMT) of middle-income groups has become an important basis for finely guiding the community life circle to create a green travel built environment. Based on the 5D of the built environment, this paper selects six indicators such as population density, land use diversity and public transport service levels to describe the built environment of a community.

Based on the 15-minute community life circle and walking speed in Technical Guide for Community Life Circle Planning, the measurement range of the community life circle is formed. The built environment is measured by combining point of interest (POI) data, road networks and other geospatial data. Taking Baoding residents’ travel behavior survey data as the source of empirical research, this paper constructs a gradient boosting decision tree (GBDT) model that considers nonlinear effects. In terms of model fitting, the GBDT model improves the adjusted R2 76% higher than the ordinary least squares (OLS) model with linear assumptions does, and performs better than machine learning models such as Support Vector Machine (SVM) and Random Forest (RF) do, indicating that the GBDT model has better adaptability in this study. In terms of the contribution of built environment indicators, the built environment attributes have a greater effect on VMT than individual social and economic attributes do, with land use mixing degree (19.10%) and distance from the city center (17.23%) contributing the most to VMT of middle-income people, which reflects the important role of reasonable land use planning in regulating middle-income people’s car use. In addition, the lowest contributor is public transportation service level (5.72%). This may be due to the fact that the middle-income group has a high travel demand with high time efficiency and high quality requirements for travel modes, and based on the preliminary research, the existing public transportation service level in the study area is still low, resulting in an inability to effectively attract the middle-income group. In terms of the nonlinear relationships of built environment indicators, the built environment factors have nonlinear relationships with VMT. The overall relationship between population density and VMT is “V-shaped”, and the lowest value of the VMT curve corresponds to a population density of 18 000 people/km2. The overall effect of community land use mixing on VMT is negative, and the threshold effect point is 0.6. When the land use mixing exceeds 0.6, it can effectively limit VMT. The influence of road density and intersection density on VMT is negatively correlated, and their threshold points are 7.6 km/km2 and 22 /buffer respectively. Overall, the impact of public transportation service levels on VMT is positively correlated, with no significant threshold relationship. For the bivariate interaction non-linear relationship, the influence of land use mixing, bus stop density and distance from the city center on VMT has a certain interaction effect with the driving age.

In general, the effect of the built environment on VMT reduces as the driving age increases. In particular, the inhibitory effect of land use mixing on VMT is substantially weak for those who have been driving for more than 10 years. In terms of public transportation service levels, for those with more than 3 years’ of driving experience, the effect on controlling the intensity of car use among middle-income people is less pronounced. In terms of distance from the city center, for those with more than 7 years’ of driving experience, the effect of reducing the distance from the city center on suppressing small car use begins to diminish. Among the nonlinear effects of the built environment on VMT, it is worthwhile for planning managers to pay attention to the inflection point of nonlinearity, which contains the optimal value range of built environment indicators on the one hand and the effect range of built environment indicators on the other hand. Therefore, community spatial planners are advised to develop a refined design of community built environment in order to effectively suppress VMT.