Nonlinear associations between design, land-use features, and active travel

The nonlinearity correlations of design and land-use features with active travel (AT) are yet to be systematically identified by recent studies. This study aims to apply the extreme gradient boosting approach to examine the nonlinear relationships between design and land-use built environment (BE) features and AT. Based on a comprehensive travel survey in Wuhan's main agglomeration with various land use and urban functions, we collected 2073 valid samples to obtain residents' daily travel patterns and socio-demographic attributes. We found that all BE features have nonlinear relationships with AT. The collective relative importance of design features has greater predictive power for AT than land-use features. The most effective range of high contributions of street connectivity and street greenery was 21-34/km(2) and 0.22-0.34, respectively. Planners should improve the street-level features and build dense, diverse land use and high street connectivity urban environments to promote AT activity.