Tao, Tao2022-12-022022-12-022022-09https://hdl.handle.net/11299/250068University of Minnesota Ph.D. dissertation. 2022. Major: Public Affairs. Advisors: Jason Cao, Greg Lindsey. 1 computer file (PDF); 107 pages.Since the 1990s, a growing number of studies have examined the relationships between built environment attributes and travel behavior and offered supportive evidence for planning policies that discourage driving and/or encourage transit and active travel modes. Due to the lack of efficient methods, most studies assume that the relationships between built environment attributes and travel behavior are (generalized) linear. The rise of machine learning approaches allows scholars to relax the assumption. Built on the recent literature, my dissertation aims to further advance the field on the nonlinear and threshold relationships between built environment attributes and travel behavior through studying three distinct but interrelated research topics. Most studies focus on only one travel mode and fall short of comparing nonlinear and threshold relationships of the built environment with different travel modes. Identifying the common thresholds for different modes enables the optimization of built environment attributes. Using regional travel survey data in the Twin Cities, the US, the study in Chapter 2 applies gradient boosting decision trees (GBDT) to examine and compare the nonlinear associations between built environment attributes and travel distances by driving, transit, and active travel. It also compares the contributions of regional and local built environment attributes. The results show that there are prevalent nonlinear associations between built environment characteristics and travel distances, informing planners of the effective ranges, within which the characteristics influence travel distances efficiently. Moreover, regional characteristics collectively have a stronger influence on all three travel distances than local characteristics. This result suggests that planners should pay more attention to metropolitan-scale planning and deploy programs that enhance regional accessibility. Few studies emphasize the nonlinear relationship between the built environment and auto use in suburban areas. However, their association in suburban areas may differ from that in urban areas, implying context-specific planning policies. The study in Chapter 3 uses GBDT to explore the nonlinear relationships between built environment attributes and driving distance in suburban areas and how the relationships differ from those in urban areas of the Twin Cities. The result shows that enhancing job accessibility and intersection density are promising for reducing driving in suburban areas. Transit supply plays a moderate role in reducing diving distance in suburban areas. However, density and land use diversity, although important in urban areas, have trivial influences in suburban areas. Previous studies using cross-sectional data to examine the nonlinear relationships usually reveal the nonlinear associations between built environment attributes and travel behavior. The study in Chapter 4 applies transport rationales, extracted by Dr. Petter Næss and his research team from a comprehensive analysis of qualitative interview data, and the GBDT approach to data from Stavanger, Norway to explore causal mechanisms for the nonlinear relationships. The results show that transport rationales for choosing activity locations and travel modes, along with configurations of the jobs and other facilities, provide causal explanations for the nonlinear and threshold effects of built environment attributes on people’s driving-related behavior. Distance to city center plays the most important role and its nonlinear relationship reflects the influence of the polycentric city structure of Stavanger on driving. For Stavanger and similar cities, compact development around the city center helps to rein auto dependence. Furthermore, the threshold relationships provide planning guidelines to support compact development policies.enActive travelLand useMachine learningSuburban areaTransitTransport rationaleThesis or Dissertation