DE-CONSTRUCTING THE D-VARIABLES: NEW METHODS TO MEASURE THE BUILT ENVIRONMENT FOR TRAVEL BEHAVIOR RESEARCH

摘要

Measures are determined by the data, technology and methods available. Untilrecently these have all limited travel behavior and built environment research to rely onzonally aggregated averages, homogeneously attributed to unique individuals.Furthermore, most studies focus on the characteristics of the origin, not the criticalcomponents of the destination (such as parking, availability, price, etc.), let alone the areain-between. But travel is an inherently linear activity with strong destination determinantsover travel choices, and thus these zonal measures of the trip origins likely miss keysubtleties of the built environment important to people traveling outside of the protectiveenclosure of an automobile (bicyclists and pedestrians).This paper presents the development and statistical testing of new methods to moreclosely align detailed measures of the built environment with the individual– in sum,more finely disaggregated data of the built environment for disaggregated analyses oftravel behavior. To ground the development of these new methods and measures, thispaper applies their development to a real-world problem: How does the urbanenvironment influence the probability that commuters will access rapid transit stationsvia green and active alternatives to driving?Through the use of predictive, multinomial logit (MNL) mode choice models andMNL model comparison methods, this paper tests whether new, linear based builtenvironment measures are an improvement over zonally aggregated "D-Variable"measures now commonly used in practice. These model comparison techniques prove thesuperiority of these new, more detailed measures over the D-Variables for linear spatialunits of analysis.