GEO-TYPOLOGY FOR TRANSPORTATION SYSTEM ANALYSIS 

Abstract
New mobility options—shared mobility, micromobility, and Mobility as a Service—are transforming transportation systems, but make it increasingly challenging to analyze transportation policies (such as pricing and allocation of right of way). There is a large gap in incorporating these modes when assessing system impacts (such as efficiency, equity, access, and externalities outcomes) of policies across all types of locations, including rural, and suburban areas.

Addressing this gap requires overcoming three challenges: 1) quantitatively characterizing the ways in which regions across the US are similar or different with respect to their potential response to transportation policies; 2) identifying the ways in which system and user costs vary systematically across different types of locations; and 3) modeling interactions between different modes, including emerging modes, under different policies across different types of locations.

The focus of this paper is the first component, the creation of a methodology to develop a set of typologies that categorize every location in the US in terms of the main regional drivers of transportation demand and supply. The typologies we develop, called geotypologies or geotypes, are the first comprehensive set of typologies for both urban and rural areas across the entire US [e.g., 1,2,3]. This sets the foundation for addressing the second two challenges listed above. This paper describes a novel two-stage clustering procedure to systematically and quantitatively characterize the ways in which regions across the nation are similar or different with respect to their potential response to policies of interest. All 73,056 US Census Tracts are clustered into 6 “microtypes,” using first factor analysis and then applying the CLARA clustering algorithm [4] with inputs made up of local characteristics, such as road configuration, population and employment density, that influence the total system and user costs of travel by different modes. Then, 934 metropolitan areas and 1,317 counties are clustered into 6 geotypes using first principal component analysis and then a k-means clustering algorithm with inputs that influence travel demand, including novel metrics developed to capture the proportion and spatial arrangement of their constituent microtypes. Regions with similar population densities are differentiated based on network features (e.g., degree of circuity of the streets, road grade, or dead-end proportion), extent of agricultural land, or amount of trucking, among other things. These geotypes are comparable across all locations, enabling a national level perspective, while maintaining sufficient heterogeneity to identify different modal opportunities for a more cost-effective and efficient transportation systems.