Building Equitable Safe Streets for All: Data-Driven Approach and Computational Tools (06-001)

Project Description: The roadway safety improvement in low-income and ethnically diverse communities in the United States has long been a major concern. Unequal allocation of transportation infrastructure, such as placing high-capacity roadway near or within low-income community of color or inadequate provision of sidewalk or bike lane in disadvantaged communities may lead to the environmental injustice, resulting in the disproportionate crash risks. This report conducted a systematic literature review and find that limited data sources and insufficient utilization of analytical approaches may hinder the investigation of environmental injustice in road accidents. To address data limitation, this report tries to fuse multiple data sources such as roadway crash data and roadway geometry, crowd-sourced pedestrian and bicyclist exposure from Strava, and sidewalk information extracted from street view images using deep learning model. We conducted two research focusing on the equity in roadway safety. Research 1 is “Investigation on the driver-victim pairs in pedestrian and bicyclist crashes by latent class clustering and random forest algorithm”. It tries to answer how the socioeconomic background of driver and victim involved in one crash are related and how this relationship is influenced by roadway infrastructure. By applying an interpretable machine learning approach with random forest algorithm, we find that low-income non-White crashes are characterized with riskier roadway environment such as higher exposure and speed limit. The data for Harris County, TX, was collected in the summer of 2022, and it includes pedestrian/bicyclist crashes data from Crash Report Information System and Roadway inventory Data from Texas Department of Transportation, bicyclist and pedestrian exposure data from Strava, and socioeconomic data from American Community Survey. Research 2 is “Disparities in Roadway Safety: Exploring Direct and Indirect Pathways Contributing to Disparities in Non-Motorist Crashes in Houston, Texas”. It tries to answer whether disadvantage communities are exposed to higher crash risks and what mechanism cause such disparity. To investigate how the disadvantaged communities exposed to higher crash risk, we developed a structural equation model incorporating neighborhood disadvantage, roadway environment, active transportation infrastructure, vehicular exposure, active transportation infrastructure and non-motorist crashes and find two pathways of how environmental injustice happened through motor vehicle mode and active transportation mode. Findings from this report can provide foundation for transportation policy and transportation planning to ensure environmental justice principle. Data used in this research for Harris County, Tx, was collected in the spring of 2023, including pedestrian/bicyclist crashes data from Crash Report Information System and Roadway inventory Data from Texas Department of Transportation, bicyclist and pedestrian exposure data from Strava, and socioeconomic data from American Community Survey, Street view image from Google maps, Bike lane data from Houston Map Viewer, traffic signal data from Transtar. Data Scope: For research 1 “Investigation on the driver-victim pairs in pedestrian and bicyclist crashes by latent class clustering and random forest algorithm”, the analysis unit is pedestrian/bicyclist crashes. There are 2,822 pedestrian crashes and 1,123 bicyclist crashes analyzed in Harris County from 2017 to 2022. The data type is csv file. For research 2 “Disparities in Roadway Safety: Exploring Direct and Indirect Pathways Contributing to Disparities in Non-Motorist Crashes in Houston, Texas”, the analysis unit is Hexagon with 1-mile side length. There are 2221 hexagons covers Harris County in the research area. The data type is csv file. Data Specification: Please see the Data Specification and Data Dictionary documents.