Development of Decision Support Tools to Assess Pedestrian and Bicycle Safety: Development of Safety Performance Functions
- PIs: Kwigizile, Oh, Van Houten – Western Michigan University
- Project Period: July 1, 2014 – June 30, 2016 (24 months)
Summary: Walking and biking are forms of transportation that offer basic mobility for all people. In totality, walking and bicycling improve quality of life in many ways, such as increasing physical activity and active lifestyles, which consequently result in health benefits. However, bicyclists and pedestrians are more likely to be killed in a crash for each trip as compared to vehicle occupants. In order to prioritize sites for safety improvements, Safety Performance Functions (SPFs) are needed. This project developed a methodology for estimating SPFs for pedestrians and bicyclists for urban collector and arterial roads intersections.
Problem: Due to lack of pedestrian and bicycle counts at intersections, development of SPFs for non-motorized traffic is very challenging. Also, while motorized traffic volume data is mostly available for intersections of major roads, it is very difficult to obtain for collectors and local roads, where pedestrians and bicyclists are commonly found. Exposure data are very essential components of SPFs as they explain most of variation of the non-motorized crashes occurring at different locations. Therefore a careful sampling plan, which captures the randomness of non-motorized crashes and inclusion of reliable proxy exposure measure for pedestrians and bicyclists, is imperative. The main purpose of this project was to develop a methodology for estimating statewide safety performance functions for pedestrians and bicyclists at intersections. A case study for this research was all urban collector and arterial roads intersections in Michigan. Specifically, the methodology addressed the following:
- Proper sampling procedure to establish an unbiased sample size for model development.
- Developing proxy measures of pedestrian and bicyclist exposure using data that are readily available at statewide level.
- Assessment of SPF performance using cross-validation technique.
Research Results: Factor analysis was used to develop pedestrian and bicycle level score using variables that are readily available at the statewide level. Latent bicyclist level score, a proxy measure of bicyclist volume, was found to increase with the presence of bicycle facility, which includes bike lanes and sidewalks, increase in percentage of people below poverty level, increase population density, lower speed limit in major and minor approach, and increase in proportion of commercial land use by area in a given census block group where the intersection is situated. Pedestrian level score was found to be related with the increase in percentage of people using the public transit in a given block group where the intersection was situated, population density, percentage of household below poverty level, number of workers commuting to their workplaces by foot per square mile, walk score index, proportion of commercial land use, and presence of pedestrian facility separated from the roadway. During development of SPFs, comparison was made across all potential count models that could fit the data. Appropriate goodness of fit tests and cross validation techniques were used in selecting the model with the best fit. The final equation for bicycle SPF is as follows:
Based on the assessment of goodness of fit measures, Zero Inflated Poisson Regression Model was selected as the final model for pedestrian SPF, expressed mathematically as: