Using Naturalistic Data for Bicycle Safety Analysis: An Application of CyclePhilly Data to Assess Wrong-Way Riding

This study focuses on wrong-way riding and associated route-choice determinants of wrong-way riding. I present this application as a case study of the many types of analysis that are possible with emerging naturalistic datasets that are not limited to conventional methods (e.g., vehicle counters). I used a dataset generated from a smartphone application, CyclePhilly, that measures rider location second-by-second in Philadelphia. The dataset covers 12,202 trips by 300 unique CyclePhilly users collected from May 2014 through April 2016. The data also includes information on socio-economics of the rider, as well as cycling experience. I merged this dataset with complementary network datasets (like speed limits and traffic levels). The data allows us to identify route choice information that includes origins, destinations, and street segments chosen. Comparing the routes travelled with the network information, allowed to assess the proportions of those trips (segments) that include wrong-way riding for each segment. From these analyses, I compared the routes with and without wrong-way riding segments and to compare the network and demographic factors that could influence wrong-way riding behavior. It was found that trips made for commute purposes were more likely to have wrong-way riding than trips made for other purposes. Different bike infrastructure showed different effect on the wrong-way riding behavior. While bike lanes and cycle tracks showed higher wrong-way riding, sharrow lanes and buffered bike lanes were found to discourage that behavior. Roads with higher average AADT also showed to have less wrong-way riding than roads with less AADT. Roads with higher number of lanes also showed more wrong-way riding. The results from this study will help engineers and planners justify the suitability of various engineering or education measures to address wrong-way riding when planning for bike facilities.