Crowdsourcing Data-driven Development of Bicycle Safety Performance Functions (SPFs): Microscopic and Macroscopic Scales

Graduation date:2017While riding bicycles has been promoted for its health, economic, and environmental benefits, it also complements other modes to complete a safe, efficient, and reliable transportation system. However, the dramatic increase of bicycle usage in the U.S. is accompanied by a growth of bicycle crashes. The U.S. Department of Transportation, therefore, is focusing on providing safer riding environments. Providing a more bicycle friendly environment means more investment in (but not limited to) bicycle infrastructure. A correct prediction of bicycle crashes can increase the return on this investment. One useful tool to understand the causality and predict crashes is Safety Performance Functions (SPFs), but no sophisticated SPFs have been established for bicycles. Therefore, the objective of this thesis is to establish SPFs for microscopic (intersection) and macroscopic (corridor) scales in medium and large size cities using crowdsourced bicycle data, with a case study in the Portland and Eugene-Springfield metropolitan, which overcomes the challenge of insufficient bicycle volume data and crash data. Specifically, in this research 1) bicycle SPFs are created for intersections and corridors that have not been sufficiently studied; 2) bicycle crash severity distributions are used the first time to predict the number of bicycle crashes with different crash severity levels; 3) affordable crowdsourced bicycle volume data – STRAVA® is chosen to solve the problem of limited data; 4) STRAVA® data was verified to be able to represent general bicyclists by comparison with automatic bike count station data; 5) a general framework for building SPFs was developed for jurisdictions