EVALUATING THE POTENTIAL USE OF CROWDSOURCED BICYCLE DATA FOR CYCLING ACTIVITIES AND SAFETY ANALYSIS
Analytics
43 views ◎22 downloads ⇓
Abstract
Cycling, as a healthier and greener travel mode, has been encouraged for short-distance trips by city planners and policymakers. Since cycling provides an efficient way to improve public health, alleviate traffic congestion, and reduce energy consumption, it is essential to analyze the contributing factors to the cycling activities on each roadway segment and bicyclist injury risk, so as to quantify the impact of certain attributes on bicycle volume as well as biking safety and further provide better cycling environment for cyclists to encourage non-motorized travels.To map ridership, data including network characteristics, sociodemographic factors, and temporal characteristics, are quite indispensable. There have been multiple bicycle volume data collection methods and the most commonly used ones include traditional manual counts, travel surveys, and crowdsourced data from the third party. Most of the previous research efforts used the first two methods mentioned above to collect the data of interest. However, such methods are expensive and time-consuming. Crowdsourced data, on the contrary, are cost effective and timesaving, and therefore they have been widely collected and used by many public agencies and private sectors in recent years. Among all the crowdsourced data, data collected from smartphone applications including Strava, CycleTracks, ORcycle, etc. have become more and more prevalent. Crowdsourcing has increased the availability of data collection and provided an efficient way to bridge the data gap for decision making and performance measures. This research concentrates on evaluating the potential use of crowdsourced bike data and comparing them with the traditional bike counting data that are collected in the city of Charlotte, NC. Using the bike data from both the Strava smartphone cycling application and the bicycle count stations, the bicycle volume models are developed. Based on the results, a bicycle volume predictive model is presented, and a map illustrating the bicycle volume on most of the road segments in the City of Charlotte is generated. In addition, to gain a better understanding of the attributes that have an impact on cycling, other supporting data are also collected and combined with the Strava bicycle count data. Multiple discrete choice models are developed to analyze the Strava users’ cycling activities. Furthermore, bicyclist injury risk analysis is also conducted to explore the impact factors affecting biking safety by developing a series of safety performance functions. Several indicators for model comparison are utilized to select the best fitting model for bicyclist injury risk modeling. Finally, recommendations are made in order to help improve the cycling environment and safety and increase the bicycle volume in the future.