Surrogate safety measures from bikesharing GPS data for proactive bicycle safety analysis

The surge in bicycle traffic worldwide has led to increased focus on bicycle safety. Beyond the limitations of prolonged analysis periods and small sample sizes, traditional road safety methods that rely on reported crash numbers suffer from underreporting. Consequently, surrogate safety measures (SSMs) that pinpoint locations with an elevated likelihood of bicycle crashes, including bicycle-bicycle and bicycle-pedestrian crashes, are being investigated. This study seeks to assess the effectiveness of an SSM derived from bikesharing GPS data for proactive, large-scale bicycle traffic safety analysis. The research examined over 900 crossings identified using OpenStreetMap and CARTO databases, utilizing one month of GPS data from a bikesharing system alongside police-reported crashes from 2016 to 2019. Among the 22 potential SSMs, the number of decelerations exceeding 0.6 m/s2 displayed the most significant correlation with actual crash counts, evaluated using the Akaike information criterion of generalized linear models. This metric underpinned the bicycle crash prediction model, which was corroborated through cumulative residual plots for a control group, revealing strong predictive accuracy. Additionally, factors such as bicycle traffic volume and average bicycle speed (estimated using bikesharing GPS data), crossing area and road class (identified using OpenStreetMap and CARTO databases) were found to have a statistically significant effect on the number of bicycle crashes. The findings offer valuable insights for sustainable urban transportation planning and injury prevention strategies.