Spatial Modeling of Bicycle Activity at Signalized Intersections

This paper presents a methodology to investigate the link between bicycle activity and built environment, road and transit characteristics and bicycle facilities while also accounting for spatial autocorrelation between intersections. This methodology includes the normalization of manual cyclist counts to average seasonal daily volumes, taking into account temporal variations and using hourly, daily and monthly expansion factors obtained from automatic bicycle count data. To correct for weather conditions, two approaches were used. In the first approach, a weather relative ridership model was generated using the automatic bicycle counts and weather data. In the second approach weather variables were introduced directly into the model. For each approach, the effect of built environment, road and transit characteristics as well as bicycle facilities on cyclist volumes was determined. This research shows that regression models accounting for spatial autocorrelation provide a better fit to the data over the standard model and the elasticities are slightly sensitive to the approach used to account for weather. It was found that employment, schools, metro stations, bus stops, parks, land mix, mean income, one-way streets, arterials, bicycle lanes, cycle tracks and length of bicycle facilities were associated with bicycle activity. Signalized intersections are critical road elements for bicycle activity. Data and models are then essential to understand the impact of interventions and serve as an input in transportation safety, air quality and travel demand analyses. Despite the satisfactory results, further work should be done to validate and improve the quality of these bicycle activity models.