Surrogate Safety and Network Screening: Modelling Crash Frequency Using GPS Data and Latent Gaussian Models

Improving road safety requires two steps of network screening and site diagnosis, which both require safety to be objectively quantified. In the screening phase, sites are identified and prioritized to maximize the efficiency of implemented countermeasures. Network screening methods commonly adopt regression techniques to estimate the expected number of crashes at sites across the network. Most existing techniques use crash-based ranking criteria which are subject to errors and omissions in collision databases, require long collection periods, and are reactive. GPS-enabled smartphones can collect reliable and spatio-temporally rich naturalistic driving data from regular drivers using an inexpensive, simple, and user-friendly tool that eliminates the need for external sensors. To date, very few studies have analyzed large volumes of smartphone GPS probe vehicle data or have considered advanced modelling techniques for screening in large road networks. The purpose of this paper is to develop a crash frequency model that incorporates surrogate safety measures (SSMs) extracted from the smartphones of regular drivers as predictive variables. After processing GPS data collected in Quebec City, Canada, several SSMs including vehicle manoeuvres (hard braking) and measures of traffic flow (congestion, average speed, and speed variation) were extracted. A Latent Gaussian Spatial Model was estimated using the INLA technique. Results showed that while negative binomial models outperformed Poisson models, the greatest improvement in model fit was achieved through a spatial model. In general, the relationships between SSMs and crash frequency established in previous studies were supported by the modelling results. Future work will include expanding the crash model to the entire Quebec City road network, comparing models estimated using INLA to those estimated using a traditional MCMC simulation, and incorporating collision severity estimation. The ability to screen the network based only on SSMs presents a substantial contribution to the field of road safety, and works towards the elimination of crash data in safety evaluation and monitoring.