Book Description
"Improving road safety requires accurate network screening methods to identify and prioritize sites to maximize effectiveness of implemented countermeasures. In screening, hotspots are commonly identified using statistical models based on historical crash data. However, collision databases are subject to errors and omissions and crash-based methods are reactive. With the arrival of Global Positioning System (GPS) trajectory data, surrogate safety methods, proactive by nature, have gained popularity. Although GPS-enabled smartphones can collect reliable and spatio-temporally rich driving data from regular drivers using an inexpensive, simple, and user-friendly tool, few studies to date have analyzed large volumes of smartphone GPS data and considered surrogate-safety modelling techniques for network screening. The main objective of this thesis is to propose and validate a GPS-based network screening modeling framework dependent on surrogate safety measures (SSMs). First, methods for collecting and processing GPS and associated data sources are presented. Data, collected in Quebec City and capturing 4000 drivers and 21,000 trips, was processed using map matching and speed filtering algorithms. Spatio-temporal congestion measures were proposed and extracted and techniques for visualizing congestion patterns at aggregate and disaggregate levels were explored. Results showed that each peak period has an onset period and dissipation period lasting one hour. Congestion in the evening is greater and more dispersed than in the morning. Congestion on motorways, arterials, and collectors is most variable during peak periods. Second, various event-based and traffic flow SSMs are proposed and correlated with historical collision frequency and severity using Spearman's correlation coefficient and pairwise Kolmogorov-Smirnov tests, respectively. For example, hard braking (HBEs) and accelerating events (HAEs) were positively correlated with crash frequency, though correlations were much stronger at intersections than at links. Higher numbers of these vehicle manoeuvres were also related to increased collision severity. Considered traffic flow SSMs included congestion index (CI), average speed (V̄), and coefficient of variation of speed (CVS). CI was positively correlated with crash frequency and showed a non-monotonous relationship with severity. V̄ was negatively correlated with crash frequency and had no conclusive statistical relationship with crash severity. CVS was positively related to increased crash frequency and severity. Third, a mixed-multivariate model was developed to predict crash frequency and severity incorporating GPS-derived SSMs as predictive variables. The outcome is estimated using two models; a crash frequency model using a Full Bayes approach and estimated using the Integrated Nested Laplace Approximation (INLA) approach and a crash severity model integrated through a fractional Multinomial Logit model. The results are combined to generate posterior expected crash frequency at each severity level and rank sites based on crash cost. Negative Binomial models outperformed alternative models based on a sample of the network, and including spatial effects showed improvement in model fit. This crash frequency model was shown to be accurate at the network scale, with the majority of proposed SSMs statistically significant at 95 % confidence. In the crash severity model, fewer variables were significant, yet the effect of all significant variables was consistent with previous results. Correlations between rankings predicted by the model and by the crash data were adequate for intersections (0.46) but were poorer for links (0.25). The inclusion of severity, which is an independent dimension of safety, is a substantial improvement over many existing studies, and the ability to prioritize sites based on GPS data and SSMs rather than historical crash data represents a substantial contribution to the field of road safety." --