Research Of Urban Road Crash Analysis Based On Multivariate Statistical Techniques

With a rapid increase in motorization level, the situation of traffic safety on urban roads is becoming a critical issue. How to decrease traffic crashes and lessen their severity levels has attracted more and more attention. In consideration of the random characteristics of crash events, multi-statistical techniques can be well used to model the statistically significant relationships between risk factors and traffic crashes based on traffic crash and related data, and accordingly the countermeasures can be proposed. In light of the serious traffic safety concerns on urban roads, this dissertation conducted a suite of research, including crash frequency analyses, crash severity analyses, and risky traffic behavior analyses.Using a number of statistical indicators, we performed a descriptive statistical analysis for urban traffic crashes that take place on different time, on different road types, at different intersections/segments, by different vehicle usage, as well as different crash categories. As a result, it was found that safety issues related to main and minor arterials, four-legged signalized intersections and road segments with unrestricted access, taxis and buses, and vehicle-bicycle crashes are worth concerning.With a review of studies on traffic crash analyses, this study developed a conception model for analyzing urban traffic crashes using multivariate statistical techniques, and designed a crash analysis framework aiming at the serious safety concerns shown above. The detailed analyses included are a crash frequency analysis of signalized intersections and road segments with unrestricted access on urban arterial network, a comprehensive analysis of vehicle-bicycle crashes, and a safety analysis of influential paths from psychology characteristics, driving behavior, to crashes for taxi and bus drivers.A total of 108 four-legged signalized intersections and 123.5 km of urban roads were selected from the arterial network of Beijing. Severe crashes which occurred at the selected sites in a period of four years, as well as traffic-related information were collected to construct a longitudinal data set. Taking one site as a modeling unit, this study used the Generalized Estimating Equations modeling technique with a Negative Binomial link function to analyze significant effects of traffic-related factors on severe crashes. Results show that the arterial roads with heavier traffic, more road lanes and higher speed limits tend to have more severe crashes. Medians are helpful in reducing severe crashes. Moreover, higher severe crash risk is generally associated with intersections with small angle and countdown signal, and road segments with higher side access density and presence of bus stops. With regard to interaction effects between non-motorist protection facilities and other factors, results reveal that a combined use of crosswalk and overpass is the most desired pedestrian crossing facility for safety, especially at sites with heavy traffic or sites located at primarily residential areas. Furthermore, barriers separating bikeway and roadway on road segments or the minor roads of intersections are found effective to improve safety.The comprehensive analysis of motor vehicle-bicycle crashes was conducted using 4 years of crash records (2004-2007) from Beijing. In this study, one crash is used as a modeling unit. Firstly, a multinomial logit model was used to associate non-compliant behaviors with demographics, geometric design, and other related factors, and it was found different non-compliant behaviors are correlated with a number of risk factors at different road locations. Secondly, with an analysis of propensities of non-compliant behaviors to different crash patterns, the results show angle collisions are the leading pattern of motor vehicle-bicycle crashes, and different non-compliant behaviors may lead to some specific crash patterns such as head-on or rear-end crashes. Thirdly, a binary logit model was employed to identify risk factors affecting bicyclist injury severity outcomes, and it was found bicyclist severe injuries are associated with head-on and angle collisions, occurrence of running over bicyclists, night without streetlight, roads without median/division, higher speed limit, heavy vehicle involvement and older bicyclists. Moreover, it was found orthokinetic scrape is more likely to result in running over bicyclists.The safety analysis of taxi and bus drivers was based on the data obtained from a questionnaire survey carried out among 248 taxi and bus drivers in Wuhan. Firstly, One-way ANOVA was used to explore the difference between taxi and bus driver groups on all measuring scales, and there were no significant differences found. Hence, the survey data related to the drivers were aggregated. Secondly, a binary logit model was employed to analyze significant factors affecting driver at-fault crash occurrence, and results show that drivers who reported more tendencies of aggressive violations and ordinary violations, and who had previously been involved in at-fault crashes are in high risks of crash involvement. Thirdly, by developing a path model to analyze the casual relationships among risk perception, risk-taking attitudes, and risky driving behaviors, it was found that drivers' attitude towards rule violations has significant impact on risky driving behaviors. Furthermore, two risk perception scales, i.e. likelihood of crash and concern, have significantly indirect effects on the risky driving behavior through their influence on drivers'attitude towards rule violations.The significant contributing factors identified in this study are expected to result in better designing of road and traffic facilities, better actualizing of traffic management for the safety, and better planning of road safety campaigns for vehicle drivers.