An investigation of pavement distress variables on crash outcomes using hierarchical generalized linear regression modeling.

摘要

Previous macro-level studies show that the condition of the pavement is associated with roadway safety. However, this premise has not been examined on a micro-level using detailed pavement distress variables (PDVs). The main focus of this research will be to expand the use of the PDV data to further understand the safety risks associated with the type of pavement and each individual PDV by analyzing the likelihood of having a rear-end and/or injurious crash, given that a crash occurs.;Sample crash data and PDV data from the Commonwealth of Virginia for the years 2007 and 2008 were used to produce a dataset that provides crash and pavement condition, type and ride quality data at the crash site and specific intervals upstream for three types of pavement. Binary logistic regression statistical modeling was used to determine if PDVs have an association with rear-end crashes and crashes with injuries. By investigating this relationship at the crash site and specific intervals upstream of the crash, this study provides valuable insights into the spatial component of pavement safety.;Additionally, there is an a priori reason that the morphology of the built-up environment could influence the severity of an accident. By including social/economic factors and applying hierarchical generalized linear modeling this study will use the hierarchical nature of crash data to examine socio-economic characteristics of the locality where the crash occurs.;The analysis of rear-end and injurious crashes resulted in PDVs that are associated with an increased risk of these types of crashes on each of the three pavement types. While this association is weak for injurious crashes, the results indicate the critical location to be upstream of the crash; for rear-end crashes, the critical location was at the crash site. It was determined the type of pavement is not significant for crashes with injuries, but it is for rear-end crashes.;The results indicate there is little benefit to using HGLM to model crashes with injuries, but the variability in rear-end crashes can be explained by the nested structure. The two socio-economic factors that reduce the odds of a rear-end crash are the average age of the driver and unemployment percentage.