Development of a Crash Prediction Model for Signalized T-Intersections in Queensland, Australia

Abstract

Vehicle crashes at signalized intersections have long been of utmost concern to the transport authorities. Some researchers have developed crash prediction models (CPMs) for roundabouts in Queensland to establish the relationship among crashes, geometric parameters, and traffic conditions. However, to date, no CPM has been developed for the signalized intersections in Queensland. In this dissertation, two CPMs for signalized T-intersections in Queensland, Australia are developed: a CPM at intersection level and a CPM at approach level. The proposed models can be used for better control/organization of the traffic and design intersection geometry/layout. After the introduction to the dissertation and overview of the relevant research, a geographic information system (GIS)-based spatial database is established in Chapter 3 to visualize the crash records in south-east Queensland. Further, a crash assignment model is developed to assign crashes at signalized T-intersections to distinct approaches on the basis of the GIS-based spatial database. In Chapter 4, a CPM at intersection level is developed to predict the crashes at intersections, given the geometric parameters and traffic conditions. It should be pointed out that the number of lanes, merging angle, and existence of slip lanes are found to be significantly related to crashes.