Accident Prediction Models for Unsignalised Urban Junctions in Ghana

The main objective of this study was to provide an improved method for safety appraisal in Ghana through the development and application of suitable accident prediction models for unsignalised urban junctions. A case study was designed comprising 91 junctions selected from the two most cosmopolitan cities in Ghana. A wide range of traffic and road data together with the corresponding accident data for each junction for the three-year period 1996-1998 was utilized in the model development process. Potential explanatory variables, which were tested were largely identified from initial analysis of the accident characteristics and associated factors. Negative Binomial models of accident frequency were developed separately for Tand X-junctions. The results showed that the best models based exclusively on traffic exposure functions (i.e. traffic flow) explained 50 per cent more of the systematic variation in accidents at T-junctions than at X-junctions. In the extended models that included road geometric and other traffic variables it emerged that the absence of street lighting and dedicated left-turning lanes and the average standard deviation of approach spot speeds of vehicles on the major road were all positively correlated with accident frequency at both Tand X-junctions. Significantly and contrary to expectation, T-junctions with YIELD control had a much lower accident potential than those with STOP control. The accident prediction models developed have a potentially wide area of application and their systematic use is likely to improve considerably the quality and delivery of the engineering aspects of accident mitigation and prevention in Ghana.