A model to predict occurrence of head-on traffic crashes along Kampala–Masaka road

Abstract

The study evaluates the influence of geometric factors and traffic factors in the causation of head-on traffic crashes on a two-lane rural highway in Uganda using Negative Binomial Regression (NBR) Model. Head-on traffic crashes data for a period of five years (2013 to 2017) was obtained from Uganda Police Traffic Accidents Register (TAR). The Traffic Accidents Register was accessed at different police stations along Kampala–Masaka road. A total of ten blackspots with the highest frequency of head-on traffic crashes were selected for this study. Geometric factors were extracted from as-built drawings obtained from Uganda National Roads Authority while others were measured using tape measure and measuring wheel. The free flow speed data was collected by mounting a tripod camera near the road for five hours to collect video recording at the study locations. Classified manual traffic counts were carried out at the ten blackspots to obtain traffic volume data. Aerial drones were flown over study points where video footages were extracted. The video footages were analysed by Kinovea software to generate passing manoeuvres data i.e. speed of the passed vehicle in the manoeuvre and time taken by passing vehicle to complete the manoeuvre. The Negative Binomial Regression Model results indicated that all the independent variables used this study were statistically significant at 95% confidence level. The performance of the model was evaluated by running the model on seven different sets of data collected at seven locations. The performance metrics used was Mean Absolute Percentage Error (MAPE) which was computed at each location. The NBR model results indicated that 85th percentile of free flow speed was the greatest predictor variable to the causation of head-on traffic crashes while time required to complete the manoeuvre was the least predictor variable to occurrence of head- on traffic crashes. Sensitivity analysis was conducted and it was determined that operating speeds greater than 105 km/hr would cause higher likelihood of occurrence of head-on traffic crashes on two-lane rural roads. It was recommended that motorists should drive at speeds less than 105 km/hr, the absolute vertical grade should be less than 8%, time required by passing vehicle to complete a manoeuvre should be less than 6 seconds and for two-lane highways with ADT value exceeding 9,800 vehs/day, an additional lane should be provided for heavy vehicles.