Crash Prediction Models and risk Factors for Two Lane Urban Roadways in Kumasi, Ghana

 

Abstract

The road network in Ghana has claimed many thousands of lives and over one hundred of thousand persons have suffered injuries in the last ten years. Prediction models can be used to improve road safety planning and prioritisation of improvement measures and funding. In Ghana, no previous work has been done on road traffic crash prediction models on urban links. In this research the primary aim was to develop prediction models after a comprehensive analysis of historical crash data.

                       Charles Adam

The specific objectives of the research were to:

1) Determine the characteristics and trends of crashes and traffic factors on two lane roadways from historical records.

2) Determine the risk factors associated with the trends in traffic and crashes and

3) Develop statistical models for the prediction of crashes on two-lane urban roads.

 

Data was collected for Traffic volumes, road traffic crashes for five years, road inventory and condition and road side environment. Risk factors were determined from trends and correlations of historical traffic, road and crash data before models were developed.

Using generalised Linear modelling techniques with binomial error structure in Statistical Analysis Software STATA software suite, prediction models were developed for Total injury crashes and two vehicle crashes. Model variables were evaluated at 95% confidence Interval for all explanatory variables after the core model with the exposure variables of Traffic and Road length had been constituted. The Akaike information criterion and the Freeman Turkey R2 were the main basis for accepting or rejecting a model and determining the quality of the models.

 

Ashanti region road networks are the most fatality prone network in Ghana. Crashes involving heavy goods vehicles and motor cycles are rising at rates far higher than their proportions in the vehicle population. Generally, most road traffic crashes have injury consequences and there is a marginal annual increase when compared with the situation a decade ago.

 

Two models have been proposed to predict Total injury crashes per year and two vehicle crashes per year on two lane roadways. The average daily traffic and section length were the main exposure variables. Average daily traffic (vehicles per day) varies linearly with both Injury crashes and two vehicle crashes. The average speed of travel and the presence of pedestrian sidewalks as a categorical variable were explanatory variables for injury crashes.

The presence of sidewalks increases the risk of injury crashes. For two vehicle crashes, average speed of vehicles (km/hr) and shoulder width are the explanatory variables for predicting crashes. Increasing shoulder width reduces the risk of incidence of two vehicle crashes.

          Charles Adams PhD

The study recommends that road safety on two lane urban roadways can be improved through the effective speed control and control of activities in the road corridor which increases pedestrian presence and concentration. Also, carefully designed wide shoulders can reduce crashes involving two vehicles on two lane roadways especially during evening peak periods. Transportation planners can apply these models to predict crashes during planning and cost benefit analysis.

Contact:

Charles Anum Adams

Senior Lecturer in Transportation

Civil Engineering Department (KNUST).
+233243788289
+233322097019

Related Links

Half Term Report on the Stewardship of Professor William Otoo Ellis, Vice-Chancellor of KNUST

On the 1st of October, 2010, Professor William Otoo Ellis was appointed as Vice Chancellor for a four year term to steer the affairs of this great institution by the Council of this University. He was inducted into office on the 2nd of October, 2010 and swore the oath of service and was formally outdoored to the University Community and to the rest of the general public.