Effect of Roadway Geometrics and Travel Characteristics on Crash Frequency for Multi-lane Divided Highway

Abstract

Rapid motorization particularly increases in use of vulnerable transportation means (motorcycles and rickshaws) has resulted into enormous increase in road traffic crashes in Pakistan. Road crash fatalities and injuries have emerged as a moral and social challenge for community with over Rs 100 billion annual economic losses to the nation. Each year approximately over 30,000 individual lose their lives and approximately 400,000 sustain injuries due to road traffic crashes. Better understanding of the factors responsible for road crashes is mandatory for selecting appropriate road safety commuter measures. Since fatalities due to road crashes are amenable to remedial measures, therefore it is possible to observe drop in annual number of fatalities of any country if effective road safety counter measures are adopted. Roadway geometry is an important factor associated with road traffic crashes. Highways with appropriate geometric characteristics can help in significant reduction in annual road traffic crashes. Thus there is need to adopt appropriate methodology that can help to identify problematic highway segments that need remedial measures. The main focus of this study is to develop a statistical model of road crash frequency using data on roadway geometrics and travel characteristics. A detailed review of past studies revealed that at national level no research effort has been made to identify geometric deficiencies responsible for road traffic crashes. At international level, numbers of studies have been carried out using advanced statistical techniques that helped to establish the relationship between crash frequency and highway geometric features. Present study using 5-year traffic accident data for 280 Km of one of the national highway of Pakistan (Grand Trunk road from Rawalpindi to Lahore) developed negative binomial regression model. Model results revealed that segment length, lane width, number of U-turns, posted speed limit, number of lanes, number of access points of highway segment, percentage of single unit truck in traffic stream, and road segments in urban area are significantly associated with road crash frequency. Comparative analysis and appropriate statistical tests revealed that negative binomial regression model is superior as compared to Poisson regression models, zero inflated negative binomial and zero inflated Poisson regression models. The results of this study xii can be used by National Highway Authority and Ministry of Communication as an input for formulation of multipronged road safety improvement policy for Pakistan and for development of effective road safety counter measures targeted at crash-prone highway segments.