Prediction of fatal crashes based on various victim types on national highways passing through urban areas of developing countries

Abstract

The issue of safety on national roads that traverse urban areas in emerging countries has been a significant cause for concern. Statistical analysis and modelling techniques for road safety evaluation in metropolitan India are still evolving due to inadequate crash data, inventory, and traffic volume data records. This work aims to formulate safety performance functions (SPF) using negative binomial (NB) count data models to pinpoint the variables influencing total fatal crashes and other individual victim types, i.e., pedestrians, motorcyclists, and single vehicles on mid-block sections. Using four years of crash data (2014–2017) from Visakhapatnam City Police, India, the applicability of the current study framework has been established. Besides the geometric design elements, segment length, speed and average daily traffic, this study also focuses on information collected from road safety audits such as the provision of service roads, land use type, median opening, side access, pedestrian crossings, sight clearance to the driver, availability of earthen shoulders, proper signage, and good road markings. The study outcome reveals that road segment length, service road presence, and land use type are significant risk variables associated with fatal crashes in Visakhapatnam City, India. The length of the road segment positively correlated with frequencies of total fatal,  pedestrian, motorcycle and single-vehicle fatal crashes. It increased the frequency of fatal crashes on each increment by 103%, 122%, 73% and 98%, respectively. Service roads increase crash frequencies, and road stretches with commercial/mixed land use types attract more crashes. This research emphasizes the essential safety precautions that transportation engineers and planners must implement to establish a more secure environment for all road users.