Assigning crashes to road segments in developing countries

Abstract

In view of the high frequency of crashes on two-lane two-way rural roads, the safety of such roads was evaluated. In developing countries like Iran, crashes on rural roads are generally recorded with low accuracy. In this work, two new methods for assigning crashes to road segments were developed. In the first method, observed crashes are assigned to segments based on the exposure index, while in the second they are assigned based on the crash prediction model (CPM) of the Highway Safety Manual (HSM). Owing to the random nature of crashes, Poisson regression was used for modelling. Evaluation and comparison of the models revealed that the best performance was obtained by the model based on observed crashes assigned by the HSM's CPM using the variables: segment length; annual average daily traffic; shoulder width; presence of residential land use on the roadside; horizontal curve radius; variance of the super-elevation required in road design codes from the existing horizontal curve super-elevation; and driveway density.