Development of a Dividing Method and Accident Estimation Models for Highway Horizontal Curve Sections Based on Geometric Characteristics

Abstract

A good understanding of the relationship between traffic accidents and geometric highway design factors is fundamental to the effective design of highway in terms of traffic safety. This study seeks to contribute towards that understanding. To do this, the following two tasks have been accomplished in this study. The first is to identify the distribution of the number of traffic accidents that is essential for developing accident prediction models. The second is to figure out the accident occurring pattern at the horizontal curve sections. For the two tasks, a method for dividing horizontal curve sections was developed. Using the dividing method, the horizontal curve sections were classified into single curve section(SCS) and spiral reverse curve section(SRCS) based on decision sight distance(DSD). For the empirical analysis, the accident data collected at Honam Highway in Korea for a five-year period from the year of 1996 to 2000 were used. Using the data, the distribution of the number of traffic accidents occurred on the two curve sections was defined and two negative binomial regression models for the curve sections were developed. Maximum likelihood method was used to estimate the unknown parameters employed in the two models.