Estimation of Safety Performance Functions (SPF) at signalized intersections in Medellín Colombia

Abstract

One of the objectives of road infrastructure sustainability is to ensure that users are treated equally and their quality of life is improved by providing better mobility and traffic safety. When designing roads, it is important to evaluate different design criteria alternatives - in this case, we look at traffic safety principles. For this, we used the Safety Performance Functions (SPF) tool to obtain the expected crash frequency. The data used were Medellín’s crash records from 2012 to 2016, as well as the geometric features and traffic conditions at signalized intersections. A negative binomial model was fitted to estimate the SPF. Exposure, geometry, and traffic volume were found to be statistically significant in determining the expected crash frequency for collisions where there was property damage only (PDO) and fatal or injury (FI). It was found that accidents were less likely on T-junctions compared to four-leg junctions, one-way approaches were found to be safer whereas right turns were found to increase collisions causing FI.

Keywords

• Safety Performance Function
• Signalized Intersections
• Traffic Safety
• Negative Binomial Regression