Assessing the transferability of the highway safety manual predictive method for urban roads in Fortaleza City, Brazil

Abstract

AASHTO's Highway Safety Manual (HSM) provides guidance to safety professionals and is becoming a standard of practice, particularly for regions that require urgent action because of persistently high crash frequencies that have resulted from an overall lack of safety investments. Differences in crash data systems, local enforcement, driving behavior, and other factors have prompted studies assessing the international transferability of the HSM predictive method. The recent increase in crash occurrences in Brazilian urban areas and the World Health Organization (WHO) Decade of Action for Road Safety initiative have motivated a nationwide joint safety research effort. One of the fundamental questions in this project is related to the applicability of HSM safety performance functions (SPFs) for the Brazilian urban environment. This paper presents the initial results regarding the transferability of the HSM SPFs for estimating the expected average crash frequency of urban intersections in Fortaleza City, Brazil. The procedure consists of two major steps: the first step investigates the estimation of the calibration factor according the procedure outlined in the HSM and is followed by a model validation effort using a different data sample; the second step is comprised of an analysis of the overdispersion parameter k, either using the original estimates from the HSM or recalibrated estimates using local conditions. The results suggest that for the Fortaleza case, the calibrated HSM baseline SPFs can be used with caution. It is, however, worthwhile to analyze the effect of the calibration sample size on the model stability, and efforts to produce SPFs using local data must be considered.