Effectiveness of Consistency Measures in Crash Prediction Models for Two-Lane Highways in Palestine
Publication Type
Conference Paper
Authors
Crash Prediction Models (CPMs) that incorporate design consistency measures for rural two-lane, two-way roadways in the West Bank of Palestine have been developed. This study aims at comparing the effectiveness of the resulting models with those, which rely solely on geometric design characteristics. Several two-lane rural highways located in the West Bank, Palestine for a total length of about 118 km were chosen for analysis based on limitations in available data. The selected highway sections encompass a variety of highway classifications, locations, terrains, and characteristics. Crash data were obtained from the Israeli Central Bureau of Statistics and included number of crashes and their type for years 2003 through 2012. The developed CPMs considered different consistency measures including variation between design speed and operating speed, absolute difference of the 85th percentile speeds between successive roadway sections, difference between side friction supplied and demanded, average radius of curvature, etc. The generalized linear regression modeling approach was used. The application of different regression models was investigated. Furthermore, the CPM proposed by the Highway Safety Manual, which is based on crash modification factor for horizontal curvature, was used for comparative analysis. A quantitative comparison between CPMs that explicitly consider design consistency with those based on geometric design characteristics was presented. A systematic approach to identify geometrically inconsistent locations using the safety-consistency factor was proposed. The results indicated that geometric consistency measures provide more reliable CPMs, which can predict collision potential more accurately.
Conference
Conference Title
World Conference on Transport Research - WCTR 2016 Shanghai
Conference Country
China
Conference Date
July 10, 2016 - July 15, 2016
Conference Sponsor
Elsevier
Additional Info
Conference Website