Safety performance functions for crash severity on undivided rural roads

摘要: The objective of this paper is to explore the effect road features two-lane rural networks on crash severity. One main goals calibrate Safety Performance Functions (SPFs) that can predict frequency per year injuries and fatalities homogeneous segments. It was found more than 2000 km study-road network annual average daily traffic, lane width, curvature change rate, length, vertical grade are important variables in explaining severity crashes. A database covering a 5-year period examined achieve (1295 injurious crashes included 2089 235 fatalities). total 1000 were used SPFs remaining reflecting geometric, functional preceding one validate their effectiveness. negative binomial regression model used. Reflecting configurations dataset maximizing validation outcomes, four sets developed as follows: (a) equation only injury for subset where non-fatal occurred, (b) two different equations fatality sub-set at least fa tality occurred together with injury, (c) casualties correlating accurate percentages obtain final expected Residual analysis confirms effectiveness SPFs.