Comprehensive evaluation of signal-coordinated arterials on traffic safety. (March 2019)

• Record Type: Journal Article
• Authors: Fan, Yingfei; Zhang, Guopeng; Ma, Jian; Lee, Jaeyoung; Meng, Teng; Zhang, Xiaoning; Jiang, Xinguo
• Is Part Of: Analytic methods in accident research. Volume 21(2019)
• Page Start: 32
• Page End: 43
• Publication Date: 2019-03
• Subjects: Signal coordination -- Crash prediction -- Joint negative binominal conditional auto-regression crash sum model -- Spatial heterogeneity; Accidents -- Research -- Methodology -- Periodicals; Accidents -- Prevention -- Periodicals; 363.100721
• DOI: 10.1016/j.amar.2019.01.001
• Languages: English
• ISSNs: 2213-6657

Abstract

Highlights: The joint negative binominal conditional auto-regression crash sum model outperforms other models in terms of the goodness-of-fit. Signalized intersections appear to have significant spatially heterogeneous effects on roadway segments, while roadway segments do not have spatially heterogeneous effects on signalized intersections. There is a strong spatially homogeneous correlation among the coordinated intersections, but not among the segments. Abstract: Many cities have adopted signal coordination schemes to improve the operational efficiency of arterials. Through providing a "green wave", the coordinated signals allow vehicles to pass consecutive intersections with fewer stops. Conventionally, studies focused mainly on the efficiency aspect of signal coordination. From a safety perspective, there are few studies devoted to investigating spatial heterogeneity and homogeneity simultaneously along arterials, particularly those with different traffic control and signal coordination schemes. Applying a joint negative binomial conditional auto-regression crash sum (JNBCS-CAR) model, this paper aims to comprehensively evaluate traffic safety on arterials with coordinated traffic signals. The performance of JNBCS-CAR is compared to those of traditional five ones (i.e., Poisson log-normal (PLN), negative binominal crash sum (NBCS), Poisson log-normal conditional auto-regression (PLN-CAR), negative binominal conditional auto-regression crash sum (NBCS-CAR), and hierarchical Poisson log-normal conditional auto-regression (HPLN-CAR)). Nine years of data are selected from the City of Ann Arbor (Michigan). Results indicate that (1) the JNBCS-CAR model outperforms other models in terms of the goodness-of-fit; (2) signalized intersections appear to have significant spatially heterogeneous effects on roadway segments under the scenario of signal coordination, while roadway segments do not have spatially heterogeneous effects on signalized intersections; and (3) there is a strong spatially homogeneous correlation among the coordinated intersections, but not among the segments. The findings serve to provide an accurate evaluation of the coordinated signalization from the safety perspective.