Citation

Wang X, Feng M. Accid. Anal. Prev. 2019; 132: e105268.

Affiliation

The Key Laboratory of Road and Traffic Engineering, Ministry of Education, China; School of Transportation Engineering, Tongji University, Shanghai, 201804, China.

Copyright

(Copyright © 2019, Elsevier Publishing)

DOI

10.1016/j.aap.2019.105268

PMID

31465932

Abstract

Single-vehicle (SV) and multi-vehicle (MV) crashes have been recognized as differing in spatial distribution and influencing factors, but little consideration has been given to these differences as related to hotspot identification. For the purpose of better hotspot identification, this study aims to analyze influencing factors of SV and MV crashes and to explore the consistency between SV and MV hotspots. Crash data, roadway geometric design features, and traffic characteristics were collected along the two directions of a 45-km freeway section in Shanghai, China. Univariate negative binomial conditional autoregressive (NB-CAR) and bivariate negative binomial spatial conditional autoregressive (BNB-CAR) models were developed to analyze the influencing factors and specifically address (1) site correlation between SV and MV crashes within the same freeway segment, and (2) spatial correlation among different freeway segments within the same direction. The modeling results showed substantial differences in the significant factors that influence SV and MV crashes, including both roadway geometric features and traffic operational factors. A non-negligible site correlation was found between SV and MV crashes. Taking into account the site correlation, the BNB-CAR model outperformed the NB-CAR model in terms of parameter estimation and model fitting. For hotspot identification, potential for safety improvement based on the empirical Bayes method was adopted to handle the crash fluctuation problem. Substantial inconsistency was found between SV and MV hotspots despite the site correlation: in the top ten hotspots, no hotspot was shared by the two crash types. This result highlights the importance of differentiating SV and MV crashes when identifying hotspots, providing insight into freeway safety analysis.

Copyright © 2019 Elsevier Ltd. All rights reserved.

Language: en

Keywords

Bivariate negative binomial spatial CAR model; Freeway safety; Hotspot identification; Multi-vehicle crashes; Single-vehicle crashes