Citation

Gharraie I, Sacchi E. Transp. Res. Rec. 2021; 2675(2): 53-64.

Copyright

(Copyright © 2021, Transportation Research Board, National Research Council, National Academy of Sciences USA, Publisher SAGE Publishing)

DOI

10.1177/0361198120961088

PMID

unavailable

Abstract

Each year, thousands of wildlife-vehicle crashes (WVCs) occur in North America with negative effects on wildlife welfare, human health, and the economy. Although previous studies have investigated factors related to WVC frequency, limited research has been conducted on factors affecting WVC severity. Using more than 10,000 WVCs occurring in the province of Saskatchewan (Canada), this study investigated the severity outcomes of WVCs and their influencing factors, using structural equation modeling (SEM) with generalized (ordered probit) links. Compared with traditional severity analysis techniques, SEM offers the added advantage of representing, estimating, and testing complex modeling structures that include both measured and latent (unmeasured) variables. Three latent variables were introduced in this study: driver's speeding attitude (SA), driver's visibility impairment (VI), and crash severity. Measured variables obtained from crash records were included in the SEM to define latent constructs, and the resulting network of relationships was tested. The results showed that crash data supported the model hypothesis well, and the measured/latent variables adequately predicted crash severity. Overall, SA and VI were demonstrated to positively affect crash severity with SA being the most influential factor. Moreover, it was demonstrated that road surface condition was the most influential factor of the SA measurement model, and weather condition was the most influential factor with respect to VI. Finally, a comparison between generalized SEM results and traditional crash severity modeling using ordered probit links was conducted. Similarities and differences between these two approaches were discussed at the end of the study.

Language: en