Modeling injury severities of single and multi-vehicle freeway crashes considering spatiotemporal instability and unobserved heterogeneity

ABSTRACT

Single and multi-vehicle (SMV) crashes remain a significant issue, causing serious safety and economic concerns, and therefore deserve more attention. Using crash data in the Beijing-Shanghai and Changchun-Shenzhen freeways over the five years (2015–2019), this paper explored the transferability and heterogeneity for crash type (single-vehicle versus multi-vehicle crashes) and spatiotemporal stability of determinants affecting the injury severity. The random parameters logit approach with heterogeneity in means and variances was used to model three possible crash injury severity outcomes (measured by the most severely injured individual in the crash) of no injury, minor injury, and severe injury and identify the determinants in terms of driver, vehicle, roadway, environment, temporal, spatial, traffic, and crash characteristics. Remarkable differences were observed in the SMV crashes, and the contributing factors also reported considerable temporal and (or) spatial instabilities. The insights of this study should be valuable to help freeway designers and decision-makers understand the contributing mechanism of the factors and develop the proper management strategies and enforcement countermeasures.

KEYWORDS:

• Spatiotemporal stability
• single and multi-vehicle crashes
• injury severity
• random parameters logit model

Acknoledgement

The authors are grateful to two anonymous reviewers for their thorough and most helpful comments.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. The temporal instability over year, season, time of day, or other periods tends to result from a wide range of factors including changes in driver decision-making, information processing, risk assessment, and safety attitudes, which suffered from changes in potential elements including vehicle interactions, and roadway-safety features, the effects of that macroeconomics, and so on. This issue has been confirmed in a large body of literature in terms of psychology, neuroscience, economics, cognitive science, and other fields (Mannering 2018).

2. The degrees of freedom equal to the summation of statistically significant parameters in each year minus the number of statistically significant parameters in the overall model (Washington, Karlaftis, and Mannering 2018), which are the same in Equations (7) and (8).

3. The results of likelihood ratio tests utilized to examine spatial and temporal stability in this study suggest that the estimated models can be transferable to other freeways, cities, or provinces in China across different time of period. By collecting dataset from more freeways and across long time series, the transferability of the estimated model can be examined in further research.

4. Behnood and Mannering (2019) also argued the temporal instability of speeding, which was found to be only statistically significant in one of four time-of-day/time-period models.

5. The results indicated considerable variations of the risk speeding behaviors. Islam and Mannering (2020) indicated that the exceeding speed limit by more than 10mi/h was statistically significantly in aggressive driving SV crash models, whereas the marginal effects also produced inconsistent signs for severe injury.

6. The roadway section was divided into segments of horizontal curves and tangents. A tangent was assigned a radius of a huge value (106 m). If a crash occurred on a segment, the relevant variables are (1) the radius and length of the curve segment on which the crash occurred ($R_{\mathit{present}}\mathit{and}{L}_{\mathit{present}}),$ (2) the length of the next curve segment ($L_{\mathit{front}}$), and (3) the length of the maximum longitudinal slope of the curve segment on which the crash occurred. The curve variables were determined from the road plan and profile of the freeway.

7. It cannot be denied that the contradictory findings were found in recent research efforts (Yu, Ma, and Shen 2021; Al-Bdairi, Behnood, and Hernandez 2020), which concluded that clear weather conditions result in higher probability of severe and minor injury. This might be explained by the higher traveling speeds under clear weather conditions, causing more severe injuries (Roque, Moura, and Cardoso 2015).

8. Similarly, the interesting findings can be seen in the study of Yan et al. (2021b) analyzing the crash-injury severities in adverse weather, the detrimental driving conditions such as slippery pavement surface illustrated complicated and opposite impacts on the crash outcomes over different periods.

9. A study by Wang et al. (2022b) demonstrated considerable differences in the rear-end and non-rear-end crashes based on the dataset of G2 Freeway from 2017 to 2019. Accounting for the complicated effects of rear-end crashes, further research efforts will be conducted by the authors to analyze the differences in the two-vehicle and multi-vehicle crashes accounting for spatiotemporal instability and unobserved heterogeneity, to explore the mechanisms of rear-end crashes.

10. As Ye and Lord (2014) concluded, the minimum sample size for the random parameter model should be approximately 5,000.

Additional information

Funding

This study was funded by the National Natural Science Foundation of China (grant number 51768063 and 51868068).