Open Collections

Modeling collision injury severity of vulnerable road users in Bangladesh

University of British Columbia

Traffic collision is one of the leading causes of death around the world; the scenario is much alarming in the low- and middle-income developing countries. This thesis investigates crash injury severity of vulnerable road users in Dhaka, Bangladesh. The study assessed collision-level and victim-level risk of different road users using statistical analysis. Pedestrians, victims in collisions involving public transit (PT), and unconventional vehicle occupants (UVOs) are identified as the vulnerable road users and considered for further analysis. This study develops advanced econometric models using a 5-year police-reported crash record from Dhaka. In the case of the pedestrian injury severity, a latent segmentation-based ordered logit (LSOL) model is developed. One of the key features of the LSOL model is to address the ordinal nature of the injury severity levels and capture unobserved heterogeneity. A similar LSOL model has been adopted to analyze the injury severity of victims in crashes involving PT. For UVOs, a hybrid model - latent segmentation-based random parameters logit (LSRPL) model has been developed. One of the unique features of the LSRPL model is to accommodate multi-dimensional heterogeneity – i.e. inter- and intra-segment heterogeneity. The LSOL model for pedestrian injury is estimated for two segments - a high-risk segment and a low-risk segment. The model results confirm that significant heterogeneity exists across segments. For instance, collisions occurring at traffic police-controlled intersections show a higher likelihood for severe pedestrian injury in the high-risk segment, and are less likely to yield severe injury in the low-risk segment. Similarly, the LSRPL model for UVOs is estimated for a high-risk segment and a low-risk segment. The model results suggest that significant heterogeneity exists across and within the segments. For example, collisions occurring in the mid-block road sections reveal a higher probability for severe injury in high-risk segment, however, mid-block sections in higher mixed land use areas might pose lower injury risk. In contrast, mid-block crashes show a negative relationship in the low-risk segment. The findings of this thesis provide insights to road safety engineers and planners in developing plans, and emphasize the need to accommodate heterogeneity for effective road safety planning.

Text

2020-08-24

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/75650

Civil Engineering

University of British Columbia

UBCO

http://creativecommons.org/licenses/by-nc-nd/4.0/