Accounting for under-reporting in wildlife-vehicle collision hotspot identification using copulas and Bayesian model...

Area of Science:

• Ecology
• Transportation Science
• Statistical Modeling

Background:

• Wildlife-vehicle collisions (WVCs) present significant safety and conservation challenges.
• Under-reporting in official crash data hinders accurate risk assessment and mitigation planning.

Purpose of the Study:

• To develop a novel hybrid copula-based framework to jointly model WVC frequency and under-reporting probabilities.
• To identify key factors influencing WVC under-reporting and collision risk across Washington State road segments.

Main Methods:

• Integration of police-reported crash data with carcass removal records.
• Application of a hybrid copula framework with Bayesian Model Averaging (BMA) to model WVCs and under-reporting.
• Utilizing Gaussian and Student-t copulas to capture dependence between reported WVCs and under-reporting.

Main Results:

• Under-reporting is influenced by factors like proximity to specific wildlife habitats (wolverine, white-tailed deer) and road characteristics.
• WVC frequency correlates with traffic volume (AADT), road design (shoulders, lanes), speed limits, and truck percentages.
• Hybrid copula models with BMA significantly improve hotspot identification and account for under-reporting biases.

Conclusions:

• The developed framework provides a more accurate assessment of WVC risk by addressing data under-reporting.
• Findings offer crucial insights for transportation agencies to prioritize effective mitigation strategies, such as wildlife crossings and speed management.
• Targeted interventions on high-risk segments, especially two-lane roads in ecologically sensitive areas, are recommended.