Application of Bayesian Space-Time interaction models for Deer-Vehicle crash hotspot identification

Area of Science:

• Transportation Safety
• Environmental Science
• Statistical Modeling

Background:

• Deer-vehicle crashes (DVCs) pose significant risks to road users and wildlife.
• Accurate identification of DVC hotspots is essential for targeted safety interventions.
• Existing methods may not fully capture the spatiotemporal dynamics of DVC occurrence.

Purpose of the Study:

• To identify and prioritize deer-vehicle crash (DVC) hotspots using five years of crash data.
• To apply and evaluate Bayesian spatiotemporal models for DVC hotspot identification.
• To understand the relationship between land use, environmental factors, and DVC frequency.

Main Methods:

• Utilized five years (2015-2019) of DVC data from Minnesota (MN).
• Employed Bayesian spatiotemporal models, specifically the Type-I spatiotemporal interaction model (Model-2), for hotspot identification.
• Aggregated data by Census Tracts (CTs), incorporating land use and transportation infrastructure variables.

Main Results:

• The Type-I spatiotemporal interaction model demonstrated superior performance in predicting DVC frequency and identifying hotspots.
• Forest, vegetation, and wetland areas were positively associated with DVC frequency.
• Developed land use was negatively associated with DVC frequency, with hotspots concentrated in suburban, mixed-use areas.

Conclusions:

• Deer population management is critical for mitigating DVC risks.
• Identified 65 high-risk CTs requiring immediate safety improvement measures.
• DVC hotspot information is vital for safety engineers and policymakers to implement effective, area-specific countermeasures.