Modeling spatiotemporal interactions in single-vehicle crash severity by road types

Area of Science:

• Traffic Safety Engineering
• Transportation Science
• Statistical Modeling

Background:

• Spatiotemporal correlations are crucial in single-vehicle (SV) crash severity analysis.
• Interactions between spatiotemporal factors in SV crashes remain underexplored.
• Existing models often fail to capture these complex interactions adequately.

Purpose of the Study:

• To propose and validate a novel spatiotemporal interaction logit (STI-logit) model for SV crash severity regression.
• To compare the performance of the STI-logit model against traditional spatiotemporal and random parameters logit models.
• To investigate the influence of various factors on crash severity across different road types.

Main Methods:

• Development of the spatiotemporal interaction logit (STI-logit) model.
• Utilizing mixture component and Gaussian conditional autoregression (CAR) for spatiotemporal interactions.
• Calibration and comparison with spatiotemporal logit and random parameters logit models.
• Separate modeling for arterial, secondary, and branch roads.

Main Results:

• The STI-logit model significantly outperforms existing models in predicting SV crash severity.
• The STI-logit model with a mixture component demonstrated superior fit compared to Gaussian CAR, irrespective of road type.
• Distracted driving, drunk driving, motorcycles, dark conditions, and fixed object collisions increase serious SV crashes.
• Trucks and pedestrian collisions decrease serious SV crash likelihood.
• Roadside hard barriers showed a significant positive association with serious crashes only on branch roads.

Conclusions:

• The STI-logit model provides a robust framework for analyzing SV crash severity by incorporating spatiotemporal interactions.
• Simultaneously modeling stable and unstable spatiotemporal risk patterns enhances model accuracy.
• Identifying specific contributing factors and their varying influence across road types is vital for targeted safety interventions.