Risk Assessment of Merging Maneuvers at Narrow Bridges: Insights from Random Effect Logistic Regression

Abstract

The safety of merging maneuvers at narrow bridges poses significant challenges due to constrained geometry, limited merging spaces, and elevated crash risks. This study examines crash risks using a random effect binary logistic regression model calibrated with field data collected at a narrow bridge on NH 37, India. After incorporating driver interaction variables such as longitudinal acceleration, time-to-collision (TTC) with lead and lag vehicles, available gaps, and relative speeds results indicate that positive longitudinal acceleration increases merging probability (β = 0.066, p < 0.001), whereas for no merging a higher TTC-based risk values were observed. Within 18m from the bridge, non-merging vehicles shows higher risk (0.65) than merging ones (0.056). The model achieves strong predictive performance Area Under Curve (AUC):0.899, precision:0.994, recall:0.936, and an F1-score:0.964 on the validation dataset. Findings highlight the importance of timely merging and suggest improvements in geometric design, signage, and early warning system.