A Traffic Conflict Prediction Model for Highway Work Zones Considering the Water Hammer Effect and Heterogeneous Behavior

Abstract The convergence of fixed and mobile bottlenecks triggers the water hammer effect in highway work zones, significantly contributing to traffic accidents. To mitigate crash risk, this paper proposes a framework (SSPF) and a microscopic prediction model (BMIM) to anticipate traffic conflict intensity at blackspots. Both the framework and the model account for the water hammer effect and heterogeneous driving behavior, enabling early warnings to highway operators based on micro-level traffic data collected upstream of work zones. Using random forest modeling and statistical analysis, we compare the predictive accuracy of the macro model (MAM), the bottleneck-containing micro model (BMIM), and the bottleneck-absent micro model (BAMIM). Further analyses reveal strong links among traffic conflict intensity, single-vehicle kinematics, bottleneck, and neighbor interaction turbulence, highlighting the adverse safety impact of the water hammer effect. Sensitivity analysis identifies “steep transition” and “high plateau” phases in conflict intensity variation, while interaction analysis identifies speed standard deviation as a key moderating variable. These findings offer actionable insights for timing risk warnings and designing control strategies.