Traffic operation risk assessment in highway tunnels based on risk field modelling and fused trajectory data

Highway tunnel traffic operation risk is difficult to quantify continuously because vehicle motion, speed variation, and car-following behaviour evolve dynamically along the tunnel. To address this issue, this study proposes a traffic operation risk assessment method for highway tunnel car-following scenarios based on fused vehicle trajectory data. Vehicle trajectories were reconstructed from millimetre-wave radar and video data collected in a 1750 m three-lane highway tunnel in Guangdong Province, China. Speed, speed difference, acceleration, and following distance were extracted to construct a car-following risk field model incorporating kinetic characteristics and longitudinal behavioural responses. Fuzzy C-means clustering was adopted to classify continuous risk values into low, medium, and high-risk states. The results indicate significant spatial heterogeneity in tunnel traffic operation risk. High-risk states were concentrated in the entrance section, accounting for 3.6%, whereas no high-risk states were observed in the middle or exit sections. The middle section showed the most stable traffic operation, with low-risk states accounting for 98.7%. In contrast, the exit section exhibited relatively stronger operational fluctuations, reflected by a higher proportion of medium-risk states than the middle section. These findings provide a quantitative basis for tunnel risk identification and proactive traffic safety management.