Abstract: To further elucidate the rules and characteristics of traffic accidents that occur in highway tunnels in mountainous regions and improve traffic safety, a case study is conducted on six tunnels of the G65 BaotouMaoming expressway. The time characteristics are examined in four dimensions of year, month, week, and hour. Furthermore, according to the tunnel type, the interval division principles are proposed; moreover, the long-and extraong tunnels are classified as emergency parking belts. The spatial distribution characteristics of the traffic accident rate in the tunnel section are also discussed. The distribution characteristics of accident morphology, vehicle type, and vehicle quantities are analyzed. With 11 potential independent variables, such as the vehicle, the road, and the environment, as the influencing factors and the accident severity as the dependent variable, SMOTE sampling methods are used to overcome unbalanced data. Based on univariate analysis, collinearity test, and binary logistic regression analysis model, the predicted results of balanced and unbalanced data are compared to identify the key factors that affect accident seve rity. The results show that the accident rate of emergency parking belts and their influencing sections is high, which seriously affects the safety of the tunnel. Furthermore, regular days (nonholidays), cloudy and rainy days, entrance/exit sections, oversized vehicle, and curved sections are significant influencing factors involved in serious accidents.

Key words: mountain highway tunnels, traffic accident, emergency parking belt, binary logistic, accidental severity