Extra-long tunnels, characterized by sudden changes in lighting and spatial configuration, exert a considerable effect on driving behavior, fuel consumption, and vehicular emissions. Despite this, investigations into the interrelationships among these factors in such environments remain scarce. In this study, driving behavior and instantaneous emissions of CO, CO 2 , and NO x in extra-long expressway tunnels were investigated using a portable emission measurement system integrated with vehicle operational sensors. A light-duty gasoline vehicle was tested across entrance, mid-tunnel, and exit sections of the Qinling and Li Jia He 3# tunnels in China. The experiment was conducted over five consecutive days, with a different designated driver completing one round trip (covering both tunnels) each day. Results showed that the entrance section exhibited frequent acceleration, highest power demand, and peak CO or CO 2 emissions. Emission peaks were closely associated with periods of unstable acceleration and deceleration, particularly at tunnel transition zones. High-emission events for CO and CO 2 predominantly occurred at 60–70 km/h, with slight acceleration (0–0.5 m/s 2 ). Furthermore, the study underscores the potential of machine-learning models for predictive emission analysis. Machine-learning models (CatBoost for CO, Random Forest for CO 2 or NO x ) predicted emissions with test-set R 2 values of 0.637 (CO), 0.557 (CO 2 ), and 0.206 (NO x ), indicating moderate predictive capability for CO and CO 2 but limited performance for NO x under the tested conditions. These results offer empirical support for optimizing tunnel design and traffic management strategies aimed at reducing emissions and enhancing safety, contributing valuable insights toward the development of sustainable transportation infrastructure.
Zhang et al. (2026) studied this question.