Systemwide variations and factors affecting mixture transit travel time distributions

Understanding transit service reliability is essential for agencies to improve their operations and passenger experiences. Transit travel times that follow mixture distributions would create an additional layer of uncertainty when studying transit reliability. This paper tries to identify segments where transit travel times follow mixture distributions at different analysis levels, namely stop pair level, route timepoint level, and service pattern level. We then identify potential factors related to them. Hartigans’ Dip Test is applied to archived transit vehicle location data from Montreal to explore the presence of mixture distributions. The results contain mixture distributions at three analysis levels, and the proportion of mixture distributions varies temporally and spatially. Then we test several classification models to identify the potential factors that affect transit travel time distributions, where we found demand variations, traffic lights, service frequency, and segment lengths have a larger effect on the results. The findings will help transit planners to later pinpoint the issues causing transit travel time variations on each segment, then create strategies to reduce the transit travel time variations thus improving the reliability of our transit system.