Understanding bus passenger crowding through origin destination inference

Abstract

Comfort is an important aspect of the transit passenger experience. Crowding can significantly decrease passenger comfort and disrupt service delivery, causing passenger travel times to increase and even resulting in passengers being unable to board an arriving vehicle. Reducing crowding is especially important to encourage ridership growth in the part of the system most attractive to customers. However, due to high marginal costs of manual data collection, crowding has not been extensively analyzed. With the advent of automatically collected data systems, it is now possible to gain a more nuanced understanding on how passengers experience crowding as well as monitor conditions as ridership increases. This thesis explores the use of passenger origin-destination inference to measure passenger crowding on buses using the Massachusetts Bay Transportation Authority (MBTA) bus network as a case study. There are three primary components of this research: vehicle trip level origin destination interchange (ODX) scaling; development of passenger centric crowding metrics, and crowding source contribution estimation. The trip level scaling process enables the reliable estimation of passenger loads (accounting for those passengers not using smart fare media) for approximately 90% of MBTA bus trips. Comparisons of ODX and Automatic Passenger Counter (APC) load estimates show that while there is some inherent variability in the ODX derived estimates, many vehicle trips have similar estimates. These ODX derived load and passenger flow estimates were used to create passenger centric crowding metrics that consider many aspects of the passenger experience. Results showed that the majority of crowding occurs on high frequency routes during the peak periods as a result of building schedules around average peak loads and slow travel speeds due to traffic congestion. Next, using a classification tree methodology, the relative contribution that different potential crowding sources have on creating crowding situations was estimated for each route/direction/30-minute-time-period combination. While there were variations between routes and time periods, most of the crowding observed appears to be derived from fixed schedules not able to account for day-to-day fluctuations in demand or service reliability problems that result in uneven headways causing loads on successive trips to vary widely. The research concludes with a review of crowding intervention/mitigation strategies including which strategies are more effective for each crowding source.