Using Smart Card Fare Payment Data To Analyze Multi-Modal Public Transport Journeys in London

This paper contributes to an emerging literature on the application of smart card fare payment data to public transportation planning. The research objective is to identify and assess complete, multi-modal journeys using “Oyster” smart card fare payment data in London. Three transfer combinations: bus-toUnderground, Underground-to-bus, and bus-to-bus are considered in order to formulate recommendations for maximum elapsed time thresholds to identify transfers between journey stages for each passenger on the London network. Recommended elapsed time thresholds for identifying transfers are: 20 minutes for Underground-to-bus, 35 minutes for bus-to-Underground, and 45 minutes for bus-to-bus, but a range of values that account for variability across the network are assessed. Key findings about bus and Underground travel in London include: an average of 2.3 daily public transportation journeys per passenger, 1.3 journey stages per public transportation journey, and 23 percent of Underground journeys include a transfer to or from a bus. The application of complete journey data to bus network planning is used to illustrate the value of new information that would be available to network planners by using smart card fare payment data. Catherine Seaborn, John Attanucci, Nigel H.M. Wilson 3 INTRODUCTION Transport for London’s (TfL) planning guidelines emphasize better system integration across modes (1). In particular, TfL’s Interchange Plan (2) categorizes 614 interchange facilities into five groups ranging from major central London termini to interchanges of local importance, and then prioritizes them for infrastructure improvement. This prioritization is based primarily on qualitative analysis and the plan states that reliable data on such basic metrics as the number of people transferring at each facility was not available. In addition to key interchange facilities, over 700 intersecting bus routes provide opportunities for on-street transfers in London. TfL currently has well-developed systems for assessing passenger demand between stops on a bus route and on parallel routes, but information on passenger demand between stops on a route and intersecting routes or Underground stations, and between ultimate origins and destinations is less well provided for. The current sources of stop-level passenger demand data are a rolling (every six years) origin-destination survey as well as a survey that records boardings, alightings and loads at 400 key bus stops on a two-year cycle. Thus, bus network planners at TfL utilize intermittent surveys and route-level Electronic Ticketing Machine data (a farebox record at point of entry on the bus), as well as experiential knowledge, to evaluate bus routing and service changes. With the ongoing challenge of developing the bus network to meet the needs of Londoners, smart card data can be used to expand information on passenger demand which might include the following: • Passenger flows between intersecting routes to provide support for direct links that reduce the need for transfers; • Transfer volumes from bus routes to an Underground station to show which routes are the most important means of accessing the station and adjust station design and/or bus routing accordingly; • Comparison of Underground-to-bus transfer times, controlling for scheduled bus frequency at an Underground station, to highlight reliability or crowding problems; • Evidence of multi-modal journeys (e.g., bus-Underground-bus) to support route redesign such as creating direct bus links that reduce the need to transfer and relieve congestion on the Underground; and • Identification of repeated daily individual passenger travel on a route to indicate strong reliance on that service. The goal of this paper is to demonstrate that smart card fare payment data can be of value in improving bus network planning by being representative of passenger demand across the TfL network. Extensive service improvements and an associated 52 percent growth in ridership between 2001 and 2007 are testimony to the high quality of bus services in London; however, this growth has been achieved with only modest advances in the methods and data systems used for network planning (3). Smart cards, such as the “Oyster” card in London, are owned by individuals and generally record the time and place of every transaction the cardholder makes on the public transportation system, for example a bus boarding or Underground station exit. Several types of analyses can be done with smart card data, including estimating origin-destination matrices, measuring passengers’ behavioural reactions to service changes, and evaluating service quality. The key contribution of this paper, however, is to develop a methodology for describing passenger transfer behaviour to, from and within the bus network in London using smart card data to identify travel patterns. The results are compared with survey data on aggregate travel patterns. The widespread adoption of the smart card fare payment system in London and the potential application of resultant data to bus network planning should be of interest to other public transportation agencies implementing smart card fare payment systems. Smart cards have been adopted by approximately 22 public transportation agencies in Europe and more than 30 cities in Asia but studies of the application of the resultant data to bus network planning are only beginning to emerge (4). A few public transportation agencies are integrating smart card data analysis into their daily operations and planning, for example the Seoul Metro Company (4) and the London Underground at TfL (5). However, smart card data has not been used to study how passengers travel across multiple modes in London, Catherine Seaborn, John Attanucci, Nigel H.M. Wilson 4 although some similar analysis has been done in Chicago (6). To the best of our knowledge, this research is the first comprehensive attempt to combine bus and Underground journey stage data derived from smart card fare payment transactions into complete journeys using informed maximum elapsed time assumptions to identify transfers. The availability of complete journey information, albeit approximate, would be an advance in knowledge for network planners in evaluating the costs and benefits of changes to the bus network. Moreover, data derived from Oyster smart cards may be less expensive (data collected for fare payment purposes), more timely (available almost immediately) and more accurate (not subject to survey errors) than data from conventional sources. PRIOR RESEARCH Analysis of automated fare collection data is an emerging theme in public transportation literature. Using the Chicago Transit Authority as an example, Utsunomiya, Attanucci and Wilson (7) discuss the potential usage of, and barriers to, increased data availability after smart card implementation in public transportation agencies, concluding that agencies need to tailor their smart card implementation plan to make the most of the increased data availability it offers and that smart card penetration as a fare payment method is the key to its effective use for the analysis of passenger behavior. Bagchi and White (8) examine three cases of smart card implementation in small bus networks in the United Kingdom. They find that the advantages of smart card data include larger samples than existing data sources and the ability to analyze travel behavior over longer periods, but there are also limitations, including in the case of bus travel in which cards are only validated upon entry to the system (i.e., bus boarding) and the absence of certain types of information such as journey purpose. They conclude that smart card data cannot replace existing survey methods for data collection but may complement them. Additionally, the authors estimate smart card turnover rates and trip rates per card, and infer the proportion of all bus boardings to linked trips (i.e., with bus-bus transfers based on a 30-minute threshold) in each network. In a similar study for a larger city, Hoffman and O’Mahony (9) use a 90minute threshold to link bus journey stages as recorded by magnetic stripe electronic ticketing technology. The highest rate of transfers in this study occurred between 18 and 28 minutes after boarding the first bus. Okamura, Zhang and Akimasa (10) define a transfer as two journey stages that are provided by different operators and occur within 60 minutes at the same location and go on to analyze transfer wait time at major transit hubs. In sum, a range of transfer time assumptions between 30-90 minutes has been used in previous studies for linking bus journey stages to form complete journeys. Trépanier et al. (11), Cui (6), Zhao et al. (12) and Chan (13) all demonstrate how to estimate the destination of individual bus or rail passengers and thus develop full origin-destination matrices using smart cards based on two assumptions: (1) a passenger’s journey stage destination is the first stop of their following journey stage, and (2) at the end of the day, passengers return to the stop where they first boarded. Further to this work, Chu and Chapleau (14) develop methods for enriching smart card data for transit demand modeling including inferring the arrival time of bus runs at the stop level using schedule constraints and linking journey stages based on both location and time constraints. Thus, they avoid the need to make arbitrary transfer time assumptions, but the methodology is complex and computationally