Estimation of transit origin destination matrices using smart card fare data

Abstract

Smart card fare systems have become a valuable source of information for public transport OriginDestination (O-D) estimation, allowing a better understanding of individual travel patterns and improving strategic public transport planning. The O-D matrix is important for transportation analysis, design, and management. It gives indispensable information on the travel demand between two different locations, which is used in many transportation applications from strategic planning to traffic control and management.High quality traffic information is required to improve the estimation of O-D matrix in public transportation. Many transit agencies around the world are now using smart card systems to replace traditional payment methods as a viable payment option. Furthermore, the smart card improves the quality of the data; increases the amount of statistics available; updates data continuously 24 hours per day; reduces boarding and alighting time and the driver's workload; saves costs in data collection and editing, eliminates human errors; and provides new opportunities for innovative and flexible fare structuring.Most Automated Fare Collection (AFC) systems, especially when implemented on buses, record passenger boarding information; however, no alighting information is recorded. The lack of alighting stop details is the result of the installed AFC systems, where passengers are not required to use their cards when alighting. Given the information limitations of many AFC systems, the accuracy level of the estimated O-D matrices is often unknown. In addition, many of the assumptions made in suchmethods are yet to be tested.Although smart card fare data has attracted consideration attention as a rich and comprehensive source of information, it has some limitations which constrain its application. One of these limitations is the missing information of passengers' trip purpose. The main reason for the missing information is due to its original use being the collection of revenue and not data collection. The smart card data analysed in the current research was obtained from Translink, the public transport authority of South East Queensland (SEQ), Australia. An important aspect of this system is that it includes both boarding and alighting times and locations, where a passenger gets on or off a public transport vehicle.The first stage of the current research deals with implementing, validating and improving the tripchaining method and its assumptions to estimate public transport O-D matrices using smart card fare data. Transfer time threshold, transfer walking distance, and the location of the last destination of a passenger in a given day, are the major assumptions investigated here. The available rich and unique ii smart card fare data allowed the current investigation and validation of the effect of different tripchaining method assumptions on the estimated matrices. The estimated O-D results are validated with the actual matrices, which are based on full boarding and alighting information. The errors distributi...

Developing a model for targeted transit advertising using smart card data

“…The buffer zone was considered as 400 m for bus stops, and 800 m for busway stations, ferry stops and train stations (Ryus et al, 2013;Alsger et al, 2018). ArcGIS software package has been used to calculate the walking distance within the network (Alsger, 2017). Since the land-use activities from SEQSTM do not match the land-use activities from HTS, which will be used to extract the temporal distribution, Table 3-2 was created by Alsger (2017) to match the two databases.…”

“…ArcGIS software package has been used to calculate the walking distance within the network (Alsger, 2017). Since the land-use activities from SEQSTM do not match the land-use activities from HTS, which will be used to extract the temporal distribution, Table 3-2 was created by Alsger (2017) to match the two databases. While the HTS dataset specified a variety of the trip purpose types, only five are distinguished in this research: work, education, shopping, recreational and home.…”

“…The average walking pace is considered 80 metres/minute based on the results from Cunningham et al (1982) and Bohannon (1997). There are also different assumptions for the maximum walking distance to board on the public transit system (Daniels and Mulley, 2013;Munizaga et al, 2014;Alsger, 2017). Without loss of generality, this paper considers 600 m around each of the alighting and next boarding stops as the walking distance (MWD), which is used in Equation 17.…”

“…Although, in this section, the spatial and temporal trip similarity measures for all passengers (disregarding having activity similarity or not) in the case study are calculated to analyse the trip similarity measure. The maximum distance is assumed as 600 meters (MD in Equation 19) based on the studies in the travel behaviour of public transit passengers and the walking speed (Daniels and Mulley, 2013;Munizaga et al, 2014;Alsger et al, 2016;Alsger, 2017). Also, the direction of the trips was defined as the angle between the two trips; if the maximum angle (MA in Equation 19) between the two trips is between 0 and 2 degrees, then the two trips will be assumed in the same direction.…”

“…It forecasts the travel demand for an average working day in the SEQ region, which includes major cities such as Brisbane, Gold Coast and Sunshine Coast. The available land-use in SEQSTM includes residential, education, industria l, commercial, hospital/medical, parkland and agricultural.Figure 3-3 shows a part of the landuse activities of the study area along with the public transit stops(Alsger, 2017).…”