Monitoring Bus Passenger Zonal Origin-Destination Matrices: Development, Validation, and Application Using Data from The Ohio State University Campus Area Bus System
Loading...
Date
2023-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
The Ohio State University
Abstract
Origin-destination (OD) flows indicate where people come from and move to within a transportation system. For bus transit, stop-to-stop OD matrices represent the number of passengers traveling from one bus stop to another for every feasible stop pair on a bus route. Stop-to-stop OD matrices can be large and are route-specific. Their large size can present a challenge in interpreting the matrices. The route-specific nature of the matrices can limit their usefulness in planning for future route changes or interpreting changes over time that result when routes are modified.
As opposed to stop-to-stop OD matrices, zonal OD matrices aggregate passenger flows across bus routes by mapping bus stops into zones. While stop-to-stop matrices are useful in monitoring passenger flows along routes, zonal OD matrices are more fundamental in representing the geography of passenger flows because they do not rely on a specific route and instead focus on movements between geographic areas. Because there are fewer zones than stops, the zonal matrices have smaller dimensions. For these reasons, zonal OD matrices can be easier to use than stop-to-stop OD matrices in representing general demand of transit passengers and in observing patterns and spatial changes over time.
The Ohio State University's (OSU's) Campus Transit Laboratory (CTL) has been estimating stop-to-stop OD matrices from automatic passenger counter (APC) data from Campus Area Bus System (CABS) buses for many years. They deliver these estimated matrices to OSU's Transportation and Traffic Management office (TTM) on a monthly basis for TTM's general monitoring and ongoing planning. Recently, CTL has also begun estimating and delivering monthly zonal OD matrices along with the stop-to-stop matrices. When considering estimated matrices, there will be differences from one month to another. Such differences can be slight, resulting from real but uninteresting variability in passenger flows or from imprecision in the estimates. However, differences can also be large and indicative of important changes in the spatial patterns of passenger flows. Therefore, it would be useful to have an automatic way to indicate when noteworthy changes occur in the matrices. Being able to automatically monitor changes in estimated zonal OD matrices would be of interest to TTM and to any transit agency that receives OD estimates on a regular basis.
In this thesis, a scalar metric was developed to allow comparisons between pairs of OD matrices in order to identify matrices that are similar over time, recurring differences in the matrices, and singular changes in the matrices. The metric was applied to pairs of 240 empirically estimated zonal OD matrices or aggregations of these matrices. The 240 matrices represent flows of passengers using CABS buses during four time-of-day (TOD) periods for each month between 01/2018 and 12/2022. This empirical application allowed an assessment of the metric's ability to detect noteworthy changes among spatial patterns in different zonal OD matrices. The application of the metric to the historical matrices also allowed for investigation and interpretation of similarities and changes in bus passenger flow patterns on OSU's campus over time.
The empirical results indicate that the metric is able to detect important changes in spatial flow patterns as well as periods of similarities in the patterns. Changes were indicated between matrices representing flow patterns in academic year months and matrices representing flow patterns in summer months. The metric was then used to identify groups of months in one year with similar flow patterns. Analysis of these monthly groups over the years showed that some stability in spatial patterns was maintained through time. However, there were large differences between matrices obtained before the impact of the COVID-19 pandemic on the OSU campus ("pre-lockdown" matrices) and matrices obtained during the period when OSU implemented important policy changes in response to the pandemic ("during-lockdown" matrices). Differences between the during-lockdown matrices and "post-lockdown" matrices were also large, while differences among pre-lockdown matrices were generally small. Differences between pre-lockdown and post-lockdown matrices indicated that post-lockdown spatial patterns are closer to pre-lockdown patterns than to during-lockdown patterns. This could reflect that conditions are gradually returning to pre-lockdown conditions. Alternatively, it could indicate a lasting structural change from both pre-lockdown and post-lockdown spatial flow patterns on the OSU campus.
The ability of the metric to represent changes in spatial flow patterns motivates its use for investigating the effects of specific changes in bus service on zonal passenger bus demand. The empirical results also motivate developing an additional measure to automatically identify noteworthy changes in zone pairs when large differences in the overall matrices are determined.
Description
Keywords
Spatial patterns, Zonal matrices, Origin-destination, Scalar metric