Unravel the impact of COVID-19 on the spatio-temporal mobility patterns of microtransit

https://doi.org/10.1016/j.jtrangeo.2021.103226Get rights and content

Abstract

Shared mobility is an essential component of the larger sharing economy. Ride-hailing, bike-sharing, e-scooters, and other types of shared mobility continue to grow worldwide. Among these services is microtransit, a new transport mode that extends transit coverage within a region. Mobile devices enable microtransit services, aggregating riders and using real-time routing algorithms to group customers traveling in similar directions. Meanwhile, the newly emerged coronavirus, COVID-19, has radically reshaped the ridership behavior of all transit services, including microtransit. While existing research evaluates the performance of microtransit pilot programs before the pandemic, there is no information concerning the spatio-temporal pattern of microtransit activities under the impact of COVID-19. The purpose of this paper is to apply eigendecomposition and k-clique percolation methods to uncover the spatio-temporal patterns of microtransit trips. Further, we used these approaches to identify underlying communities using data from a pilot program in Salt Lake City, Utah. The resulting research offers insight into how COVID-19 altered travel behavior. Specifically, eigendecomposition delineated the homogeneity and heterogeneity of travel patterns across temporal dimensions. We identified first mile/last mile trips as a major source of variance in both pre- and post-COVID periods and that transit-dependent users prove to be inelastic despite the threat of COVID-19. The k-clique percolation method detected possible community formations and tracked how these communities evolved during the pandemic. In addition, we systematically analyzed overlapping communities and the network structure around shared nodes by using a clustering coefficient. The workflow developed in this research broadly is generalizable and valuable for understanding the unique spatio-temporal patterns of microtransit. The framework can also help transit agencies with performance evaluation, regional transport strategies, and optimal vehicle dispatching.

Introduction

Sustainable transport systems are crucial for interregional and intraregional mobility, commerce, and the socioeconomic stability of the communities they serve. Public transit is an essential component of modern multimodal transportation systems but suffers from various pressures (Wei et al., 2018; Zhou et al., 2020). For example, increasing operating costs and decreasing ridership continue to stress public transit systems throughout the United States (FTA, 2019). While many public systems continue to struggle, microtransit is emerging as an agile alternative for personal mobility. Microtransit is technology-enabled shared transportation that operates in-between fixed-route transit and ride-hailing. It leverages rider aggregation and routing algorithms to group customers traveling within the service zone in similar directions in real-time. Moreover, microtransit often expects customers to walk a short distance to common pick-up/drop-off locations. Thus, the service is transit-like but more nimble when compared to traditional public transit. As a technology-enabled on-demand service, microtransit shares many similarities with other services such as ride-hailing and paratransit. The platform collects requests from personal devices like smartphones then dynamically dispatches available vehicles to fulfill those requests. However, one substantial difference exists. The design of microtransit allows for integration into the current public transit system. Users often take advantage of microtransit to complete first mile/ last mile connections to the larger transit system.

For similar on-demand services, studies of dispatching algorithms are numerous. For example, there have been many tests of different optimization algorithms and heuristics for ride-sharing (Agatz et al., 2011; Agatz et al., 2012; Aissat and Oulamara, 2014). Agatz et al. (2011) proposed an optimization framework based on a rolling horizon strategy to solve the dynamic ride-share problem. The authors tested this using a simulation environment based on the field data from the Atlanta Regional Commission (Agatz et al., 2011). Agatz et al. (2012) systematically reviewed the issues in ride-sharing and assessed relevant optimization models. Aissat and Oulamara (2014) modeled dynamic ride-sharing with intermediate locations and presented one enumerate algorithm and two heuristics to solve it. They later tested the approaches on real networks consisting of 3.5 million nodes and 8.7 million edges. Chen et al. (2019) designed an agent-based model to simulate dynamic ride-sharing in a multimodal network then tested it on the classic Sioux Falls network.

For microtransit, the global launch of pilot programs during the past five years (Haglund et al., 2019; Westervelt et al., 2018) seeks to provide (and improve) first/last mile connections to fixed transit stops and stations, replace underperforming bus routes, provide coverage in areas without fixed-route service, and extend the hours of operation for existing bus services. Many studies analyze the overall performance of microtransit pilot programs (Haglund et al., 2019; Volinski (2019) or microtransit vehicles (Ongel et al., 2019). For example, Haglund et al. (2019) systematically evaluated the performance of the Kutsuplus pilot in Helsinki, Finland. The evaluation framework used aggregated measures and spatio-temporal metrics, including the average annual number of passengers, annual price per journey, user age class, distribution of hourly departure/arrival trips for analysis. Volinski (2019) provided a case-based review and synthesis of more than 20 transit agencies that had implemented or intended to launch a microtransit service. This review included evaluations of underlying motivations, planning, design, marketing strategies, technology, and performance metrics. Ongel et al. (2019) evaluated the impacts of novel vehicle technologies on vehicle acquisition costs. This evaluation included lifecycle and end-of-life cost estimates for electric microtransit vehicles and conventional buses operating in Singapore.

Meanwhile, a new challenge has emerged for public transit. COVID-19, a novel coronavirus disease, became a global pandemic in, 2020. Nearly 90% of the American adults reported that COVID-19 impacted their personal lives, and 44% of them claimed their lives had changed dramatically (Pew Research Center, 2020). Due to its collective nature, public transit has been hit even harder (Liu et al., 2020; Wilbur et al., 2020; Yi et al., 2021). In New York City, the average subway and commuter rail ridership declined by 80%, and bus ridership dropped by 50% (Gao et al., 2020a, Gao et al., 2020b) in the first week of July 2020 compared to 2019. In Washington DC, subway and bus ridership declined by 90% and 75%, respectively, by the end of March, 2020 (WMATA, 2020) compared to their typical values. In Utah, three major public transit modes - bus, FrontRunner, and TRAX - have witnessed a massive decline in the total ridership upon the pandemic outbreak (Dillman and Posvistak, 2020). The week after the state of emergency was declared, average ridership has declined by 56% compared to the previous week (Dillman and Posvistak, 2020). Similarly, there was a substantial downturn in microtransit use throughout Utah after the COVID-19 outbreak.

Previous studies help deepen our understanding of microtransit and its dynamics. However, there is very little work concerning the spatio-temporal patterns of microtransit trips and the causal factors contributing to these patterns, especially under the impacts of COVID-19. The purpose of this research is to leverage trip data from a microtransit pilot in the State of Utah for developing a methodological framework that unravels the spatio-temporal patterns of microtransit activities in the region. The framework utilizes eigendecomposition to uncover the rhythms and structures of microtransit trips. Using 7-months of microtransit data, we constructed the spatiotemporal patterns of microtransit activities in pre- and post-COVID periods, respectively. Then, we systematically analyze how these patterns deviate from the average pattern in both periods and what possibly caused such variation. We use eigendecomposition to unravel the hidden temporal structures and k-clique percolation theory to explore the potential spatial communities formed in the service region. Also, for both periods, we intend to determine which locations are connected, how strong the connections are, what roles shared nodes (by different communities) play in different network structures, and how patterns evolve as COVID-19 progresses.

This study is important for three reasons. First, because microtransit helps fill gaps between fixed-route systems and ride-hailing, it is essential to understand its underlying spatio-temporal patterns for a community. In short, does the service provide connectivity to the places that people want to go and when they want to get there? Second, the costs of delivering microtransit services can be substantial, and there are no guarantees that the service will attract riders. For example, the now-bankrupt Bridj service in Kansas City served only 1480 riders during its year of operation, with the Kansas City Area Transportation Authority (KCATA) spending $1.5 million to subsidize the service. Considering that the first ten rides were free for users of Bridj, this translated into a subsidy of $1000 per ride (Schmitt, 2018). Thus, there are real financial implications for communities offering microtransit services. Developing a framework that can provide the geospatial intelligence required for improving system performance is crucial for service sustainability. Third, while the influence of COVID-19 on microtransit is easily observable in Utah, there is no analysis of the overall effects. This research focuses on the underlying travel patterns associated with microtransit and their changes during the pandemic. Our findings could help transit agencies understand the decline in microtransit ridership and the relationships between public health crisis and microtransit demand.

We organize the remainder of this paper as follows. Section 2 presents a literature review where we discuss elements of the impacts of COVID-19, spatio-temporal analysis, the application of eigendecomposition, and k-clique percolation methods. Section 3 describes the data used in this study and its pre-processing. Section 4 presents our methodological framework for uncovering the spatio-temporal patterns of microtransit activities. Finally, we offer the results in Section 5 and conclude with a summary of our findings and key contributions of the study.

Section snippets

COVID-19 related analysis

COVID-19 has been transforming current society in various aspects. Recent studies have found that COVID-19 exerts a substantial impact on the global economy (McKibbin and Fernando, 2020), education (Pragholapati, 2020), mental health (Xiong et al., 2020), public transit (Liu et al., 2020; Wilbur et al., 2020), and many other domains. Researchers have also been studying the reasons why these impacts have taken place. Take public transit as an example. Before the pandemic, a significant

Data source

UTA partnered with Via transportation to launch a microtransit pilot program beginning November 2019 in South Salt Lake. Salt Lake County funded the project. This on-demand, shared-ride pilot is designed to expand access to the transit service throughout the service zone, improve mobility for all users, and provide a quality customer experience. UTA conducted this pilot to see whether microtransit provides a valuable and cost-effective service and whether a future deployment of microtransit

Eigendecomposition

To uncover the spatiotemporal trip pattern for the microtransit program, we developed a methodological framework to delineate the variation and the homogeneity/heterogeneity in trips across spatial and temporal dimensions. Specifically, we employ eigendecomposition to achieve this. As detailed previously, eigendecomposition is well suited for this as it is good at uncovering hidden structures of spatiotemporal patterns. Compared to other variation extraction methods, such as factor analysis and

Eigendecomposition

There are 16,522 pre-COVID trips and 12,062 post-COVID trips in the 77 active TAZs used for analysis. These counts translate to an average of 229.5 and 86.2 trips per day, respectively. Such daily usage of microtransit demonstrates a sharp decrease in activities since the outbreak of COVID-19. Fig. 8 presents the overall temporal patterns (sum of each hour) for departure and arrival trips in pre- and post-COVID regimes, respectively. Both regimes show a two-peak distribution, one in the morning

Conclusions

The worldwide expansion of microtransit continues, bringing many opportunities and challenges for service providers. As a new demand-responsive transport mode, it has the advantage of increasing service coverage, flexible routing, and enhancing transit accessibility. With proper design and execution, microtransit can supplement traditional transport methods to help reduce road pressure and create first mile/last mile connections. However, microtransit projects often suffer from poor marketing

Acknowledgments

This article is based upon work partially supported by the National Science Foundation under Grant No. 2051226, and partially supported by the Mountain-Plains Consortium (MPC) of the U.S. Department of Transportation University Transportation Center (MPC-608). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s).

References (51)

  • J. Xiong et al.

    Impact of COVID-19 pandemic on mental health in the general population: a systematic review

    J. Affect. Disord.

    (2020)
  • Y. Xu et al.

    Unravel the landscape and pulses of cycling activities from a dockless bike-sharing system

    Comput. Environ. Urban. Syst.

    (2019)
  • H. Abdi et al.

    Principal component analysis

    WIREs Comp Stat.

    (2010)
  • D. Achlioptas et al.

    Explosive percolation in random networks

    Science

    (2009)
  • C. Aissat et al.

    A priori approach of real-time ride-sharing problem with intermediate meeting locations

    J. Artif. Intellig. Soft Comp. Res.

    (2014)
  • M.J. Alonso-González et al.

    The potential of Demand-responsive transport as a complement to public transport: an assessment framework and an empirical evaluation

    Transp. Res. Rec.

    (2018)
  • A. Barrat et al.

    The architecture of complex weighted networks

    Proc. Natl. Acad. Sci.

    (2004)
  • S.R. Broadbent et al.

    Percolation processes: I. Crystals and mazes

  • I. Derényi et al.

    Clique percolation in random networks

    Phys. Rev. Lett.

    (2005)
  • M. Dillman et al.

    COVID-19 and Public Transportation in Utah. Inquiry of the Public Sort

    (2020)
  • Federal Transit Administration (FTA)

    National Transit Summaries and Trends

  • Federal Transit Administration (FTA)
  • J. Gao et al.

    The effects of the COVID-19 pandemic on transportation systems in new york city and seattle, USA

  • J. Gao et al.

    Initial impacts of COVID-19 on transportation systems: a case study of the US epicenter, the New York metropolitan area

  • S. Goffri et al.

    Multicomponent semiconducting polymer systems with low crystallization-induced percolation threshold

    Nat. Mater.

    (2006)
  • Cited by (0)

    View full text