Exploring behavioral heterogeneities of metro passenger’s travel plan choice under unplanned service disruption with uncertainty

Highlights

• Uncertain duration indicated by the time interval is integrated into the utility.

• Passengers are divided into two significant classes based on the latent class model.

• Each class has a different perception of the disruption duration’s uncertainty.

• The impact of occurrence period and passenger’s location on behavior are quantified.

Abstract

Understanding metro passenger’s travel plan choice behavior under unplanned service disruptions is vital for transit agencies. It allows capturing the changes in the demands of the passengers, adopting measures aimed at minimizing the impact on the transit system, and ensuring the safety of the interrupted passengers. Different from planned metro service disruptions, unplanned service disruptions cannot be known in advance and have high uncertainty. However, little is known regarding the role of the uncertain duration in the decision-making process and the taste heterogeneity in the behavior under unplanned metro service disruptions. To fill this gap, we first established a time interval in each scenario of a stated preference questionnaire to indicate the uncertain duration and conducted a web-based survey. Based on the survey data collected at Guangzhou, China, we developed an error component latent class model for travel plan choice behavior considering uncertainty and heterogeneity. The model result showed that the population can be classified into two classes, uncertainty pessimists and uncertainty optimists, who have a strong and weak perception of the uncertainty of the disruption duration, respectively, based on socio-demographic and travel characteristic attributes. Concurrently, the perception of uncertainty in both the classes is enhanced during peak hours and when a passenger enters a station. The findings of this study provide more insights into passenger’s travel behavior under unplanned service disruptions with uncertainty. Moreover, they can assist transit agencies in adopting effective management strategies, which, in turn, will aid in improving the service quality for its passengers.

Introduction

Owing to its high capacity and high operating speed, the metro has become the main part of urban transit services. However, worldwide, its high-intensity operations and large-scale network result in frequent occurrence of unplanned service disruptions caused by vehicle breakdowns and infrastructure failures. Some example countries are the US (Kaviti et al., 2018, Rahimi et al., 2019), Canada (Louie et al., 2017, Itani et al., 2019), Australia (Pender et al., 2014, Currie and Muir, 2017), Netherlands (Durand, 2017), France (Adelé et al., 2019), Austria (Sarker et al., 2019), Singapore (Jin et al., 2015), and China (Dai et al., 2016, Sun and Guan, 2016). Particularly, in the top three Chinese cities of Beijing, Guangzhou, and Shanghai, in 2015, the recorded unplanned metro service disruptions were 189, 348, and 143, respectively (Li et al., 2020). Different from its response to the planned service disruptions caused by personnel strikes and maintenance, a transit agency cannot inform passengers of the duration of an unplanned service disruption in advance. Thus, once an unplanned service disruption occurs, the transit agency needs to identify and repair the fault(s) causing the service disruption as well as publish information about the disruption and alternative travel plans at the earliest. Simultaneously, the passengers must choose one of these plans to reach their destination station.

Understanding the behaviors of passengers under such scenarios can enable a transit agency to accurately grasp the change in the passenger travel demand. It can also assist a transit management department to adopt measures to minimize the effects on the urban transportation system and ensure the safety of the passengers. Therefore, travel behavior analysis under unplanned service disruptions has been attracting attention of increasing number of scholars over the past few years. To our best knowledge, these studies were frequently focused on the travel behavior under service disruptions of certain durations. However, the duration of an unplanned service disruption is uncertain because the process of dealing with the fault(s) is nonroutine and nonstandard. Among the information obtained by the passengers, this duration is typically an interval with lower and upper values, instead of a fixed value. However, uncertainty was ignored in previous studies, which may have led to incorrect travel behavior analysis. Thus, the proposed management strategies for unplanned service disruptions do not exactly match actual travel demands, causing secondary congestion of passengers and waste of transportation resources (Saadatseresht et al., 2009, Li et al., 2020). Therefore, this research gap needs to be filled to better understand passenger travel behavior under unplanned service disruptions. On the other hand, the existing research suggest widespread heterogeneity in choice behavior (Hess, 2005); nevertheless, its study in travel behavior analysis under unplanned service disruptions is limited. Particularly, the variation in the weights of travel plan’s attributes of passengers of different classes under different scenes (e.g., the disruption occurrence period and the location of the passenger) has not been studied thoroughly.

To fill the gap, in this study, we conduct travel behavior analysis under unplanned service disruptions considering uncertainty and taste heterogeneity. Specifically, we apply a time interval composed of the minimum disruption duration and the uncertain disruption duration to suggest the unplanned service disruption duration in a questionnaire survey to achieve similarity to the scenario when passengers encounter unplanned service disruptions in reality. Subsequently, the heterogeneity is tested using the latent class model (LCM) for passenger travel plan selection under unplanned service disruptions, which is estimated using the collected questionnaire data. Furthermore, based on the willingness to pay theory, the values of travel time, minimum disruption duration, uncertain disruption duration, and transfer times for the various classes with different occurrence periods and locations of the passenger are calculated and compared. The findings of this study can extend the existing knowledge on passenger travel plan choice behavior under unplanned disruptions. Moreover, they can offer further insights into unplanned service disruption management, which, in turn, will assist in improving the safety and service quality for passengers. The main contributions of this study are as follows:

• Uncertain disruption duration represented by a time interval is integrated into the utility function to fill the gap in the existing research arising from ignoring the uncertainty and assuming the duration as a fixed value.

• The population is categorized into two significant latent classes based on the LCM, which verifies the heterogenous travel plan choice behavior under unplanned service disruptions and aids in the adoption of effective management strategies.

• The impacts of the disruption occurrence period and the location of the passenger on the travel plan choice preferences are analyzed quantitatively using interaction variables integrated into the model.

The remainder of this paper is organized as follows. Section 2 presents an overview of the state-of-the-art studies on travel behavior under unplanned service disruptions. Section 3 describes the web-based questionnaire survey designed for this study and the data collection process conducted at Guangzhou. Section 4 provides the descriptive statistics and the travel plan selection results distribution of groups with different attributes in the choice dataset collected from the survey. Section 5 introduces the error component LCM used to test the heterogeneity in travel plan choice behavior, followed by the specifications of the attributes integrated into the model. Section 6 presents the estimation results of the LCM as well as the discussion and implications. Finally, the conclusions and future works are summarized in Section 7.