Virtual Capacity Expansion of Stations in Bikesharing System: Potential Role of Single Station-Based Trips ^†

Abstract

Bikeshare systems usually relocate bikes to respond to a mismatch between demand and bike supply, imposing substantial costs to operators despite the effort to encourage users to participate in voluntary rebalancing. This study initiates a search for a new strategy that can involve single station-based (SSB) riders and consider their bikes as the reserve of the current bike balance, resulting in the virtual expansion of station capacity. Thus, the behaviors of bike riders related to SSB trips are compared to investigate the potential applications. The results from analyzing the data of Citi Bike in New York City indicate that 13.4% of total trips were SSB, and the average trips per origin and destination (OD) pair was 2.6 times higher. Also, distinctive characteristics such as mean trip time regarding user groups and bike types were statistically significant within numerous OD pairs, implying the need for separate policies for both groups. Based on the analysis, stations with the highest expected benefit are identified.

1. Introduction

Bikeshare systems (BSSs) have been iconic urban mobility services in many global cities since they were introduced as an alternative to transportation modes depending on internal combustion engine vehicles. Meanwhile, it is natural that users of a BSS fall into two heterogenous groups: residents who conduct daily travels, and visitors who temporarily stay. We can guess their behavioral differences would be substantial due to the distinct circumstances that lead them to use bikeshares. One interesting point is the proportion of users who check out and return bikes at the same station. It is expected that short-term customers will be the majority of them, as they are more likely to ride bikes to look around nearby. These “single station-based (SSB)” trips can play an important role in BSS station capacity management, functioning as an additional storage of bikes. This study, therefore, demonstrates data analyses that can verify the benefit of considering SSBs as one of the alternatives to manage dock capacity. Annual members and casual users are separately grouped to take into account their behavioral differences.

2. Relevant Literature and Research Gap

Bikeshare users choose between annual membership and a casual pass according to attributes such as their length of stay, trip purposes, or “willingness-to-cycle”, affecting discrepancies in trip behaviors. Thus, many studies have shown the heterogeneity in BSS user groups and distinguished riders by the passes they held. While several studies included them as factors influencing the result [1], some addressed direct comparisons of two user types as the main research objective to derive meaningful insights [2,3].

BSS station capacity, usually equivalent to the number of docks, affects the convenience of users [4]. However, excessive capacity may lead to an increasing need for rebalancing bikes since the station retains more bikes to be relocated to other stations unless it is one of the popular pickup points. Therefore, planners designate the location and specification of bike stations by expected demand distribution at the planning stage [5]. Although a rebalancing strategy based on portable stations that can function as independent stations or be appended to existing ones was proposed [6], operators need to pay attention to the relocation of portable stations and become reluctant to actively apply capacity management to their system.

Instead of modifying the number of docks, it may be possible to “virtually” keep bikes at stations by incorporating bike fleets for SSB trips as the “reserve”. Stations with many SSB trips can have the potential to attract more demand by providing larger bike fleets if the number of bikes to be returned is appropriately predicted. In this context, this study can provide a background for the development of a BSS station capacity management scheme that ensures simple implementation and modification without physical adjustment.

3. Methodology

3.1. Preprocessing of Citi Bike Historical Usage Dataset

Monthly usage data with abnormal records and daily data associated with the precipitation history from NOAA Online Weather Data (NOWData) were deleted. There remained four bike-user groups since there are two pass types and two bike types: members with classic bikes (MC), members with e-bikes (ME), casual users with classic bikes (CC), and casual users with e-bikes (CE). First, pass type is the only clue that can differentiate the customer type between casual users and annual members, assuming that they relate with them. Second, different bike types affect the trip duration, one of the core factors compared in this study.

3.2. Single Station-Based Trip Identification

The objective of this part is to verify if the composition of bike-user types in SSB trips is differentiated from that in trips between different stations regarding both the amount and proportion of trips. Among several possible reasons to return to the point where they checked bikes out, riding a bike as a recreational activity such as physical exercise or jaunts may also be a primary purpose. Then, it can be expected that casual users may perform this kind of trip pattern more frequently than annual members. Although members would take more recreational trips in total, the proportion may be higher with casual users. The contingency table approach can be adequately applied to this question [7].

4. Results and Discussion

Table 1. Number of trips per trip types with non-zero trips.

Trip Type	Annual Members		Casual Users		Total
	Classic Bikes	Electric Bikes	Classic Bikes	Electric Bikes
SSB trips (1306 pairs)	224,239 (43.7%)	109,164 (21.3%)	128,934 (25.1%)	50,429 (9.8%)	512,766 (100.0%)
Avg. trips per OD pair	171.7	83.6	98.7	38.6	392.6
Normal trips (21,622 pairs)	1,902,012 (57.5%)	551,973 (16.7%)	617,423 (18.7%)	238,598 (7.2%)	3,310,006 (100.0%)
Avg. trips per OD pair	88.0	25.5	28.6	11.0	153.1
Total	2,126,251	661,137	746,357	289,027	3,822,772

shows the composition of trips. Only OD pairs with more than six trips were chosen to conduct the two-sample Kolmogorov–Smirnov (K-S) test, a non-parametric statistical test based on empirically cumulated density function, which requires at least six samples [8]. Figure 1 indicates the rejection rate of tests per paired type in different trip patterns.

The higher value means a larger discrepancy. First, trip duration distributions of classic and electric bike trips by casual users become similar if their trip pattern is an SSB OD pair. Second, for users who rode the same bike type, their trip duration distributions were more similar if their trips were not SSB. This result implies that casual users have higher chances to perceive riding the bike itself as a recreational activity and are less attentive to reducing their trip duration. From the results, annual members and casual users seem to show distinct bike usage behaviors in terms of spatiotemporal distribution and potential purpose of bike usage. This heterogeneity should be considered when designing a voluntary bike rebalancing program for casual users which has never been implemented in reality.

The most promising benefit of the proposed approach is clear: capacity expansion without additional massive capital investment and rebalancing effort. The system can induce SSB riders to bring their bikes back to the stations where they checked them out. A future research direction can aim to quantify the impact of the proposed virtual station capacity management. The system can add a function that asks whether a user would return the bike to the origin station. Collecting this information in advance can support a more sophisticated estimation of the number of available bikes. Moreover, additional surveys may reveal the relationship between compensation and participation rate, which should differ by user group.