A comparison in travel patterns and determinants of user demand between docked and dockless bike-sharing systems using multi-sourced data

Highlights

• The difference in travel characteristics between docked and dockless bike-sharing are revealed, including riding distance, riding time, usage frequency and spatio-temporal patterns.

• Geographically and temporally weighted regression (GTWR) outperforms ordinary least squares (OLS) and geographically weighted regression (GWR), when explaining the spatiotemporal data.

• The relationship between the bike-sharing usage and its determinants (socio-demographic and land use factors) is discussed from both temporal and spatial dimensions.

• Policy implications and suggestions are proposed to improve the performance of docked and dockless bike-sharing systems.

Abstract

The co-existence of traditional docked bike-sharing and emerging dockless systems presents new opportunities for sustainable transportation in cities all over the world, both serving door to door trips and accessing/egressing to/from public transport systems. However, most of the previous studies have separately examined the travel patterns of docked and dockless bike-sharing schemes, whereas the difference in travel patterns and the determinants of user demand for both systems have not been fully understood. To fill this gap, this study firstly compares the travel characteristics, including travel distance, travel time, usage frequency and spatio-temporal travel patterns by exploring the smart card data from a docked bike-sharing scheme and trip origin–destination (OD) data from a dockless bike-sharing scheme in the city of Nanjing, China over the same spatio-temporal dimension. Next, this study examines the influence of the bike-sharing fleets, socio-demographic factors and land use factors on user demand of both bike-sharing systems using multi-sourced data (e.g., trip OD information, smart card, survey, land use information, and housing prices data). To this end, geographically and temporally weighted regression (GTWR) models are built to examine the determinants of user demand over space and time. Comparative analysis shows that dockless bike-sharing systems have a shorter average travel distance and travel time, but a higher use frequency and hourly usage volume compared to docked bike-sharing systems. Trips of docked and dockless bike-sharing on workdays are more frequent than those on weekends, especially during the morning and evening rush hours. Significant differences in the spatial distribution between docked and dockless bike-sharing systems are observed in different city areas. The results of the GTWR models reveal that hourly docked bike-sharing trips and dockless bike-share trips influence each other throughout the week. The density of Entertainment points of interest (POIs) is positively correlated with the usage of dockless bike-sharing, but negatively correlated with docked bike-sharing usage. On the contrary, the proportion of the elderly has a positive association with the usage of docked bike-sharing, but a negative association with the usage of dockless bike-sharing. Finally, policy implications and suggestions are proposed to improve the performance of docked and dockless bike-sharing systems, such as increasing the flexibility of docked bike-sharing, designing and promoting mobile applications (APP) for docked bike-sharing, improving the quality of dockless shared bikes, and implementing dynamic time-based pricing strategies for dockless bike-sharing.

Introduction

As a short-term bike rental service, bike-sharing has become common in many cities around the world during the last decades. It has not only been regarded as an economical, flexible, convenient and sustainable travel mode, but also as a means to mitigate problems like air pollution and traffic congestion, to promote a healthy lifestyle by involving more physical activity benefits and to support multimodal transport connections (Maizlish et al., 2013, Yang et al., 2016). Bike-sharing systems enable smooth door-door transport by itself or by serving as access/egress mode of public transport (Van Mil et al., 2020; Shelat et al., 2018). This will increase the catchment areas of rail transit stations and thereby increasing ridership (Brand et al., 2017). By October 2019, more than 2080 bike-sharing schemes are already in operation and 360 others are under construction in more than 50 countries (Meddin and Demaio, 2019).

Currently, the bike-sharing systems operated worldwide can be divided into two categories: docked bike-sharing and dockless bike-sharing (Liu et al., 2018b). In the docked bike-sharing system, users have to rent bikes from designated docking stations and then return them to the available lockers in docking stations. As a result, the docked bike-sharing system cannot provide door-to-door services. Another challenge in operating a docked bike-sharing system is that the numbers of bikes and docks required at some stations are often insufficient to satisfy the corresponding cycling demand (Szeto and Shui, 2018). Compared with the traditional docked systems, the dockless bike-sharing system is connected to the internet with a mobile phone application (APP) to help users rent dockless bikes (Shaheen et al., 2010). The dockless bike-sharing system allows the users to park the bikes in the physical or geo-fencing designated parking areas (Pal and Zhang, 2017). Without being constrained by docking-station infrastructure, it saves users walking distance between bike stations and origin/destination locations (Cheng and Gao, 2018) and strengthens the seamless connection with public transport (Ai et al., 2018, Shelat et al., 2018). However, dockless bike-sharing could also bring negative societal effects. Because it is often lacking an adequate demand estimation, dockless bike-sharing system often experiences oversupply of bike fleets in high population density areas, which hurts its economic sustainability, occupies urban space resources, harms the urban transport system and causes visual pollution (Du and Cheng, 2018). While in low population density areas, dockless bike-sharing system often experiences low bike utilization levels (Shen et al., 2018).

The joint deployment of traditional docked and emerging dockless bike-sharing systems presents new opportunities for sustainable transportation in cities all over the world. In order to provide users with better services and to help operators enhance the bike fleet reallocation, it is necessary to compare and comprehend the travel characteristics and influential factors between docked and dockless bike-sharing (Gu et al., 2019, Shen et al., 2018). However, most of previous studies have separately examined different aspects of docked and dockless bike-sharing schemes, whereas the investigation of the similarities and differences in travel patterns between the two systems is scarce, due to the difficulty of acquiring historical trip data of both systems over the same spatio-temporal dimension. The aim of this paper is to better understand the difference between the travel patterns of bike-sharing users of the two systems and to examine the influence of bike-sharing fleets, socio-demographic factors and land use factors on user demand. This is achieved by using a state-of-the-arts regression model - geographically and temporally weighted regression (GTWR), based on multi-sourced data (e.g., trip origin–destination (OD) data, smart card, survey, land use information, Gross Domestic Product (GDP) and housing prices data) from the city of Nanjing, China. Particularly, the main contributions of this paper lie in:

• Revealing the difference in travel characteristics, including riding distance, riding time, usage frequency and spatio-temporal usage patterns by mining smart card data from docked bike-sharing and the trip OD data from dockless bike-sharing over the same spatio-temportal dimension;

• Establishing a GTWR model to analyze the influential factors (bike-sharing fleets, socio-demographic, points of interest (POIs), etc.) associating with user demand of docked and dockless bike-sharing.

The remainder of this paper is structured as follows. In Section 2, a literature review of the evolution history, usage patterns and influential factors of both docked and dockless bike-sharing is provided. In Section 3, the study area and dataset are introduced. In Section 4, the basic frameworks of GTWR model and its counterpart models (ordinary least squares model and geographically weighted regression model) used in the study are described. In Section 5, the results of historical data analysis are discussed, and the results of three regression models are compared. Next, the regression coefficients of the GTWR model are analyzed in detail temporally and spatially. The conclusions and suggestions for future research are summarized in the last part of the paper.