CHEMICAL ENGINEERING TRANSACTIONS 

VOL. 89, 2021 

A publication of 

The Italian Association 
of Chemical Engineering 
Online at www.cetjournal.it 

Guest Editors: Jeng Shiun Lim, Nor Alafiza Yunus, Jiří Jaromír Klemeš

Copyright © 2021, AIDIC Servizi S.r.l. 
ISBN 978-88-95608-87-7; ISSN 2283-9216 

Changes in Shared Bicycle Usage by COVID-19 

Juhyeon Kwaka, Haram Oha, Ilho Jeongb, Seungheon Shinb, Donggyun Kub, 

Seungjae Leeb,*

a Department of Transportation Engineering/Department of Smart Cities, University of Seoul, Korea 
b Department of Transportation Engineering, University of Seoul, Korea 

sjlee@uos.ac.kr 

The Seoul Metropolitan Government is promoting and implementing eco-friendly transportation policies covering 

the revitalization of public transportation, establishment of green transportation promotion areas, and the 

expansion of shared-bicycle use. With the prevalence of the COVID-19 pandemic, cities worldwide have been 

paralyzed by lockdowns, the strict implementation of social-distancing policies, and the shutdown of different 

industry sectors. People's movement and traffic volume have decreased significantly. The pandemic has forced 

to people change their travel-related decision-making tendencies. Against this backdrop, this study analysed 

the changes in shared-bicycle traffic in Seoul during the pandemic. This study analyzed the relationship between 

social distancing levels and shared bicycle usage using regression analysis. This study conducted an analysis 

of spatially formed clusters through Moran's I test and hotspot analysis. The results indicate that an important 

correlation between the spread of COVID-19 and shared bicycle use has been found. When social distancing 

levels increased by one level, the number of shared bicycle rental per day increased by about 10,000 times. As 

a result of spatial analysis, changes in shared bicycle usage are spatially forming clusters, and Moran's I test 

showed statistically significant results with a p-value very close to zero. The study results indicate that under 

special circumstances, government policy plays an important role in promoting green transportation. The use of 

green transportation is directly related to falling atmospheric pollution in the city. Such transportation can 

significantly benefit the environment. The government can produce more effective results by considering local 

characteristics to revitalize eco-friendly transportation. 

1. Introduction

The atmospheric environment in cities is continually worsening because of the pollution due to increasing 

numbers of motor vehicles. Various studies have been conducted toward reducing the environmental pollution 

due to vehicular transport. Ku et al. (2020) studied low emission zones (LEZs) in Europe and found that while 

European cities outperform other cities in their application of environmental policies, there was no significant 

improvement in the application of LEZ policies alone. As regards vehicular energy usage, Yoo et al. (2021) 

developed a gradient-based path-exploration algorithm to maximize the energy efficiency of electric vehicles 

(EVs) as an eco-friendly means of transportation. In this regard, shared bicycles in Seoul have emerged as eco-

friendly means of transportation that can reduce air pollution, and consequently, study has been conducted on 

the effectiveness and cost of such facilities. Kim et al. (2020) proposed a cost-estimation method in a feasibility 

study of facilities for eco-friendly transportation (such as walking and cycling), and Ku et al. (2021) proposed a 

methodology to consider health benefits for analysing the effectiveness of bicycle-only road projects. In 2020, 

drastic social changes occurred with the advent of COVID-19. Education has changed so that it can be done 

online (Novak-Pintarič et al, 2020). Owing to the government's lockdown policy, vehicular traffic reduced 

drastically. Giacobbe et al. (2020) studied conformity assessment methods using information and 

communication technologies (ICT) in covid-19 pandemic situations. In Seoul, people adapted to social changes 

resulting from the social-distancing policy. Against this backdrop, in this study, based on the public-bicycle data 

available with the Seoul Metropolitan Government, this study analyzed the factors that affected the use of public 

bicycles after the COVID-19 outbreak and examined the cycle-use characteristics of each region in the city. 

Throughout this study, spatial characteristics that affect the use of public bicycles were analyzed. The purpose 

 DOI: 10.3303/CET2189029 

Paper Received: 22 April 2021; Revised: 16 September 2021; Accepted: 23 November 2021
Please cite this article as: Kwak J., Oh H., Jeong I., Shin S., Ku D., Lee S., 2021, Changes in Shared Bicycle Usage by COVID-19, Chemical 
Engineering Transactions, 89, 169-174  DOI:10.3303/CET2189029

169



of this study is to analyze the spatial characteristics and to find ways to activate public bicycles as an eco-

friendly means of transportation. Figure 1 shows the flow of this study. 

Figure 1: Framework of Study 

2. Literature Review

Many researchers have studied the relationships between the various variables related to bicycle use. The main 

factors affecting bicycle traffic and usage are weather, accessibility to public transportation, and bicycle 

infrastructure. Since the spread of COVID-19, the use of bicycles has significantly increased. In this regard, Hu 

et al. (2021) studied the spatiotemporal changes in shared-bicycle use due to the epidemic and compared these 

data with the counterpart data for other transportation means. Traffic movement was estimated from January to 

July 2020 using mobile-phone data. It was found that traffic declined significantly during COVID-19. The scale 

of transport reduction was not constant. When compared with passenger cars, public transportation, walking, 

and shared bicycles exhibited the smallest decline and fastest recovery rates. The traffic volume increased 

because of a steady rebound. Kim (2018) investigated the impact of external factors such as weather and 

holidays on public-bicycle use in Daejeon. A higher temperature corresponded to increased bicycle use. The 

number of bicycle rentals reduced when the maximum temperature was above 30 °C. There was no significant 

difference in the number of bicycle rentals between weekends and weekdays, although the number of bicycle 

rentals decreased on holidays. Sohrabi et al. (2020) developed generalized extreme value (GEV) count models 

using large digital data that continuously record the status of PBSSs. They considered the time of day, weather, 

deployment environment, infrastructure, and spatiotemporal dependencies, and predicted the departure and 

arrival of the number of bicycles per hour at a given station. In this context, Yang et al. (2019) analysed the use 

of dockless shared bicycles in Nanchang, China, following the opening of new subway lines. They found that 

the opening of new subway lines increased the demand for dockless shared bicycles by 28 %. The 

spatiotemporal travel patterns of bicycle travel distance and traffic have changed over time. Zhao et al. (2015) 

analysed the travel time and trip chain patterns of shared bicycles in Nanjing, China, based on gender and the 

day of the week using smart-card-based data. They found that residential areas formed the origin points, railway 

stations formed the destinations. It was also found that more women used a trip chain of shared bicycles than 

men on weekdays. Xu et al. (2019) reconstructed the visual usage patterns of shared bicycles across different 

locations using four months of GPS data collected from Singapore's major operators of shared bicycles. 

According to their analysis, construction-related environmental indicators such as residential density, 

commercial density, and intersection numbers were found to be correlated with temporal usage patterns, land 

use and the length of bicycle paths were less affected by these indicators. Wang et al. (2021) compared the 

shared-bicycle traffic patterns in New York from September 2019 to September 2020 to analyse changes in 

bicycle-usage patterns. They determined that the number of users gradually increased from April to September 

2020, exceeding the number of users before COVID-19. They also found that business travel decreased, leisure 

travel increased. 

3. Methodology

3.1 Open Data 

The government of Seoul contributes to research by making available data across various fields to the general 

public. Park et al. (2020) investigated the impact of social distancing on the public bicycle-sharing system 

(PBSS) during the COVID-19 outbreak using public data from the Seoul Metropolitan Government. In this study 

as well, open data from the Seoul Metropolitan Government has been used. The data comprised the public-

bicycle-rental history and other relevant information provided by the government. The time range of the data is 

from January to December 2020, and the spatial range is for Seoul. The relationship between the variables was 

analyzed using weather-related data and social-distancing policy levels. Various methodologies were used to 

process these variables in this study. Regression analysis was used to analyze the relationship between the 

170



variables. Moran's I method was applied to verify the spatial correlation, and local characteristics were explored 

via hotspot analysis. Table 1 summarizes the data used in the analysis.  

Table 1: Summary of Data 

Category Variable Unit Information 

Shared-Bicycle Number of Rental per Day Rental/day Mean: 64,767.81 SD: 31,054.87 

COVID-19 Social Distancing Level - Level: 0, 0.5, 1, 1.5, 2, 2.5, 3 

Weather Precipitation mm/h Mean: 4.51 SD: 14.59 

Wind Speed km/h Mean: 8.51 SD: 2.52 

Sunlight Time h Mean: 6.18 SD: 3.97 

Day of the week properties Weekend/Weekday - Weekend: 1 Weekday: 0 

Social Index Number of Office - Spatial Data 

3.2 Regression Analysis 

Regression analysis is a statistical methodology for estimating the relationships between a dependent variable 

and one or more independent variables. Linear regression is a statistical technique that models the relationship 

between a dependent variable and one or more independent variables as linear relationships. The case of one 

independent variable is called simple linear regression; for more than one, the process is called multiple linear 

regression. The basic model for multiple linear regression can be expressed as Eq(1) for each observation. 

𝑌𝑖 = 𝛽0 + 𝛽1𝑋𝑖1 + 𝛽2𝑋𝑖2 + ⋯ + 𝛽𝑝𝑋𝑖𝑝 + 𝜖𝑖  , (𝑖 = 1, … , 𝑛) (1) 

In the formula Eq (1), variables consider n observations of one dependent variable and p independent variables. 

Yi is the ith observation of the dependent variable, Xij is ith observation of the jth independent variable, j = 1, 2, ..., 

p. The values βj represent parameters to be estimated, and ϵi is the ith independent identically distributed normal 

error. 

Feng et al. (2017) combined historical usage patterns with weather data to predict the demand for rental bicycles 

available as per the Capital Bikeshare program in Washington, D.C. Demand forecasting for conventional 

bicycles was conducted using a multi-linear regression model, which included a random forest model and a 

GBM (Generalized Boosted Regression Models), along with the variables of time and season. The multiple 

regression model was found to significantly improve the prediction accuracy. Noland et al. (2016) applied 

Bayesian regression models to investigate the impact of shared-bicycle station utilization on New York's bike 

infrastructure, population and employment, land use, and user types. Shared-bicycle stations in crowded 

subway stations and places with many bicycle infrastructure facilities are highly utilized. They found that in 

densely populated, high-employee-number areas, there were more rentals. 

3.3 Spatial Analysis 

Moran's I test is a statistical methodology that verifies the spatial autocorrelation of any variable and can be 

expressed as Eq(2): 

𝐼 =
𝑛×(∑ ∑ 𝑤𝑖𝑗

𝑛
𝑗=1 ∙(𝑥𝑖−�̅�)(𝑥𝑗

𝑛
𝑖=1 −�̅�))

(∑ ∑ 𝑤𝑖𝑗
𝑛
𝑗=1 )×(∑ (𝑥𝑖−�̅�)

2)𝑛𝑖=1
𝑛
𝑖=1

  (2) 

where n denotes the number of observation points represented by positions i and j, xi and xj the observations at 

positions i and j, �̅� the mean of x, and wij the matrix of weights represented based on the spatial proximity of 

positions i and j. The I value usually ranges between -1 and 1, representing negative autocorrelation for negative 

numbers and positive autocorrelation for positive numbers. For I = 0, it has been shown that arbitrary spatial 

patterns can exist (Das et al., 2017). 

Hotspot analysis is a technique for testing the cluster trends of neighbouring regions within a certain range, and 

the method can be used to analyse regions with high and low values of certain variables. Parameter Gi* can be 

represented as Eq(3): 

𝐺𝑖
∗ =

∑ 𝑤𝑖,𝑗
𝑛
𝑗=1 𝑥𝑗 − �̅� ∑ 𝑤𝑖,𝑗

𝑛
𝑗=1

𝑆√
𝑛 ∑ 𝑤𝑖,𝑗

2𝑛
𝑗=1 − (∑ 𝑤𝑖,𝑗

𝑛
𝑗=1 )

2

𝑛 − 1

 
(3) 

where xj denotes the property value at position j, wi,j the spatial weight between positions i and j, n the total 

number of properties, and �̅�  and S are defined as per Eq(4) and Eq(5). Gi* does not require additional 

computation to obtain the z-score. 

171



�̅� =
∑ 𝑥𝑗

𝑛
𝑗=1

𝑛
 (4) 

𝑆 =
∑ 𝑤𝑖,𝑗

𝑛
𝑗=1 𝑥𝑗 − �̅� ∑ 𝑤𝑖,𝑗

𝑛
𝑗=1

√
∑ 𝑥𝑗

2𝑛
𝑗=1

𝑛
− (�̅�)2

 
(5) 

Positive large values of Gi* correspond to a larger “hotspot” region, on the other hand, negative small Gi* values 

correspond to smaller “coldspot” regions (Zhang and Tripathi, 2018).  

Researchers in various fields are applying these methods to conduct spatial analysis. In this context, Yuan et 

al. (2017) identified spatial clusters and spatial outliers using Moran's I technique to determine the distribution 

patterns of rare materials in London soil. Concha et al. (2017) used Moran's I technique to determine the spatial 

autocorrelation of pH and Cr in natural water regions. Zhang and Tripathi (2018) used methods such as Global 

Moran's I and Getis-Ord Gi* statistics to determine PM2.5 hotspots in eastern Thailand and identify spatial 

correlations of lung cancer cases, mortality, and PM2.5 regions. 

4. Results  

4.1 Regression Analysis 

In this study, ‘R Studio’ software was used for regression analysis. Simple regression analysis was performed 

by setting the number of public-bicycle rentals as the dependent variable and the level of social distancing as 

the independent variable. The results of the simple regression analysis showed that whenever the level of social 

distancing policy increases by one, the number of public-bicycle rentals increases by a factor of 10,140, which 

positively affects the environment. The P-value was also shown to be statistically significant. Table 2 shows the 

results of simple regression analysis. 

Table 2: Result of Simple Regression Analysis 

 Estimate Std.Error t value Pr(>|t|) 

Intercept 51,566 2,639 19.542 <2e-16 *** 

Level of Social Distancing policy 10,140 1,641 6.179 1.73e-09 *** 

 

Bicycle rental and use are affected by various variables such as weather and infrastructure. Multiple regression 

analysis was performed by adding the related variables. The data were analysed by applying stepwise 

techniques that selected only high explanatory variables. Precipitation was considered to have a constant effect 

on bicycle use beyond a certain threshold. So, precipitation above the threshold is reflected in the same effect. 

The analysis results showed that the social-distancing level, precipitation level, and average wind speed were 

significant variables, and even when other variables were added, the social-distancing level still acted as a 

significant variable. Table 3 shows the results of multiple regression analysis. 

Table 3: Results of Multiple Regression Analysis 

 Estimate Std.Error t value Pr(>|t|) 

Intercept 67,880.5 5,210.0 13.029 <2e-16 

Social Distancing Level 9,976.9 1,518.3 6.571 1.74e-10 

Precipitation -30,540.5 4,184.7 -7.298 1.86e-12 

Wind Speed -1,252.7 584.4 -2.143 0.0327 

4.2 Spatial Analysis 

In this study, ‘ArcGIS’ software was used for spatial analysis. By applying Moran's I test, spatial autocorrelation 

in increasing shared bicycle usage in 2020 has been checked. This can identify whether changes in the use of 

shared bicycles form a spatial cluster. Given a z-score of 12.2645 and p-value of 0, there is a <1 % likelihood 

that this clustered pattern could be the result of random. There is a spatial correlation and spatially clustered 

pattern in the use of shared bicycles. Figure 2 show the results of the Moran's I test.  

172



 

Figure 2: Result of Moran’s I Test 

Local characteristics were analyzed through hotspot analysis. In particular, the use of shared bicycles has rarely 

increased in Gangnam-gu and Jung-gu, but it has increased significantly in Yeouido and other western regions. 

Gangnam-gu and Jung-gu are known to contain concentrated work areas with a particularly large number of 

workers and businesses. When the same analysis was conducted based on the number of businesses, the 

'hotspot' of the number of offices was distributed similarly to the 'coldspot' of shared bicycle use. Figure 3 shows 

the results of hotspot analysis. Yeouido Park, which lies in the western part of Seoul, has a large population and 

allows outdoor leisure activities. The Han River flows through this area, which showed a particular increase in 

the use of public bicycles. In areas with a large number of businesses, the use of public bicycles did not increase 

significantly since the COVID-19 outbreak. This result indicates that social changes and government policies do 

not significantly affect the use of public bicycles in work districts. 

  

Figure 3: Results of hotspot analysis (a) Difference in public-bicycle rental and (b) Number of offices 

5. Conclusions 

According to this analysis, the level of social distancing policy acted as a statistically significant variable affecting 

the use of shared bicycles in Seoul. The results of the regression show that the average number of shared 

bicycle rental per day increases by 9,977 as the social distancing level increases by one step. The change in 

public-bicycle usage in 2020 (when compared with 2019) exhibited a spatially clustered pattern. It is suggesting 

the existence of local characteristics. The unusual social situation relating to the pandemic and government 

policies influenced the use of public bicycles, and this use varied depending on the spatial characteristics of 

each region. The effect of policies on bicycle use was found to be greater in areas with many businesses and 

in areas with large populations that allowed leisure activities. These results suggest that government policies 

based on social situations can significantly impact the choice of transportation. Active government intervention 

can significantly contribute to the revitalization of a city's atmospheric environment and promote eco-friendly 

transportation. The government's establishment of policies considering local characteristics to promote eco-

friendly transportation can yield significant environmental benefits. When the government considers the regional 

(a) (b) 

173



characteristics of a city and suggests policies appropriate for the spatial characteristics of each region, eco-

friendly transportation in the city can become more effective, leading to the development of eco-friendly cities. 

Acknowledgments 

This work was supported by the Ministry of Education of the Republic of Korea and the National Research 

Foundation of Korea (2019K1A4A7A03112460). 

References 

Castro C. C., Portacio T. G., Mendoza D., Pajaro C. M., Herrera R. F., Lambis-Miranda H. A., 2017, Spatial 

analysis about structural patterns of fate of chromium (VI) in relation to ph at natural sea water bodies, 

Chemical Engineering Transactions, 57, 391-396. 

Das M., Ghosh S. K., 2017, Measuring Moran's I in a cost-efficient manner to describe a land-cover change 

pattern in large-scale remote sensing imagery, IEEE Journal of Selected Topics in Applied Earth 

Observations and Remote Sensing, 10(6), 2631-2639. 

Feng Y., Wang S., 2017, A forecast for bicycle rental demand based on random forests and multiple linear 

regression, In 2017 IEEE/ACIS 16th International Conference on Computer and Information Science (ICIS), 

101-105. 

Giacobbe F., Bemporad E., 2020, Conformity Assessment of Pressure Equipment during Coronavirus Disease 

Emergency: Risks and Opportunities, Chemical Engineering Transactions, 82, 55-60.  

Hu S., Xiong C., Liu Z., Zhang L., 2021, Examining spatiotemporal changing patterns of bike-sharing usage 

during COVID-19 pandemic, Journal of transport geography, 91, 102997. 

Kim S., Na Y., Na S., Ku D., Lee S., 2020, Cost Estimation Methods in Feasibility Studies on Eco-Friendly 

Transportation Facilities, Chemical Engineering Transactions, 78, 247-252. 

Kim K., 2018, Investigation on the effects of weather and calendar events on bike-sharing according to the trip 

patterns of bike rentals of stations, Journal of transport geography, 66, 309-320. 

Ku D., Bencekri M., Kim J., Lee S., Lee S., 2020, Review of European Low Emission Zone Policy, Chemical 

Engineering Transactions, 78, 241-246. 

Ku D., Kwak J., Na S., Lee S., Lee S., 2021, Impact Assessment on Cycle Superhighway Schemes, Chemical 

Engineering Transactions, 83, 181-186. 

Lee J. W., Lee W. J., 2007, A study on the characteristics of disparity between autonomous districts in Seoul, 

Journal of Korea Planning Association, 42, 81-94. 

Noland R. B., Smart M. J., Guo Z., 2016, Bikeshare trip generation in New York city, Transportation Research 

Part A: Policy and Practice, 94, 164-181. 

Novak-Pintarič Z., Kravanja Z., 2020, The Impact of the COVID-19 Pandemic in 2020 on the Quality of STEM 

Higher Education, Chemical engineering transactions, 81, 1315-1320. 

Sohrabi S., Paleti R., Balan L., Cetin M., 2020, Real-time prediction of public bike sharing system demand using 

generalized extreme value count model, Transportation Research Part A: Policy and Practice, 133, 325-336. 

Seoul Institute, 2018, Regional gap by sector in Seoul (5) Businesses and workers, Seoul Institute 

<www.si.re.kr/node/60810> accessed 20.07.2021. 

Wang H., Noland R., 2021, Changes in the pattern of bikeshare usage due to the covid-19 pandemic, Findings, 

18728. 

Xu Y., Chen D., Zhang X., Tu W., Chen Y., Shen Y., Ratti C., 2019, Unravel the landscape and pulses of cycling 

activities from a dockless bike-sharing system, Computers, Environment and Urban Systems, 75, 184-203. 

Yang Y., Heppenstall A., Turner A., Comber A., 2019, A spatiotemporal and graph-based analysis of dockless 

bike sharing patterns to understand urban flows over the last mile, Computers, Environment and Urban 

Systems, 77, 101361. 

Yoo N., Ku D., Choi M., Lee S., 2021, An Eco-friendly Path Guidance Algorithm for Electric Vehicle, Chemical 

Engineering Transactions, 83, 217-222. 

Yuan Y., Cave M., Zhang C., 2018, Using Local Moran's I to identify contamination hotspots of rare earth 

elements in urban soils of London, Applied geochemistry, 88, 167-178. 

Zhang H., Tripathi N. K., 2018, Geospatial hot spot analysis of lung cancer patients correlated to fine particulate 

matter (PM2. 5) and industrial wind in Eastern Thailand, Journal of Cleaner Production, 170, 407-424. 

Zhao J., Wang J., Deng W., 2015, Exploring bikesharing travel time and trip chain by gender and day of the 

week, Transportation Research Part C: Emerging Technologies, 58, 251-264. 

174