International Journal of
Environmental Research
and Public Health
Article
A Framework for Assessing Resilience in Urban Mobility:
Incorporating Impact of Ridesharing
Dawei Li 1,2,3 , Yiping Liu 1,2,3, * , Yuchen Song 1,2,3 , Zhenghao Ye 1,2,3 and Dongjie Liu 1,2,3
Citation: Li, D.; Liu, Y.; Song, Y.; Ye,
Z.; Liu, D. A Framework for
Assessing Resilience in Urban
Mobility: Incorporating Impact of
Ridesharing. Int. J. Environ. Res.
Public Health 2022, 19, 10801.
https://doi.org/10.3390/
ijerph191710801
Academic Editors: Paul
B. Tchounwou and
María-del-Mar Alonso-Almeida
Received: 14 July 2022
Accepted: 29 August 2022
Published: 30 August 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Int. J. Environ. Res. Public Health 2022, 19, 10801. https://doi.org/10.3390/ijerph191710801
1 Jiangsu Key Laboratory of Urban ITS, Southeast University, 2 Dongnandaxue Rd, Nanjing 211189, China
2 Jiangsu Province Collaborative Innovation Center of Modern Urban Traffic Technologies,
2 Dongnandaxue Rd, Nanjing 211189, China
3 School of Transportation, Southeast University, 2 Dongnandaxue Rd, Nanjing 211189, China
* Correspondence: 220205021@seu.edu.cn
Abstract: To a certain degree, the resilience of the transportation system expresses the safety of the
transportation system, because it reflects the ability of the system to maintain its function in the face
of disturbance events. In the current research, the assessment of the resilience of urban mobility
is attractive and challenging. Apart from this, the concept of green mobility has been popular in
recent years. As a representative way of shared mobility, the implementation of ridesharing will
affect the level of urban mobility resilience to a certain extent. In this paper, we use a data low-
intensity method to evaluate the urban traffic resilience under the circumstance of restricted car use.
In addition, we incorporate the impact of ridesharing services. The research in this paper can be
regarded as an evaluation framework, which can help policy makers and relevant operators to grasp
the overall resilience characteristics of cities in emergencies, identify weak sectors, and formulate the
best response plan. This method has been successfully applied to two cities in China, demonstrating
its potential for practice. Finally, we also explored the relationship between urban traffic resilience
and the pattern of population distribution. The analysis shows that population density has an impact
on the level of transportation resilience. And the incorporation of ridesharing will bring an obvious
increment in resilience of most areas.
Keywords: resilience; system safety; urban mobility; ridesharing; population density
1. Introduction
Over the years, due to the frequent occurrence of some extreme weather conditions
and the fluctuation of transportation fuel prices, various negative impacts have occurred
on the urban transportation system [1,2], for instance, breakdown of the road network,
increase on the cost of motor vehicle trips, decrease in the capacity of the road link, and
so on. This has seriously affected the safety of transportation system operation and the
negative influence of these events show that the traffic reliability on cars is fragile. Apart
from this, many nations and organizations, such as the European Union [3], and China [4],
are accelerating the process of reducing greenhouse gas emission, which contribute to
the social and economic uncertainty in relation to urban mobility. Moreover, due to its
characteristics of eco-friendliness and convenience, shared mobility has been promoted
in cities. As a representative form of shared mobility, benefiting from the development of
the online hailing platform, ridesharing services have developed rapidly in recent years.
Considering the situation, we apply a concise framework to analyze the resilience of urban
mobility under the circumstance that car trips are restricted, the specific disturbance event
can be various, for instance, an energy crisis, economic unrest. This paper only focuses
on the final influence of these circumstances on transportation system. In this framework,
we incorporated two active modes, including cycling and walking, which would be less
affected by the circumstances, and the impact of ridesharing services. Referring to study
https://www.mdpi.com/journal/ijerph
Int. J. Environ. Res. Public Health 2022, 19, 10801 2 of 20

of Martins et al. [5], we introduced a data low-intensity (which means the data needed
are moderate) method to assess the degree of dependence on cars of one city under the
assumed circumstance, through calculating the proportion of car trips able to be converted
to active modes. We improve the original methodology by considering ridesharing, which
is playing an important role in daily transportation. The main purpose of the method is to
help transportation managers and decision makers assess the system’s ability to maintain
its function when car use is restricted and to provide suggestions for operators about
making ridesharing strategies.
The concept of resilience was first introduced into the ecosystem in the early 1970s [6].
It is defined as the ability of the system to maintain the durability and stability of its rela-
tionship in the face of external disturbances, and the idea of quantifying it is preliminarily
put forward. Its applications later in the transportation field are mostly used as the objective
function of system optimization to evaluate the cost level [7,8]. In reality, the definition of
resilience is not uniform. For example, Fisher proposed more than 70 methods to display
the definition of resilience [9]. However, the definition of transportation system resilience in
existing studies can be summarized as the ability to resist disturbance (robustness, vulnera-
bility) [10,11], and the ability to restore the function after disturbance (recoverability) [12].
In this paper, the concept of resilience is defined as the ability of transportation system
to maintain its function during the disturbance events. This definition can reflect the
robustness and vulnerability in some degree. In our study, we mainly focus on the system’s
function of satisfying people’s travel demand during the event.
In recent decades, with more appeal for greener mobility, many innovative transporta-
tion modes and techniques are promoted in daily life, such as electric vehicles and shared
mobility, and many studies have been conducted on them [13,14]. Benefiting from the
development of communication technologies and computer science, ridesharing, as one
representative types of the shared mobility, has developed very rapidly in recent years.
Ridesharing is defined as travelers with similar origin and destination or partial same path
are aggregated to share the same vehicle so that they can share travel expenses equally or
make a profit. From the perspective of travelers, the adoption of the ridesharing system
can reduce travel costs and parking costs, while from the perspective of traffic systems, this
travel mode can effectively decrease urban traffic load. As early as the late 1980s, studies
showed that ridesharing accounted for a considerable share of the daily travel [15]. In
China, the scale of ridesharing has also been expanded [16]. The promotion of ridesharing
services is conducive to the alleviation of urban traffic congestion [17], the solution of
urban parking space problems [18], and has a positive impact on the environment [19].
With the emergence of online platforms, scholars pay attention to the efficiency of different
ridesharing matching strategies for taxi operators [20].
The purpose of this study is to propose a framework for combining the ridesharing
service with active travel modes in the evaluation of urban traffic resilience. Our work
is a new iteration of an existing study [5]. We assume the circumstance that the car trip
is restricted because of natural or socio-economic events, and we successfully apply the
framework to two distinctive cities in China. In this study, we stress the impact of different
ridesharing service strategies. Apart from converting replaceable car trips into active modes,
the hailing platform can also try different ridesharing matching strategies to decrease the
car trip demand so that the negative influence can be lowered. This method could be
applied with merely OD data. The work can help decision makers have a whole view of
the transportation system during the event so that they can grasp the weakness of urban
mobility and try to formulate an optimal traffic resources allocation scheme. In general, this
can help creating management measures both in strategic and operational layers. Moreover,
we have explored the relationship between the resilience results and the spatial distribution
of urban population through analyzing it specifically in administrative divisions.
Next, we will briefly review the framework of resilience and associated methodolo-
gies to quantify this and the evolution of research in ridesharing. The motivations and
contributions of this paper will also be illustrated.
Int. J. Environ. Res. Public Health 2022, 19, 10801 3 of 20

1.1. Literature Review

Resilience is a definition originated from the area of civil engineering to evaluate the
endurance of external force, which is then applied to the ecosystem. The conceptual frame-
work of resilience was originally developed since 2003 [21], its main principle incorporates
‘reducing the possibility of failure’, ‘reducing the consequences of failure’, and ‘reducing
the time to recover from failure’. Thenceforth, the concept was broadly applied to the field
of transportation, and many scholars presented their own interpretations on the definition
of transportation resilience. Ref. [22], introduced that the four aspects of transportation
resilience are: latitude, resistance, precariousness, and panarchy. Ref. [23], proposed that a
characteristic of resilience is the ability of the system to operate and recover under abnormal
conditions. Ref. [24], defined traffic resilience as the ability of a transportation system to
absorb risk, maintain function, and recover from disruptions.
Due to the frequent occurrence of extreme weather and some other social and economic
unrest, resilience has been a trendy topic in the related transportation research area. Tonn
et al. recently proposed research about the evaluation of transportation infrastructure
resilience under the influence of social policy [25], and this is a tendency to connect resilience
assessment with policy decision. Quite a few studies focus on the relationship between
natural hazards and resilience [26,27]. Apart from these, some scholars focused on the social
and economic level. Mattioli conducted a study about the impact of fuel price fluctuation
on vulnerability [28], Buinevich forecasted the resilience of ITS under cyber-attacks [29].
Many approaches have been proposed to assess the transportation resilience, both
qualitative and quantitative ones. Shekar conducted simulation with sumo to evaluate
the resilience of a simple road network. The evaluation index adopted was the delay
time of the vehicle [30]. Wang used an end-to-end deep learning method to assess the
resilience of Shenzhen’s transportation network under extreme weather conditions [31].
The evaluation index is the integration of capacity retention ability and recovery ability
of transportation network. For resilience of multi-modal transportation system, Baggag
adopted the supra-Laplacian matrix to evaluate the coverage of the urban road network
by coupling different modes in the transportation network [32], and thereafter obtained
an evaluation of the resilience of the entire city. A method is presented that considers the
conversion between active travel and passive travel, and applied OD data to calculate the
proportion of trips which can be converted to active travel as an indicator to evaluate urban
transportation resilience [5]. However, this method only considers the modes of cycling
and walking, and we have boosted the consideration of ridesharing to this method.
Most of the literature on ridesharing focus on its impact on traffic flow or its operational
layers. Some of these studies are based on an approach of constructing a mathematic
model, for example, Ma constructed a stochastic ridesharing equilibrium considering
elastic demand [33], and Di formed a model based on link and node to get a ridesharing
equilibrium [34]. While some other researchers rely on big data, Veve categorized trips in
accordance to regularity and similarity [35]. After modeling different clusters, thereby the
number and location of the service demand can be predicted. Zhong and Sun analyzed
travelers’ behavior when using hailing platform by applying float car data [36].
With regard to the literature related to ridesharing strategies, most of the current
studies focus on the impact of different ridesharing strategies on traffic demand and
travelers’ behavior. Fiedler’s research combines travelers with similar travel trajectories and
limits the waiting time boundary, and measures the impact of this service on urban traffic
demand [37]. Ma formulated the dynamic ridesharing charging policy from the perspective
of space and time and assessed its impact [38]. Ehsani considered the weighted influence
of ridesharing service detour distance and waiting service duration on travelers [39].
Based on the above research, one thing that cannot be ignored is that the mode of
ridesharing is playing an increasingly important role in our transportation system, while
almost no literature has paid attention to its influence on transportation resilience. The data
demand of some methods is quite complicated, which means managers may not be able to
get a preliminary analysis quickly during the emergency events. For this reason, the use of
 almost no literature has paid attention to its influence on transportation resilience. The
data demand of some methods is quite complicated, which means managers may not be
able to get a preliminary analysis quickly during the emergency events. For this reason,
the use of comparative low-intensity data to get a resilience analysis result and the incor-
Int. J. Environ. Res. Public Health 2022,poration
19, 10801 of ridesharing in resilience assessment are necessary to adapt to the requirement
4 of 20
of reality.

1.2. Contributions
comparative low-intensity data to get a resilience analysis result and the incorporation of
ridesharing in resilience
The main purposeassessment are necessary
of this paper is to maketoan
adapt to the
overall requirementofofurban
assessment reality.traffic re-
silience under the scene that motor vehicle use is restricted, and the available data is non-
1.2. Contributions
intensive. This paper proposes a framework which focuses on the impact of ridesharing
The main purpose of this paper is to make an overall assessment of urban traffic
services on urban traffic resilience. In view of the motivations, this paper presents the
resilience under the scene that motor vehicle use is restricted, and the available data is non-
following contributions:
intensive. This paper proposes a framework which focuses on the impact of ridesharing
•services
Incorporating consideration
on urban traffic resilience. of
Inridesharing service in a methodology
view of the motivations, to evaluate
this paper presents the the
resilience
following of urban transportation network, which is innovative in this area;
contributions:
•• Through applying the method
Incorporating consideration to two different
of ridesharing cities,
service in we testify to
a methodology the feasibility
evaluate the of the
work proposed,
resilience of urbanwhich can provide
transportation suggestions
network, for relatedindecision
which is innovative this area;makers and op-
• erators;
Through applying the method to two different cities, we testify the feasibility of
• the work proposed,
Exploring which can
the relationship provideregional
between suggestions for related
resilience anddecision makers
population and
density.
operators;
• The rest of
Exploring thethis paper isbetween
relationship organized as follows.
regional Section
resilience 2 specifically
and population introduces the
density.
methodology of the framework and the way to incorporate ridesharing services into this.
The rest of this paper is organized as follows. Section 2 specifically introduces the
Section 3 illustrates
methodology of the two cases ofand
framework Shenzhen
the wayand Haikou, respectively,
to incorporate ridesharingand makes
services compar-
into
ative analysis3 on
this. Section their behavioral
illustrates two cases ofand spatial and
Shenzhen features. Finally,
Haikou, Sectionand
respectively, 4 concludes
makes the
whole paper and outlooks the future research.
comparative analysis on their behavioral and spatial features. Finally, Section 4 concludes
the whole paper and outlooks the future research.
2. Methodology
2. Methodology
In this section, we give a detailed introduction to our methodology and provide an
In this section, we give a detailed introduction to our methodology and provide an
example to illustrate this. The structure of the method is shown as followed in Figure 1.
example to illustrate this. The structure of the method is shown as followed in Figure 1.

Figure 1. The structure of the method.

The original method evaluated the resilience by calculating the ratio of motorized
mobility in the city which can be converted to cycling and walking when the car is restricted
to use [5]. The definition of resilience in our study is the ability of the transportation
Int. J. Environ. Res. Public Health 2022, 19, 10801 5 of 20

system to maintain its function under specific circumstances. In our method, the negative
influence due to car restriction can be decreased by two ways: ridesharing reduces car trip
demands, while MPD scenarios can transfer car trips. Then we can judge that this method
is reasonable in this paper, because in the scene of limited car travel, the high percentage
of car trips which can be converted to non-motorways means that the transportation
system can meet more individual travel needs in this case. This method establishes various
scenarios by adjusting the maximum potential distance (MPD) for cycling and walking.
The analysis of results explores the relationship between urban transportation resilience
and MPD. Depending on this, in our augmented method, before evaluating the relationship
between resilience and MPD, we first considered the possible impact of a ridesharing
service through adopting various matching strategies. By means of presetting parameters
of the ridesharing service, the total amount of travel is processed first. Afterwards, the
proportion of trips which can be converted to an active mode under specific scenarios is
calculated. This method uses urban OD travel data and divides the transportation analysis
zones of the study area by rasterization to determine clusters with similar trips.

2.1. Quantitative Way for Resilience

The maximum possible distance (MPD) is defined as the farthest travel radius of a
certain travel mode. In this paper, we use the MPD of walking and cycling to construct the
scenarios. It is worth noting that considering the realistic situation, the MPD of walking
should be less than that of riding. The quantitative way of the resilient level is to examine
the proportion of motor vehicle travel distance within the MPD of walking and riding. In
other words, when motor vehicle is limited, what proportion of these kinds of trips can be
satisfied and converted to walking and cycling decides the resilient level of urban mobility,
which are more active modes.
Based on the definition of [5], we complement it and the motor vehicle trips can be
divided into the following categories:
• Persistent—travel distance less than MPD of walking;
• Adaptable—travel distance exceeding MPD of walking but is less than MPD of cycling;
• Transformable—travel distance exceeding MPD of cycling;
• Flexible—travel demand reduced due to ridesharing.
On the basis of the existing research, a ridesharing service can significantly reduce
the amount of urban trips [40]. Consequently, in this paper what is distinctive from [5],
is that we specified the reduced travel demand caused by adopting different ridesharing
strategies as flexible trips. In all categories of motor vehicle travel, persistent, adaptable,
and flexible trips are classified as resilient trips. The formula for calculating the resilient
level is shown as followed:
Persistent trips + Adaptable trips + Flexible trips
Resilient level =
All trips

The resilient level can be explained as a proportion of car trips which can be converted
to active modes or decreased due to the ridesharing service in the circumstance of restricted
car use. This indicator can represent the degree of negative influence a city may experience
in the circumstance, alternatively, the ability of transportation to maintain its function of
satisfying travelers’ travel demand. After conducting analysis of various scenarios, we can
obtain the proportion of trips which can be converted into active modes. We divided the
resilient level into five groups: very low (0–20.0%), low (20.1–40.0%), medium (40.1–60.0%),
high (60.1–80.0%), and very high (80.1–100.0%).

2.2. Division and Encoding of Traffic Analysis Zone (TAZ)

The division of TAZ in this paper adopts the method of rasterization. Rasterizing the
research area to obtain unit square areas, we consider these unit areas as TAZs and then we
encoded each unit area to complete division. However, this kind of division is not fixed.
Instead, the area size of each TAZ is adjusted according to the various ridesharing strategies
Int. J. Environ. Res. Public Health 2022, 19, 10801 6 of 20

for further analysis. After the processing is completed, pieces of individual trips are capable
of being mapped to each single unit area. The data required in this paper can only be
datasets including information of OD. Of note, for the consideration of travel distance
in each trip, this paper has made a simplification, only accounting for the straight-line
distance of each segment, and the MPD value to be adopted in the following part is also
based on such principle.

2.3. Characteristics of Ridesharing Strategies

The main distinction between ridesharing strategies lies in the difference between
similar trips matching criteria. In this place, we defined a concept of similar trips (ST) as:
trips with similar origin, destination, and departure time. Only when trips are all in a group
of ST can they select to adopt ridesharing service. Obviously, to distinguish STs, we need
two parameters, one spatial parameter is used to confirm similar origins and destinations,
and one temporal parameter to confirm similar departure. The spatial parameters are
reflected in the side length of the TAZs we divided. We segmented the time of one day into
groups with the same time interval, and the temporal parameters correspond to different
time intervals. Different trips are classified as ST when their origins and destinations are in
the same TAZ and their departures are in the same time interval.
In addition to these two parameters, the merging coefficient of trips selecting rideshar-
ing service also needs to be considered. We define the merging coefficient as M. M reflects
the willingness of travelers to select the ridesharing service which is equal to the proportion
of STs choosing this service. This coefficient involves travelers’ consideration of ridesharing,
such as delay, subsidy, comfort, and so on. It should be noted that M used in this paper is
not a constant value, but one that can be adjusted and controlled by managers under the
scenario of restricting motor vehicle travel. Considering the vehicle seat restrictions, the
maximum setting of M is 0.4, which means 10 STs can be merged into 4 at least.
In this context, all the ridesharing strategies are determined by the spatial parameters
and temporal parameters. We have given a simple example to illustrate one of the strategies
in Figure 2. Assuming that an area is composed of four TAZs (encoded as 1, 2, 3, 4) and
each TAZ has the same size with a side length of 0.5 km. The OD matrix among the TAZs
is given. Noting that the departure of trips in the OD matrix all in the same time group
Int. J. Environ. Res. Public Health 2022, 19, x FOR PEER REVIEW 7
of 10 min intervals. OD travel matrix considering ridesharing service can be obtained via
processing, as is shown in the left half of the figure.

Figure 2. Adopting ridesharing

Figure service
2. Adopting among TAZ
ridesharing 1,2,3,4.
service among TAZ 1,2,3,4.

2.4. Combination with Scenarios of MPD.

The approach to setting MPD scenarios depends on adjustment of MPD for cyc
and walking. Regarding the adjustment of parameters in scenarios, we adopted
method to increase the MPD value of walking and cycling with incremental step len
of 0.5 km to construct different scenarios. Likewise, for different ridesharing strateg
Int. J. Environ. Res. Public Health 2022, 19, 10801 7 of 20

2.4. Combination with Scenarios of MPD.

The approach to setting MPD scenarios depends on adjustment of MPD for cycling
and walking. Regarding the adjustment of parameters in scenarios, we adopted the
method to increase the MPD value of walking and cycling with incremental step length of
0.5 km to construct different scenarios. Likewise, for different ridesharing strategies, with
consideration of realistic ridesharing service, we adopted the incremental step of 5 min for
the time interval, and 0.5 km for side length.
In this paper, we have adopted a simple way of writing. The ridesharing strategies
have two parameters, we used T and L to represent these two ones. For example, if the
side length is 0.5 km and the interval of time division groups is 5 min, this strategy can be
described as T5L0.5. T0L0 means no strategies are adopted. The labeling method for MPD
scenarios is the same as [5]. We used W and C to represent MPD for walking and cycling.
For instance, W1.5C2 means that in this scenario, the MPD value for walking is 1.5 km and
the MPD value for cycling is 2 km.

2.5. Spatial Distribution of Resilience and Combination with Socioeconomic Statistics

After simulating different scenarios, the results obtained can be mapped to the research
area to get the spatial distribution. As a result, we can see the resilient level of each region,
so as to enable city managers to compare and draw some conclusions. Apart from this, it
can also be combined with some socioeconomic statistics to analyze the relevance among
them. In this paper, we adopted population distribution data to analyze the relationship
with resilient level. Afterwards, the relevant conclusions can provide useful suggestions
to policy and decision makers on the topic of urban transportation management and
development [41]. This analysis will display the most critical areas and relatively weak
areas for improving the overall resilience of the city so that the safety of transportation
system is increased as well.

3. Applications
3.1. The Case Studies
In this paper, we used data from two cities in China: Shenzhen and Haikou as our
analysis samples. Due to the limitations of data acquisition, data used in this paper only
includes taxi trip data, which can reflect the characteristics of urban mobility to a certain
extent. It should be noted that the contribution of this paper is to improve the framework
of resilience assessment. When there is more complete data, the method proposed in this
paper can be directly used for analysis without other changes. We chose these two cities for
demonstration due to their distinctive characteristics of amount of trips, trip distribution,
and city scale.
Shenzhen has one of the most developed road networks in China and has a high
population density. The total area of Shenzhen is about 1997 square kilometers, and its
permanent population is 10.63 million. It is divided into ten administrative districts, and
the population density of each district varies greatly. As is shown in Figure 3, this is a port
city approaching Hong Kong and it is with considerable number of ports and plenty of
floating population China. This city has a well-developed internet industry and financial
industry, which provides a plenty of related jobs.
The characteristics of Haikou are quite different. The total area of Haikou City is about
2145 square kilometers, and the permanent population is about 2.27 million. Haikou has
four administrative districts. As is shown in Figure 4, this city is also a port city. However,
this city is a famous tourist city and its pillar industry is tourism.
The selected Shenzhen data is the itinerary data of taxis collected on 22 October 2013,
which was a Tuesday. The selected Haikou data is the itinerary data of taxis collected on 16
May 2017. These two data sets consist of labels of latitude and longitude of OD and the
departure of each trip. The data sets of the two selected cities are both a representative
working day. Due to the limitation of data conditions, although we only have taxi data, taxi
trips can reflect the urban mobility to a certain extent [35]. The sample size of the Shenzhen
 tion, and city scale.
Shenzhen has one of the most developed road networks in China and has a h
Int. J. Environ. Res. Public Health 2022,population
19, x FOR PEERdensity.
REVIEW The total area of Shenzhen is about 1997 square kilometers, 8 of 21 and
permanent population is 10.63 million. It is divided into ten administrative districts,
Int. J. Environ. Res. Public Health 2022, 19, 10801
the population density of each district varies greatly. As is shown in Figure8 of 3,20this is a p
city
extent.approaching Hongthat
It should be noted Kong and it is with
the contribution considerable
of this numberthe
paper is to improve offramework
ports and plenty
of resilience assessment. When there is more complete data, the
floating population China. This city has a well-developed internet industrymethod proposed in this
and finan
data is 443,815
paper can be and that
directly of Haikou is 72,134.
used fora analysis One
without thing
other we need to supplement here is that,
changes. We chose these two cities
industry, which provides plenty of related jobs.
due to limitations of data acquisition, we cannot make the analysis among datasets with
for demonstration due to their distinctive characteristics of amount of trips, trip distribu-
a relatively
tion, and cityshort
scale.time gap. However, the most significant innovation of this study is to
propose a framework to assess resilience when incorporating the impact of ridesharing
Shenzhen has one of the most developed road networks in China and has a high
strategies, and when more suitable datasets are available, no obvious change will need to be
population density. The total area of Shenzhen is about 1997 square kilometers, and its
done to replace the original ones. This method only needs low-intensity data, and this will
permanent population is 10.63 million. It is divided into ten administrative districts, and
not be difficult to get for most city transportation managers. Apart from this, a considerable
the population density of each district varies greatly. As is shown in Figure 3, this is a port
part of this study is to make comparative analysis between two cities, considering that
city approaching Hong Kong and it is with considerable number of ports and plenty of
there is a magnitude gap between Haikou and Shenzhen in terms of population, economy,
floating population China. This city has a well-developed internet industry and financial
and other social characteristics [42–45], although features of one city may change in several
industry, which provides a plenty of related jobs.
years, this will not obviously affect the comparison results between the two cities.

Figure 3. Region of Shenzhen, background Google Earth.

The characteristics of Haikou are quite different. The total area of Haikou Cit
about 2145 square kilometers, and the permanent population is about 2.27 million. Hai
has four administrative districts. As is shown in Figure 4, this city is also a port city. H
ever, this city is a famous tourist city and its pillar industry is tourism.
Figure 3.
Figure Region of
3. Region of Shenzhen,
Shenzhen, background
background Google
Google Earth.
Earth.

The characteristics of Haikou are quite different. The total area of Haikou City is
about 2145 square kilometers, and the permanent population is about 2.27 million. Haikou
has four administrative districts. As is shown in Figure 4, this city is also a port city. How-
ever, this city is a famous tourist city and its pillar industry is tourism.

Figure 4. Region of Haikou, background Google Earth.

After simple processing, the OD trip information of the two data sets can correspond
to the map. As shown in the Figure 5, we provide an example to rasterize the map and
 encode every unit as TAZ which has a side length of 1.5 km. Apart from this, we have
distributed all the trips on the map, and the distribution of trips and itineraries is obvi-
ously unbalanced, regionalized, and centralized.
For consideration of temporal and spatial parameters cited in Section 2 in these two
cities, we listed a total of 17 ridesharing strategies, which can be seen in Table 1. It is9 as-
Int. J. Environ. Res. Public Health 2022, 19, 10801 of 20
sumed that the M value of has been determined to be 0.5, which reflects a comparative
high willingness to accept ridesharing services. As stated in the existing research [46,47],
both passengers and drivers’ recognition of ridesharing service will be notably affected
encode every unit as TAZ which has a side length of 1.5 km. Apart from this, we have
by waiting time and detour distance. Considering this, we set the maximum TAZ side
distributed all the trips on the map, and the distribution of trips and itineraries is obviously
length as 2 km and the maximum time interval as 20 min.
unbalanced, regionalized, and centralized.

Int. J. Environ. Res. Public Health 2022, 19, x FOR PEER REVIEW 10 of 21

(a)

(b)
Figure
Figure5.5.OD
ODdistribution
distributionmap
mapof
oftwo
twocities
citieswith
withside
sidelength
lengthof
of1.5
1.5km
km(a):
(a):Shenzhen
Shenzhen(b):
(b):Haikou.
Haikou.

Table For
1. List of ridesharingofstrategies.
consideration temporal and spatial parameters cited in Section 2 in these two
cities, we listed a total of 17 ridesharing strategies, which can be seen in Table 1. It is
Time Interval (min)
assumed that the M value of has been determined to be 0.5, which reflects a comparative
Side Length
high willingness of TAZ
to accept (km) services. As0 stated 5in the existing
ridesharing 10 15
research 20
[46,47],
0 1 - -
both passengers and drivers’ recognition of ridesharing service will be notably affected - - by
waiting time and detour 0.5distance. Considering this,- we set2the maximum3 TAZ 4 side length
5
as 2 km and the maximum 1 time interval as 20 min.- 6 7 8 9
1.5 we can calculate the total
For all strategies above, - number10 of trips11 in the 12two cities13after
adopting the corresponding 2 ones. We have presented - the results
14 in Table15 2. Before16 adopting
17
any ridesharing strategy, the total number of one-day trips in Shenzhen is 443,815 and that
in Haikou
For allisstrategies
72,134. above, we can calculate the total number of trips in the two cities
As a resultthe
after adopting of the ridesharing strategy,
corresponding ones. Wesimilar trips in daily
have presented thetrips willinbeTable
results merged, which
2. Before
means the amount of trips will be scaled down. In Table 2, by simulating
adopting any ridesharing strategy, the total number of one-day trips in Shenzhen is strategies of
ridesharing, the total number of
443,815 and that in Haikou is 72,134. trips can be reduced to a large extent. For example, in the
context of L2T20, the number of trips in Shenzhen decreased by about 40% compared with
Table 2. Amount of Shenzhen (Haikou) trips after adopting ridesharing strategies.

Time Interval (min)

Side Length of TAZ (km) 0 5 10 15 20
443,815
Int. J. Environ. Res. Public Health 2022, 19, 10801 10 of 20

L0T0, and that of Haikou can also reach about 35%. It is worth noting that changes in the
number of trips seem to be more sensitive to the spatial parameters of the strategy. For
example, by adopting the strategy from L0.5T5 to L2T5, due to the increase of qualified ST,
the number of trips in Shenzhen and Haikou decreased by 30% and 22%. Correspondingly,
from L0.5T5 to L0.5T20, the number of trips in Shenzhen and Haikou decreased by only
about 3%. This shows that the spatial homogeneity of urban travel is more significant than
the temporal homogeneity.

Table 1. List of ridesharing strategies.

Time Interval (min)

Side Length of TAZ (km) 0 5 10 15 20
0 1 - - - -
0.5 - 2 3 4 5
1 - 6 7 8 9
1.5 - 10 11 12 13
2 - 14 15 16 17

Table 2. Amount of Shenzhen (Haikou) trips after adopting ridesharing strategies.

Time Interval (min)

Side Length
0 5 10 15 20
of TAZ (km)
443,815
0 - - - -
(72,134)
432,980 426,625 421,329 416,734
0.5 -
(65,332) (63,323) (62,659) (61,402)
403,008 382,376 368,574 358,155
1 -
(60,311) (57,499) (56,362) (54,987)
346,105 321,405 307,347 297,639
1.5 -
(54,664) (53,113) (51,453) (48,437)
304,129 281,348 269,219 261,081
2 -
(47,482) (46,994) (45,132) (43,169)

3.2. Resilience Level under Different Ridesharing Strategies and MPD Scenarios
In this part, based on the 17 ridesharing strategies that have been delineated, the
MPD scenarios are further subdivided. Considering the characteristics of the Shenzhen
and Haikou data used in this paper, an increase of 0.5 km is used as the iteration step of
MPD. Regarding the characteristics of urban travel distance, the maximum walking MPD
in Shenzhen is 4 km, and that of cycling is 40 km. A total of 456 scenarios are considered.
The maximum MPD for walking in Haikou is 4 km and that for cycling is 21 km. A total of
228 scenarios are considered. The list of MPD scenarios is presented in Table 3.
After delineating all scenarios of MPD, different ridesharing strategies need to be
added to further analyze the resilience level under these scenarios. Table 4 provides an
example to evaluate resilient level of two cities by adopting the L1T10 strategy. It should
be added that the MPD of riding and walking in Table 4 does not necessarily represent the
value of real travel, but only to reflects the conditions that need to be met when different
levels of resilience are achieved. When making a certain assessment, the MPD value should
be determined in combination with the travel survey of the city. Initially when the MPD of
riding and walking is 0 km, the resilience level of Haikou is 13.79% and that of Shenzhen is
13.84%, which means the proportion of trips that the adoption of L1T10 strategy reduces in
the two cities. With the increase of MPD, the two cities show a similar changing pattern.
The change of resilience can be divided into three stages. When the value of MPD is less
than 3 km, the resilient level increases very rapidly (from C0W0 to C3W2), especially when
the value of MPD is 2 km. When the value of MPD exceeds 3 km and less than 5.5 km, the
Int. J. Environ. Res. Public Health 2022, 19, 10801 11 of 20

growth of the resilience level is relatively gentle, but it is still significant, and the urban
resilience reaches a high level. When the value of MPD exceeds 6 km, especially exceeding
10 km, the resilience level increases very slowly, and the resilience of the city reaches a very
high level. This shows that the trips within the city are mostly concentrated in short- and
medium-distance trips (less than 10 km). When the value of MPD reaches a certain degree,
even if it continues to increase, the effect will not be significant.

Table 3. List of MPD Scenarios.

MPD for Walking (km)

MPD for Cycling (km) 0 0.5 1 1.5 2 2.5 3 3.5 4
0 1 - - - - - - - -
0.5 - 2 - - - - - - -
1 - - 3 - - - - - -
1.5 - - - 4 - - - - -
2 - - - 5 6 - - - -
2.5 - - - 7 8 9 - - -
3 - - - 10 11 12 13 - -
3.5 - - - 14 15 16 17 18 -
21 - - - 223 224 225 226 227 228 *
39 - - - 439 440 441 442 443 444
39.5 - - - 445 446 447 448 449 450
40 - - - 451 452 453 454 455 456
* Last scenario tested for Haikou.

Table 4. Resilience level of various MPD Scenarios.

HaiKou ShenZhen
Scenario MPD (km) Scenario MPD (km)
Cycling Walking Resilience (%) Cycling Walking Resilience (%)
1 0 0 16.80% 1 0 0 13.84%
2 0.5 0.5 19.12% 2 0.5 0.5 16.90%
3 1 1 25.92% 3 1 1 26.91%
4 1.5 1.5 35.77% 4 1.5 1.5 37.83%
5 2 1.5 46.72% 5 2 1.5 46.90%
6 2 2 46.72% 6 2 2 46.90%
7 2.5 1.5 57.34% 7 2.5 1.5 54.19%
8 2.5 2 57.34% 8 2.5 2 54.19%
10 3 1.5 64.23% 10 3 1.5 60.14%
11 3 2 64.23% 11 3 2 60.14%
12 3 2.5 64.23% 12 3 2.5 60.14%
13 3 3 64.23% 13 3 3 60.14%
14 3.5 1.5 71.93% 14 3.5 1.5 64.90%
15 3.5 2 71.93% 15 3.5 2 64.90%
36 5 4 84.28% 36 5 4 75.04%
37 5.5 1.5 85.33% 37 5.5 1.5 77.47%
38 5.5 2 85.33% 38 5.5 2 77.47%
222 20.5 4 98.77% 450 39.5 4 99.91%
223 21 1.5 99.34% 451 40 1.5 99.92%
224 21 2 99.34% 452 40 2 99.92%
225 21 2.5 99.34% 453 40 2.5 99.92%
226 21 3 99.34% 454 40 3 99.92%
227 21 3.5 99.34% 455 40 3.5 99.92%
228 21 4 99.34% 456 40 4 99.92%

In addition, it is worth noting that when the MPD of cycling is less than 6 km, the
resilience level of Haikou is obviously higher than that of Haikou. This shows that Shenzhen
is more affected by medium- and long-distance travel.
 Int. J. Environ. Res. Public Health 2022, 19, 10801 12 of 20

Int. J. Environ. Res. Public Health 2022, 19, x FOR PEER REVIEW 13 of

In order to show the impact of different ridesharing strategies on urban transportation

Int. J. Environ. Res. Public Health 2022, 19, x FOR PEER REVIEW 13 of
resilience, we have compared and analyzed the result by adjusting the MPD for cycling in
Figure 6 (Haikou) and Figure 7 (Shenzhen). The MPD for walking of these two cities is set
to be 1.5 km.

L1T10
(%) (%)

60% L1T20
Resilience

L1T10
60% L0.5T15
L1T20
Resilience

L1T15
L0.5T15
L1.5T15
L1T15
L2T15
L1.5T15
30% L2T15
1.5 4.5 13.5
30% MPD for cycling (km)
1.5 4.5 13.5
MPD for cycling (km)
Figure 6. Comparison of resilience level under different ridesharing strategies, by adjusting th
MPDFigure
for 6. Comparison
cycling of resilience level under different ridesharing strategies, by adjusting the MPD
(Haikou).
for cycling (Haikou).
Figure 6. Comparison of resilience level under different ridesharing strategies, by adjusting th
MPD for cycling (Haikou).

L1T5
(%) (%)

L1T10
60% L1T5
Resilience

L1T20
L1T10
60%
Resilience

L0.5T15
L1T20
L1T15
L0.5T15
L1.5T15
L1T15
L2T15
L1.5T15
30%
1.5 4.5 13.5 L2T15
30% MPD for cycling (km)
1.5 4.5 13.5
Figure 7. Comparison of resilience level under different ridesharing strategies, by adjusting the MPD
MPD for cycling (km)
for cycling
Figure (Shenzhen). of resilience level under different ridesharing strategies, by adjusting th
7. Comparison
MPD for cycling (Shenzhen).
Figure In7. the figure, various
Comparison line segments
of resilience level represent different
under different ridesharingstrategies,
ridesharing strategies. by
In this
adjusting th
place, we selected several representative strategies from all 17 ones. Each node represents
MPDIn forthe
cycling (Shenzhen). line segments represent different ridesharing strategies. In th
figure,
the MPD scenariovarious
with the value for cycling corresponding to the horizontal axis. The
place,
corresponding vertical axisrepresentative
we selected several strategies
is the resilience level from all 17 ones. Each node represen
in the scenario.
In
the MPD the figure,
Thescenario various
with the
entry resilience line segments
levelvalue
of the for represent
cycling
strategy the different
withcorresponding ridesharing strategies.
to the horizontal
lowest performance (L0.5T15) In Th
inaxis. th
place, we selected
Haikou is aboutvertical
corresponding several
32%, andaxis representative
the corresponding
is the resilience strategies
value of the
level from
instrategyall 17 ones. Each node
with the best performance
the scenario. represen
the(L2T15)
MPD scenario
is 50%. Withwith
the the value
increase in for
MPD cycling
for corresponding
cycling,
The entry resilience level of the strategy with the lowest the impact toperformance
the horizontal
of changing ridesharing axis. Thi
(L0.5T15)
corresponding
Haikou is aboutvertical axisthe
32%, and is the resilience level
corresponding in the
value scenario.
of the strategy with the best perfo
mance The entry is
(L2T15) resilience
50%. Withlevel
the of the strategy
increase in MPDwith the lowest
for cycling, the performance (L0.5T15)
impact of changing ride
Haikou is about 32%, and the corresponding value of the strategy with the
haring strategy on resilience becomes weaker. When the value of riding MPD is less tha best perfo
3mance (L2T15)
km, the is 50%.
difference With the
between increase
various in MPDisfor
strategies cycling, the
significant, impact
while of changing
when the MPD ofride
cy
Int. J. Environ. Res. Public Health 2022, 19, 10801 13 of 20

strategy on resilience becomes weaker. When the value of riding MPD is less than 3 km,
the difference between various strategies is significant, while when the MPD of cycling
exceeds 4.5 km, the difference in the resilience level between the different strategies can
even be ignored, and it cannot be affected even if the parameters of ridesharing strategy
are adjusted.
The entry resilience level of the strategy with the lowest performance (L0.5T15) in
Shenzhen is about 35%, and the corresponding value of the strategy with the best perfor-
mance (L2T15) is 55%. This data are obviously higher than the corresponding value of
Haikou. Similar to Haikou, the impact of different strategies on resilience in Shenzhen
tends to be insignificant with the increase of MPD for cycling, but this distinction almost
disappears when it reaches about 6 km, which is greater than that of Haikou. This may be
due to the larger proportion of short-distance trips in Shenzhen.
The average resilience level of Shenzhen under different strategies is higher than that
of Haikou, which shows that ridesharing is more suitable under the travel characteristics of
Shenzhen and has a broader development prospect in Shenzhen. Another observed finding
is that the optimal strategy of Shenzhen and Haikou is (L2T15), while the worst strategy is
(L0.5T15), which further shows that for urban mobility, the key to ridesharing strategy lies
in its spatial parameters. This again verifies the previous conclusion.

3.3. Resilience Level Pattern under Various Merging Factors and Ridesharing Strategies
In this section, this paper will show the changing patterns of the resilience levels of
the two cities under different ridesharing strategies with various M, as shown in Figure 8.
The merging coefficient M reflects, to a certain extent, the willingness of travelers to
accept ridesharing under the scenario of restricting motor vehicle travel, which depends on
government subsidies, the comfort of ridesharing services, waiting time, and so on. In this
paper, three Ms (0.4, 0.6, and 0.9) are selected to represent the three levels of willingness for
ridesharing (high willingness, medium willingness, and low willingness). We set the MPD
for cycling to be 4.5 km and the MPD for walking to be 1.5 km.
As can be seen from Figure 8, under the condition of low willingness for ridesharing,
the resilience level of Haikou is higher than that of Shenzhen. However, with the improve-
ment of willingness, the increment of the resilience level of Shenzhen will be significantly
greater than that of Haikou. Apart from this, the resilience level distribution pattern of
both Shenzhen and Haikou shows that when the willingness changes from medium to
high, the gain of resilient level will be more obvious, compared with the change from
low to medium. The analysis in this section will provide some references for urban traffic
managers to promote ridesharing service.

3.4. Spatial Distribution of Resilience and Population Density

In this section, we explored the relationship between the spatial distribution of pop-
ulation based on administrative divisions and the level of resilience. To ensure legibility,
we mapped the spatial distribution of population of these two cities in Figure 9 to identify
the characteristics.
Shenzhen has ten administrative districts. The population density of Shenzhen is
relatively large, and the density varies greatly among regions. Futian District, which has
the highest population density, has reached the level of 21,000 people per square kilometer.
In comparison, Haikou has a relatively sparse population distribution, and the population
density levels among its four administrative regions are relatively similar. The Longhua
District with the highest population density has only 2300 people per square kilometer.
Figure 10 shows the distribution of trips of each administrative region in Haikou under
the selected four representative ridesharing strategies. The given MPD value of cycling is
3 km and that of walking is 2 km. It can be seen from the figure that with adopting different
strategies, the composition of trips of each region is almost the same, and the changes
of composition are also very similar. This may be because the population distribution of
 finding is that the optimal strategy of Shenzhen and Haikou is (L2T15), w
strategy is (L0.5T15), which further shows that for urban mobility, the key
strategy lies in its spatial parameters. This again verifies the previous con
Int. J. Environ. Res. Public Health 2022, 19, 10801 14 of 20
3.3. Resilience Level Pattern under Various Merging Factors and Ridesharing St
In this section, this paper will show the changing patterns of the res
Haikou
the twoiscities
relatively average
under and sparse,
different so the degree strategies
ridesharing of time and space
withhomogeneity
various M, of as sho
travel in each region is also very similar.

(a)

(b)
Figure
Figure 8.8.Comparison
Comparison of resilience
of resilience levels
levels with variouswith various
M under variousM under various
ridesharing strategiesridesharin
(a).
Shenzhen (b). Haikou.
Shenzhen (b). Haikou.
 3.4. Spatial Distribution of Resilience and Population Density
In this section, we explored the relationship between the spatial distribution of pop-
ulation based on administrative divisions and the level of resilience. To ensure legibility,
we mapped the spatial distribution of population of these two cities in Figure 9 to identify
Int. J. Environ. Res. Public Health 2022, 19, 10801 15 of 20
the characteristics.

Int. J. Environ. Res. Public Health 2022, 19, x FOR PEER REVIEW 16 of 21

Figure 9.
Figure Spatial distribution
9. Spatial distributionofofthe
thepopulation
populationdensity (Left:
density Shenzhen
(Left: Right:
Shenzhen Haikou).
Right: Haikou).

Shenzhen has ten administrative districts. The population density of Shenzhen is rel-
atively large, and the density varies greatly among regions. Futian District, which has the
highest population density, has reached the level of 21,000 people per square kilometer.
In comparison, Haikou has a relatively sparse population distribution, and the population
density levels among its four administrative regions are relatively similar. The Longhua
District with the highest population density has only 2300 people per square kilometer.
Figure 10 shows the distribution of trips of each administrative region in Haikou un-
der the selected four representative ridesharing strategies. The given MPD value of cy-
cling is 3 km and that of walking is 2 km. It can be seen from the figure that with adopting
different strategies, the composition of trips of each region is almost the same, and the
changes of composition are also very similar. This may be because the population distri-
bution of Haikou is relatively average and sparse, so the degree of time and space homo-
geneity of travel in each region is also very similar.

Figure10.
Figure Thedistribution
10.The distributionofoftrips
tripsininHaikou,
Haikou,under
underdifferent
differentridesharing
ridesharingstrategies.
strategies.

Figure 11 shows the impact on the composition of trips of regions in Haikou by

Figure 11 shows the impact on the composition of trips of regions in Haikou by ad-
adjusting the MPD for cycling under the determined ridesharing strategy (L1.5T10). Similar
justing the MPD for cycling under the determined ridesharing strategy (L1.5T10). Similar
to Figure 10, the proportions of resilient and non-resilient trips in different regions are
to Figure 10, the proportions of resilient and non-resilient trips in different regions are
very similar, and the impact of adjusting MPD for cycling is almost the same. This further
very similar, and the impact of adjusting MPD for cycling is almost the same. This further
reflects the homogeneity of characteristics of travel distances in various regions of Haikou.
reflects the homogeneity of characteristics of travel distances in various regions of Haikou.
Figure 12 shows the resilience level of Shenzhen’s administrative divisions under
four ridesharing strategies. It can be observed that in the southern administrative regions,
where the population is relatively concentrated, the proportion of flexible travel and the
level of resilience are significantly higher than other regions. Meanwhile, with the change
of the adopted strategy, the proportion of resilient travel in these areas also increase more
significantly. The findings may be due to the fact that the southern area, as the central area
of Shenzhen, has a high demand for travel. This means that ridesharing services are easier
to find potential target groups in these regions, so appropriate ridesharing strategies are
more effective in improving the resilience of these regions. Some administrative regions
in the north and east have significantly lower levels of resilience and are insensitive to
 Figure 10. The distribution of trips in Haikou, under different ridesharing strategies.

Figure
Int. J. Environ. Res. Public Health 2022, 19, 10801 11 shows the impact on the composition of trips of regions in Haikou by ad-
16 of 20
justing the MPD for cycling under the determined ridesharing strategy (L1.5T10). Similar
to Figure 10, the proportions of resilient and non-resilient trips in different regions are
very similar,
changes anddifferent
under the impactST of adjustingThis
scenarios. MPD for be
may cycling
owingis almost the that
to the fact same. This
the further
density of
reflects the homogeneity of characteristics of travel distances in various regions of Haikou.
trips in these areas is low, and the space and time homogeneity of trips is not significant.

Int. J. Environ. Res. Public Health 2022, 19, x FOR PEER REVIEW 17 of 21

of the adopted strategy, the proportion of resilient travel in these areas also increase more
significantly. The findings may be due to the fact that the southern area, as the central area
of Shenzhen, has a high demand for travel. This means that ridesharing services are easier
to find potential target groups in these regions, so appropriate ridesharing strategies are
more effective in improving the resilience of these regions. Some administrative regions
in the north and east have significantly lower levels of resilience and are insensitive to
changes under different ST scenarios. This may be owing to the fact that the density of
trips
Figurein
Figure these
11.
11. The areas is low,
Thedistribution
distribution ofand
of tripsthe
trips in space and
inHaikou,
Haikou, withtime
with homogeneity
different
different MPD of for
MPD values
values trips
for is not significant.
cycling.
cycling.

Figure 12 shows the resilience level of Shenzhen’s administrative divisions under

four ridesharing strategies. It can be observed that in the southern administrative regions,
where the population is relatively concentrated, the proportion of flexible travel and the
level of resilience are significantly higher than other regions. Meanwhile, with the change

Figure 12. The distribution of trips in Shenzhen, under different ridesharing strategies.

Figure 13
Figure 13 shows
shows the
the distribution and composition
distribution and composition ofof trips
trips in
in different
different regions
regions ofof
Shenzhen by
Shenzhen by adjusting
adjusting the
the MPD
MPD forfor cycling
cycling under
under the
the determined
determined ridesharing strategy
ridesharing strategy
(L1.5T10). It
(L1.5T10). It can
can be
be observed
observed that
that with
with the
the increase
increase in
in MPD
MPD for
for cycling,
cycling, the
the resilience
resilience level
level
of the administrative regions with higher population density increases more
of the administrative regions with higher population density increases more obviously. obviously.
For example, when the MPD for cycling increases from 5 km to 7.5 km, the resilience of
southern administrative region has been greatly improved, reaching a higher level of re-
silience, while the eastern and northern regions with lower population density have not
changed much. This finding shows that the average travel distance in densely populated
central regions is shorter, so the trips in these regions are easier to switch to active mode,
Int. J. Environ. Res. Public Health 2022, 19, 10801 17 of 20

For example, when the MPD for cycling increases from 5 km to 7.5 km, the resilience
of southern administrative region has been greatly improved, reaching a higher level of
resilience, while the eastern and northern regions with lower population density have not
changed much. This finding shows that the average travel distance in densely populated
Int. J. Environ. Res. Public Health 2022, 19, x FORregions
central PEER REVIEW
is shorter, so the trips in these regions are easier to switch to active18mode,
of 21
while the travel distance in sparsely populated regions is generally longer, resulting in a
minor proportion of resilient trips.

Figure
Figure13.
13.The
Thedistribution
distributionofoftrips
tripsininShenzhen,
Shenzhen,with
withdifferent
differentMPD
MPDvalues
valuesfor
forcycling.
cycling.

Therefore, we reach an interesting conclusion. For populous city, dense regions tend
4. Conclusions and Limitations
to have a higher level of resilience in the face of restricted use of motor vehicles. Taking
some This paper proposes
measures for dense aareas
framework
such asfor evaluating
adopting the overall
suitable urban
ridesharing transportation
strategies, re-
increasing
silience in the case that the use of cars may be restricted or threatened. The
cycling facilities can more significantly improve the overall resilience of the city. For non- resilience index
adopted is the proportion
dense regions, of car trips
they are relatively that can be
vulnerable to reduced
the impact or of
converted
events, so into active
it is modes.to
necessary
Afocus
major innovation of this paper is that the impact
on improving its connection with the central area of the city. of ridesharing services is considered
in the proposed framework. By determining similar trips with similar origin, destination
4. Conclusions
and departure and and Limitations
integrating various ridesharing willingness parameters, we can get the
amount of car trips that can
This paper proposes a framework be reduced byfor ridesharing
evaluatingservices,
the overalland urban
finallytransportation
the impact of
ridesharing services on urban transportation resilience. The framework
resilience in the case that the use of cars may be restricted or threatened. The resilience established in this
paper can be integrated with urban economic and social data. In this
index adopted is the proportion of car trips that can be reduced or converted into active paper, the relation-
ship between
modes. A major the resilience
innovation level of urban
of this paperadministrative
is that the impact divisions and the population
of ridesharing services is
distribution
considered of in those is explored.
the proposed framework. By determining similar trips with similar origin,
Numerical
destination andexamples
departureofand twointegrating
cities in China
variousare ridesharing
provided towillingness
show the feasibility
parameters, of the
we
proposed
can get the method.
amountThe analysis
of car trips shows
that canthatbe under
reduced thebycondition
ridesharing thatservices,
the MPD andremains
finally
unchanged,
the impact ridesharing
of ridesharing services can on
services effectively improve the overall
urban transportation resilience
resilience. The level of the
framework
city under the
established in condition
this paperofcan restricting car use,
be integrated withbecause
urban they can greatly
economic reduce
and social theIntotal
data. this
demand forrelationship
paper, the car trips. From the perspective
between the resilienceof operators and policymakers,
level of urban administrativeincreasing
divisions theand
the population
radiation range distribution
of vehicles in ofridesharing
those is explored.
service can effectively increase the amount of
Numerical
trips that meet the examples
conditionsof two cities in China
of carpooling, are provided
thereby improving to the
show the feasibility
contribution of the
of rides-
proposed method. The analysis shows that under the condition
haring services to reducing car trips. However, the loose ridesharing matching strategy that the MPD remains
unchanged,
may lead to theridesharing
decline of services can effectively
ridesharing willingness,improve
whichtheis overall
also be resilience
an important levelfactor
of the
city under
affecting thethe condition
effect of restricting
of carpooling car This
services. use, because
paper alsothey can greatly
analyzes this.reduce
Therefore,the total
the
demand forproposed
framework car trips. in From
this the perspective
paper shows the of potential
operatorsofand policymakers,
providing reference increasing
for policy the
makers to formulate the optimal ridesharing service strategy, which also involves many
other factors such as subsidy setting.
The comparative analysis of the two cities in this paper shows that the resilience level
distribution in Shenzhen, which is a city with a larger and more concentrated population,
Int. J. Environ. Res. Public Health 2022, 19, 10801 18 of 20

radiation range of vehicles in ridesharing service can effectively increase the amount of trips
that meet the conditions of carpooling, thereby improving the contribution of ridesharing
services to reducing car trips. However, the loose ridesharing matching strategy may lead
to the decline of ridesharing willingness, which is also be an important factor affecting
the effect of carpooling services. This paper also analyzes this. Therefore, the framework
proposed in this paper shows the potential of providing reference for policy makers to
formulate the optimal ridesharing service strategy, which also involves many other factors
such as subsidy setting.
The comparative analysis of the two cities in this paper shows that the resilience level
distribution in Shenzhen, which is a city with a larger and more concentrated population,
is more unbalanced. In contrast, the resilience level distribution in a city with a sparse
and average population is more even, such as Haikou. However, due to the various city
sizes, the MPD demand of Shenzhen reaching a high resilience level is obviously higher
than that of Haikou, indicating that long-distance travel accounts for a larger proportion in
Shenzhen. These findings indicate that this framework can provide a valuable reference
for policy makers to formulate policies and allocate traffic resources according to different
regions under specific circumstances cited above.
The framework proposed in this paper takes into account the impact of ridesharing
services by considering the impact of MPD, which is different from existing research. The
supplement makes this framework more universal. The data requirement of this framework
is OD data, which is not difficult to obtain with current easier access to transportation big
data (for example, mobile phone signaling data, RFID data, etc.), which shows the potential
to be widely used in cities of various sizes. In addition, this framework also shows the
potential of analyzing urban travel behavior characteristics in combination with other social
and economic data.
Nevertheless, this paper still has some shortcomings. Firstly, this paper merely used
the OD data of taxis, the future research should obtain more complete urban overall motor
vehicle trips data as far as possible. In addition, in order to simplify the determination of the
merging coefficient, this paper only uses a rough value, which lacks more detailed research.
Therefore, in this paper only time and spatial factors are considered in the consideration
of ridesharing, and more social and demographic factors can be added in future research,
such as in the division of TAZ. Finally, this study focuses on the travelers with similar
origin, destination, and departure, while we do not consider the circumstance that travelers
may only share partial trips in the ridesharing service. Future research can consider this to
enrich the hypothesis of this paper.

Author Contributions: Conceptualization, D.L. (Dawei Li) and Y.L.; methodology, D.L. (Dawei Li)
and Y.L.; validation, Y.L., Y.S. and Z.Y.; formal analysis, Y.L. and Y.S.; investigation, Z.Y.; resources,
D.L. (Dawei Li); writing—original draft preparation, Y.L. and Y.S.; writing—review and editing, Y.L.
and D.L. (Dongjie Liu); supervision, D.L. (Dongjie Liu); funding acquisition, D.L. (Dawei Li). All
authors have read and agreed to the published version of the manuscript.
Funding: This work was supported by the National Key Research and Development Program of
China (No. 2019YFB1600200), the National Natural Science Foundation of China (Nos. 71971056,
51608115), the six talent peaks project in Jiangsu Province (No. XNYQC-003), the Science and
technology Project of Jiangsu Province, China (BZ2020016).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Acknowledgments: The authors are grateful for comments made by anonymous referees.
Conflicts of Interest: The authors declare no conflict of interest.
Int. J. Environ. Res. Public Health 2022, 19, 10801 19 of 20

References
1. Sathiaraj, D.; Punkasem, T.-O.; Wang, F.; Seedah, D.P.K. Data-driven analysis on the effects of extreme weather elements on traffic
volume in Atlanta, GA, USA. Comput. Environ. Urban Syst. 2018, 72, 212–220. [CrossRef]
2. Zhang, T.; Burke, P.J. The effect of fuel prices on traffic flows: Evidence from New South Wales. Transp. Res. Part A Policy Pract.
2020, 141, 502–522. [CrossRef]
3. Förster, H.; Nieese, C.; Siemons, A.; Renders, N.; Dael, S.; Sporer, M.; Tomescu, M. Trends and Projections in Europe 2021; EEA
Report: Copenhagen, Denmark, 2021.
4. Huaxia. China Contributes to Global Carbon Emissions Reduction. Xinhua 2021. Available online: http://www.news.cn/english/
2021-11/09/c_1310300766.htm (accessed on 9 November 2021).
5. Martins, M.C.D.M.; da Silva, A.N.R.; Pinto, N. An indicator-based methodology for assessing resilience in urban mobility. Transp.
Res. Part D Transp. Environ. 2019, 77, 352–363. [CrossRef]
6. Holling, C.S. Resilience and Stability of Ecological Systems. Annu. Rev. Ecol. Syst. 1973, 4, 245–256. [CrossRef]
7. Suwansirikul, C.; Friesz, T.L.; Tobin, R.L. Equilibrium Decomposed Optimization: A Heuristic for the Continuous Equilibrium
Network Design Problem. Transp. Sci. 1987, 21, 254–263. [CrossRef]
8. Watling, D. Stability of the stochastic equilibrium assignment problem: A dynamical systems approach. Transp. Res. Part B
Methodol. 1999, 33, 281–312. [CrossRef]
9. Fisher, L. More than 70 ways to show resilience. Nature 2015, 518, 35. [CrossRef]
10. Xu, X.; Qu, K.; Chen, A.; Yang, C. A new day-to-day dynamic network vulnerability analysis approach with Weibit-based route
adjustment process. Transp. Res. Part E Logist. Transp. Rev. 2021, 153, 102421. [CrossRef]
11. Gu, Y.; Fu, X.; Liu, Z.; Xu, X.; Chen, A. Performance of transportation network under perturbations: Reliability, vulnerability, and
resilience. Transp. Res. Part E Logist. Transp. Rev. 2020, 133, 101809. [CrossRef]
12. Baroud, H.; Barker, K.; Ramirez-Marquez, J.E.; Rocco, C.M. Inherent Costs and Interdependent Impacts of Infrastructure Network
Resilience. Risk Anal. 2014, 35, 642–662. [CrossRef]
13. Sun, D.J.; Zheng, Y.; Duan, R. Energy consumption simulation and economic benefit analysis for urban electric commercial-
vehicles. Transp. Res. Part D Transp. Environ. 2021, 101, 103083. [CrossRef]
14. Ma, J.; Xu, M.; Meng, Q.; Cheng, L. Ridesharing user equilibrium problem under OD-based surge pricing strategy. Transp. Res.
Part B: Methodol. 2020, 134, 1–24. [CrossRef]
15. Teal, R.F. Carpooling: Who, how and why. Transp. Res. Part A Gen. 1987, 21, 203–214. [CrossRef]
16. Cai, H.; Wang, X.; Adriaens, P.; Xu, M. Environmental benefits of taxi ride sharing in Beijing. Energy 2019, 174, 503–508. [CrossRef]
17. Dewan, K.K.; Ahmad, I. Carpooling: A step to reduce congestion (A case study of Delhi). Eng. Lett. 2007, 14, 61–66.
18. Shaheen, S.; Cohen, A.; Chan, N.; Bansal, A. Sharing Strategies: Carsharing, Shared Micromobility (Bikesharing and Scooter Sharing),
Transportation Network Companies, Microtransit, and Other Innovative Mobility Modes; Elsevier Inc.: Amsterdam, The Netherlands, 2019.
19. Javid, R.J.; Nejat, A.; Hayhoe, K. Quantifying the environmental impacts of increasing high occupancy vehicle lanes in the United
States. Transp. Res. Part D Transp. Environ. 2017, 56, 155–174. [CrossRef]
20. Zhang, W.; He, R.; Ma, C.; Gao, M. Research on Taxi Driver Strategy Game Evolution with Carpooling Detour. J. Adv. Transp.
2018, 2018, 2385936. [CrossRef]
21. Bruneau, M.; Chang, S.E.; Eguchi, R.T.; Lee, G.C.; O’Rourke, T.D.; Reinhorn, A.M.; Shinozuka, M.; Tierney, K.; Wallace, W.A.; Von
Winterfeldt, D. A Framework to Quantitatively Assess and Enhance the Seismic Resilience of Communities. Earthq. Spectra 2003,
19, 733–752. [CrossRef]
22. Walker, B.; Holling, C.S.; Carpenter, S.R.; Kinzig, A. Resilience, Adaptability and Transformability in Social-ecological Systems.
Ecol. Soc. 2004, 9, 5. [CrossRef]
23. Murray-tuite, P.M.; Tech, V. 2006 Winter Simulation Conference; Perrone, L.F., Wieland, F.P., Liu, J., Lawson, B.G., Nicol, D.M.,
Fujimoto, R.M., Eds.; IEEE: Piscataway, NJ, USA, 2006; pp. 1398–1405.
24. Wan, C.; Yang, Z.; Zhang, D.; Yan, X.; Fan, S. Resilience in transportation systems: A systematic review and future directions.
Transp. Rev. 2017, 38, 479–498. [CrossRef]
25. Tonn, G.; Reilly, A.; Czajkowski, J.; Ghaedi, H.; Kunreuther, H. US transportation infrastructure resilience: Influences of insurance,
incentives, and public assistance. Transp. Policy 2021, 100, 108–119. [CrossRef]
26. Fang, C.; Chu, Y.; Fu, H.; Fang, Y. On the resilience assessment of complementary transportation networks under natural hazards.
Transp. Res. Part D Transp. Environ. 2022, 109, 103331. [CrossRef]
27. Singh, P.; Amekudzi-Kennedy, A.; Woodall, B.; Joshi, S. Lessons from case studies of flood resilience: Institutions and built
systems. Transp. Res. Interdiscip. Perspect. 2021, 9, 100297. [CrossRef]
28. Mattioli, G.; Philips, I.; Anable, J.; Chatterton, T. Vulnerability to motor fuel price increases: Socio-spatial patterns in England.
J. Transp. Geogr. 2019, 78, 98–114. [CrossRef]
29. Buinevich, M.; Vladyko, A. Forecasting Issues of Wireless Communication Networks’ Cyber Resilience for An Intelligent
Transportation System: An Overview of Cyber Attacks. Information 2019, 10, 27. [CrossRef]
30. Shekar, V.; Fiondella, L.; Chatterjee, S.; Halappanavar, M. Quantitative assessment of transportation network vulnerability with
dynamic traffic simulation methods. In Proceedings of the 2017 IEEE International Symposium on Technologies for Homeland
Security, HST 2017, Greater Boston, MA, USA, 25–26 April 2017. [CrossRef]
Int. J. Environ. Res. Public Health 2022, 19, 10801 20 of 20

31. Wang, H.-W.; Peng, Z.-R.; Wang, D.; Meng, Y.; Wu, T.; Sun, W.; Lu, Q.-C. Evaluation and prediction of transportation resilience
under extreme weather events: A diffusion graph convolutional approach. Transp. Res. Part C Emerg. Technol. 2020, 115, 102619.
[CrossRef]
32. Baggag, A.; Abbar, S.; Zanouda, T.; Srivastava, J. Resilience analytics: Coverage and robustness in multi-modal transportation
networks. EPJ Data Sci. 2018, 7, 14. [CrossRef]
33. Ma, J.; Meng, Q.; Cheng, L.; Liu, Z. General stochastic ridesharing user equilibrium problem with elastic demand. Transp. Res.
Part B Methodol. 2022, 162, 162–194. [CrossRef]
34. Di, X.; Ma, R.; Liu, H.X.; Ban, X. A link-node reformulation of ridesharing user equilibrium with network design. Transp. Res. Part
B Methodol. 2018, 112, 230–255. [CrossRef]
35. Veve, C.; Chiabaut, N. Estimation of the shared mobility demand based on the daily regularity of the urban mobility and the
similarity of individual trips. PLoS ONE 2020, 15, e0238143. [CrossRef]
36. Zhong, S.; Sun, D.J. Taxi hailing choice behavior and economic benefit analysis of emission reduction based on multi-mode travel
big data. In Logic-Driven Traffic Big Data Analytics; Springer: Berlin/Heidelberg, Germany, 2022; pp. 227–254.
37. Fiedler, D.; Čertický, M.; Alonso-Mora, J.; Čáp, M. The impact of ridesharing in mobility-on-demand systems: Simulation case
study in Prague. In Proceedings of the 2018 21st International Conference on Intelligent Transportation Systems (ITSC), Maui,
HA, USA, 4–7 November 2018; pp. 1173–1178.
38. Ma, H.; Fang, F.; Parkes, D.C. Spatio-temporal pricing for ridesharing platforms. Oper. Res. 2022, 70, 1025–1041. [CrossRef]
39. Ehsani, P. Ridesharing Using Adaptive Waiting Time. Ph.D. Thesis, Concordia University, Montréal, QC, Canada, 2020.
40. Li, R.; Liu, Z.; Zhang, R. Studying the benefits of carpooling in an urban area using automatic vehicle identification data.
Transp. Res. Part C Emerg. Technol. 2018, 93, 367–380. [CrossRef]
41. Cariolet, J.-M.; Vuillet, M.; Diab, Y. Mapping urban resilience to disasters—A review. Sustain. Cities Soc. 2019, 51, 101746.
[CrossRef]
42. Haikou Municipal Bureau of Statistics. GDP: Hainan: Haikou. 2021. Available online: https://www.ceicdata.com/en/china/
gross-domestic-product-prefecture-level-city/cn-gdp-hainan-haikou (accessed on 1 June 2022).
43. Haikou Population 2022. World Population Review. 2022. Available online: https://worldpopulationreview.com/world-cities/
haikou-population (accessed on 1 July 2022).
44. Shenzhen Population 2022. World Population Review. 2022. Available online: https://worldpopulationreview.com/world-cities/
shenzhen-population (accessed on 1 July 2022).
45. Shenzhen Municipal Bureau of Statistics. GDP: Guangdong: Shenzhen. 2021. Available online: https://www.ceicdata.com/en/
china/gross-domestic-product-prefecture-level-city/cn-gdp-guangdong-shenzhen (accessed on 1 June 2022).
46. van der Waerden, P.; Lem, A.; Schaefer, W. Investigation of Factors that Stimulate Car Drivers to Change from Car to Carpooling
in City Center Oriented Work Trips. Transp. Res. Procedia 2015, 10, 335–344. [CrossRef]
47. Malodia, S.; Singla, H. A study of carpooling behaviour using a stated preference web survey in selected cities of India. Transp. Plan.
Technol. 2016, 39, 538–550. [CrossRef]