Content uploaded by Liu Yang
Author content
All content in this area was uploaded by Liu Yang on Dec 10, 2017
Content may be subject to copyright.
Case Study
Urban Form and Travel Patterns by Commuters:
Comparative Case Study of Wuhan and Xi’an, China
Liu Yang1; Yuanqing Wang2; Qiang Bai3; and Sunsheng Han4
Abstract: This study explores the relationship between monocentric and polycentric urban forms and commuting patterns in Chinese cities
with dense populations and rapid motorization and urbanization, with Wuhan (three subcenters separated by rivers since the city’s formation)
and Xi’an (a monocentric city) as case cities. The analysis uses statistical methods and logit models based on 1,194 and 1,501 surveyed
households in Wuhan and Xi’an, respectively. Compared with Xi’an and other polycentric cities, Wuhan has shorter commuting distances in
the outer areas (3.9 km), smaller car mode share (16.6%), larger nonmotorized mode share (36.8%), more commutes inside the subcenters
(91.3%), more public transit use when commuting among the subcenters (67.3%), and less commuting CO2emissions. The reason lies in its
polycentric structure with strong industries in the subcenters and the limited number of and capacity for traffic corridors among the sub-
centers. These findings can inform sustainable transportation developments in Chinese cities and city cluster areas and can provide empirical
evidence and reference values for other global cities. DOI: 10.1061/(ASCE)UP.1943-5444.0000417.© 2017 American Society of Civil
Engineers.
Author keywords: Polycentric and monocentric urban forms; Travel patterns and behavior; Transportation and urban planning; Wuhan;
Xi’an; China.
Introduction
There has been considerable interest in the relationship between
polycentrism and travel behavior and transport emissions. Some
researchers found that polycentric urban structures bring about
shorter commutes, more public transit use, and lower transport
CO2emissions (Knaap et al. 2016;Grunfelder et al. 2015;Yin
et al. 2015;Alqhatani et al. 2014;Alpkokin et al. 2005;Veneri
2010;Cirilli and Veneri 2014;Gordon et al. 1991). However, some
studies show contrary results, namely that polycentric urban struc-
tures lead to increasing commuting distance, more dependence on
cars, and less reduction of transport CO2emissions (Burgalassi and
Luzzati 2015;Lee and Lee 2014;Aguilera 2005;Schwanen et al.
2001,2003;Vandersmissen et al. 2003;Aguilera and Mignot 2004;
Cervero 1995;Adolphson 2009;Cervero and Wu 1998;Giuliano
and Small 1993;Prevedouros and Schofer 1991). As such, there
have not been consistent findings about whether polycentricity
is associated with more ecofriendly travel patterns or not. Further-
more, existing studies of the relationship between polycentrism and
travel behaviors and transport emissions are mainly conducted in
the cities of United States and Europe, but they are still unknown in
Chinese cities with higher densities, larger urban sizes, and heavy
traffic. The urban densities of Chinese provincial cities (more than
approximately 10,000 people=km2) are higher than those of devel-
oped Western cities on average, and they become much higher in
the central urban areas (Ding 2004). The urban areas of Chinese
cities are also on average much larger than those of European cities.
Chinese cities are now experiencing a rapid process of economic
growth, urbanization, and motorization, and thus there has been an
increasing amount of travel demands and more intense traffic in
urban areas, especially during peak hours on workdays. These char-
acteristics of Chinese cities make them different from developed
Western cities and thus they deserve scholarly attention. Therefore
this paper explores the commuting patterns and commuting CO2
emissions in a polycentric city of China and compares these
characteristics with a monocentric Chinese city and other global
polycentric cities. This comparative case study examines typical
polycentric and monocentric Chinese cities, Wuhan and Xi’an, re-
spectively. Wuhan is a midland central city in China and Xi’an is
a western central city in China. Like many other cities in China,
Wuhan and Xi’an have experienced fast economic growth and rapid
motorization and urbanization, as well as the development of
three ring roads spreading from the city center to the outer regions.
However, Wuhan is characterized by a polycentric urban form, with
three towns separated by the Yangtze and Han rivers since the city’s
formation. The traffic capacities of its bridges are limited and car
restrictions and tolling policies exist on the bridges. In addition,
many lakes are scattered inside the urban areas of Wuhan’s three
towns. By contrast, the city of Xi’an has been associated with a
monocentric urban form with rapid sprawling (tandabing) through-
out its history.
This study provides an initial analysis of the relationship
between the monocentric/polycentric urban structures and the
transport outcomes in rapidly growing Chinese cities with high
densities, large urban sizes, and intensive traffic compared with de-
veloped Western cities. It supplements the existing knowledge
about urban forms and associated travel patterns. The findings
and implications of this study will help sustainable transportation
1Lecturer, Dept. of Urban and Rural Planning, College of Urban and
Environmental Sciences, Northwest Univ., Xi’an, Shaanxi 710127,
P.R. China.
2Professor, Dept. of Traffic Engineering, School of Highway, Chang’an
Univ., P.O. Box 487, Middle Section of South 2nd Ring Rd., Xi’an,
Shaanxi 710064, P.R. China (corresponding author). E-mail: wyq21@
vip.sina.com
3Associate Professor, Dept. of Traffic Engineering, School of Highway,
Chang’an Univ., Xi’an, Shaanxi 710064, P.R. China.
4Professor, Faculty of Architecture, Building and Planning, Univ. of
Melbourne, Parkville, VIC 3010, Australia
Note. This manuscript was submitted on December 22, 2016; approved
on August 3, 2017; published online on December 8, 2017. Discussion
period open until May 8, 2018; separate discussions must be submitted
for individual papers. This paper is part of the Journal of Urban Planning
and Development, © ASCE, ISSN 0733-9488.
© ASCE 05017014-1 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
and urban development in Chinese cities and city cluster areas, and
provide empirical evidence and reference value for other cities
around the world.
Definition and Measure of Polycentricity and Two
Case City Models
The term polycentricity basically refers to the plurality of urban
centers in a given territory (Burger and Meijers 2012;Parr 2004).
Kloosterman and Musterd (2001) pointed out that polycentricity
refers to the multinodal development of any human activity.
Typically, population or employment distribution is considered
in this regard. In addition, the concept of polycentricity is scale-
dependent, and it may be understood differently when measured
at different scales (Vasanen 2012;Davoudi 2008). Traditionally,
the concept of polycentricity has been applied at the intraurban
scale (Vasanen 2012). Furthermore, the definition of polycentricity
has both morphological and functional approaches. In strict mor-
phological terms, the concept of polycentricity refers to several ad-
jacent centers that are located in the same urban system (Vasanen
2012). Some studies emphasized that functional linkages between
the nodes in an urban system are also required in order to deem it
polycentric. Furthermore, empirical research on functional poly-
centricity has typically included the measuring of flows between
the centers (Green 2007;Burger et al. 2011;Vasanen 2012).
Following the existing research, this study defines polycentricity
as the multinodal development of human activities, typically con-
sidered to be population and employment aggregated at an intra-
urban scale with flows among the nodes.
For centricity measurement, many studies used density-based
approaches and identified limits of employment density and total
employment of the census tracts or traffic analysis zones (Knaap
et al. 2016;Cervero and Wu 1998;Giuliano and Small 1991).
Another approach for measuring subcenters is based on flow data,
such as commuting flows, which identify the threshold of entry
flows of commuters (Bourne 1989;Burns et al. 2001). This paper,
following the aforementioned studies, defines subcenters as com-
prising a contiguous set of traffic analysis zones with at least 15%
of the total urban traffic flows attracted to the subcenters.
Fig. 1models the two case cities, Xi’an and Wuhan. The city
model of Xi’an has some similarities to the pattern of concentric
zones proposed by Burgess (1925), but the difference lies in the
existence of industry development zones on the transportation axes
outside the urban core areas. The attracted traffic trips inside the
urban core areas of Xi’an constitute approximately 49% of the total
trips, and in other industry development zones, the proportion of
the attracted traffic trips are less than 15%. Wuhan’s city model
shows that Wuhan has three central business districts (CBDs) in
the initial stage of urban development, which is a result of separa-
tion by two rivers and no bridge construction before the mid-
twentieth century. Each of the three towns of Wuhan forms an
independent urban system, and the attracted traffic trips in the three
towns of Wuhan constitute approximately 18–47% of the total trips.
Wuhan’s urban growth radiates from each of the three CBDs to the
outer areas, and thus this study considers Wuhan’s urban growth as
wedge-shaped urban growth.
Related Work
In North America, Knaap et al. (2016) found that in the Baltimore–
Washington region, with 23 economic centers, the polycentric
urban structure promoted public transit ridership and suggested
promoting polycentric employment development and balancing
jobs and housing within a center’s public transit commute shed.
Gordon et al. (1991) maintained the colocation hypothesis, which
refers to firms and households periodically readjusting spatially to
achieve balanced average commuting distances and durations.
Their study results were based on 20 of the largest metropolitan
areas in the United States from 1980 to 1985 and suggested that
polycentric or dispersed metropolitan structures are especially fa-
vorable to short commutes. Lee and Lee (2014) found that in the
125 largest urbanized areas in the United States, polycentric struc-
tures had only a moderate impact on reducing transportation CO2
emissions. On the one hand, polycentric structures shorten com-
muting distances, but on the other hand, they can make serving
urban activities by public transportation more difficult. The reason
for this may lie in the need to provide public transportation routes
and facilities in each subcenter, which causes more dispersed public
transportation routes, more investments in public transportation,
and lower efficiency of public transportation services compared
with a monocentric urban pattern. Cervero and Wu (1998) found
that in the San Francisco Bay Area, with 22 employment centers, as
with many large U.S. metropolitan areas and contrary to the colo-
cation hypothesis, there was a substantial increase in the average
commuting vehicle kilometers travelled during the decade of rapid
suburban employment growth that began in 1980. Giuliano and
Small (1993) found that in the polycentric urban structure of
Los Angeles there was also an imbalance between the residential
and working areas. Vandersmissen et al. (2003) found that in the
Québec metropolitan area from 1977 to 1996, the shift from a
monocentric structure to a dispersed city form was responsible
for increasing commuting times.
In the cities of Europe, Grunfelder et al. (2015) found that in two
Danish city regions in 1982 and 2002, interurban commuting trips
increased whereas intraurban commuting trips decreased. Further-
more, commuting distances were shortest in the polycentric region.
Alpkokin et al. (2005) found that from 1985 to 1997 in Istanbul,
Turkey, polycentric centers assisted in reducing commuting travel
times and maintained the trip share of public transport, and the pub-
lic transit share attracted to the subcenters was approximately 60%.
Research involving 82 Italian metropolitan areas showed that a
higher degree of polycentricity can lead to lower commuting times
and per capita CO2emissions (Veneri 2010), and the study by
Cirilli and Veneri (2014) of 111 Italian urban areas showed similar
Fig. 1. City models: (a) Xi’an; (b) Wuhan
© ASCE 05017014-2 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
results—that less monocentric areas were associated with lower
levels of CO2per commuter. However, some contrary views exist.
Burgalassi and Luzzati (2015) found that in Italy there is no evi-
dence that polycentricity reduces emissions, which can possibly be
attributed to the analysis being focused on the provincial level.
Aguilera (2005) found increased growth in commuting distances
in the French polycentric metropolitan areas of Paris, Lyon, and
Marseille from 1990 to 1999, and that the majority of the people
living in the subcenters tended to work outside their subcenters of
residence. Research in seven French urban areas from 1990 to 1999
showed that subcenters were not able to resist the growing distance
between housing and places of work (Aguilera and Mignot 2004).
Driving times were higher in most polycentric systems in the
Netherlands (Schwanen et al. 2003), and the deconcentration of
urban land use and the development of polycentric urban structures
encouraged driving and discouraged the use of public transport, as
well as cycling and walking (Schwanen et al. 2001). In the multi-
centered metropolis of Stockholm, Cervero (1995) found that
few of the new town residents worked locally and that most
new town workers were imported from the outside. Adolphson
(2009) showed that there was an increase in relative accessibility
by car and a decrease in relative accessibility by public transport
modes in Stockholm County from 1991 to 2004, with the goal of a
higher share in public transport possibly not being fulfilled.
Overall, the previous study results were sometimes contradic-
tory and it is difficult to draw a uniform conclusion about the re-
lationship between monocentricity/polycentricity and transport
outcomes. One possible reason for the inconsistent results is that
there still exist factors other than monocentricity or polycentricity
which could impact travel behavior, such as economic levels, car
ownership, the supply of transportation facilities, and job–housing
imbalances. For instance, high income, car availability, and weak
public transit services can lead to an increase in car use (Yang
et al. 2017a;Wang et al. 2017). Cities that promote public transit–
oriented developments can increase public transport use and reduce
car use, such as in Curitiba, Brazil; Copenhagen, Denmark; and
Tokyo, Japan (Cervero 1998). Job–housing imbalances can cause
long commuting distances and more transport CO2emissions, such
as in the inner suburbs of Beijing, China (Yang et al. 2017b).
Another possible reason lies in the different contexts of the urban
form across the cities, in terms of population densities and urban
sizes, which could have effects on travel patterns. Furthermore,
studies using different analysis scales and different periods of urban
development could produce different results, because polycentric-
ity is scale-dependent and a city’s polycentric development and
evolution is long-term in nature.
In addition, the case studies were mainly of American and
European cities, with little known about Chinese cities with dense
populations, large urban areas, intensive traffic, and rapid growth—
factors which are noticeably different from developed Western
cities. In this study, the cities of Wuhan and Xi’an have both
high-density and compact urban forms, and they have polycentric
and monocentric patterns, respectively, in their initial stages of ur-
ban development. Furthermore, they are megacities with sound and
comprehensive transportation systems and are in the process of fast
sprawl and development of city clusters. This study will enrich the
research of the monocentricity/polycentricity and associated travel
patterns, and will provide references for other cities.
This study analyzes the relationship between monocentricity/
polycentricity, urban forms, commuting patterns, and mode choice
behavior and compares these factors in Wuhan and Xi’an. Previous
studies have found that socioeconomic characteristics have an im-
pact on mode choice behavior, with the results showing that people
with cars, higher incomes, longer distances for travel, who are
middle-aged, and have better educational backgrounds preferred
the car mode, whereas people with lower incomes, without cars,
and who were younger preferred public transit modes (Wang et al.
2012,2013;Yang et al. 2015;Scheiner 2010;Santos et al. 2013;
Ewing et al. 2004;Dargay and Hanly 2007;Zhang et al. 2008;
Xianyu and Juan 2008;Chen et al. 2008;Pinjari et al. 2007;
Kim and Ulfarsson 2008;Patterson et al. 2005). This study exam-
ined the impact of urban forms in terms of the following factors on
mode choice behaviors: the household location separated by the
ring roads, straight-line distance from the household to the city
center or to the subcenters, and whether the trips cross rivers or
not (trips among or inside the subcenters). There are several reasons
for choosing these three innovative factors. The first factor repre-
sents the common characteristic of urban sprawl in Chinese cities,
with several ring roads spreading from the city center to the outer
regions. Areas separated by adjacent ring roads were developed in
similar periods and display similar urban form characteristics in
terms of transportation infrastructure, densities, block sizes, plot
ratios of buildings, building styles, and land-use diversities. The
second factor is another quantitative method of measuring city
sprawl, and the third factor is unique to Wuhan city. Two rivers
separate the city into three towns, and each town formed a subcen-
ter in the beginning of the city’s formation, with limited traffic
capacity on the bridges. This is expected to have an impact on mode
choice behavior. In addition to these three factors, the mode choice
models also examined the type of work unit to reflect whether dif-
ferences exist in the mode choice behaviors among the commuters
of state-owned companies, government offices, public institutions,
private companies, and foreign companies. Because previous stud-
ies have found that trip distance and socioeconomic characteristics
can have an impact on mode choice behavior, these factors were
included in the models as control variables.
Data Collection
Household surveys were carried out in the urban areas of Wuhan
and Xi’an to collect data on commuting trips in the years of
2010 and 2012, respectively. The statistical method suggested
by Meyer and Miller (2001) was used to determine the sample size
n¼½Z1−ð1=2Þα2ð1−pÞ
r2pð1Þ
where r= margin of error or precision and is assumed to be 0.05
(assuming an estimate of the sample size within 5% of the real
value 95% of the time); p= observed value of the proportion of
the commuting trips in urban passenger transport; and Z1−ð1=2Þα=
standard normal statistic corresponding to the (1−α)level.
According to this method, a minimum of 1,085 and 1,476
observations of commuting trips were needed in Wuhan and Xi’an,
respectively.
The simple random sampling method was implemented in each
zone in the urban areas of the two cities. In total, 1,194 households
and 2,050 commuters were surveyed in Wuhan, and 1,501 house-
holds and 2,449 commuters were surveyed in Xi’an. Thus the
sample sizes of the two cities exceed the required minimum obser-
vations. Fig. 2shows the location and distribution of the samples
and gives a general description of Wuhan’s and Xi’an’s main urban
areas. Table 1lists the descriptive statistical results of the socioeco-
nomic characteristics of the samples in Wuhan and Xi’an. The cities
share some common characteristics: most commuters work in
enterprises and more than half of the commuters graduated from
junior college. In terms of differences between the two cities, there
are more commuters working in state-owned enterprises in Wuhan
© ASCE 05017014-3 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
Fig. 2. Sampled locations of the household survey and general description of the main urban areas of Xi’an and Wuhan: (a) Xi’an main urban area;
(b) Wuhan main urban area
Table 1. Socioeconomic Characteristics and Distributions by Ring Roads of the Samples in Wuhan and Xi’an
Variable Level
Wuhan Xi’an
N(%) N(%)
Age <35 335 17.2 1,378 47.9
35–55 1,571 80.8 1,364 47.4
>55 38 2.0 137 4.8
Work unit Government 101 5.3 113 4.2
Public institution/school/hospital/
research or design institute
504 26.7 511 19.0
Foreign enterprise 54 2.9 32 1.2
Private enterprise 443 23.5 1,120 41.7
State-owned enterprise 415 22.0 453 16.9
Others 372 19.7 454 16.9
Education background Graduated from middle school 289 14.9 307 10.7
Graduated from high school or
technical secondary school
615 31.7 671 23.4
Graduated from junior college 400 20.6 664 23.1
Bachelor’s degree 529 27.2 1,029 35.9
Master’s degree 91 4.7 167 5.8
Ph.D. degree 19 1.0 31 1.1
Household traffic vehicles Car availabilitya461 38.6 817 54.4
Household annual income <US$2,000 10 0.9 9 0.6
US$2,000–6,000 233 21.7 77 5.2
US$6,000–10,000 414 38.6 326 22.1
US$10,000–20,000 305 28.4 891 60.4
US$20,000–40,000 86 8.0 139 9.4
>US$40,000 25 2.3 34 2.3
Household location
by ring roads
Inside first ring road 145 12.1 93 6.20
Between first and second ring road 432 36.2 469 31.25
Between second and third ring road 409 34.3 848 56.50
Outside third ring road 208 17.4 91 6.1
aIn Xi’an, household car availability refers to the household owning car or planning to buy a car.
© ASCE 05017014-4 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
than in Xi’an, and Wuhan has a lower proportion of car availability
per household (this can be attributed to the fact that car availability
in the questionnaire survey in Xi’an referred to the household
owning a car or planning to buy car). There are more commuters
with household annual incomes above US$10,000 in Xi’an than
in Wuhan.
Methodology
Statistical and comparative analyses were conducted to explore
the relationship between urban forms and commuting patterns in
Wuhan and Xi’an. Binary logit models were established to analyze
and compare mode choice behavior in the two cities
lnpðxÞ
1−pðxÞ¼αþX
K
k¼1
βkxkð2Þ
where pðxÞ= probability of mode choice; xk= independent
variables, consisting of the socioeconomic characteristics of the
commuters and the characteristics of urban forms; βk= coeffi-
cients; and α= constant.
The maximum likelihood estimation method was used for the
model estimation. The potential independent variables of the binary
logit models included socioeconomic characteristics of the com-
muters and households (age, education level, household income,
household car availability, and type of work unit), commuting dis-
tance, and the characteristics of the urban forms (polycentricity,
household location separated by ring roads, whether the trips
crossed rivers or not, and the straight-line distance from the house-
hold to the city center or to the subcenters). In the modeling
process, all the potential independent variables were considered.
Then, based on the statistical test results, less-significant indepen-
dent variables were removed. The best models of the commuting
mode choices were established with all significant variables.
In order to make the parameter estimations more accurate, boot-
strapping technique were applied during the modeling process.
Because a sound estimation of the confidence intervals of the
parameters requires more than 500 or 1,000 bootstrap replications
(Efron and Tibshirani 1993), this study applied 1,000 bootstrap
replications.
General Description of Wuhan and Xi’an
Wuhan and Xi’an, the midland and western central cities of China,
respectively, are associated with geographical and economic influ-
ences. Fig. 2provides a general description of the main urban areas
of Wuhan and Xi’an.
Wuhan has a built-up urban area of 520 km2, a total population
of 10.12 million, and 1.037 million private motor vehicles, and it
ranks ninth in gross domestic production (GDP) among Chinese
cities. The leading industries in Wuhan are iron and steel, automo-
bile and mechanical equipment, optoelectronic information,
and petrochemicals (Han 2015;Huang et al. 2015;LCL 2012;
WLRPB and WTDSI 2012;WSB 2014). The Yangtze and Han
Rivers separate Wuhan city into three towns—Wuchang, Hankou,
and Hanyang—and many lakes are scattered inside each town.
Wuhan has three ring roads spreading from the city center to
the outer regions in its main urban area. Because of the separation
created by the rivers, industry and self-contained developments
have existed inside each of the three towns since the beginning
of Wuhan’s urban development. Hankou is the commercial center
of the city; Wuchang’s industries are characterized by education,
administration and high-technology; and Hanyang’s industries
are mainly related to automobiles and mechanical equipment.
The districts near Jianghan Street in Hankou have formed the
CBD since the twentieth century (Li 2005). Wuchang and Hanyang
formulated their CBDs in the Fruit Lake area and in the area from
the Wangjiawan area to the Zhongjia Village area between the first
and second ring roads, respectively. The area within the second ring
road comprises the main commercial and official areas and high-
density residential land-use area, commonly referred to as the cen-
tral urban area (Huang et al. 2015). Between 1956 and 2010, nine
bridges and one tunnel were built across the two rivers in Wuhan’s
main urban area. These bridges and tunnels facilitated the moving
of people and goods among the three towns (Han and Wu 2004)
and also stimulated Wuhan’s rapid urban sprawl along and away
from the rivers in the three towns (Huang et al. 2015). Since
1999, the rapid growth of the outer areas in the three towns has
been associated with large new residential areas, industrial parks,
and economic development zones linked by the third ring road
(Huang et al. 2015). Three national economic development zones
are successful: Zhuankou in Hanyang, East Lake Zone in Wuchang,
and Wujiashan in Hankou. These three development zones have
their own core industries, and many high-technology enterprises
and light or heavy industry. Some enterprise headquarters in the
adjacent smaller cities have settled in these three zones, which have
attracted many residents to work and live there. The three zones
have developed better and also have more potential for future de-
velopment capacity than do the economic development zones in
Xi’an. In addition, the Qingshan Industrial Zone, established in the
1950s with its residential areas on the west side, is well-known for
its iron and steel industries. Wuhan has long suffered from high
transportation infrastructure investment, limited traffic capacities
of the bridges, traffic congestion on the bridges, and longer travel
distances caused by the separation of the rivers and lakes. The rel-
atively higher transportation infrastructure investment in Wuhan is
caused by constructing more bridges and more roads under the
polycentric urban form—three towns being separated by two rivers.
Because of the increasing travel demands across the two rivers and
the limited capacity of the bridges, a car restriction policy (an odd-
and-even license plate rule for passenger cars) has been promoted on
the two oldest bridges (No. 1 Yangtze Bridge and No. 1 Jianghan
Bridge), which are part of the first ring road. The separation created
by the lakes and rivers, the limited capacity of the bridges, and the
higher transportation infrastructure investment have created a poly-
centric urban structure since the Wuhan’sformationandwedge-
shaped urban growth in each of the three towns.
On the other hand, Xi’an has a built-up urban area of 449 km2,a
total population of 8.55 million, and 1.174 million private motor
vehicles, and ranks 29th in GDP among all Chinese cities. Its lead-
ing industries are cultural tourism, high-technology, and equipment
manufacturing (Wang and Liu 2015;XBS and NBSXIT 2013;LCL
2012). Xi’an sprawls in a concentric pattern with a bell tower at its
city center and spreads out across its first, second, and third ring
roads. The central urban area has an area of 10.78 km2, with com-
mercial, provincial government, and old residential areas clustered
together, and this area is surrounded by the city wall. There are
396,600 people living in the central urban area, with population
density of 36,800 people=km2. The first ring road is just outside
the city wall. Because there are historical relics inside the first ring
road, the building height has a limit in each zone—approximately
36 m (XMG 2002). To protect the ancient style of Xi’an, strict land
development controls were implemented in 2005 inside the first
ring road (XMG 2005). The area (the annular belt) between the
first and second ring roads was developed between the 1950s
and the 1990s. It has an area of 64.51 km2with a population
of 1,324,400 (20,500 people=km2). The area (the annular belt)
© ASCE 05017014-5 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
between the second and third ring roads was developed during the
late 1990s–2000. It has an area of 271.57 km2with a population of
1,791,300 (6,600 people=km2). Some areas outside the third ring
road were developed after the year 2000. High-technology industry
development zones and cultural and tourism industry development
zones (including Jingkai, Gaoxin, and Qujiang) were developed
outside the second ring road. Generally, more mature and stronger
industrial developments are in the central and southern parts of
the city. The public transit line densities inside the first ring road,
in the area between the first and second ring roads, and in the area
between the second and third ring roads are 45.73, 28.68, and
10.58 km=km2, respectively, with 300-m bus stop coverage rates
of 83.8, 61.3, and 36.8%, respectively.
Commuting Pattern and CO2Emissions
Commuting Mode Shares and Distance Distributions
The commuting mode shares of cars and public transit in Xi’an
(29.3 and 42.1%, respectively) were larger than those in Wuhan
(16.6 and 36%, respectively); the share of nonmotorized trips
(walking and bicycle trips) in Wuhan (36.8%) was larger than that
in Xi’an (20.4%) [Fig. 3(a)]. In addition, there was a great differ-
ence in the mode shares between the trips across the rivers and not
across the rivers (trips between and within the subcenters) in
Wuhan. For trips across the rivers (trips between the subcenters),
public transit trips constituted 67.3% of trips, car trips constituted
19.5%, and nonmotorized and motor/electric bicycle trips only
constituted 7.6 and 5.7%, respectively. Public transit was the most
commonly used mode for trips across the rivers. For the trips not
across the rivers (trips within the subcenters), nonmotorized and
motor/electric bicycle trip shares constituted 39.4 and 11.2%, re-
spectively, and public transit and car shares constituted 33.1 and
16.3%, respectively. Nonmotorized modes and public transit were
the most-used modes for trips not across the rivers. Fig. 3(b) shows
that the percentage of trips continuously decreased as the commut-
ing distance increased in Wuhan. Fig. 3(c) shows the cumulative
percentage of trips with respect to commuting distance. Wuhan
had more short-commute trips (less than about 3 km) than did
Xi’an.
Compared with Xi’an, more nonmotorized mode shares, decreas-
ing trip frequency as distances increased, and more short-commute
trips in Wuhan indicated the more sustainable transportation
development patterns when a city experiences rapid growth. The
commuting mode shares in Wuhan were different from those in
previous studies of developed Western nations that indicated that
polycentric urban structures discourage public transit uses, cycling,
and walking, and instead encourage car use (Schwanen et al. 2001;
Adolphson 2009). In addition, the increased nonmotorized mode
shares in Wuhan are different from those in previous studies of
0 2 4 6 8 10 12 14 16
0
5
10
15
20
25
Commuting Distance (km)
Percentage (%)
Xian
WuHan
0 2 4 6 8 10 12 14 16
0
10
20
30
40
50
60
70
80
90
100
Commuting Distance (km)
Cumulative Percentage (%)
Xian
WuHan
(a) (b)
(c)
Fig. 3. Commuting mode share and distance frequency distributions of Wuhan and Xi’an: (a) commuting mode share; (b) frequency of the
commuting distance; (c) cumulative frequency of the commuting distance
© ASCE 05017014-6 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
the state of Maryland in the United States and Istanbul, Turkey,
which maintained that polycentric urban structures promote public
transit ridership (Knaap et al. 2016;Alpkokin et al. 2005).
Person Kilometers Traveled and Average Commuting
Distance by Mode
Figs. 4(a and b) indicate that the person kilometers traveled (PKT)
of public transit and nonmotorized traffic in Wuhan were larger
than those in Xi’an. The PKT of cars in Wuhan was smaller than
that in Xi’an. Fig. 4(c) shows that Wuhan had a slightly longer aver-
age commuting distance (4.1 km) than did Xi’an (3.8 km). The
longer average commuting distance in Wuhan can be explained
by its larger urban size (Han 2015;Schwanen 2002) and the many
lakes scattered across and the two rivers that run through Wuhan
city. For trips across the rivers, the average commuting distance of
all modes was much longer than the trips not across the rivers and
the total trips in Wuhan [Fig. 4(c)]. However, the trips across the
rivers and their PKT only accounted for a small part of total trips
and total PKT (7.9 and 18%, respectively). Furthermore, trips
across the rivers in Wuhan used more public transit (67.3%) than
other modes. Fig. 4(c) also shows that, except for trips across the
rivers, the average commuting distances of the public transit trips
were longer than that of car trips in Wuhan. In Xi’an, it was the
opposite: the car trip distance was longer than the public transit
trip distance. The longer average public transit trip distance than
the car trip distance in Wuhan indicates that commuters used the
public transit mode more often than the car mode for long-
distance trips.
Compared with Xi’an, the larger PKT of the nonmotorized and
public transit modes, the smaller PKT of the car mode, and the
much less frequent and smaller PKT of the trips between the sub-
centers in Wuhan provide further evidence of a more environmen-
tally friendly commuting pattern. These results are different from
those of previous studies on developed Western cities which
showed that polycentric urban structures bring about more car
use with long distances, imbalances in job–housing, and more com-
muters working outside their subcenter of residence (Grunfelder
et al. 2015;Aguilera 2005;Aguilera and Mignot 2004;Cervero
and Wu 1998;Cervero 1995;Giuliano and Small 1993).
Commuting Mode Share Changes between Ring Roads
The motorized traffic shares (car and public transit) increased and
the nonmotorized traffic shares (walking and bicycle) decreased
from the city center to the third ring road in Xi’an [Fig. 5(a)].
Outside the third ring road there was a sharp increase in trips
by motor/electric bicycles and a sharp decrease in car trips, which
is related to household income. Inside the third ring road of Xi’an,
families have higher incomes than do those located outside the third
ring road, which is an urban–rural fringe area of low-income fam-
ilies. Figs. 5(b–d) show that, generally, there were more walking
and bicycle trip shares and fewer car trip shares in all three towns
of Wuhan than in Xi’an. Fig. 5(b) indicates that the nonmotorized
(a) (b)
(c)
Fig. 4. Person kilometers traveled and average commute distance by mode in Wuhan and Xi’an: (a) PKT by mode; (b) percentage of PKT by mode;
(c) average commuting distance by mode
© ASCE 05017014-7 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
traffic shares (walking and bicycle) generally decreased and the
motorized traffic shares (car and public transit) increased from
the city center to the third ring road in Hankou, Wuhan, which
is similar to the trends in Xi’an. These outcomes can be explained
by the facts that there are no large lakes in Hankou, the center
of this area is near the Yangtze River in Jianghan Street, and the
urban sprawl is in the direction away from the rivers. Unlike the
tendency found in Xi’an, there was a sharp increase in the car trip
shares outside the third ring road in Hankou. This tendency can be
explained by the higher household income and the good living and
driving conditions in this area, which attract many families with
higher incomes and more car use. Weak public transit service in
the outer areas is another reason. There were different mode share
changes inside the third ring road in Hanyang, Wuhan compared
with Hankou and Xi’an [Fig. 5(c)]. The public transit shares in
the area between the second and third ring roads sharply reached
60% and the car trip shares became much smaller. On the one hand,
this tendency can be explained by the lower level of household
income in the area between the second and third ring roads. On
the other hand, Wuhan has a strip urban development pattern along
the radial traffic corridors, and the origins and destinations of the
commuting trips can be matched with public transit services along
the traffic corridors to a greater extent. Therefore there was more
public transit use among the residents located along the traffic cor-
ridors, especially in this area with one main radial traffic corridor
from the northeast to the southwest linking the inner and outside
areas. Outside the third ring road in Hanyang, there was a sharp
increase in car trip shares. This is related to the higher household
(a) (b)
(c) (d)
Fig. 5. Mode shares by ring roads in Wuhan and Xi’an: (a) Xi’an; (b) Hankou; (c) Hanyang; (d) Wuchang
© ASCE 05017014-8 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
income, good driving conditions, and the Zhuankou industry zone,
which mainly focuses on automobiles and spare parts and encour-
ages employees’high car availability and increased car use for
commuting. Fig. 5(d) shows a sharp increase in the walking shares
(around 50%) outside the third ring road in Wuchang, although
there is higher household income. This result is due to more
short-distance trips in this area, which is the center of the Jiangxia
district; 47% of these commuters’workplaces are outside the third
ring road, which is the same area as their residences.
Commuting Distance Changes between Ring Roads
Figs. 6(a and b) show different trends in the changes in average
commuting distance between the ring roads in Wuhan and Xi’an.
In Wuhan there was not much increase in commuting distance from
the first ring road to the third ring road, and the commuting distance
decreased outside the third ring road, especially in terms of car-trip
distance. There was only a slight increase in commuting distance by
public transit mode outside the third ring road, which was due to the
long-distance cross-river trips by public transit. The car-trip dis-
tance had a decreasing tendency from the area between the first
and second ring roads to the area outside the third ring road. Xi’an
had a slightly shorter commuting distance on average; however, the
average commuting distance greatly increased from the area inside
the first ring road to the area outside the third ring road, especially
in terms of car-trip distance.
Generally, the decrease in the average commuting distance out-
side the third ring road in Wuhan was due to the better job–housing
balance in the high-technology economic development zones and
fewer long-distance commuting trips across the rivers because of
the limited traffic capacity of the bridges, the car-use restriction
policy on two of the oldest bridges, and the tolls on the newly built
bridges. However, in Xi’an, the increase in average commuting
distance was due to more through traffic among the districts, espe-
cially in the north–south and outer–inner directions. This is caused
by the sprawling urban development, land development controls
inside the first ring road that protect historic heritage buildings,
decreasing employment density from the center district to the outer
regions, and the more mature and stronger industry development
in the central and southern areas.
Commuting CO2Emissions
The commuting CO2emissions were calculated based on an emis-
sion factor (by mode, fuel type, and occupancy) multiplied by the
trip distance (Wang et al. 2017;Yang et al. 2017b)
C¼EF ×Lð3Þ
where C¼CO2emissions (kg=passenger=km); EF = emission
factor by mode, fuel type, and occupancy (Table 2); and L= com-
muting distance (km).
The well-to-wheel (WTW) CO2emission factors by mode
were calculated based on the WTW CO2intensity of various fuels,
local fuel consumption based on driving experience, and the vehicle
occupancy in Wuhan and Xi’an (Table 2). The fuel consumed
(e.g., liters of gasoline consumed per 100 km) by vehicles is asso-
ciated with uncertainty because it is affected by several factors,
such as driving speed, and the fact that actual vehicle occupancy
is not a fixed value (Wang et al. 2017). Therefore the authors col-
lected local data on the ranges of these values in Wuhan and Xi’an
through surveys and from the related literature (cf. notes to Table 2).
Because more buses in Wuhan use diesel, the CO2emissions from
diesel buses were greater than those in Xi’an. On the other hand,
CO2emissions from compressed natural gas (CNG) buses in
Wuhan were less than those in Xi’an. The larger share of routine
coaches in Wuhan caused the larger CO2emissions from routine
coaches in Wuhan than in Xi’an. Because of the fewer car trips
in Wuhan, the CO2emissions of cars in Wuhan were much less
than those in Xi’an.
Although the average commuting distance in Wuhan is slightly
longer than in Xi’an, the total emissions in the two cities were
similar. Residents in a city may have longer commuting distances
simply because that city is large in terms of geographical coverage
(Han 2015;Schwanen 2002). Considering the impact of urban
(
a
)(
b
)
Fig. 6. Average commuting distance by ring roads in Wuhan and Xi’an: (a) Xi’an; (b) Wuhan
© ASCE 05017014-9 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
size on commuting distance, the total commuting CO2emissions in
Wuhan were less than those in Xi’an. The fewer car uses and in-
creased nonmotorized trips in Wuhan were the main contributors
to the smaller commuting CO2emissions. The relatively lesser com-
muting CO2emissions in Wuhan differs from the insights of previous
studies in Italy and the United States which found that polycentric
urban structures have no effect on reducing transportation CO2
emissions (Burgalassi and Luzzati 2015;Lee and Lee 2014).
Commuting Mode Choice Behavior
Commuting Mode Choice Models for Walking
Table 3shows the walking mode choice models of Wuhan and
Xi’an. In both cities, households located in the inner areas preferred
walking. This was due to the shorter distance in the inner areas of
the cities. In addition, the coefficient of the constant in the model of
trips not across the rivers in Wuhan was larger than others, indicat-
ing that these trips used more walking for commuting. Other than
this, in both cities, commuters with car availability and longer dis-
tances did not prefer walking, and in Xi’an, commuters with high
incomes also did not prefer walking.
Commuting Mode Choice Models for Cars
The factor of commuting trips across the rivers (trips between the
subcenters) was found to be significant in the car mode choice
model in Wuhan (Table 4). The negative coefficient shows that
commuters across the rivers did not prefer the car mode, and
the negative and much smaller constant in the Wuhan model of
the trips across the rivers also demonstrates this character. The type
of work unit was also found to have effects on car mode choice
behavior in both cities. In both cities, commuters working in public
institutions did not prefer using cars. In Xi’an, commuters working
in foreign companies and private companies did not prefer using
cars. Furthermore, home–subcenter distance was found to be a sig-
nificant factor in mode choice behavior in Wuhan. The farther the
commuter was located from the subcenter, the larger was the prob-
ability that the commuter would choose the car mode. The common
factors determining the choice of car mode in Wuhan and Xi’an
were car availability, commuting distance, and household annual
income, which is a similar finding to previous studies (Dargay
and Hanly 2007;Pinjari et al. 2007;Chen et al. 2008;Zhang et al.
2008;Xianyu and Juan 2008;Scheiner 2010;Santos et al. 2013).
In both cities, commuters with car availability, longer commuting
distances, and higher household annual incomes (more than US
$40,000) preferred using cars, whereas commuters with household
annual incomes of less than US$10,000 did not prefer to use cars.
Commuters between 35 and 55 years old and commuters with
master’s or doctorate degrees were also found to be significant fac-
tors in Xi’an and Wuhan, respectively, indicating that middle-aged
commuters preferred using cars in Xi’an and commuters with high
education levels preferred to use cars in Wuhan.
Table 2. Well to Wheel (WTW) CO2Emission Factor and Commuting Trip CO2Emissions of Xi’an and Wuhan Samples
Fuel type
WTW CO2
emission factor
(tons CO2equivalent/unit of fuel)a
Range of/average
fuel consumption
(L) per 100 kmb
Range of/average
occupancy
(persons/vehicle)c
Commuting CO2emissions (kg)
Xi’an Wuhan Wuhan’d
Bus (CNG) 2.76 per 1,000 m3ð36–44Þ=40 m3(60–100)/50 78.18 25.78 22.26
Bus/routine coach (diesel) 3.94 per ton ð39.12–42.38Þ=40L(20–50)/35 5.01 109.71 94.73
Car (gasoline) 3.87 per ton ð7.80–10.45Þ=9L(1–3)/1.4 458.53 322.49 278.46
Metro/rail (electricity) 0.83 per 1,000 kWh 3,348.15 kWhe(1,200–1,800)/
1,500 per train
7.95 2.50 2.16
Electric-bike/motor
(electricity)
0.83 per 1,000 kWh ð1.2−1.67Þ=1.4kWhfð1–2Þ=1.44.47 6.81 5.88
Sample size factorg————1.01g1.01g
Population factorg————1.18g1.18g
Total commuting
CO2emissions
———554.14 556.92 480.88d
aData from Huo et al. (2012).
bData from Huo et al. (2011), Zhang et al. (2014), Liu and Hou (2009); fuel consumption considered the factor of vehicle speed during peak hours; values in
parentheses are range of fuel consumption during commuting.
cData from Xi’an Transport Development Annual Report (XCTMCC 2012) and Wuhan Transportation Annual Report (WLRPB and WTDSI 2012); values in
parentheses are range of vehicle occupancy during commuting.
dWuhan commuting CO2emission calculation also considered the impact of the urban size on the longer commuting distances; distances by mode were
adjusted by the two cities’urban sizes.
eData from the survey of Xi’an and Wuhan Metro Co., Ltd.
fData from CSBQTS (1999) and questionnaire surveys of electric-motor/bicycle users in Xi’an and Wuhan.
gSample size and population factors were calculated according to effective samples of households’two heads (1952 and 1947 samples in Xi’an and Wuhan)
and the cities’urban population.
Table 3. Mode Choice Models for Walking in Wuhan and Xi’an for 1,000
Bootstrap Samples
Variable Wuhan (total)
Wuhan (trips not
across rivers) Xi’an
Constant 1.138 (0.000) 1.207 (0.000) 0.518 (0.002)
Car availability −0.646 (0.000) −0.678 (0.000) −0.225 (0.083)
Commuting
distance (km)
−0.889 (0.000) −0.923 (0.000) −0.792 (0.000)
Household annual income
>US$40,000 ——−0.974 (0.086)
Household location by ring roads
Inside first
ring road
0.646 (0.001) 0.666 (0.001) 0.358 (0.163)
First–second
ring road
0.381 (0.005) 0.369 (0.006) —
Log likelihood −739.42 −694.87 −689.58
χ2156.26 155.18 105.27
Prob >χ20.00 0.00 0.00
Pseudo R20.35 0.33 0.23
Observations 1,947 1,700 1,914
Note: Significance levels are in parentheses.
© ASCE 05017014-10 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
Commuting Mode Choice Models for Public Transit
The farther the commuters were located from the subcenters, the
smaller was the probability they would choose the public transit
mode in Wuhan (Table 5). In the Xi’an model, household location
separated by the ring roads also showed a similar effect on public
transit mode choice behavior. Commuters located outside the third
ring road did not prefer the public transit mode. This was partly
related to the weak public transit service in the outer areas of the
two cities. In contrast to the results in Table 4, the factor of trips
across the rivers, with a coefficient of 0.448, had significant
Table 4. Mode Choice Models for Car in Wuhan and Xi’an for 1,000 Bootstrap Samples
Variable Wuhan (total) Wuhan (trips not across rivers) Wuhan (trips across rivers) Xi’an
Constant −4.584 (0.000) −4.289 (0.000) −6.204 (0.001) −4.572 (0.000)
Age
35–55 years old —— —0.405 (0.002)
Education background
Master’s degree 0.970 (0.018)
Car availability 4.298 (0.000) 4.281 (0.000) 3.685 (0.000) 3.972 (0.000)
Commuting distance (km) 0.166 (0.000) 0.155 (0.000) 0.185 (0.175) 0.156 (0.000)
Household annual income
US$2,000–6,000 −0.916 (0.023) −1.122 (0.012) −0.780 (0.128)
US$6,000–10,000 −0.537 (0.011) −0.557 (0.012) −2.754 (0.001) −0.721 (0.000)
>US$40,000 0.751 (0.138) 0.999 (0.008)
Work unit type
Foreign company —— —−1.210 (0.056)
Public institution −0.496 (0.030) −0.503 (0.026) —−0.243 (0.166)
Private company —— —−0.232 (0.108)
Home–subcenter distance 0.059 (0.023) 0.046 (0.094) 0.276 (0.107) —
Trips across rivers −1.576 (0.000) —
Log likelihood −382.97 −349.08 −29.92 −765.08
χ2384.36 367.27 28.73 231.50
Prob >χ20.00 0.00 0.00 0.00
Pseudo R20.52 0.52 0.54 0.32
Observations 1,844 1,618 142 1,877
Note: Significance levels are in parentheses.
Table 5. Mode Choice Models for Public Transit in Wuhan and Xi’an for 1,000 Bootstrap Samples
Variable Wuhan (total) Wuhan (trips not across rivers) Wuhan (trips across rivers) Xi’an
Constant −1.562 (0.000) −1.730 (0.000) 1.349 (0.000) −1.013 (0.000)
Age
<35 years old 0.419 (0.005) 0.484 (0.002) 0.275 (0.012)
Education background
Middle school 0.618 (0.000) ———
High school/technical secondary school 0.398 (0.003) ———
Junior college —— —0.241 (0.082)
Bachelor’s degree —— —0.339 (0.008)
Car availability −1.384 (0.000) −1.327 (0.000) −1.841 (0.000) −1.469 (0.000)
Commuting distance (km) 0.154 (0.000) 0.194 (0.000) 0.176 (0.000)
Household annual income
US$6,000–10,000 0.279 (0.049) 0.262 (0.084) ——
US$10,000–20,000 0.246 (0.104) 0.308 (0.080) ——
US$20,000–40,000 —— —−0.645 (0.002)
>US$40,000 —— —−1.025 (0.049)
Work unit type
Government ——−1.106 (0.175)
Foreign company −0.809 (0.115) ——1.263 (0.010)
Public institution —0.271 (0.096) —0.783 (0.000)
Private company —0.361 (0.020) —0.706 (0.000)
State-owned company −0.348 (0.028) ——
Home–subcenter distance −0.088 (0.000) −0.086 (0.000) −0.155 (0.055) —
Trips across rivers 0.448 (0.037) ———
Household location by ring roads
Outside third ring road —— —−0.695 (0.018)
Log likelihood −897.34 −774.15 −82.24 −1,087.49
χ2188.85 138.66 25.85 281.11
Prob >χ20.00 0.00 0.00 0.00
Pseudo R20.13 0.11 0.15 0.14
Observations 1,841 1,615 142 1,878
Note: Significance levels are in parentheses.
© ASCE 05017014-11 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
positive effects on public transit mode choice behavior in Wuhan.
Furthermore, the positive constant in the model of trips across
the rivers in Wuhan also indicated that trips across the rivers used
more public transit, whereas other trips in Wuhan and trips in Xi’an
did not. This study also found that commuters working in public
institutions and private companies preferred the public transit in
both cities. Commuters working in a foreign company did not ap-
pear to have consistent public transit mode choice behavior in the
two cities. In the Xi’an model this factor showed a positive effect,
whereas in Wuhan model the effect was negative. Commuters
working in the government and state-owned companies also did
not prefer the public transit mode in Wuhan. In the models of both
cities, commuters with long distances and those less than 35 years
old preferred the public transit mode, whereas commuters with
car availability and high incomes did not prefer the public transit
mode—this is similar to results from previous studies (Dargay and
Hanly 2007;Pinjari et al. 2007;Chen et al. 2008;Zhang et al. 2008;
Xianyu and Juan 2008;Scheiner 2010;Santos et al. 2013). In ad-
dition, the model results in Wuhan show that commuters with
middle school and high school education backgrounds preferred
the public transit mode, and the model results in Xi’an show that
commuters with junior college and bachelor’s degree education
background preferred the public transit mode.
From the model results of mode choice behavior in Wuhan and
Xi’an, it was found that no matter whether a city was monocentric
or polycentric, when commuters were located farther from the sub-
centers or city centers, they preferred the car mode. This was related
to the weak public transit services and better road conditions for
driving in the outer areas of Xi’an city and in the areas farther from
the three town subcenters in Wuhan. It could also be related to
the much longer commuting distances in Xi’an city’s outer areas.
Furthermore, commuters located inside the central urban areas
preferred the walking mode, which can be attributed to the shorter
commuting distances in the central urban areas in both cities. Trips
across the rivers were likely to be in public transit mode and not
car mode in the polycentric city of Wuhan. This was due to the
limited traffic capacities of the bridges, traffic congestion, and
car restrictions and tolls on the bridges, which make car trips incon-
venient, time-consuming, and costly. This is different from devel-
oped Western polycentric cities, in which commuters who work
outside the subcenters more often use cars for traveling (Lee and
Lee 2014;Schwanen et al. 2001;Cervero and Wu 1998).
Effect of Polycentricity on Mode Choice Behavior
In order to test whether the effect of polycentricity on mode choice
behavior is statistically significant, this study pooled the survey
data of Wuhan and Xi’an to establish mode choice models for walk-
ing, public transit, and car for both Wuhan and Xi’an, and a new
variable was defined, named polycentricity, which had a value of 1
for commuters in Wuhan and a value of 0 for commuters in Xi’an.
All the potential variables were considered in the modeling process,
including the socioeconomic characteristics of the commuters and
households (age, education level, household income, household car
availability, and type of work unit), commuting distance, and the
characteristics of the urban forms (polycentricity, household loca-
tion separated by the ring roads, and the straight-line distance from
the household to the city center or to the subcenters). The insignifi-
cant variables were removed during the modeling process. The best
mode choice models for walking, public transit, and car were
established with all significant variables (Table 6). It was found that
polycentricity was the significant factor in the walking and car
mode choice model. This factor had a positive effect (with a coef-
ficient of 0.771) on walking mode choice, whereas it had a negative
effect (with a coefficient of −0.848) on the car mode choice.
The results indicate that polycentricity can promote walking and
decrease car uses.
Discussion and Conclusion
The significant differences in commuting patterns between Wuhan
and Xi’an are related to the cities’polycentric and monocentric
urban structures. In Wuhan, the three towns’polycentric pattern,
Table 6. Mode Choice Models of Pooled Data of Wuhan and Xi’an for 1,000 Bootstrap Samples
Variable Walking Public transit Car
Age
<35 years old —0.370 (0.000) −1.093 (0.000)
Car availability −0.157 (0.106) −1.114 (0.000) 2.006 (0.000)
Commuting distance (km) −0.730 (0.000) ——
Household annual income
US$2,000–6,000 ——−2.095 (0.000)
US$6,000–10,000 ——−0.144 (0.000)
US$10,000–20,000 —0.132 (0.058) —
US$20,000–40,000 —−0.543 (0.001) —
>US$40,000 —−1.356 (0.005) 1.166 (0.000)
Work unit type
Public institution ——−1.291 (0.000)
Private company ——−1.421 (0.000)
State-owned company —−0.669 (0.000) —
Polycentricity 0.771 (0.000) −1.061 (0.000)
Household location by ring roads
Inside first ring road 0.652 (0.000) —−0.848 (0.000)
First–second ring road 0.348 (0.000) ——
Outside third ring road —−1.197 (0.000) —
Log likelihood −1,465.33 −2,187.12 −1,407.72
χ2571.55 469.87 1,185.66
Prob >χ20.00 0.00 0.00
Observations 3,887 3,617 3,612
Note: Significance levels are in parentheses.
© ASCE 05017014-12 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
with their separation by two rivers, was formed in the beginning of
the city’s development. The limited capacities of the bridges have
encouraged a car-use restriction policy on two of the oldest bridges.
In addition, a tolling policy was also implemented on newly built
bridges. These have had a significant impact on the commuting
trips between the subcenters, which only account for a small per-
centage of the total (7.9%), with more public transit use (67.3%)
instead of cars. Three national high-technology economic develop-
ment zones with strong industry development and large housing
communities in Hanyang, Hankou, and Wuchang have also at-
tracted many people to work and live in these areas, because cross-
ing the rivers costs time and money. As a result, there are more
nonmotorized trips and many fewer car trips in Wuhan. In addition,
there has not been much of an increase in the average commuting
distances from the inner areas to the outer regions during rapid ur-
ban growth phases in Wuhan, and the average car-trip distance de-
creases greatly outside the third ring road. Furthermore, commuters
more often used public transit than cars for long-distance trips, and
the trip frequency decreased as the distance increased. Less car use
and more nonmotorized trips in Wuhan greatly contribute to the
lower levels of commuting CO2emissions. These all indicate that
more ecofriendly commuting patterns exist in Wuhan.
However, the situation is different in the monocentric city of
Xi’an with rapid sprawling (tandabing). Commuters use more cars
and fewer nonmotorized modes than in Wuhan. There are sharply
increasing trends in the motorized mode use and commuting
distances from the inner areas to the outside regions. In addition,
commuters in Xi’an more often use cars than public transit for
long-distance trips. In Xi’an, there are land development controls
inside the first ring road, large-scale housing developments with
better living conditions in the outer areas of the city, decreasing
employment densities from the center area to the outside region,
and more mature and stronger industry developments in the central
and southern parts of the city. These have inevitably encouraged
more through traffic among the districts, with long distances
and more car use, especially in the north-south and outer-inner
directions.
Previous studies in developed Western cities maintained that
there is a limited effect of polycentricity on ecofriendly travel
patterns. Furthermore, studies from the United States and Europe
found that polycentricity can promote public transit uses. The re-
sults in Wuhan are different. There are ecofriendly commuting
patterns and less CO2emissions in terms of shorter distances in
the outer areas, more public transit use between the subcenters,
more trips inside each subcenter area, greater nonmotorized mode
share, and less car use. One reason for these differences lies in the
separation of the three subcenters in Wuhan caused by two large
rivers, which have caused more inconvenience for travelling be-
tween the three subcenters. Another reason is the compact and
high-density developments in each subcenter in Wuhan. This con-
trasts with the more dispersed urban development patterns and a
decentralization of employment and household locations in devel-
oped Western polycentric case cities. The latter leads to longer
commuting distances, difficulty in serving urban activities by pub-
lic transit, weak public transit services, and increased car use. These
implications indicate that in order to develop sustainable transpor-
tation, cities should develop more compactly and set some barriers
to limit the automobile capacities on the traffic corridors connecting
the subcenters.
Continuous economic growth and rapid urbanization and
motorization in Chinese cities have stimulated much more traffic
demand in the urban areas and have brought about the prevalent
use of cars and long-distance trips during commuting. Furthermore,
more road and urban rail transit construction have occurred to
meet these demands, which have become prevalent trends in the
urban and transportation development of many Chinese cities in
recent decades. In the context of the continuous economic and ur-
ban growth in Xi’an and the majority of monocentric Chinese
cities, car trips of long distances, especially originated from the
city’s outer areas, will continue to exist. To cope with this, it is
important to foster strong industries in a city’s subcenters to form
a polycentric urban structure combined with car restrictions and
public transit promotion policies implemented on the major traffic
corridors connecting a city’s subcenters simultaneously. Otherwise,
polycentric urban structures will be difficult to create. In the pro-
cess of cities’cluster developments in China, the travel demands
between the central city and other smaller cities will substantially
increase in the future. Wuhan, the midland central city of China,
can support much denser developments inside each of its three
towns and can form wedge-shaped urban growth on a large scale
in the metropolitan area, because it has decentralized traffic due to
its polycentric pattern. However, the western central city of Xi’an
will face much more concentrated traffic and even traffic paralysis
in the central urban areas. Other Chinese cities should thus be on
the alert for the aforementioned situations.
In Wuhan, although there are fewer long-distance trips across
the rivers by cars and there are environmentally friendly commut-
ing patterns, the public transit services outside the third ring road,
farther from the three towns’subcenters, should be improved to a
large extent because there are more short-distance car trips in
these areas.
This study is an initial analysis of the relationship between the
monocentric/polycentric urban forms and transport outcomes in
rapidly developing countries with dense populations, fast motori-
zation and urbanization, and intensive traffic. The findings and im-
plications in this paper can help inform low-carbon and sustainable
transportation development in the majority of Chinese cities
and city cluster areas, through better spatial planning and traffic
demand management rather than merely expanding roads and ur-
ban rail transit infrastructure. Furthermore, this study contributes to
existing knowledge on urban forms and associated travel patterns,
and can provide empirical evidence and reference values for other
cities around the world.
Acknowledgments
This study was funded by the Australian Research Council
(ARCDP1094801), Asia Pacific Network for Global Change Re-
search (ARCP2011-07CMY-Han), National Natural Science Foun-
dation of China (51178055-E0807), The Fundamental Research
Funds for the Central Universities of China (310821172201),
and The Fundamental Research Funds for the Central Universities
of China (310821172202). The authors appreciate the work of Prof.
Jingnan Huang of Wuhan University collecting data for Wuhan.
The authors also thank Zhen Wang and Zihe Zhang, postgraduate
students at Chang’an University, and Chao Li, Ph.D. student at
Shanghai Jiao Tong University, for providing research assistance.
Finally, the authors thank Candice Tan at the University of
Melbourne for the English editing of this work.
References
Adolphson, M. (2009). “Estimating a polycentric urban structure. Case
study: Urban changes in the Stockholm region 1991–2004.”J. Urban
Plann. Dev.,10.1061/(ASCE)0733-9488(2009)135:1(19), 19–30.
© ASCE 05017014-13 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
Aguilera, A. (2005). “Growth in commuting distances in French polycen-
tric metropolitan areas: Paris, Lyon and Marseille.”Urban Stud., 42(9),
1537–1547.
Aguilera, A., and Mignot, D. (2004). “Urban sprawl, polycentrism and
commuting. A comparison of seven French urban areas.”Urban Public
Econ. Rev.,1,93–113.
Alpkokin, P., Hayashi, Y., Black, J., and Gercek, H. (2005). “Polycentric
employment growth and impacts on urban commuting patterns: Case
study of Istanbul.”J. East. Asia Soc. Transp. Stud., 6, 3835–3850.
Alqhatani, M., Setunge, S., and Mirodpour, S. (2014). “Can a polycentric
structure affect travel behaviour? A comparison of Melbourne,
Australia and Riyadh, Saudi Arabia.”J. Modern Transp., 22(3),
156–166.
Bourne, L. S. (1989). “Are new urban forms emerging? Empirical tests for
Canadian urban areas.”Can. Geogr., 33(4), 312–328.
Burgalassi, D., and Luzzati, T. (2015). “Urban spatial structure and envi-
ronmental emissions: A survey of the literature and some empirical
evidence for Italian NUTS 3 regions.”Cities, 49, 134–148.
Burger, M. J., Goei, B. D., Laan, L. V. D., and Huisman, F. J. M. (2011).
“Heterogeneous development of metropolitan spatial structure: Evi-
dence from commuting patterns in English and Welsh city-regions,
1981–2001.”Cities, 28(2), 160–170.
Burger, M. J., and Meijers, E. (2012). “Form follows function? Linking
morphological and functional polycentricity.”Urban Stud., 49(5),
1127–1149.
Burgess, E. W. (1925). The growth of the city, R. E. Park, E. W. Burgess,
and R. D. Mc-Kenzie, eds., University of Chicago Press, Chicago.
Burns, M. C., Boix, M., and Roca, J. (2001). “Contrasting indications of
polycentrism within Spain’s metropolitan urban regions.”8th European
Estate Society Conf., European Real Estate Society, Alicante, Spain.
Cervero, R. (1995). “Sustainable new towns: Stockholm’s rail-served
satellites.”Cities, 12(1), 41–51.
Cervero, R. (1998). The transit metropolis: A global inquiry, Island Press,
Washington, DC.
Cervero, R., and Wu, K. L. (1998). “Sub-centring and commuting:
Evidence from the San Francisco Bay area, 1980–90.”Urban Stud.,
35(7), 1059–1076.
Chen, C., Gong, H., and Paaswell, R. (2008). “Role of the built environ-
ment on mode choice decisions: Additional evidence on the impact of
density.”Transportation, 35(3), 285–299.
Cirilli, A., and Veneri, P. (2014). “Spatial structure and carbon dioxide
(CO2) emissions due to commuting: An analysis of Italian urban areas.”
Reg. Stud., 48(12), 1993–2005.
Dargay, J., and Hanly, M. (2007). “Volatility of car ownership, commuting
mode and time in the UK.”Transp. Res. Part A: Policy Pract., 41(10),
934–948.
Davoudi, S. (2008). “Conceptions of the city-region: A critical review.”
Urban Des. Plann., 161(2), 51–60.
Ding, C. R. (2004). “Is the population density of Chinese cities high?”
City Plann. Rev., 28(8), 43–48 (in Chinese).
Efron, B., and Tibshirani, R. J. (1993). An introduction to the bootstrap,
CRC Press, Boca Raton, FL.
Ewing, R., Schroeer, W., and Greene, W. (2004). “School location and stu-
dent travel analysis of factors affecting mode choice.”Transp. Res. Rec.,
1895, 55–63.
GAQSIQPRC (General Administration of Quality Supervision, Inspection
and Quarantine of the People’s Republic of China). (1999). “National
standard of the People’s Republic of China: Electric bicycles—General
technical requirements.”GB17761-1999, Standards Press of China,
Beijing.
Giuliano, G., and Small, K. A. (1991). “Subcenters in the Los Angeles
region.”Reg. Sci. Urban Econ., 21(2), 163–182.
Giuliano, G., and Small, K. A. (1993). “Is the journey to work explained by
urban structure?”Urban Stud., 30(9), 1485–1500.
Gordon, P., Richardson, H. W., and Jun, M. J. (1991). “The commuting
paradox evidence from the top twenty.”J. Am. Plann. Assoc., 57(4),
416–420.
Green, N. (2007). Functional polycentricity: A formal definition in terms of
social network analysis.”Urban Stud., 44(11), 2077–2103.
Grunfelder, J., Nielsen, T. A. S., and Groth, N. B. (2015). “Changes to
urban form and commuting patterns: Trends in two Danish city
regions.”Danish J. Geogr., 115(2), 73–87.
Han, S. S. (2015). “Urban form and household CO2emissions A four-city
comparison.”Towards low carbon cities in China, urban form and
greenhouse gas emissions, S. Han, R. Green, and M. Y. Wang, eds.,
Routledge, London.
Han, S. S., and Wu, X. (2004). “Wuhan.”Cities, 21(4), 349–362.
Huang, J. N., Ding, W., and Han, J. Y. (2015). “Wuhan.”Towards low car-
bon cities in China, urban form and greenhouse gas emissions, S. Han,
R. Green, and M. Y. Wang, eds., Routledge, London.
Huo, H., et al. (2011). “Fuel consumption rates of passenger cars in China:
Labels versus real-world.”Energy Policy, 39(11), 7130–7135.
Huo, H., et al. (2012). “Projection of energy use and greenhouse gas emis-
sions by motor vehicles in China: Policy options and impacts.”Energy
Policy, 43, 37–48.
Kim, S., and Ulfarsson, G. F. (2008). “Curbing automobile use for sustain-
able transportation: analysis of mode choice on short home-based trips.”
Transportation, 35(6), 723–737.
Kloosterman, R. C., and Musterd, S. (2001). “The polycentric urban region:
Towards a research agenda.”Urban Stud., 38(4), 623–633.
Knaap, E., Ding, C., Niu, Y., and Mishra, S. (2016). “Polycentrism as a
sustainable development strategy: Empirical analysis from the state
of Maryland.”J. Urbanism: Int. Res. Placemaking Urban Sustainabil-
ity, 9(1), 73–92.
LCL (Livable City Laboratory). (2012). “China’s urban GDP ranking in
2012.”〈http://www.elivecity.cn/html/jingjifz/731.html〉(Dec. 9, 2014).
Lee, S., and Lee, B. (2014). “The influence of urban form on GHG emis-
sions in the U.S. household sector.”Energy Policy, 68, 534–549.
Li, J. (2005). The evolution of urban spatial form in modern times in
Wuhan, Changjiang Press, Wuhan, China (in Chinese).
Liu, L. Z., and Hou, K. (2009). “Study on gas consumption index for CNG
bus.”Gas Heat, 29(1), 20–22 (in Chinese).
Meyer, M., and Miller, E. (2001). Urban transport planning: A decision-
oriented approach, 2nd Ed., McGraw-Hill Book Co., Singapore.
Parr, J. (2004). “The polycentric urban region: A closer inspection.”Reg.
Stud., 38(3), 231–240.
Patterson, Z., Haider, M., and Ewing, G. (2005). “A gender-based analysis
of work trip mode choice of suburban Montreal commuters using stated
preference data.”Transp. Res. Rec., 1924, 85–93.
Pinjari, A. R., Pendyala, R. M., Bhat, C. R., and Waddell, P. A. (2007).
“Modeling residential sorting effects to understand the impact of the
built environment on commute mode choice.”Transportation, 34(5),
557–573.
Prevedouros, P. D., and Schofer, J. L. (1991). “Trip characteristics
and travel patterns of suburban residents.”Transp. Res. Rec., 1328,
49–57.
Santos, G., Maoh, H., Potoglou, D., and von Brunn, T. (2013). “Factors
influencing modal split of commuting journeys in medium-size
European cities.”J. Transp. Geogr., 30, 127–137.
Scheiner, J. (2010). “Interrelations between travel mode choice and trip
distance: Trends in Germany 1976–2002.”J. Transp. Geogr., 18(1),
75–84.
Schwanen, T. (2002). “Urban form and commuting behaviour: A cross-
European perspective.”Tijdschrift Voor Economische En Sociale
Geografie, 93(3), 336–343.
Schwanen, T., Dieleman, F. M., and Dijst, M. (2001). “Travel behaviour in
Dutch monocentric and polycentric urban systems.”J. Transp. Geogr.,
9(3), 137–186.
Schwanen, T., Dieleman, F. M., and Dijst, M. (2003). “Car use in Nether-
lands daily urban systems: Does polycentrism result in lower commute
times?”Urban Geogr., 24(5), 410–430.
SCXMPC (The Standing Committee of Xi’an Municipal People’s
Congress). (2002). “Regulations on the protection of historical and cul-
tural city of Xi’an”Xi’an, China (in Chinese).
Vandersmissen, M.-H., Villeneuve, P., and Thériault, M. (2003). “Analyz-
ing changes in urban form and commuting time.”Prof. Geogr., 55(4),
446–463.
© ASCE 05017014-14 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.
Vasanen, A. (2012). “Functional polycentricity: examining metropolitan
spatial structure through the connectivity of urban sub-centres.”Urban
Stud., 49(16), 3627–3644.
Veneri, P. (2010). “Urban polycentricity and the costs of commuting:
Evidence from Italian metropolitan areas.”Growth Change, 41(3),
403–429.
Wang, Y. Q., Li, L., Wang, Z. C., Lv, T., and Wang, L. (2013). “Mode shift
behavior impacts from the introduction of metro service: Case study of
Xi’an, China.”J. Urban Plann. Dev.,10.1061/(ASCE)UP.1943-5444
.0000148, 216–225.
Wang, Y. Q., and Liu, Y. Y. (2015). “Xi’an.”Towards low carbon cities in
China, urban form and greenhouse gas emissions, S. Han, R. Green,
and M. Y. Wang, eds., Routledge, London.
Wang, Y. Q., Wang, Z. C., Li, Z. Z., Staley, S. R., Moore, A. T., and
Gao, Y. N. (2012). “Study of modal shifts to bus rapid transit in Chinese
cities.”J. Transp. Eng.,10.1061/(ASCE)TE.1943-5436.0000523,
515–523.
Wang, Y. Q., Yang, L., Han, S. S., Li, C., and Ramachandra, T. V. (2017).
“Urban CO2emissions in Xi’an and Bangalore by commuters: Impli-
cations for controlling urban transportation carbon dioxide emissions
in developing countries.”Mitigation Adaptation Strategies Global
Change, 22(7), 993–1019.
WLRPB and WTDSI (Wuhan Land Resources and Planning Bureau
and Wuhan Transportation Development Strategy Institute). (2012).
“Wuhan transportation annual report of the year 2012.”Wuhan, China
(in Chinese).
WSB (Wuhan Statistics Bureau). (2014). “Wuhan national economy and
society developed statistical communiqué of the year 2013.”〈http://
www.stats-hb.gov.cn/wzlm/tjgb/ndtjgb/whs/whs/108937.htm〉(Apr. 10,
2014).
XBS and NBSXIT (Xi’an Bureau of Statistics and National Bureau of
Statistics Xi’an Investigation Team). (2013). Xi’an statistical yearbook,
China Statistics Press, Xi’an, China (in Chinese).
XCTMCC (Xi’an City Traffic Management Committee and Chang’an
University). (2012). “Documentation of Xi’an transport development
annual report of the year 2012.”Xi’an, China (in Chinese).
Xianyu, J. C., and Juan, Z. C. (2008). “Travel behavior and mode choice of
urban motorcycle.”J. Transp. Syst. Eng. Inf. Technol., 8(5), 136–140
(in Chinese).
XMG (Xi’an Municipal Government). (2005). “Imperial Renaissance
Project of Xi’an.”Xi’an, China (in Chinese).
Yang, L., Wang, Y. Q., Han, S. S., Li, C., Liu, Y. Y., and Ren, Q. (2017a).
“Carbon dioxide emissions from commuter traffic in Xi’an, China.”
Proc. Inst. Civil Eng. Transp., 170(1), 8–18.
Yang, L., Wang, Y. Q., Han, S. S., and Liu, Y. Y. (2017b). “Urban transport
carbon dioxide (CO2) emissions by commuters in rapidly developing
cities: The comparative study of Beijing and Xi’an in China.”Transp.
Res. Part D: Transp. Environ, in press.
Yang, L., Wang, Y. Q., and Li, C. (2015). “A mode choice behaviour study
of bicycle travelling.”Advances in civil engineering and building
materials IV, CRC Press, Boca Raton, FL, 401–404.
Yin, Y., Mizokami, S., and Aikawa, K. (2015). “Compact development and
energy consumption: Scenario analysis of urban structures based on
behavior simulation.”Appl. Energy, 159, 449–457.
Zhang, J., Fujiwara, A., and Soe, T. (2008). “Capturing travelers’stated
mode choice preferences under the influence of income in Yangon city,
Myanmar.”J. Transp. Syst. Eng. Inf. Technol., 8(4), 49–62.
Zhang, S. J., et al. (2014). “Real-world fuel consumption and CO2
emissions of urban public buses in Beijing.”Appl. Energy, 113,
1645–1655.
© ASCE 05017014-15 J. Urban Plann. Dev.
J. Urban Plann. Dev., 2018, 144(1): 05017014
Downloaded from ascelibrary.org by Yuanqing Wang on 12/08/17. Copyright ASCE. For personal use only; all rights reserved.