You are on page 1of 14

Journal of Transport Geography xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Journal of Transport Geography


journal homepage: www.elsevier.com/locate/jtrangeo

Connectivity of intercity passenger transportation in China: A multi-modal


and network approach
Zhenran Zhua, Anming Zhanga,c,⁎, Yahua Zhangb
a
Sauder School of Business, University of British Columbia, 2053 Main Mall, Vancouver, BC V6T 1Z2, Canada
b
School of Commerce, University of Southern Queensland, Toowoomba, Queensland, Australia
c
China Academy of Financial Research, Shanghai Jiao Tong University, China

A R T I C L E I N F O A B S T R A C T

Keywords: This paper proposes a dynamic weighted model to measure the connectivity of intercity passenger transportation
Connectivity in China. We consider both quality and quantity of the connections of two transport modes: air and rail. Among
Intercity passenger transportation the 23 major cities selected, Shanghai is revealed to have the highest connectivity level, leading in both air and
Air and rail transport rail connectivity. Hong Kong, Kunming, and Urumqi are the three cities that predominantly rely on air
Node connectivity
transportation whose contribution to the connectivity exceeds 80%. This research also suggests that the
Link (route) connectivity
connections between international cities and China's domestic network are highly concentrated on a few cities,
namely, Hong Kong, Shanghai, Beijing, and Guangzhou, and that Seoul is the best connected international city in
terms of its transport links with China. Shanghai-Nanjing has been found to be the best-connected city pair,
primarily due to the significant contribution from high-speed rail (HSR) service. Our study shows that the
contribution from train service is more than 80% for 19 of the 20 top-ranking domestic routes measured by
connectivity. In addition, HSR has become a preferred and dominant option over air on a number of long-
distance routes up to 1,300km. This finding has significant policy implications for transportation infrastructure
planning and investment.

1. Introduction There is a strong relationship between transportation connectivity


and regional economic development. It is a widely held view that as an
The landscape of China's transportation network has been con- input into many economic activities including tourism, trade and
stantly shaped by the fast-growing air transportation industry and the investment, an efficient transportation system has been an important
continually evolving rail system in the last three decades. The change component in achieving economic development and welfare enhance-
has been revolutionary in recent years with the introduction of the ment. For example, air transportation is particularly important to
high-speed rail (HSR) services and the proposed “one belt one road” distant and remote regions where there is no close substitute for this
(OBOR) initiative, which involves heavy investment and planning in transport mode due to the tyranny of distance. A large volume of
transportation infrastructure including HSR, international rail, ports, literature has reported the causality relationship between transporta-
and airports, with an aim to build better connections between China tion infrastructure and local economy (e.g., Li and Qi, 2016, and studies
and the rest of the world. Many Chinese cities are embracing this idea cited therein), and air accessibility has a significant impact on GDP,
and seizing the opportunity to redesign their long-term planning and employment, regional development, and foreign direct investment
investment strategies, in the hope of aligning themselves with the (FDI) (e.g., Brueckner, 2003; Basile et al., 2006; Zhang, 2012; Banno
OBOR blueprint and thus attracting more transportation infrastructure and Redondi, 2014). Tanaka (2016) notes that despite the advances in
investment funds. However, the resources are scarce and limited. It is information and communication technology that have reduced the
important first to assess the accessibility and connectivity of the barriers to acquiring codified and explicit information on foreign
existing transportation network, and to understand the current status market, in-person meetings are still crucial for building business
that each city plays to make sure that the resources will be put to more relationships and managing production activities. Banno and Redondi
efficient use in promoting the connectivity of domestic cities and (2014) argued that the introduction of a new route would lead to a
building a better connected transportation network. reduction in total transportation costs and facilitate knowledge flow,


Corresponding author at: Sauder School of Business, University of British Columbia, 2053 Main Mall, Vancouver, BC V6T 1Z2, Canada.
E-mail addresses: anming.zhang@sauder.ubc.ca (A. Zhang), yahua.zhang@usq.edu.au (Y. Zhang).

http://dx.doi.org/10.1016/j.jtrangeo.2017.05.009
Received 18 January 2017; Received in revised form 16 April 2017; Accepted 23 May 2017
0966-6923/ © 2017 Elsevier Ltd. All rights reserved.

Please cite this article as: Zhu, Z., Journal of Transport Geography (2017), http://dx.doi.org/10.1016/j.jtrangeo.2017.05.009
Z. Zhu et al. Journal of Transport Geography xxx (xxxx) xxx–xxx

thereby increasing the likelihood of FDI exchange between newly business opportunities (e.g. if a city is ranked low, there may be
connected areas. Cristea (2011) provided evidence for the importance opportunities for improvements and investments), and comparing
of in-person business meetings in international trade, and pointed out alternatives and choosing the best option.
that interactions among trade partners generate relationship capital, Given that connectivity has been an important performance indi-
which adds bilateral specific value to the traded products. It is thus cator for an airport or a city (like efficiency and productivity), in this
expected that improved transportation connectivity of a city generates article we attempt to develop a dynamic connectivity measure to
significant economic benefits not only to individual travelers but also to capture the aggregated connectivity of multi-transport modes including
businesses and local economy. both air and rail services. Our connectivity measure incorporates both
In transportation, the concept of connectivity was first introduced to quality and quantity of the connections, which will be used to evaluate
evaluate the importance of an airport in terms of its connection to other the performance of 23 major Chinese cities and 13 international cities.
airports. Scholars have developed various airport connectivity mea- The rest of the article is organized as follows. Section 2 presents the
sures, representing the degree to which an airport is connected to other research methodology and data. Section 3 reports the results and
airports in a given network. This will help policy makers monitor the discusses, in particular, the domestic city (node) connectivity and route
network performance against that of other airports, airline networks (link) connectivity. The last section contains concluding remarks and
and regions, and design strategies to improve the competitive positons policy implications.
of airports (Burghouwt and Redondi, 2013). Passengers tend to have
different preferences over travel time, travel distance and destinations,
2. Methodology and data
as well as other travel dimensions. Therefore, different connectivity
measures have been developed, taking travel purposes and many other
2.1. Background and the choice of cities for this study
factors into consideration with more detailed and quantitative data to
make the measure objective. Burghouwt and Redondi (2013) conducted
Both the airline and rail markets in China are experiencing fast
a good review of the various measures for the air connectivity,
expansion, underpinned by a huge population, rapid economic growth,
including the shortest path length accessibility model (Shaw, 1993;
and policy reforms (e.g., Lei and O'Connell, 2011; Fu et al., 2012; Zhang
Shaw and Ivy, 1994; Malighetti et al., 2008), the quickest path length
and Zhang, 2016). In 2015 the length of new railways put into
accessibility model (Malighetti et al., 2008; Paleari et al., 2010), the
operation totaled 9531 km, of which HSR measured 3306 km. By the
weighted number of connections model (WNX) (Burghouwt and de Wit,
end of 2015, the length of China's railways reached 121, 000 km,
2005), and the Netscan connectivity unit (NCU) model (Veldhuis, 1997;
ranked the second largest rail network in the world after the United
Burghouwt and Veldhuis, 2006; Veldhuis and Kroes, 2002; Matsumoto
States (US), while the HSR system, whose length amounted to
et al., 2008). Boonekamp and Burghouwt (2017) note that more
19,000 km, ranked number one (in effect, it was greater than the HSR
recently developed models incorporate the connection quality, fre-
system of the rest of the world combined). The volumes of passenger
quency, importance of the destination, etc.
and freight traffic delivered by the rail system were also the largest in
However, most of the existing studies only focus on the connectivity
the world. In 2015, eight new airports were put into use, bringing the
of a single transport mode, while in many countries, there exist several
total number of China's civil aviation airports to 210. In 2015 China's
main transport modes which complement each other. This is especially
air transportation industry handled 435.6 million passengers and
so in China where air transportation and rail services are the two
6.3 million tons of air cargo, an 11.1% and 5.2% increase from the
dominant transport modes for long-distance travels. Many studies have
previous year, respectively. The International Air Transport Association
shown that HSR and air transport are potential substitutes. In the case
(IATA) forecasts that China will overtake the US as the largest air
of China, for example, Fu et al. (2012) indicated that the demand for
passenger market by around 2030 as measured by traffic to, from and
airline flights between Wuhan and Shanghai/Nanjing/Hangzhou fell
within a country.
sharply following the launching of HSR between these cities in 2009. As
Considering the large size of the air and train services data, we only
a result, airlines cut services between Wuhan and the Yangtze River
select 23 domestic cities, which include Hong Kong Special
Delta area by about one-third.1 It is, therefore, necessary to consider
Administration Region, and assess their connectivity on the basis of
both transport models when considering a city's connectivity. A
their significance and influence to the surrounding areas. Hong Kong is
connectivity measure based on any one model alone may not provide
a leading international hub in terms of air passengers handled and the
accurate and useful information to decision makers. Furthermore,
world's number one measured by international air cargo volume (Wan
government initiatives such as OBOR will have an impact on the spatial
and Zhang, 2017). The other 22 cities have well-developed multi-modal
transformation of transport systems in the form of, e.g., connectivity
transport system serving their home province and neighboring pro-
changes. To see such an impact, measuring connectivity is a starting
vinces: they together accounted for 31.5% of China's GDP (excluding
point: Connectivity between two cities shows the convenience of
Hong Kong) and 17.2% of its population in 2014.2 Together, these 23
transportation between the city pair and also, reflects how much the
cities and the flight and train links between them form a representative
two cities rely on (and relate to) each other in an economy, whereas
network of all Chinese major cities. The locations of the 23 cities are
connectivity between a city and all other cities in the network shows
shown in Fig. 1. A map of China's planned HSR network including
the degree of the city's importance in the network.
completed and planned projects is shown in Fig. 2.
Quantifying, and visualizing the overwhelming connection data
In addition, 13 international cities are selected based on the ranking
with dynamic models will trace the spatial transformation of transport
of their air passenger throughput, locations, and the economic and
systems, which often is influenced by various government initiatives.
regulatory connections with China.3 These international cities are
They in turn will assist governments for 1) better monitoring and
included to provide the show case of international connectivity.
planning of their initiatives; and 2) optimal management and control of
Table 1 explains the rationale of the choice.
the transformation. The implications for management and investors
include: facilitating networking and routing decisions, identifying
2
A city's GDP and population include the statistics of the districts, counties and county-
level cities that this city governs.
1 3
See also Wang et al. (2017). For works on HSR, as well as HSR-air transport As discussed in Zhang and Chen (2003), the international airline markets used to be
interactions, in Japan and European countries, see, e.g., Gutierrez (2001), Givoni (2006), heavily, and unevenly, regulated between China and other countries. Fortunately, after
Yamaguchi et al. (2008), Dobruszkes (2011), Ha et al. (2011), Behrens and Pels (2012), several rounds of liberalization in bilateral air service agreements, by 2016 (our sample
Givoni and Dobruszkes (2013), Clewlow et al. (2014), Fu et al. (2014), Cheng et al. year) the markets are fairly liberalized in the 3rd and 4th freedom rights for the sample
(2015), Perl and Goetz (2015), and Wan et al. (2016). international cities (Lei et al., 2015) which facilitates our analysis with these cities.

2
Z. Zhu et al. Journal of Transport Geography xxx (xxxx) xxx–xxx

Fig. 1. Locations of selected domestic cities.

Fig. 2. Planned HSR network in China.


Adapted from SDPC (2016).

The data for the flight frequency were obtained from the website of source for the train services is the website of 12306, the only official
FEEYO, managed by the Civil Aviation Resource Net of China. FEEYO is train service online booking website operated by the Ministry of China
the largest online platform specializing in air passenger service and Railways. The data were extracted from the two websites on May 11,
flight data analysis in China. By cross-checking with other major flight 2016 for all the routes out of the cities under study. R was used for data
information sources, such as Ctrip, Qunar, and AliTrip, we found that screening and calculation. For code-shared flights, we only consider the
FEEYO's schedule data are the most complete and accurate among those operating airlines.
available to us when we started this project. The flight information also For air connections, when FEEYO indicates that a flight has an
contains stops and on-time (punctuality) performance rates. We later intermediate stop, this flight will be counted as an indirect flight from
doubled checked the flight schedule information in IATA's Airport origin to destination, and a direct flight from origin to the intermediate
Intelligence and confirmed the reliability of the FEEYO data. The data point, as well as a direct flight from the intermediate point to the final

3
Z. Zhu et al. Journal of Transport Geography xxx (xxxx) xxx–xxx

Table 1
Cities chosen for this study.

City Choice rationale

Beijing, Tianjin Beijing and Tianjin are both municipalities directly under the central government. The two cities are linked by HSR that
allows the intercity travel to be within 30 min. They both have well developed airports and rail systems connecting other
parts of China.
Dalian, Shenyang Dalian and Shenyang are two of the major cities in Northeast China. Both of them are sub-provincial cities, which means
that their importance is only slightly lower than Beijing, Tianjin, Shanghai and Chongqing, which are directly controlled
by the central government. Both Shenyang and Dalian have well-connected air transport and rail systems.
Shanghai, Nanjing, Hangzhou, Ningbo These four cities are major cities in the Yangtze River Delta in East China, the most developed and wealthy area in China.
Shanghai is China's economic and financial center. Nanjing, Hangzhou and Ningbo are all sub-provincial cities.
Guangzhou, Shenzhen, Hong Kong They are the leading cities in the Pearl River Delta in South China, another well-developed and wealthy zone in China.
Guangzhou and Shenzhen are both sub-provincial cities. Hong Kong is the center of finance, transportation and logistics in
Asia.
Zhengzhou, Xi'an, Wuhan, Changsha They are respectively the capital cities of Henan Province, Shaanxi Province, Hubei Province and Hunan Province. These
cities are linked by HSR. They are also key air transportation hubs in central China.
Chongqing, Chengdu Both cities are important commercial cities in West China with a long history. Chongqing is the largest city in China in
terms of the population and land area. Chengdu is the capital city of Sichuan Province, and a sub-provincial city.
Guiyang, Kunming, Nanning They are the capital cities of Guizhou Province, Yunnan Province, and Guangxi Province, respectively.
Xiamen Xiamen is a sub-provincial city, a famous tourism city in Southeast China on the west side of Taiwan Straights
Urumqi, Yining Urumqi is the capital of Xinjiang Uyghur Autonomous Region. It's the largest city in Xinjiang and the only large city that is
close to China's Northwest border. It has close commercial, cultural and transportation links with Central Asia. Yining is a
relatively young and small city, while its location is strategically important as it is the gateway to Central Asia and Europe.
Seoul, Taipei, Tokyo, Singapore, Osaka, Nagoya All of them are important cities in East Asia with strong cultural and economic links with China. These cities are also the
main tourism destinations for Chinese tourists. Connectivity of these cities with China shows the passenger travel pattern
between China and the neighborhood region of East Asia.a
New York, Los Angeles, Atlanta, London, Paris, These cities are major international air transportation hubs. Connectivity of these cities with China represents the
Frankfurt, Dubai passenger travel pattern between China and global areas on long-distance routes.

a
We acknowledge that the choice of international cities is still somewhat arbitrary, although we have considered their geographic locations, economic significance and economic ties
with China that have been found to be important for transport connectivity (e.g., Gutierrez, 2001; Geurs and Van Wee, 2004). However, one of the main purposes of this paper is to
present a new connectivity model. Using a simplified transport network eases the data calculation burden and the simplified network still gives a good representation of the reality.

destination. In our flight data, 655 flights out of 6588 have one which is a standard practice in the literature. Similarly, transfers
intermediate stop.4 between trains5 or between air and rail are not considered in this
study. This is a realistic assumption as these two practices are not
common in China.6 It is understood that without considering these
2.2. Assumptions made for constructing our connectivity model
transfers, the connectivity level of a city, especially those cities relying
on another hub city, is underestimated. However, we believe that this
Any connectivity measure involves some realistic assumptions so as
bias is minor for the network considered in this paper: the under-
to simplify the model such as the maximum transit time and maximum
estimation for the connectivity between domestic cities is slight,
number of steps allowed (Burghouwt and Redondi, 2013; Boonekamp
because all the cities in the domestic network under study are regional
and Burghouwt, 2017). The following assumptions are made in
hubs and the contribution of indirect connections to the overall
constructing our connectivity measure.
connectivity is limited, owing in part to the poor quality of such
Frist, we only consider the passengers traveling by air and by train
indirect connections.
although intercity passenger transport modes include means of air, rail,
Fifth, we assume that a rail connection between two major cities has
highway, and water. However, for long-distance travel, the vast
a capacity equal to that of Airbus 380 (A380). Normally the capacity of
majority of Chinese passengers use air and train services. Also due to
a train is 2 to 4 times as large as that of an A380. There are 506 seats on
the unavailability of data for highway and water transportation, these
a typical A380. A basic unit of China Railway High-speed (CRH), which
transportation means are not included in our connectivity measure.
is the mainstream train type for HSR services, has 494 to 610 seats. A
Second, we only consider a network formed by the 23 domestic
CRH train (i.e., D-train or G-Train)7 between major cities usually
cities and 13 international cities, meaning that only connections
operates with 2 units, i.e., 988 to 1220 seats per train. However, as
between the 23 domestic cities and connections between the 23
there are multiple stops between the origin city and destination city,
domestic cities and the 13 international cities are considered.
and we are unable to obtain the accurate information about the
Connections between international cities are not considered.
capacity used for transporting people between origin and destination,
Third, for the schedules of flights and trains, it is assumed that they
are cyclical with a cycle of one week. It is recognized that there are
frequent delays in air transportation. However, due to the unavail- 5
Rail-rail transfers (especially for K-train services with a speed not more than 120 km/
ability of flight delay data, flights and trains are all considered to h) is unusual for the following two reasons. Firstly, purchasing indirect ticket bundles is
depart/arrive on time as scheduled. not supported by the official platform. Tickets must be bought separately for every
section. Secondly, end-to-end services are available for most city pairs as long-haul
Fourth, for indirect travel by air, we only consider scheduled
transportation is the major concern for China Rail. Short-haul end-to-end trains are only
transfers noted in the database of FEEYO. Transfers that are arranged applied to a few routes, like Beijing-Tianjin, Shanghai-Nanjing, etc. However, with the
by passengers themselves (so-called “self-connecting”) are not counted, extension of the HSR network, more and more time-sensitive passengers choose to travel
to their destinations via HSR with one or two transfers even though the direct K-train
services are available between the origin and destination. This change should be
4
Our results suggest that flights with one stop account for 9.94% of the total flights considered in future research. Without considering transfers between trains, rail
counted by flight number, 8.96% of the total flights counted by frequency, 5.87% of the connectivity of some cities especially those such as Wuhan and Zhengzhou will be
total air connectivity considering quality discount of aircraft size and travel length. The underestimated.
6
importance of indirect connectivity for one specific route is significantly correlated with In particular, the unpopularity of air-rail transfers is due largely to the lack of
the natural logarithm connectivity level of the route, with p-value = 0.000 and Pearson integration of the two transport modes (Chen and Lin, 2016; Xia and Zhang, 2016b).
correlation = −0.765. Indirect connections are more important for weakly connected 7
D-train operates at a speed between 200 and 250 h while the maximum speed of HSR
routes, although its impact on the overall connectivity is negligible. train (G-train) is usually at 300 km/h.

4
Z. Zhu et al. Journal of Transport Geography xxx (xxxx) xxx–xxx

it is reasonable to assume that half of the G- or D-train capacity, We first retrieved the connection data from the websites of FEEYO
equivalent to the capacity of A380, is allocated to the origin and and 12306, including flight number/train number, aircraft type, origin
destination cities.8 Other main types of trains operating in China such city, destination city, scheduled departure time, and scheduled arrival
as K-train9 usually operate with 16 cars, with 110 to 128 seats per car, time. Next, we assign discount factors, ranging between 0 and 1, to
that is, 1760 to 2048 seats per train. As these types of trains have more every connection. The discount factors for air connections include
A
stops between two cities, we still assume that the capacity used between capacity discount (denoted as DxyaC), travel time discount (DxyaT ), and
the two cities is equivalent to that of an A380. velocity discount (DijV). The discount factors for train connections are
T
Finally, it should be noted that prices are not considered in the travel time discount (DxybT ), and velocity discount.
C
model, which means that G-trains (more expensive but faster) are Dxya is used to normalize the effect of different sizes of the aircraft
always preferred to K-trains (cheaper but slower). In reality, passengers on the capacity of the connection. A flight using aircraft A380 and a
will not always choose the faster connection. Their utility may be likely flight using aircraft A320 provide different levels of capacity. DxyaC is
influenced by factors such as the prices. calculated based on the aircraft type used in the connection and the
number of seats offered, as shown in Eq. (1) below. For flight a from
2.3. Dynamic weighed model measuring connectivity city x to city y, DxyaC is the square root of the ratio of the number of
seats on this flight to the number of seats of the largest aircraft. This
As reviewed in Burghouwt and Redondi (2013), the shortest path function form allows routes with higher frequency to have higher
length accessibility model uses the average number of steps needed to connectivity, ceteris paribus. That is that a flight with 200 seats
reach any other airport in the network to measure the connectivity. It generates smaller connectivity than two flights with 100 seats each.
only considers the steps, but ignores all other factors that could affect When flight a operates with the biggest aircraft, the capacity discount
the accessibility, such as flight duration and ticket availability, etc. for flight a is 1. The capacity discount is 1 for all train connections as
Also, it only considers the shortest path, while in fact, the other feasible the capacity of train connections between two cites is assumed to be
paths also contribute to the connectivity of that route. In this article, we equivalent as an A380.
use travel time instead of steps as the major indicator. Seatxya
C
The quickest path length accessibility model uses travel time needed Dxya =
max Seatxya
to reach other cities to measure the connectivity. Similar to the shortest x , y, a (1)
path length accessibility model, it only considers the best connection
option from origin to destination while ignoring the second-best The travel time discount is to normalize the effect of different travel
options. Also, it uses the throughput of the airport to measure the length within a single travel mode. A four-hour connection from
airport's connectivity instead of the actual flow traffic on a particular Shanghai to Beijing provides a better accessibility level than an 8-hour
route. Without measuring the route level capacity, this model does not connection on the same route. Passengers usually select connections
capture the bilateral connectivity between cities. In the model proposed based on the travel length (in the sense of total transportation costs).
in this article, apart from using time as an important indicator to However, the travel time discount is calculated based on the adjusted
represent the quality of the connection, all connections as well as route travel length instead of the actual travel length of a connection. The
level capacity from origin to destination are considered. adjusted travel length is made up of travel duration, transfer time, and
The weighted number of connections model (WNX) weighs each extra time needed at airport/station. Even though in some cities, the
connection by its quality in terms of transfer and detour time. The WNX distance of the HSR station from the city center is similar to that of the
has considered the disadvantage of additional time for transfer for a airport, air passengers normally need to spend more time at airports
connection, which is also included in our model. However, the WNX because they need to arrive early to go through more procedures, such
does not account for the different capacity sizes for the connections and as check-in, baggage drop, security check, etc. They also need to spend
the relative time quality of non-stop connections for the same route, extra time at destination airport to pick up baggage. Therefore, 1.5 h is
which will be considered in our model. added to all air connections as most airlines require passengers to
The Netscan connectivity unit (NCU) model has included discount report to the check-in counter 2 h before departure.10 Furthermore, it is
for the actual time of a connection compared with the theoretical direct well recognized that the time spent at a transfer stop is more
flight time. However, the theoretical direct flight time is very roughly uncomfortable to passengers than the time spent in a moving plane
estimated, which may not reflect the true minimum travel time for most because of the risk of losing baggage or missing connecting flight and
of the cases. In this article, we use the best connection of the route as the inconvenience of physical movements to get to a different gate
the dynamic benchmark which varies for every route. It enables us to (Burghouwt and de Wit, 2005; Burghouwt and Veldhuis, 2006;
capture a more realistic state of the connection. The measure result Burghouwt and Redondi, 2013). Therefore, the value of time for time
evolves dynamically when a better connection is built for one route. spent at transit airports is higher for passengers. In this research, we
More specifically, a dynamic weighted model (DWM) is developed assume an extra penalty of 1 h for each stop that a flight makes to
in this study to quantify the quality of a connection. This model is an represent the inconvenience of flights with intermediate stops. The
extension of the NCU model. The NCU model only considers one shortest adjusted travel length by air from city i to city j is denoted as tA
ij .

dimension of quality for flights, which is time, while our DWM includes The shortest adjusted travel length by train from city i to city j is
capacity as an additional dimension measuring connectivity quality. It denoted as tTij. The calculations of tAij and tTij are shown in Eqs. (2) and
A T

is also the first time that the NCU model is applied to measure train (3), respectively. The travel time discount (DxyaT , DxybT ) is calculated
connectivity. A connection is converted to the dynamic weighed by comparing the adjusted travel length of a flight (txyaA) or a train
connectivity unit (DWCU) by applying various discount factors which (txybT) with the shortest adjusted travel length of that mode (tijT, tijA), as
quantify different dimensions of its quality. The construction of the shown in Eqs. (4) and (5). The travel time discount is 1 for the
prototype of DWM and calculation of numeric results are as follows. connection on a route with the shortest adjusted travel length.

tijA = min txya


A

8
x = i, y = j (2)
We understand that the assumption of the capacity for trains between two cites is
again somewhat arbitrary due to the lack of data. We hope that when the ticket pre-
allocation data becomes available in the future, the capacity of trains will be better
10
estimated. This means that we do not consider each individual city's actual locations of their
9
K-train is slower than D-train and HSR train, with the maximum speed being around airport and train station. Future research should account for this for a more accurate
120 km/h. connectivity measure.

5
Z. Zhu et al. Journal of Transport Geography xxx (xxxx) xxx–xxx

tijT = min txyb


T
Connectivityx = ∑ Connectivityij + ∑ Connectivityij
x = i, y = j (3)
j=x i=x (12)

T A tijA In our DWM, we assume that all the discount factors and weights are
Dxya = A
txya (4) quasi-concave functions, that is, we assume a diminishing marginal
effect. For example, we consider the connectivity of a flight with 400
T tijT seats contributes less than 4 times that of a flight with 100 seats does.
T
Dxyb = T All the discount factors are calculated by square-rooting the ratio,
txyb (5)
because the square root keeps the relative magnitude when a quality
The velocity discount (DijV) is to normalize the difference of the dimension is referred to and compared with for more than once, e.g.,
connection quality between different routes. The velocity discount is travel time. The use of square root avoids repeated connectivity scaling
necessary to reveal the actual connectivity level. When the connection down when taking the product of mode weighing factor, travel time
time for two cities of 10 km apart is the same as the connection time for discount, and velocity discount.
two cities of 100 km apart, the latter has a better connection. The
velocity discount is calculated by comparing the route's highest average 3. Results and analysis
speed with the global highest average speed, as shown in Eq. (6):

distanceij min (tijT , tijA) 3.1. Domestic city connectivity


DijV =
⎛ distance ⎞
max ⎜⎜
ij
⎟⎟ The proposed DWM considers multiple transport modes. The
i, j ⎝ min (tijT , tijA ) ⎠ connectivity of a city shows its status in the city network. A city with
(6)
a high connectivity represents a high level of importance in the
The single-mode DWCU is calculated by taking the product of all network. Naturally, a city with a high connectivity is more likely to
discount factors and the frequency of connections per time unit (day, serve as a transit hub. The domestic city global (DCG) connectivity is
week, or year). As shown in Eq. (7), the single-mode DWCU from city i calculated by aggregating the connectivity of all the routes connecting
to city j by flight a is calculated first. The single-mode DWCU from city i the city with all other cities in our domestic city network and the
to city j by air transportation (ConnectivityijAir) is calculated by international city network. The DCG connectivity shows the compre-
aggregating the single-mode DWCU of all flights from city i to city j. hensive connectivity level for a city. The ranking of the 23 domestic
Similarly, in Eq. (8), the single-mode DWCU from city i to city j by train cities in terms of the DCG connectivity is shown in Table 2 and Fig. 3.
b (ConnectivityijTrain) is calculated first. The single-mode DWCU from Shanghai, Beijing, and Guangzhou, China's first-tier cities in terms
city i to city j by rail is the aggregate of the single-mode DWCU of all of income, are the three most important nodes in our domestic city
trains from city i to city j. network. They are evenly distributed in North, East and South China
ConnectivityijAir = C T A along the east coast line. Shanghai beats Beijing by 3.45% in rail
∑ Frequencyxya × Dxya × Dxya
x = i, y = j (7) connectivity and 10.79% in air connectivity. Shanghai and Beijing
exhibit a huge advantage over Guangzhou. Shanghai's rail connectivity
T
ConnectivityijTrain = ∑ T
Frequencyxyb × Dxyb is only 11.74% higher than Guangzhou, while Shanghai performs
x = i, y = j (8) 96.71% better than Guangzhou in air connectivity. This is fully
expected, because Guangzhou Airport faces intensive air competition
The final DWCU, which combines the DWCU of multi-modes, is
from airports in Hong Kong and Shenzhen. In the Pearl River Delta
calculated by taking weighed sum of the single-mode DWCUs. The
region, within a circle of 140 km diameter, three airports have very
reason for taking a weighed sum is that the DWCU of different modes
close connectivity levels (1752.54 for Guangzhou, 1963.49 for Hong
has been calculated against different standards of travel time. The
Kong, and 1483.31 for Shenzhen), while Beijing and Shanghai have
weight brings different modes to the same frame of reference. The
achieved outstanding performance in air connectivity compared with
weight is calculated by comparing the shortest travel time within
the nearby airports in their respective regions (3447.44 for Shanghai,
different modes, which ensures a uniform discounting standard in travel
but only 1073.84 for Hangzhou and 778.86 for Nanjing; 3111.79 for
time. As shown in Eqs. (9) and (10), when air transportation has shorter
Beijing, but only 652.46 for Tianjin). Hong Kong has a historical
adjusted travel length than rail transportation on a route, the weight of
dominant advantage in international connectivity, and Shenzhen is
the air mode (WijAir) will be 1. The weight of the rail mode (WijTrain) will
adjacent to Hong Kong, which gives Shenzhen a small advantage in
be accordingly smaller than 1.
being a transit hub for domestic passengers heading for Hong Kong.
⎛ tijT ⎞ Shenzhen and Hong Kong, two strong competitors, have undoubtedly
WijAir = min ⎜⎜ A , 1⎟⎟ constrained the growth of air transport in Guangzhou.
⎝ tij ⎠ (9) Wuhan, Nanjing, Changsha, Hangzhou, Shenzhen, and Zhengzhou
are in the second echelon in terms of the overall DCG connectivity. All
⎛t A ⎞
WijTrain = min ⎜⎜ T , 1⎟⎟
ij these cities are connected by high-speed rail with each other and with
⎝ tij ⎠ (10) other major cities. As a result, a noticeable feature for these cities is that
rail connectivity accounts for a higher percentage than air in their
The aggregated route DWCU shows the performance of one route, overall connectivity. All of them are key transport hubs in the HSR
and of the level of connectivity between an origin-destination pair. The network. Wuhan is at the junction of Beijing-Guangzhou HSR and
multi-mode DWCU from city i to city j can be calculated with Eq. (11): Shanghai-Chengdu HSR, while the Beijing-Guangzhou HSR and
Connectivityij = DijV × (WijAir × ConnectivityijAir + WijTrain × ConnectivityijTrain ) Shanghai-Kunming HSR meet at Changsha. Wuhan is only 1.5 h away
from Changsha by HSR and 3–4 h away from Guangzhou. Nanjing is the
(11) intersection of the Beijing-Shanghai HSR and Shanghai-Chengdu HSR.
The aggregation of the DWCU from and to one city shows the Hangzhou has extensive HSR links to Shanghai, Nanjing, Ningbo,
connectivity level between the city and the rest of the cities in the Kunming, etc. The HSR lines have connected Shenzhen to Guangzhou
network. It also shows the importance of the city (node) in the network. and Xiamen, and more HSR lines under construction.
When cities have multiple airports, connectivity is aggregated for all Zhengzhou is the capital city of Henan Province which is tradition-
airports. The multi-mode DWCU of city x is calculated with Eq. (12): ally regarded as the geographical center of China (a position similar to

6
Z. Zhu et al. Journal of Transport Geography xxx (xxxx) xxx–xxx

Table 2
Domestic city global (DCG) connectivity values of a city.

Ranking City DCG connectivity Air Rail Air percentage Rail percentage

1 Shanghai 7324.283 3447.444 3876.839 47.07% 52.93%


2 Beijing 6859.250 3111.790 3747.460 45.37% 54.63%
3 Guangzhou 5222.059 1752.545 3469.514 33.56% 66.44%
4 Wuhan 4231.772 649.492 3582.280 15.35% 84.65%
5 Nanjing 4173.083 778.859 3394.224 18.66% 81.34%
6 Changsha 4139.328 599.265 3540.063 14.48% 85.52%
7 Hangzhou 4030.799 1073.837 2956.962 26.64% 73.36%
8 Shenzhen 3934.420 1483.306 2451.114 37.70% 62.30%
9 Zhengzhou 3419.829 638.527 2781.302 18.67% 81.33%
10 Tianjin 2924.357 652.460 2271.897 22.31% 77.69%
11 Hong Kong 2396.591 1963.487 433.104 81.93% 18.07%
12 Xian 2190.101 1112.199 1077.903 50.78% 49.22%
13 Chengdu 2012.041 1310.900 701.141 65.15% 34.85%
14 Shenyang 2004.061 688.329 1315.731 34.35% 65.65%
15 Chongqing 1941.790 1123.267 818.522 57.85% 42.15%
16 Xiamen 1615.640 901.225 714.415 55.78% 44.22%
17 Kunming 1573.660 1321.971 251.689 84.01% 15.99%
18 Ningbo 1527.691 255.286 1272.405 16.71% 83.29%
19 Guiyang 1316.484 479.285 837.199 36.41% 63.59%
20 Dalian 1232.488 731.376 501.112 59.34% 40.66%
21 Nanning 985.908 418.296 567.612 42.43% 57.57%
22 Urumqi 923.469 752.963 170.506 81.54% 18.46%
23 Yining 48.945 45.895 3.050 93.77% 6.23%

Kansas City in the US). Zhengzhou is an inland transport hub with two tributes little to Hong Kong's overall connectivity. There are only 12
major trunk railway lines, Longhai Railway from east to west, and trains from Hong Kong to Guangzhou every day. Trains from Hong
Beijing-Guangzhou from north to south meeting here. Beijing- Kong to Shanghai and Beijing are operated for leisure trips, with one
Guangzhou and Shanghai-Xi'an HSR lines also meet at Zhengzhou and train every two days. Therefore, air is the major transport mode
both have been in operation. Several HSR lines linking Zhengzhou- connecting Hong Kong to other parts of China.
Chongqing, Zhengzhou-Hefei, Zhengzhou-Taiyuan, Zhengzhou-Jinan For Kunming, the dependence on air is mainly due to its well-
and Zhengzhou-Lanzhou are being constructed or soon to be con- developed tourism and a lack of HSR operations. Its home province,
structed (Wang et al., 2017). Eventually, Zhengzhou will be a crisscross Yunnan, is geographically large and mountainous. Air transport links to
base of an HSR network. Two cargo-rail lines to Europe cross through other parts of China and Southeast Asia are fairly developed. Kunming
Zhengzhou with one going north to Russia and the other going west serves as an air transit hub and gateway to many scenic spots in its
through Kazakhstan. As China's HSR network continues to expand, it is home province for passengers from other provinces. Due to the long-
expected that the rail connectivity of these second-echelon cities will be distance between Yunnan and some wealthy provinces in East and
even higher in the near future. South China, tourists usually use air services when traveling Yunnan.
Hong Kong, Kunming, and Urumqi are the three cities that The HSR connecting Shanghai and Guiyang has been extended to
predominantly rely on air transportation. Due to historical reason, Kunming and the HSR will be in operation in early 2017. It will then
Hong Kong's rail system is relatively independent from mainland China. take only nine hours from Kunming to Shanghai. There should be a
Although Hong Kong is connected to Shenzhen by subway, and to significant rise in Kunming's connectivity by then. Similar to Kunming,
Guangzhou, Beijing, and Shanghai by regular rail services, rail con- Urumqi is far away from China's major cities. Urumqi is located in the

DCG Connectivity Ranking


100.00%
7200
6800 90.00%
6400
6000
DCG Connectivity Units

80.00%
5600
5200 70.00%
Rail Percentage

4800
4400 60.00%
4000
3600 50.00%
3200
2800 40.00%
2400 30.00%
2000
1600 20.00%
1200
800 10.00%
400
0 0.00%

Air Rail Rail Percent

Fig. 3. Domestic city global (DCG) connectivity ranking.

7
Z. Zhu et al. Journal of Transport Geography xxx (xxxx) xxx–xxx

Table 3 connectivity only.


Domestic city global (DCG) connectivity rankings with different modes. Using connectivity as the weight of edges and cities as nodes, we can
obtain the network graph with Gephi shown in Fig. 5. It can be seen
City Air DCG ranking Rail DCG ranking Combined DCG ranking
that Beijing-Tianjin, Guangzhou-Shenzhen, and Shanghai-Hangzhou-
Shanghai 1 1 1 Nanjing-Ningbo, have formed tightly connected clusters, respectively.
Beijing 2 2 2 We turn now to the domestic city domestic (DCD) connectivity,
Hong Kong 3 20 11
which is calculated by aggregating route connectivity of all the routes
Guangzhou 4 5 3
Shenzhen 5 9 8 connecting the city with other domestic cities. The DCD connectivity
Kunming 6 21 17 represents a city's domestic radiation. Similarly, the domestic city
Chengdu 7 17 13 international (DCI) connectivity only calculates the connectivity of
Chongqing 8 15 15 international routes, and shows the international accessibility of a
Xi'an 9 13 12
domestic city. The rankings of the DCD and DCI connectivity are
Hangzhou 10 7 7
Xiamen 11 16 16 presented in Figs. 6 and 7, respectively.
Nanjing 12 6 5 When considering DCD connectivity only, the ranking of Hong Kong
Urumqi 13 22 22 drops substantially. This is not surprising as Hong Kong has an
Dalian 14 19 20
extensive international air network. Its international connectivity
Shenyang 15 11 14
Tianjin 16 10 10 accounts for 46.42% of its DCG connectivity, while this figure is only
Wuhan 17 3 4 12.14% for Shanghai. When considering DCI connectivity only, Hong
Zhengzhou 18 8 9 Kong ranks the 1st with a huge lead over other cities. Shanghai, Beijing,
Changsha 19 4 6 and Guangzhou are in the 2nd, 3rd, and 4th places, respectively (see the
Guiyang 20 14 19
blue bars in Fig. 7). All other cities' international connectivity is at a
Nanning 21 18 21
Ningbo 22 12 18 much lower level compared with the top four. However, Hong Kong is
Yining 23 23 23 normally regarded as an international city rather than a domestic city
by most Chinese. When we exclude Hong Kong from our domestic
The underlined numbers show that a city's air DCG ranking and its rail DCG ranking can network, the DCI connectivity ranking changes significantly (see the red
be substantially different.
bars in Fig. 7). Shenzhen rises to number 3 from number 5. Xiamen rises
from 11 to 9. Hangzhou rises from 12 to 6. Nanjing rises from 13 to 10.
Chongqing rises from 15 to 12. Wuhan rises from 17 to 14. Ningbo rises
from 21 to 18. These changes indicate that these cities have good
far northwest of China. A train trip between Urumqi and Beijing takes connections with Hong Kong. As Hong Kong is well connected to many
about 30–39 h. Air is a natural choice for many travelers. international airports, these cities could use Hong Kong as a transit hub
For the reasons mentioned above, when only air connectivity is to increase their indirect connectivity to the outside world (Wan and
considered, Hong Kong ranks the 3rd out of 23 (see Table 3 and Fig. 4). Zhang, 2017).
However, it ranks the 20th place in terms of rail DCG connectivity, and
the 11th for the combined modes. Kunming ranks the 6th place in air, 3.2. International city connectivity
21st in rail, and 17th for the combined DCG connectivity. As noted
earlier, Kunming did not have HSR until early 2017. In contrast, Wuhan An international city domestic (ICD) connectivity is calculated by
and Changsha are located in Central China with an extensive HSR aggregating the connectivity of all the routes connecting this interna-
network. The well-developed HSR system has given them the 4th and tional city to China's domestic cities. The ranking of the 13 interna-
6th places for the combined DCG connectivity, respectively. However, tional cities in ICD connectivity is presented in Fig. 8.
they come in the 17th and 19th, respectively, when measured by air When considering Hong Kong as a domestic city, Seoul is the

Figure 4 DCG connectivity ranking changes considering


different modes
25

20

15
Ranking

10

Rank_Air Rank_Rail Rank_Agg

Fig. 4. Domestic city global (DCG) connectivity ranking with transport modes.

8
Z. Zhu et al. Journal of Transport Geography xxx (xxxx) xxx–xxx

Fig. 5. Graph network showing the clusters of Beijing & Tianjin, Guangzhou & Shenzhen, Shanghai & Hangzhou & Nanjing & Ningbo.

DCD Connectivity Ranking


6500 100.00%
6000 90.00%
5500
DCD Connectivity unit

80.00%
5000

Rail Percentage
4500 70.00%
4000 60.00%
3500
50.00%
3000
2500 40.00%
2000 30.00%
1500
20.00%
1000
500 10.00%
ZHENGZHOU

0 0.00%
SHENZHEN

SHENYANG
CHANGSHA

CHONGQING
GUANGZHOU

HONG KONG

CHENGDU
HANGZHOU
SHANGHAI
BEIJING

GUIYANG
NANJING

TIANJIN

NINGBO

NANNING
KUNMING

YINING
DALIAN
XIAMEN

URUMQI
WUHAN

XIAN

Air Rail Rail Percent

Fig. 6. Domestic city domestic (DCD) connectivity ranking.

Domesc City Internaonal Connecvity Ranking


1400

1200

1000

800

600

400

200

HK Domesc HK Internaonal

Fig. 7. Domestic city international (DCI) connectivity ranking. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)

9
Z. Zhu et al. Journal of Transport Geography xxx (xxxx) xxx–xxx

International City Domestic Connectivity Ranking


1300.00
1200.00
1100.00
1000.00
900.00
800.00
700.00
600.00
500.00
400.00
300.00
200.00
100.00
0.00

ICD Connectivity(HK Domestic) ICD Connectivity(HK Int'l)

Fig. 8. International city domestic (ICD) connectivity ranking. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)

International City Domestic Connectivity Structure Analysis


650 100.00%
90.00%
520 80.00%
70.00%
390 60.00%
50.00%
260 40.00%
30.00%
130 20.00%
10.00%
0 0.00%

Beijing Shanghai Guangzhou Hong Kong All BSGH%

Fig. 9. International city domestic (ICD) connectivity structure analysis.

international city that is best connected with China's domestic network, is not counted as a Chinese domestic city. In contrast, Paris, Frankfurt,
followed by Tokyo, Taipei and Singapore (see the blue bars in Fig. 8).11 Seoul, Tokyo and Osaka have better connections with mainland China
Each week, there are 1447 flights connecting Taipei to China's in the international network as the changes in the size of the blue and
domestic network, and 1255 flights between Tokyo and China. red bars are not dramatic.
However, after applying the discount factors, Taipei's ICD connectivity Our results suggest that the connections between the international
is 12.46% lower than that of Tokyo. This is because flights connecting cities and China's domestic network are highly concentrated on a few
Taipei to mainland China have to detour around the Taiwan Strait cities, namely, Hong Kong, Shanghai, Beijing, and Guangzhou. As
middle line, which results in a significant increase in flight distance, shown in Fig. 9, the cumulative ICD connectivity with Beijing,
especially for short-haul flights. As the direct distance is used to Shanghai, Guangzhou and Hong Kong (BSGH) accounts for at least
calculate the velocity discount, the detour implies a smaller value of 59% of the overall ICD connectivity for every international city. It even
this discount factor. In fact, the average velocity discount for routes accounts for more than 95% for New York, London, and Atlanta.
connecting Taipei to China's domestic network is 0.601, while the
average velocity discount for all international routes is 0.749.
When Hong Kong is regarded as an international city, it is the Table 4
absolute winner measured by ICD connectivity, beating all other Ranking of city pairs by flight frequency.

international cities (see the red bars in Fig. 8). The connectivity values City A City B Weekly flights
for New York and London drop by 50.31% and 51.10%, respectively,
which means that about half of New York and London's connections Shanghai Beijing 793
with China are via Hong Kong. Taipei, Atlanta and Dubai's connection Shanghai Shenzhen 751
Shanghai Hong Kong 695
with China also heavily rely on Hong Kong. The destination of 41% Shanghai Guangzhou 576
(595 weekly) of the flights from Taipei to China is Hong Kong, resulting Shanghai Kunming 527
in the ranking of Taipei being overtaken by Singapore when Hong Kong Beijing Hong Kong 523
Beijing Chengdu 485
Beijing Shenzhen 465
Shanghai Chengdu 450
11
When a city has multiple airports, the connectivity measures the aggregate of all Shanghai Xi'an 437
airports.

10
Z. Zhu et al. Journal of Transport Geography xxx (xxxx) xxx–xxx

Table 5 Table 5. Shenzhen-Hong Kong holds the advantage over other city-pairs
Ranking of city pairs measured by train frequency. in rail connections. The subway system connects Shenzhen and Hong
Kong at Lok Ma Chau Station and Lo Wu Station, with trains arriving/
City A City B Weekly trains
departing every 8–12 min. However, other city-pairs are connected by
Shenzhen Hong Kong 3486 HSR or intercity trains rather than subways.
Shanghai Nanjing 3374 The route connectivity measured by the aggregated DWCU gives a
Guangzhou Shenzhen 2611
ranking list for 40 domestic routes in China (see Table 6) that differs
Shanghai Hangzhou 2324
Beijing Tianjin 2324 greatly from those in Tables 4 and 5. Shenzhen-Hong Kong ranks the
Guangzhou Changsha 2002 17th with the measure of DWCU, mainly due to the low speed of
Wuhan Changsha 1946 subway between the two cities and the extra time spent on immigration
Nanjing Hangzhou 1414 and customs clearance. Kunming-Shanghai ranks the 42nd with DWCU
Zhengzhou Wuhan 1400
instead of the 5th in terms of the air frequency. The drop in ranking is
Hangzhou Ningbo 1337
largely due to the absence of HSR and the high percentage of indirect
flights. In fact, 31.88% of the flights connecting Kunming and Shanghai
3.3. Route connectivity are indirect flights, while the indirect flights only account for 10.27% in
the whole domestic network. Beijing-Tianjin has the same train
The information of the connectivity of the routes can help disclose frequency as Shanghai-Hangzhou, while Beijing-Tianjin's connectivity
the main travel channels in the domestic city network, just like the is 11.12% higher than the latter. This is because only 65 out of the 332
arteries in the human body. trains between Beijing and Tianjin are K-trains (low speed trains), while
The top 10 city-pairs with the most connecting flights are listed in 97 of the 332 trains between Shanghai and Hangzhou are K-trains.
Table 4. When considering the raw connections, Shanghai-Beijing takes Among all the city pairs, Shanghai-Nanjing has the highest con-
the lead. nectivity, mainly due to the significant contribution from the HSR
The top ten city-pairs with the most connecting trains are listed in services between the two cities. Beijing-Shanghai is the best-connected
pair by air.
It appears that the connectivity between city-pairs follows the
Table 6 Pareto principle quite well. It can be seen from Fig. 10 that the top
Route connectivity ranking. 10% routes contribute more than 50% of the total network connectiv-
ity. The first 30% routes contribute more than 80% of the total network
Rank City pair DWCU Air Rail Air% Rail%
connectivity.
1 Shanghai-Nanjing 1343.04 2.96 1340.09 0.22% 99.78% Looking closely at the top-ranking routes, we can find some
2 Beijing-Tianjin 1029.79 0.00 1029.79 0.00% 100.00% similarities.
3 Guangzhou-Shenzhen 996.42 0.00 996.42 0.00% 100.00%
As shown in Fig. 11, the top routes benefit greatly from train
4 Shanghai-Hangzhou 926.70 0.00 926.70 0.00% 100.00%
5 Guangzhou-Changsha 888.92 13.30 875.61 1.50% 98.50% connections. Train services contribute more than 70% of the route
6 Wuhan-Changsha 808.04 0.00 808.04 0.00% 100.00% DWCU for 26 out of the top-40 routes. This trend is more obvious
7 Beijing-Shanghai 625.89 321.23 304.66 51.32% 48.68% among the top-20 routes, with train contributing more than 80% of the
8 Zhengzhou-Wuhan 599.17 0.00 599.17 0.00% 100.00% route DWCU for 19 of them. The only exception is Shanghai-Beijing,
9 Guangzhou-Wuhan 591.27 50.50 540.76 8.54% 91.46%
with train contributing 48.68%. However, considering Shanghai-Beij-
10 Beijing-Zhengzhou 559.98 16.36 543.62 2.92% 97.08%
11 Hangzhou-Ningbo 517.39 0.00 517.39 0.00% 100.00% ing is the longest route among the top-20 (the direct distance is
12 Nanjing-Hangzhou 513.51 0.00 513.51 0.00% 100.00% 1178 km), the role that train plays is still significant. Although there
13 Beijing-Nanjing 494.31 53.05 441.27 10.73% 89.27% are K- or D-trains operating on these top-20 routes, the vast majority of
14 Shenyang-Dalian 412.75 0.00 412.75 0.00% 100.00%
the train services are provided by the HSR, especially during the
15 Beijing-Wuhan 402.85 76.64 326.21 19.03% 80.97%
16 Zhengzhou-Changsha 394.88 0.00 394.88 0.00% 100.00% daytime. Although the HSR has been well recognized as the dominant
17 Shenzhen-Hong Kong 383.86 0.00 383.86 0.00% 100.00% transport mode only for short- and medium-haul routes with a distance
18 Zhengzhou-Xi'an 360.10 0.00 360.10 0.00% 100.00% below 700 km (e.g., Yamaguchi et al., 2008; Adler et al., 2010; Yang
19 Shenyang-Tianjin 353.17 3.61 349.56 1.02% 98.98% and Zhang, 2012; Fu et al., 2014; Wan et al., 2016; Xia and Zhang,
20 Shenzhen-Changsha 331.13 8.15 322.98 2.46% 97.54%
2016a), with people building in confidence in its safety and punctuality,
21 Shanghai-Shenzhen 317.04 264.61 52.44 83.46% 16.54%
22 Beijing-Changsha 311.76 95.14 216.62 30.52% 69.48% the HSR has become a better and dominant option against air for some
23 Beijing-Shenyang 305.59 61.02 244.57 19.97% 80.03% medium- and long-haul routes up to 1300 km. For example, rail
24 Beijing-Guangzhou 293.35 199.36 93.99 67.96% 32.04% transport contributes 83.08% of the connectivity for Guangzhou-
25 Tianjin-Nanjing 284.90 1.85 283.05 0.65% 99.35%
Zhengzhou, whose direct distance is 1294 km. It contributes 80.97%
26 Shanghai-Tianjin 281.34 79.17 202.18 28.14% 71.86%
27 Shanghai-Guangzhou 280.61 226.66 53.95 80.77% 19.23%
of the connectivity for Beijing-Wuhan, with direct distance being
28 Chengdu-Chongqing 274.66 0.00 274.66 0.00% 100.00% 1133 km, and 89.27% of the connectivity for Beijing-Nanjing whose
29 Beijing-Xi'an 273.28 145.14 128.13 53.11% 46.89% direct distance is 981 km. Therefore, when studying the connectivity of
30 Guangzhou- 271.51 45.95 225.56 16.92% 83.08% a route, it is necessary to add rail to the passenger transportation
Zhengzhou
network, especially the HSR, which allows a better understanding of
31 Shanghai-Hong Kong 262.69 260.76 1.92 99.27% 0.73%
32 Beijing-Shenzhen 261.09 211.91 49.18 81.16% 18.84% how cities are connected.
33 Changsha-Hangzhou 251.19 17.51 233.68 6.97% 93.03% As one anonymous referee correctly pointed out, the relative level of
34 Shanghai-Wuhan 246.75 78.57 168.18 31.84% 68.16% ticket price, passengers' time value and income are the factors that may
35 Beijing-Hangzhou 239.44 130.67 108.77 54.57% 45.43%
have made HSR more attractive than air in the case of China. The
36 Beijing-Chengdu 237.46 213.13 24.33 89.75% 10.25%
37 Beijing-Hong Kong 235.67 233.50 2.17 99.08% 0.92%
published fare for G-train is around 0.46 yuan/km and the fare for D-
38 Shanghai-Xiamen 235.18 126.40 108.78 53.75% 46.25% train is 0.3 yuan/km (Sina, 2017). For airlines, the published airfare is
39 Guangzhou-Nanning 234.74 14.56 220.18 6.20% 93.80% 0.75 yuan/km. By offering various discounts, domestic airlines' actual
40 Shanghai-Changsha 233.02 73.42 159.60 31.51% 68.49% yield has been around 0.55 in recent years. For example, China
Southern's yield was 0.58 yuan/km in 2014, and 0.55 in 2015 in the
domestic market. Therefore, the HSR services are still relatively
cheaper. Furthermore, Wang et al. (2017) report the HSR ticket prices

11
Z. Zhu et al. Journal of Transport Geography xxx (xxxx) xxx–xxx

Fig. 10. Route connectivity distribution.

Fig. 11. Route connectivity ranking and train proportion.

among China, Japan and Europe for routes of comparable distance. The highest connectivity level, leading in both rail and air connectivity.
HSR ticket prices in China are much cheaper that its counterparts in Hong Kong, Kunming, and Urumqi are the three cities that predomi-
Japan and Europe. In addition, flight delays have become a norm at the nantly rely on air transportation whose contribution to the connectivity
mega-airports. Taken together, it may not be very surprising that HSR exceeds 80%. Beijing-Tianjin, Guangzhou-Shenzhen, and Shanghai-
has been a preferred choice, even on some longer routes: HSR in China Hangzhou-Nanjing-Ningbo, have formed tightly connected clusters,
is still competitive against air transportation on some long-haul routes respectively, as revealed by the graph network. This research also
with distance even up to 1300 km, as compared to other countries suggests that the connections between the international cities and
where HSR is the dominant transport mode only on short-haul and China's domestic network are highly concentrated on a few cities,
medium-haul routes with distance below 700 km. namely, Hong Kong, Shanghai, Beijing, and Guangzhou. To build better
connections with the rest of the world and implement the OBOR
4. Concluding remarks strategy, other cities will need to improve their international air
connectivity.
This paper has developed a DWM to measure and compare the Shanghai-Nanjing has been found to be the best-connected city-pair,
network performance of 23 major cities in China and 13 major mainly due to the significant contribution from the HSR services.
international cities that are of economic/location significance. The Beijing-Shanghai is the best-connected pair by air. Seoul is the best
DWM takes into account the connections by both flights and trains. connected international city in term of its transportation links with
Apart from the quantities, three aspects of the connection quality, China. Our study also shows that train connections have made
namely, capacity, travel time, and velocity, are incorporated into the considerable contributions to the route connectivity. The contribution
model to account for the utility function of travelers. To our best from train services is more than 80% for 19 of the 20 top-ranking
knowledge, this is the first study that uses such a dynamic model to domestic routes. Although most studies suggest that the HSR is the
consider both quality and quantity of the connections of two transport dominant transport mode only for short- and medium-haul routes with
modes. distance below 700 km, this research has found that the HSR in China
Among the 23 cities selected, Shanghai is revealed to have the has become a better and dominant option against air for routes up to

12
Z. Zhu et al. Journal of Transport Geography xxx (xxxx) xxx–xxx

1300 km. This finding has significant policy implications for transpor- References
tation infrastructure planning and investment in China.
The multi-modal connectivity information is valuable to airport and Adler, N., Pels, E., Nash, C., 2010. High-speed rail and air transport competition: game
HSR planning. In 2017, the Director of the Civil Aviation engineering as tool for cost-benefit analysis. Transp. Res. B Methodol. 44 (7),
812–833.
Administration of China (CAAC) announced that some flights from Banno, M., Redondi, R., 2014. Air connectivity and foreign direct investments: economic
the second- and third-tier cities to Beijing will be shifted to Tianjin and effects of the introduction of new routes. Eur. Transp. Res. Rev. 6, 355–363.
Shijiazhuang in the future and these airports will be linked to Beijing Basile, R., Benfratello, L., Castellani, D., 2006. Attracting foreign direct investments in
Europe: are Italian regions doomed? In: Malagarini, M., Piga, G. (Eds.), Capital
via HSR. Also in 2017, National Development and Reform Commission Accumulation Productivity and Growth. Palgrave Macmillan, Monitorino.
and the CAAC jointly issued a notice on national civil aviation airport Behrens, C., Pels, E., 2012. Intermodal competition in the London-Paris passenger market:
network planning in which constructing new airports in West China and high-speed rail and air transport. J. Urban Econ. 71 (3), 278–288.
Boonekamp, T., Burghouwt, G., 2017. Measuring connectivity in the air freight industry.
integrating airports with other transport modes such as HSR were top J. Air Transp. Manag. 61, 81–94.
priorities. These moves have shown that the Chinese government Brueckner, J.K., 2003. Airline traffic and urban economic development. Urban Stud. 40
understands the importance of coordination of the development of (8), 1455–1469.
Burghouwt, G., de Wit, J., 2005. The temporal configuration of airline networks in
airports and HSR. Although these two modes are substitutable in
Europe. J. Int. Logist. Trade 11 (3), 185–198.
general, they could possibly complement each other (e.g., Givoni and Burghouwt, G., Redondi, R., 2013. Connectivity in air transport networks: an assessment
Banister, 2006; Jiang and Zhang, 2014; Román and Martín, 2014; Xia of models and applications. Journal of Transport Economics and Policy 47 (1), 35–53.
and Zhang, 2016b; Wang et al., 2017). Understanding the overall Burghouwt, G., Veldhuis, J., 2006. The competitive position of hub airports in the
Transatlantic market. J. Air Transp. 11 (1), 106–130.
connectivity of a city/region is key to implement these proposals and Chen, X., Lin, L., 2016. The integration of air and rail technologies: Shanghai's Hongqiao
move forward the OBOR agenda. Integrated Transport Hub. J. Urban Technol. 23 (2), 23–46.
However, this research only considers 23 important domestic cities. Cheng, Y.S., Loo, B.P., Vickerman, R., 2015. High-speed rail networks, economic
integration and regional specilization in China and Europe. Travel Behav. Soc. 2 (1),
In future research, the DWM developed in this paper can be used to 1–14.
examine a larger network by including more domestic cities, and of Clewlow, R.R., Sussman, J.M., Balakrishnan, H., 2014. The impact of high-speed rail and
course more international cities, which will reduce the bias of the low-cost carriers on European air passenger traffic. Transp. Policy 33, 136–143.
Cristea, A.D., 2011. Buyer-seller relationship in international trade: evidence from U.S.
connectivity measure resulting from the small sample size. Indirect states' exports and business-class travel. J. Int. Econ. 84, 207–220.
connections arranged by passengers may also need to be incorporated Dobruszkes, F., 2011. High-speed rail and air transport competition in Western Europe: a
into the model for a more complete analysis of the connectivity. The supply-oriented perspective. Transp. Policy 18 (6), 870–879.
Fu, X., Zhang, A., Lei, Z., 2012. Will China's airline industry survive the entry of high
DWM can also be used to assess the possible vulnerability and resilience speed rail? Res. Transp. Econ. 35, 13–25.
of intercity passenger transportation network and thereby develop an Fu, X., Oum, T.H., Yan, J., 2014. An analysis of travel demand in Japan's intercity market:
emergency response strategy for possible transportation accidents, or empirical estimation and policy simulation. Journal of Transport Economics and
Policy 48, 97–113.
national transportation safety and security planning.12 Moreover, our
Geurs, K.T., Van Wee, B., 2004. Accessibility evaluation of land-use and transport
DWM can be extended and applied in freight transportation research, strategies: review and research directions. J. Transp. Geogr. 12 (2), 127–140.
which will produce useful results for the transportation and logistics Givoni, M., 2006. Development and impact of the modern high-speed train: a review.
industry. Finally, transport geographers and other researchers are Transp. Rev. 26 (5), 593–611.
Givoni, M., Banister, D., 2006. Airline and railway integration. Transp. Policy 13 (5),
interested in how contextual factors such as the particularities of the 386–397.
urban, economic and institutional system shape transportation system Givoni, M., Dobruszkes, F., 2013. A review of ex-post evidence for mode substitution and
outcomes in general, and connectivity in particular (e.g., Gutierrez, induced demand following the introduction of high-speed rail. Transp. Rev. 33 (6),
720–742.
2001; Geurs and Van Wee, 2004; Lin, 2012; Matsumoto et al., 2016; Wu Gutierrez, J., 2001. Location, economic potential and daily accessibility: an analysis of
et al., 2016; Zhang et al., 2017). For example, Zhang et al. (2017) the accessibility impact of the high-speed line Madrid-Barcelona-French border. J.
attempt to identify the underlying drivers (e.g. urban, economic and Transp. Geogr. 9, 229–242.
Ha, H.K., Yoshida, Y., Zhang, A., 2011. Social efficiency benchmarking of Japanese
institutional system) of the variation in Chinese airport connectivity domestic transport services: a comparison of rail and air. Transp. Res. Part D: Transp.
over the 2005–2016 period. We consider that incorporating our multi- Environ. 16, 554–561.
modal approach with the underlying drivers would be a natural and Jiang, C., Zhang, A., 2014. Effects of high-speed rail and airline cooperation under hub
airport capacity constraint. Transp. Res. B Methodol. 60, 33–49.
interesting extension, but perhaps beyond the scope of the present
Lei, Z., O'Connell, J.F., 2011. The evolving of Chinese aviation policies and impact of a
paper. deregulating environment on Chinese carriers. J. Transp. Geogr. 19, 829–839.
Lei, Z., Yu, M., Chen, R., O'Connell, J.F., 2015. Liberalization of China-US air transport
market: assessing the impacts of the 2004 and 2007 protocols. J. Transp. Geogr. 50,
24–32.
Acknowledgement Li, K.X., Qi, G., 2016. Transport connectivity and regional development in China. J. Int.
Logist. Trade 14 (2), 142–155.
We are very grateful to three anonymous referees and the Associate Lin, J., 2012. Network analysis of China's aviation system, statistical and spatial structure.
J. Transp. Geogr. 22, 109–117.
Editor (Peter Hall) whose detailed and thoughtful comments have led to Malighetti, P., Paleari, S., Redondi, R., 2008. Connectivity of the European airport
a large improvement of the paper. We also thank Guillaume Burghouwt, network: “self-help hubbing” and business implications. J. Int. Logist. Trade 14 (2),
Yaap de Wit, Jun-yeop Lee, Kevin Li, Becky Loo, and participants at the 53–65.
Matsumoto, H., Veldhuis, J., de Wit, J., Burghouwt, G., 2008. Network Performance, Hub
seminar at the University of Hong Kong, Kyoto University, Hiroshima Connectivity Potential, and Competitive Position of Primary Airports in Asia/Pacific
University, the World Aviation Hub Conference (Incheon), the 5th Region. Paper Presented at the Air Transport Research Society Conference, July 6–10,
European Aviation Conference (Amsterdam), and the “One Belt, One Athens.
Matsumoto, H., Domae, K., O'Connor, K., 2016. Business connectivity, air transport and
Road” Conference at RMIT (Melbourne) for helpful comments. Partial the urban hierarchy: a case study in East Asia. J. Transp. Geogr. 54, 132–139.
financial support from the UBC Centre for Transport Studies (to Paleari, S., Redondi, R., Malighetti, P., 2010. A comparative study of airport connectivity
Zhenran Zhu) and the Social Sciences and Humanities Research in China, Europe and US: which network provides the best service to passengers?
Transportation Research Part E: Logistics and Transportation Review 46 (2),
Council of Canada (435-2015-0845) (to Anming Zhang) is gratefully
198–210.
acknowledged. A. Zhang would like to thank Shanghai Advanced Perl, A.D., Goetz, A.R., 2015. Corridors, hybrids and networks: three global development
Institute of Finance, Shanghai Jiao Tong University, for hospitality as strategies for high speed rail. J. Transp. Geogr. 42, 134–144.
Román, C., Martín, J.C., 2014. Integration of HSR and air transport: understanding
a visiting professor when the paper was revised.
passengers' preferences. Transportation Res. E Logist. Transportation Rev. 71,
129–141.
SDPC, 2016. Medium- and long-term planning of Chinese railway network. In: National
Development and Reform Commission. No. 2016-1536, (Beijing).
12
Shaw, S.-L., 1993. Hub structures of major US passenger airlines. J. Transp. Geogr. 1 (1),
We thank an anonymous referee for providing this insightful point.

13
Z. Zhu et al. Journal of Transport Geography xxx (xxxx) xxx–xxx

47–58. high-speed rail in developing China's future supercities. Transp. Res. Rec.: J. Transp.
Shaw, S.-L., Ivy, R.L., 1994. Airline mergers and their effect on network structure. J. Res. Board 2546, 78–87.
Transp. Geogr. 2 (4), 234–246. Xia, W., Zhang, A., 2016a. High-speed rail and air transport competition and cooperation:
Sina, 2017. Why will HSR fare be raised-70% of the truth found here. 18/02/2017. Sina a vertical differentiation approach. Transp. Res. B Methodol. 94, 456–481.
Finance. Available at. http://finance.sina.com.cn/china/gncj/2017-02-18/doc- Xia, W., Zhang, A., 2016b. Effects of air and high-speed rail transport integration on
ifyarrcf4596275.shtml. profits and welfare: the case of air-rail connecting time. Accepted by J. Air Transp.
Tanaka, K., 2016. Do international flights promote FDI? The role of face-to-face Manag. Available at SSRN: https://ssrn.com/abstract=2868581.
communication. In: IDE Discussion Paper No. 554. Institute of Developing Economies Yamaguchi, K., Ohashi, H., Hihara, K., 2008. Competitive Analysis of Airfare in Japan:
Available at. www.ide.go.jp/English/Publish/Download/Dp/pdf/554.pdf. Competition Among Air Carriers and Rivalry With the High-speed Railway System.
Veldhuis, J., 1997. The competitive position of airline networks. J. Int. Logist. Trade 3 Paper Presented at the 13th WCTR World Conference, Athens, Greece. (Unpublished
(4), 181–188. manuscript, University of Tokyo).
Veldhuis, J., Kroes, E., 2002. Dynamics in Relative Network Performance of the Main Yang, H., Zhang, A., 2012. Effects of high-speed rail and air transport competition on
European Hub Airports. Paper Presented at the European Transport Conference, prices, profits and welfare. Transp. Res. B Methodol. 46, 1322–1333.
September 9–11, Cambridge. Zhang, A., 2012. Airport improvement fees, benefit spillovers, and land value capture
Wan, Y., Zhang, A., 2017. Air cargo transport and logistics in Hong Kong and southern mechanisms. In: Ingram, Gregory K., Hong, Yu-Hung (Eds.), Value Capture and Land
China. In: Zhang, J., Fung, C.-M. (Eds.), Handbook of Transport in Asia. Routledge Policies. Lincoln Institute of Land Policy, Cambridge, MA, pp. 323–348 (Ch. 13).
(forthcoming). Zhang, A., Chen, H., 2003. Evolution of China's air transport development and policy
Wan, Y., Ha, H.-K., Yoshida, Y., Zhang, A., 2016. Airlines' reaction to high-speed rail towards international liberalization. Transp. J. 42, 31–49.
entries: empirical study of the Northeast Asian market. Transp. Res. A Policy Pract. Zhang, Y., Zhang, A., 2016. Determinants of air passenger flows in China and gravity
94, 532–557. model: deregulation, LCCs, and high-speed rail. J. Transp. Econ. Policy 50 (3),
Wang, K., Xia, W., Zhang, A., 2017. Should China further expand its high-speed network? 287–303.
Consider the low-cost carrier factor. Transp. Res. A Policy Pract. 100, 105–120. Zhang, Y., Zhang, A., Zhu, Z., Wang, K., 2017. Connectivity at Chinse Airports: The
Wu, Q., Perl, A., Sun, J., 2016. Bigger and different: beginning to understand the role of Evolution and Drivers. (unpublished manuscript).

14

You might also like