Identification of multi-temporal urban growth patterns with a modified urban growth index: Case study of three fast growing cities in the Greater Mekong Subregion (GMS)

Abstract

Identification of urban growth patterns is essential to better understand urbanization and its eco-environmental consequences. Landscape indices, such as the simple urban growth index (S) and the landscape expansion index (LEI), are commonly used to identify urban growth patterns, but most of them are unable or not robust enough to recognize multi-temporal patterns. In this study, we proposed the modified urban growth index (Sm) with a 4-step approach and identified the urban growth patterns of three fast growing cities, namely, Xishuangbanna, Phnom Penh and Ho Chi Minh City. The results of Sm and S in identifying multi-temporal urban growth patterns were compared, and the multi-temporal outlying patches were always classified as edge-expansion patches or infilling patches in the results of S. Based on Sm, the spatial and temporal changes of urban growth patterns in the case cities were presented. Generally, Sm performed better in identifying multi-temporal urban growth patterns than S. However, like many other indices, it has limitations in certain situations. We further discussed the potential applications of urban growth pattern identification using Sm. First, the identified multi-temporal urban growth patterns validated the diffusion-coalescence theory. Outlying patches usually tended to decrease during the transition period from diffusion to coalescence, while edge-expansion growth and infilling growth always became dominant urban growth patterns at the end of the coalescence process. The dynamics of urban spatial structure could also be recognized by highlighting the multi-temporal growth of outlying clusters. In conclusion, with the identification of multi-temporal urban growth patterns, the modified urban growth index Sm is able to delineate the diffusion-coalescence process of urban growth and reveal the evolving characteristics of urban spatial structure.

Full text

Ecological Indicators 142 (2022) 109206
1470-160X/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
Original Articles
Identication of multi-temporal urban growth patterns with a modied
urban growth index: Case study of three fast growing cities in the Greater
Mekong Subregion (GMS)
Hui Cao
a
,
b
, Jianglong Chen
a
,
b
, Cheng Chen
a
,
b
, Pingxing Li
a
,
b
,
*
a
Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, Nanjing 210008, China
b
Key Laboratory of Watershed Geographic Sciences, Chinese Academy of Sciences, Nanjing 210008, China
ARTICLE INFO
Keywords:
Urban growth patterns
Landscape index
Diffusion and coalescence
Urban spatial structure
ABSTRACT
Identication of urban growth patterns is essential to better understand urbanization and its eco-environmental
consequences. Landscape indices, such as the simple urban growth index (S) and the landscape expansion index
(LEI), are commonly used to identify urban growth patterns, but most of them are unable or not robust enough to
recognize multi-temporal patterns. In this study, we proposed the modied urban growth index (S
m
) with a 4-
step approach and identied the urban growth patterns of three fast growing cities, namely, Xishuangbanna,
Phnom Penh and Ho Chi Minh City. The results of S
m
and S in identifying multi-temporal urban growth patterns
were compared, and the multi-temporal outlying patches were always classied as edge-expansion patches or
inlling patches in the results of S. Based on S
m
, the spatial and temporal changes of urban growth patterns in the
case cities were presented. Generally, S
m
performed better in identifying multi-temporal urban growth patterns
than S. However, like many other indices, it has limitations in certain situations. We further discussed the po-
tential applications of urban growth pattern identication using S
m
. First, the identied multi-temporal urban
growth patterns validated the diffusion-coalescence theory. Outlying patches usually tended to decrease during
the transition period from diffusion to coalescence, while edge-expansion growth and inlling growth always
became dominant urban growth patterns at the end of the coalescence process. The dynamics of urban spatial
structure could also be recognized by highlighting the multi-temporal growth of outlying clusters. In conclusion,
with the identication of multi-temporal urban growth patterns, the modied urban growth index S
m
is able to
delineate the diffusion-coalescence process of urban growth and reveal the evolving characteristics of urban
spatial structure.
1. Introduction
Urbanized area supported 55 % of the world’s population in 2018,
and by 2050, the proportion of the world’s urban population is expected
to be 68 % (United Nations, 2018). Although urban expansion occurs on
less than 5 % of the Earth’s terrestrial surface, it accounts for many types
of global environmental changes (Grimm et al., 2008). In recent years,
Asia, especially the Greater Mekong Subregion (GMS), has quickened
the urbanization process and led to unprecedented expansion of urban
areas (Florian and Januar, 2016; Shazadeh-Moghadam et al., 2019).
However, numerous environmental and ecological problems, such as
climate change (Fan and Zhou, 2019; Lemoine-Rodriguez et al., 2022),
losses of habitats and biodiversity (Ahrends et al., 2015; Yang et al.,
2022), land fragmentation (Zou et al., 2022), air pollution (Qiu et al.,
2019; Wang et al., 2018), urban heat island phenomena (Li et al., 2018;
Zhou and Chen, 2018), losses of ecosystem services (Delphin et al.,
2016; Xie et al., 2018), and deterioration of the water environment (Fan
et al., 2022; Tam and Nga, 2018), have emerged in the process of dra-
matic urban expansion.
Cities can be considered dynamic environments that are associated
with duration, intensity, volatility, and location (Batty, 2002). Quanti-
fying city’s landscape patterns is an essential step to understand the
dynamics of urban expansion and its effects on environmental and
ecological processes (Cui and Wang, 2015; Dupras et al., 2016; Green
et al., 2022). Most of the early theories or studies provided descriptive
approaches for the understanding of urban growth patterns, such as
* Corresponding author at: 73 East Beijing Road, Nanjing 210008, China.
E-mail addresses: hcao@niglas.ac.cn (H. Cao), jlchen@niglas.ac.cn (J. Chen), chchen@niglas.ac.cn (C. Chen), pxli@niglas.ac.cn (P. Li).
Contents lists available at ScienceDirect
Ecological Indicators
journal homepage: www.elsevier.com/locate/ecolind
https://doi.org/10.1016/j.ecolind.2022.109206
Received 21 December 2021; Received in revised form 4 July 2022; Accepted 20 July 2022

Ecological Indicators 142 (2022) 109206
2
concentric zone theory (Burgess et al., 1925), sector theory (Hoyt,
1939), and multiple nuclei theory (Harris and Ullman, 1945). However,
they were not yet capable of quantifying the spatial and temporal
characteristics of urban growth until the development of new tech-
niques, especially remote sensing and geographic information system
(GIS) (Batty et al., 2001; Goodchild, 1992). With the application of these
techniques, many researchers have tried to dene urban growth patterns
from different perspectives or phenomena, such as urban sprawl, urban
density, compactness, location centrality, spatial relationships, and
developing directions (Camagni et al., 2002; Ewing et al., 2002; Herold
et al., 2005; Marquez and Smith, 1999; Reis et al., 2015).
Since landscape indices have been widely used in characterizing
urban morphology, such as urban sprawl (Ewing and Hamidi, 2015;
Getu and Bhat, 2021; Jia et al., 2022) and land fragmentation (Irwin and
Bockstael, 2007; Shrestha et al., 2012; Wei et al., 2020), some re-
searchers have developed a variety of landscape indices to characterize
urban growth patterns (Table 1). Basically, urban growth can be sum-
marized into three main patterns: edge-expansion growth, inlling
growth and outlying growth (Berling-Wolff and Wu, 2004; Forman,
1995; Wilson et al., 2003). Xu et al. (2007) proposed the simple urban
growth index (S) to distinguish urban growth patterns. This index offers
an effective way to identify urban growth patterns and has been widely
used in urban growth or urban structure studies (Fei and Zhao, 2019;
Huang et al., 2017; Wu et al., 2015; Zhao et al., 2015). Liu et al. (2010)
developed the landscape expansion index (LEI), which uses the
overlapping buffer area to characterize urban growth patterns. The au-
thors mentioned that the robustness of the index depended on the se-
lection of buffer distance and suggested setting the buffer distance
smaller than the data resolution. S and LEI use a similar criterion to
dene urban growth patterns, although they utilize different approaches
to classify urban growth patches. However, these approaches describe
only urban growth patterns of two time phases since they cannot char-
acterize the dynamics of continuously grown urban patches from multi-
temporal urban land use maps. Accordingly, Jiao et al. (2015) proposed
a multi-order landscape expansion index (MLEI) to recognize multi-
temporal urban growth patterns. The MLEI solved the problem that
both S and LEI can describe only urban growth patterns of two-time
phases, but it was still constrained by the selection of buffer distance.
Jiao et al. (2018) tried to address this issue by proposing the proximity
expansion index (PEI) with multiple buffers. Xia et al. (2020) developed
a shape weighted landscape evolution index (SWLEI), and argued that
the buffer based indices were affected not only by spatial resolution, but
also by the shape of newly grown patches (i.e., patch size, perimeter or
intersection with old patches). Based on this argument, the urban scale
or the shape of old urban patches may also be considered when selecting
buffer distance. All the conditions and discussions mentioned above
indicated that buffer distance selection is a complicated issue. In gen-
eral, S provides a simple way to distinguish urban growth patterns but
can only be applied for two-time phases. The buffer based indices, such
as the LEI, MLEI, PEI, and SWLEI, must address the issue of buffer dis-
tance selection.
Since all these indices were applied based on urban land use data,
most studies ignored another critical issue, namely, the existing urban
growth patterns. Even the rst phase of the urban area is the result of
urban growth at the previous stage. Although inlling patches and edge-
expansion patches cannot be distinguished in the rst phase, outlying
patches can be recognized by dening the main urban area. In this study,
we aimed to provide a simple and easily understood index to identify
urban growth patterns from the perspective of the entire process of
spatial and temporal urban patch development. To achieve this objec-
tive, some key principles were taken into consideration, including
simplicity, ease of computation, and distinctiveness (clear denitions for
different patterns) (Bhatta et al., 2010). Since the selection of buffer
distances is a quite complicated and debated issue, we proposed a
modied urban growth index S
m
on the basis of S. Compared with S, S
m
provided proper results in the identication of multi-temporal urban
growth patterns. Details of this index and a 4-step GIS approach are
described in Section 2. The results of urban growth patterns in the case
cities and the comparison between S
m
and S are presented in Section 3.
Section 4 discusses the advantages and limitation of S
m
, and applications
of multi-temporal urban growth pattern identication in characterizing
the diffusion-coalescence process and highlighting the evolution of
outlying clusters. Conclusions are summarized in the last section.
2. Materials and methods
2.1. Study area
The Greater Mekong Subregion (GMS) consists of China (Yunnan and
Guangxi), Thailand, Myanmar, Laos, Cambodia and Vietnam. It has been
one of the fastest developing regions in the world since the early 1990 s.
As an important international river in Asia, the Lancang-Mekong River
runs through the whole Greater Mekong Subregion. Its upper reach
within China is called the Lancang River, and its lower reach outside
China is called the Mekong River (Fig. 1).
Xishuangbanna Dai Autonomous Prefecture (hereinafter referred to
as Xishuangbanna) lies in Yunnan Province, Southwest China. It is a key
node city linking the Lancang River and Mekong River. Xishuangbanna
covers more than 19000 km
2
, and its altitude ranges from approximately
500 m to 2430 m. Xishuangbanna is the second largest rubber produc-
tion base in China, and tourism, tea and rubber are its economic pillars.
Table 1
Landscape indices for the identication of urban growth patterns.
Indices Equation Explanations
Simple urban growth
index (S)
S=Lc/P L
c
represents the length of the
shared edge of a newly
generated urban patch and the
old urban patches, and P
represents the perimeter of this
newly generated patch. S ranges
from 0 to 1, and the newly
grown urban patch will be
recognized as inlling if S is
greater than or equal to 0.5,
edge-expansion if S is less than
0.5 and larger than 0, and
outlying if S is equal to 0.
Landscape
expansion index
(LEI)
LEI =100 ×A0
A0+Av
A
0
represents the common area
between the buffer of newly
generated urban patches and the
old urban patches, A
v
is the area
of the rest buffer. The newly
generated urban patch will be
identied as inlling if LEI ≥50,
edge-expansion if 0 <LEI less
than 50, and outlying if LEI =0.
Multi-order
landscape
expansion index
(MLEI)
MLEI(t)
i=
∑j=1
m(MLEI(t−1)
j×aij )
Ai
MLEI(t)
i represents the value of
MLEI of the newly generated
urban patch i at phase t, m
represents the amount of the old
urban patches at phase t −1 that
intersect patch i, MLEI(t−1)
j
represents the value of MLEI of
patch j that intersects patch i at
phase t −1, a
ij
represents the
shared area of patch i and patch
j, and A
i
represents the buffer
area of patch i. The initial value
MLEI0
i =100.
Proximity
expansion index
(PEI)
PEI =1
N+ (1−Ai
An)
N represents the total number of
buffers of the newly generated
urban patches; A
n
represents the
area of the Nth buffer; and A
i
represents the common area
between the Nth buffer and the
old urban patches.
H. Cao et al.

Ecological Indicators 142 (2022) 109206
3
Since the reform and opening up policy in 1978, Xishuangbanna has
experienced rapid urbanization. The urban population of Xishuang-
banna has grown from approximately 50 thousand in 1978 to more than
480 thousand in 2014 (Cao et al., 2017).Sitting at the conuence of the
Mekong River, Tonle Sap River and Bassac River, Phnom Penh was
developed on a ood plain. The annual average temperature in Phnom
Penh ranges from 28 ℃ to 34 ℃. As the capital city, Phnom Penh is the
political, economic and cultural center of Cambodia. The Khmer Rouge
regime (1975–1979) almost ruined the urban development of Phnom
Penh, and it was not until the late 1980 s that this city started to be
gradually rehabilitated and reconstructed (Shatkin, 1998). The popu-
lation of Phnom Penh now exceeds 2 million. Ho Chi Minh City sits close
to the Mekong Delta and covers more than 2000 km
2
. As the largest city
of Vietnam, Ho Chi Minh City is also the most important port city and
economic center in southern Vietnam. In the decade following the civil
war (1975–1985), Ho Chi Minh City showed hardly any urban devel-
opment. The implementation of Doi Moi policy (similar to the Reform
and Opening up policy of China) in the late 1980 s marked the entry of
Ho Chi Minh City into a new era of fast urban development (Nguyen
et al., 2016). The population of Ho Chi Minh City was approximately 12
million by 2014.
2.2. Data processing
To characterize urban growth patterns using the modied urban
growth index, Landsat MSS/TM/ETM+/OLI images (https://earth
explorer.usgs.gov) of Xishuangbanna, Phnom Penh and Ho Chi Minh
City were rst downloaded. All the Landsat images were L1T products
that had been processed with systematic, radiometric, geometrical, and
topographical correction. Of these images, MSS images had lower res-
olution (approximately 60 m) compared to TM/ETM+/OLI images
(approximately 30 m). To ensure that the classication results are
comparable with each other, all the images were resampled to a ground
resolution of 30 m. Through a decision tree process combined with vi-
sual modication in eCognition Developer 8.7, the remote sensing
images were interpreted as land use maps (Cao et al., 2019). Table 2
shows the overall classication accuracies which were veried by
topographic maps, aerial photos, and Google Earth (Shazadeh-Mog-
hadam et al., 2021). Urban land use data were extracted from the land
use maps. The identication and analyses of urban growth patterns were
then performed in ArcGIS 10.8.
Like most Chinese cities, Xishuangbanna has experienced fast urban
development since the Reform and Opening up in 1978. Therefore,
seven phases of urban land use data in Xishuangbanna (1976, 1990,
1995, 2000, 2005, 2010, and 2015) were produced to identify the multi-
temporal urban growth patterns. Interpreted from MSS images, the land
use map in 1976 showed lower overall accuracy compared with other
phases. The small extent of urban area in 1976 allowed us to implement
careful visual modication based on the existing literature and data.
Phnom Penh and Ho Chi Minh City have long histories of urban growth
that are beyond the acquisition ability of satellite technology. Since the
urban development of both cities was revived in the early 1990 s, the
urban area of these two cities in 1990 can be regarded as the initial
phase in the study of urban growth patterns. Furthermore, no adequate
remote sensing images corresponding to the rst phase of Xishuang-
banna are found in Phnom Penh or Ho Chi Minh City. With the above
considerations, we produced six phases of urban land use data for
Phnom Penh and Ho Chi Minh City, namely, 1990, 1995, 2000, 2005,
2010, and 2015.
2.3. The modied urban growth index to identify multi-temporal urban
growth patterns
According to the denition of S in Table 1, the urban growth patterns
are determined by L
c
and P, where L
c
is the length of the shared edge of a
newly generated urban patch and the old urban area; L
u
is the length of
the unshared edge of the newly generated urban patch; and P is the
perimeter of the newly generated urban patch and consists of L
c
and L
u
(Fig. 2a). When a new urban patch is generated inside the old urban
area, P is equal to L
c
(Fig. 2b), and the S value will be 1 at this time. By
Fig. 1. Location of case studies.
H. Cao et al.

Ecological Indicators 142 (2022) 109206
4
identifying the newly generated urban patches and old urban are at each
time period, S is easy to compute in the GIS software. However, since all
the newly generated urban patches would be regarded as old urban area
in the next time phase, S could not properly recognize the growth
pattern of continuously grown urban patches (Fig. 3), and, thus, was
unable to identify multi-temporal urban growth patterns.
In this study, we proposed the modied urban growth index S
m
to
identify multi-temporal urban growth patterns. Different from S, S
m
calculates the length of the shared edge of a newly generated urban
patch based on the main urban area, not the entire old urban area. The
Table 2
Classication accuracies of Xishuangbanna, Phnom Penh and Ho Chi Minh City.
City 1976 1990 1995 2000 2005 2010 2015
Xishuangbanna 87.52 % 92.12 % 92.50 % 93.85 % 94.42 % 91.73 % 94.23 %
Phnom Penh ———— 87.93 % 89.82 % 88.21 % 89.82 % 90.00 % 93.18 %
Ho Chi Minh City ———— 87.94 % 91.62 % 87.50 % 88.53 % 91.91 % 91.41 %
Fig. 2. Illustration of the denition of S.
Fig. 3. Example of multi-temporal urban growth.
H. Cao et al.

Ecological Indicators 142 (2022) 109206
5
main urban area is dened as a spatially continuous urban patch that
includes the central urban area.
Sm=Lm/P(1)
Where L
m
is the length of the shared edge of a newly generated urban
patch and the main urban area; L
u
is the length of the unshared edge of
the newly generated urban patch; and P is the perimeter of the newly
generated patch and consists of L
m
and L
u
(Fig. 3). For an outlying
growth patch, P is represented by L
u
, and L
m
is equal to 0. S
m
ranges from
0 to 1 and follows the criterion of S to dene urban growth patterns. The
newly grown urban patch will be recognized as inlling if S
m
is greater
than or equal to 0.5, edge-expansion if S
m
is less than 0.5 and larger than
0, and outlying if S
m
is equal to 0.
Fig. 3 shows an example of multi-temporal urban growth patterns. At
phase t, patch b and patch c can be recognized as outlying patches, and
patch a represents the main urban area. At phase t +1, patch d and patch
g exhibit outlying growth. However, patch g has always been catego-
rized as an edge-expansion patch by index S. In fact, patch g, together
with patch c, is part of a developing outlying cluster in the process of
urban expansion. Patch e and patch f are identied as inlling growth
and edge-expansion growth, respectively. They become part of the main
urban area during this period. At phase t +2, patch i and patch h are
outlying patches developed around patch d, while they will be catego-
rized as edge-expansion growth or inlling growth by index S. Since
patch b is connected to the main urban area through patch j, both patch
b and patch j are merged into the main urban area. Patch j is conse-
quently identied as inlling growth.
To identify the multi-temporal urban growth patterns using the
modied urban growth index S
m
, a 4-step GIS approach is provided
based on urban land use data.
Step 1: Dene the main urban area in the rst phase of the urban land
use map (e.g., patch a at phase t in Fig. 3) and assign patches isolated
with the main urban area as outlying patches (e.g., patch b and patch c in
Fig. 3).
Step 2: For the next phase of the urban land use map, if the newly
generated patches are connected to both the main urban area and the old
outlying patches, the old outlying patches are merged into the main
urban area. For example, in Fig. 3, patch j connects both patch a and
patch b at phase t +2, and patch b should be assigned as part of the main
urban area.
Step 3: Based on the denition of index S
m
, identify each urban
growth pattern and then merge both inlling patches and edge-
expansion patches into the main urban area.
Step 4: Repeat steps 2 and 3 until all phases of urban growth patterns
are identied.
Using this index, the urban growth patterns of each time period in
Xishuangbanna, Phnom Penh and Ho Chi Minh City were identied
accordingly.
2.4. The aggregation index to detect the diffusion-coalescence process
Landscape indices, such as the contagion index, are commonly used
to detect the diffusion-coalescence process ((Dietzel et al., 2005b; Tian
et al., 2011; Zhao et al., 2014). In this study, the aggregation index was
introduced to validate the relationship between the diffusion-
coalescence process and urban growth patterns. Compared with the
contagion index, which measures the overall landscape aggregation, the
aggregation index (AI) is more precise in calculating the aggregation
level since it is class-specic (He et al., 2000; McGarigal et al., 2012).
AI =[eii
max→eii]×100 (2)
Where e
ii
represents the total edges shared by class i itself and max →
e
ii
is the largest likely number of shared edges with the same area of class
i. The aggregation index ranges from 0 to 100, and it will achieve the
maximum value if there is only one single, compact patch. Like the the
contagion index, the aggregation index usually shows the highest value
at the very beginning of diffusion and when the urban area is completely
built out. In the early stage of diffusion, the aggregation index will
decrease with the dispersion of urban patches. When the urban land-
scape transforms from diffusion to coalescence, the aggregation index
tends to increase because the urban patches will usually aggregate
during this period.
3. Results
3.1. Comparison of S
m
and S
To assess the performance of S
m
in identifying multi-temporal urban
growth patterns, we compared the results of urban growth pattern
recognition using S
m
and S (Fig. 4).
In Xishuangbanna, the result of S
m
showed that outlying growth was
always the dominant urban growth pattern before 2010. The area and
patch number of newly generated edge-expansion patches rst
decreased before 2005 and increased signicantly thereafter. It became
the dominant urban growth pattern between 2010 and 2015. The total
area and number of edge-expansion, inlling, and outlying patches from
1976 to 2015 were approximately 20.8 km
2
, 6 km
2
, 22.3 km
2
and 142,
150, 456, respectively. The result of S showed that edge-expansion
growth had been the dominant urban growth pattern since 1990. The
total area and number of edge-expansion, inlling, and outlying patches
from 1976 to 2015 became 34.1 km
2
, 6.8 km
2
, 8.1 km
2
and 369, 179,
200, respectively. In Phnom Penh, the result of both S
m
and S showed
that edge-expansion growth was always the dominant urban growth
pattern from 1990 to 2015. However, the area and number of outlying
patches identied by S were less than those identied by S
m
, and the
total area of inlling patches was larger than that of outlying patches
determined by S. In Ho Chi Minh City, the result of S
m
indicated that
outlying growth and the other two urban growth patterns alternately
dominated the urban growth process from 1990 to 2015. As a result of S,
edge-expansion and inlling growth have always been the dominant
urban growth patterns in recent decades.
Generally, since S categorized urban growth patterns based on the
shared boundary between the newly generated patches and the old
urban area, urban patches grown along or inside the existing outlying
clusters would be identied as edge-expansion patches or inlling
patches. Therefore, edge-expansion and inlling patches identied by S
would always be dominant in both area and patch number in these case
cities.
3.2. Changes in urban growth patterns
Fig. 5 and Table 3 show the spatial distribution, area and patch
number of urban growth patterns in each time period in Xishuangbanna
from 1976 to 2015. The signicant increasing trend of both the area and
number of newly generated urban patches after 2005 indicates that
Xishuangbanna has experienced an accelerating process of urban
expansion in recent years.
The main urban area of Xishuangbanna in 1976 was quite small and
mostly distributed along the south bank of the Lancang-Mekong River.
From 1976 to 1990, the area and number of newly generated urban
patches were 7.77 km
2
and 145, respectively. Outlying growth domi-
nated in both area and patch number. The remaining newly generated
urban patches were mostly edge-expansion patches, and the area of
inlling patches was close to zero. The spatial distribution of outlying
patches almost outlined the urban extent of Xishuangbanna in the next
decades.
Both the area and number of newly generated urban patches
decreased by half from 1990 to 1995 compared with the last period.
Outlying growth continually dominated in both area and number, fol-
lowed by edge-expansion growth. The area of outlying patches on the
H. Cao et al.

Ecological Indicators 142 (2022) 109206
6
north bank of the Lancang River comprised approximately 55 % of the
total area of outlying patches. Edge-expansion growth mainly occurred
south of the main urban area.
The area of newly generated urban patches between 1995 and 2000
was close to that in the period from 1976 to 1990. However, the number
of newly generated urban patches in the former was only approximately
2/3 of that in the latter. Outlying growth was still the dominant urban
growth pattern, while inlling patches exceeded edge-expansion
patches in both area and number during that time. Over 70 % of the
inlling growth occurred to the southeast of the main urban area. The
north bank of the Lancang-Mekong River was still a signicant outlying
cluster that accounted for more than 40 % of the total area of outlying
patches during this period.
From 2000 to 2005, the number of newly generated urban patches
increased slightly compared with the last period, while the area of newly
generated urban patches decreased signicantly. Outlying growth was
still the dominant urban growth pattern, and the area and number of
inlling patches were continually greater than those of edge-expansion
patches during that time. More than 80 % of the edge-expansion growth
occurred south of the main urban area and the bank of the Lancang-
Mekong River. Inlling patches were mainly distributed throughout
the west and northwest of the main urban area. Nearly 3/4 of the
outlying growth occurred in the direction between southeast and
southwest around Xishuangbanna International Airport and the Man-
longfeng area.
After 2005, Xishuangbanna experienced faster urbanization. The
area and number of newly generated urban patches increased by more
than 10 km
2
and 170, respectively, from 2005 to 2010. Outlying growth
continually dominated in number during this period. Although the
number of outlying patches was more than 5 times that of edge-
expansion patches, the area of outlying patches was slightly larger
than that of edge-expansion patches. The number of edge-expansion
patches was close to that of inlling patches, while the area of edge-
expansion patches was approximately 12 times that of inlling
growth. Since Xishuangbanna International Airport and Jinghong In-
dustrial Park were integrated with the main urban area, the newly
generated edge-expansion patches mostly distributed throughout the
southwest alongside airport express highways, in the northwest in
Jinghong Industrial Park, and along the south bank of the Lancang-
Mekong River. The Manlongfeng area southeast of the main urban
area, experienced signicant outlying development during that time,
and the area of outlying patches in this region comprised almost 40 % of
the total area of outlying patches.
The area of the newly generated urban patches continued to increase
signicantly from 2010 to 2015, while the number of newly generated
urban patches was almost identical to that between 2005 and 2010.
Edge-expansion growth rst became the dominant urban growth
pattern, and the area of edge-expansion patches comprised approxi-
mately 70 % of the total area of the newly generated urban patches.
Outlying growth decreased dramatically in both area and patch number
during that time. Edge-expansion growth mainly occurred in the
southeast between the Lancang-Mekong River and Liusha River, and in
the northwest around Jinghong Industrial Park. Since the central urban
area was almost saturated, inlling growth mainly occurred in the
southwest around the Xishuangbanna International Airport and Man-
longfeng areas. Approximately 40 % of the outlying growth occurred in
the southwest in the Gasa area, 9 ~ 12 km from the urban center.
Fig. 6 and Table 4 show the spatial distribution, area and patch
Fig. 4. Comparison of the area and number of urban growth patterns in the three cities identied by S
m
and S. (a) Area of urban growth patterns in Xishuangbanna;
(b) number of urban growth patterns in Xishuangbanna; (c) area of urban growth patterns in Phnom Penh; (d) number of urban growth patterns in Phnom Penh; (e)
area of urban growth patterns in Ho Chi Minh City; (f) number of urban growth patterns in Ho Chi Minh City.
H. Cao et al.

Ecological Indicators 142 (2022) 109206
7
number of urban growth patterns in each time period in Phnom Penh
from 1990 to 2015. In general, edge-expansion growth was the domi-
nant urban growth pattern in the past decades, accounting for over 50 %
of the total area of newly generated urban patches. Edge-expansion
growth mainly occurred toward the west along the central urban area
and the Phnom Penh International Airport, and the edge-expansion
patches always had larger sizes compared with the patches of other
urban growth patterns. Outlying growth mainly occurred in the north-
ern peninsula, southern peninsula, the Phnom Penh Special Economic
Zone (SEZ), and some other satellite cities outside the central urban
area.
In 1990, the urban area of Phnom Penh mainly consisted of the
central urban area and the Phnom Penh International Airport. From
1990 to 1995, the area and number of newly generated urban patches
were 14.3 km
2
and 94, respectively. The patch numbers of urban growth
patterns were close to each other. Of these patches, the area of newly
generated edge-expansion patches was the largest, accounting for
almost half of the total area of newly generated urban patches.
From 1995 to 2000, both the area and patch number of the newly
generated urban patches slightly increased compared with those in the
previous period. Outlying growth exceeded edge-expansion growth and
became the largest urban growth pattern at that time. Most of the newly
generated outlying patches were located in the northern and southern
peninsulas, while edge-expansion growth mainly occurred at the south
of the main urban area and around the Phnom Penh International
Airport.
Phnom Penh experienced the most signicant urban growth between
2000 and 2005, and the total area of the newly generated urban patches
exceeded 32 km
2
. The large size edge-expansion growth contributed to
more than 2/3 of the overall urban growth during that period. The
newly generated edge-expansion patches were mainly distributed to the
west and extended along the National 4th Road.
Although the patch number of the newly generated urban patches
slightly increased from 2005 to 2010, the area of the newly generated
urban patches decreased dramatically to less than 10 km
2
. Edge-
expansion growth contributed more than half of the overall urban
Fig. 5. Spatial variation in urban growth patterns in Xishuangbanna from 1976 to 2015.
Table 3
Changes in the area and patch number of urban growth patterns in Xishuangbanna.
Period Edge-expansion Inlling Outlying Total
Area(km
2
) Number Area(km
2
) Number Area(km
2
) Number Area(km
2
) Number
1976–1990 1.883 29 0.025 14 5.864 102 7.773 145
1990–1995 1.227 12 0.219 10 2.132 48 3.578 70
1995–2000 1.028 11 1.663 18 3.812 61 6.503 90
2000–2005 0.728 9 0.804 16 3.339 79 4.871 104
2005–2010 4.870 24 0.423 20 5.724 126 11.017 170
2010–2015 11.066 57 2.886 72 1.387 40 15.339 169
H. Cao et al.

Ecological Indicators 142 (2022) 109206
8
growth and mainly occurred along the north and south of the main
urban area, while outlying growth had the largest patch number and
mainly occurred at the northern peninsula and the Phnom Penh SEZ.
From 2010 to 2015, Phnom Penh showed fast urban growth again.
The total area of the newly generated urban patches exceeded 25 km
2
,
and the number of the newly generated urban patches was almost 2
times of that between 2005 and 2010. The Phnom Penh SEZ and south
peninsula experienced signicant outlying growth at that time, while
newly generated edge-expansion patches were distributed surrounding
the main urban area.
Fig. 7 and Table 5 show the spatial distribution, area and patch
number of urban growth patterns in each time period in Ho Chi Minh
City from 1990 to 2015. Ho Chi Minh City also showed fast urban
growth in recent decades, especially between 1995 and 2010. Compared
with Xishuangbanna and Phnom Penh, the newly generated urban
patches in Ho Chi Minh City were more fragmented. The urban growth
patterns mainly include outlying growth toward Thu Duc County and
District 9 in the east, toward Districts 2 and 7 in the southeast, and to-
ward Cu Chi County in the northwest, as well as edge-expansion and
inlling growth toward Thu Duc County in the east, and along the north
and west of the main urban area.
From 1990 to 1995, the area and number of newly generated urban
patches were 42.2 km
2
and 578, respectively. Outlying growth was the
dominant urban growth pattern in both area and patch number. The
newly generated outlying growth patches were mainly distributed
throughout Thu Duc County and District 7 in the east and southeast.
After 1995, the urban growth rate of Ho Chi Minh City was accel-
erated. Between 1995 and 2000, the area and number of newly gener-
ated urban patches reached 109.8 km
2
and 1221, respectively. Since Thu
Duc County integrated into the main urban area during that time, edge-
expansion and inlling growth together exceeded outlying growth. Most
of the newly generated edge-expansion and inlling patches were
distributed toward Thu Duc County, and to the north and west of the
main urban area, while outlying growth mainly occurred to the south of
the main urban area.
From 2000 to 2005, the area of the newly generated urban patches
decreased to 84.3 km
2
, while the number of the newly generated urban
patches increased slightly. Outlying growth became the dominant urban
growth pattern again during that time. The newly generated outlying
patches were mainly distributed in Districts 9, 2, and 7, from east to
south. Edge-expansion and inlling growth continually occurred toward
Thu Duc County, and along the north and west sides of the main urban
area.
The area of the newly generated urban patches between 2005 and
2010 was slightly larger than that in the previous period, while the
number of the newly generated urban patches signicantly decreased.
The area of edge-expansion growth increased dramatically and was close
to that of the outlying growth. Outlying growth mainly occurred in
District 2, District 7, and Cu Chi County, while the newly generated
edge-expansion patches were mostly distributed north and west of the
Fig. 6. Spatial variation in urban growth patterns in Phnom Penh from 1990 to 2015.
Table 4
Changes in the area and patch number of urban growth patterns in Phnom Penh.
Period Edge-expansion Inlling Outlying Total
Area(km
2
) Number Area(km
2
) Number Area(km
2
) Number Area(km
2
) Number
1990–1995 6.981 26 4.175 35 3.147 33 14.304 94
1995–2000 6.042 32 2.278 20 7.404 45 15.724 97
2000–2005 22.074 21 5.027 41 5.204 50 32.306 112
2005–2010 4.928 39 2.303 35 2.163 48 9.395 122
2010–2015 11.847 49 5.530 74 8.164 106 25.540 229
H. Cao et al.

Ecological Indicators 142 (2022) 109206
9
Fig. 7. Spatial variation in urban growth patterns in Ho Chi Minh City from 1990 to 2015.
Table 5
Changes in the area and patch number of urban growth patterns in Ho Chi Minh City.
Period Edge-expansion Inlling Outlying Total
Area(km
2
) Number Area(km
2
) Number Area(km
2
) Number Area(km
2
) Number
1990–1995 9.415 85 7.356 86 25.464 407 42.235 578
1995–2000 29.587 252 32.976 247 47.209 722 109.772 1221
2000–2005 13.541 248 25.573 365 45.137 743 84.251 1356
2005–2010 36.331 249 13.027 267 40.124 577 89.482 1093
2010–2015 14.688 372 18.008 509 23.623 506 56.319 1387
Fig. 8. Differences in urban growth patterns identied by S
m
and S in the Manlongfeng area in Xishuangbanna. (a) Growth patterns identied by S
m
from 2000 to
2005; (b) growth patterns identied by S from 2000 to 2005; (c) growth patterns identied by S
m
from 2005 to 2010; (d) growth patterns identied by S from 2005 to
2010; (e) growth patterns identied by S
m
from 2010 to 2015; (f) growth patterns identied by S from 2010 to 2015.
H. Cao et al.

Ecological Indicators 142 (2022) 109206
10
main urban area.
The urban growth from 2010 to 2015 was the largest; however, the
area of the newly generated urban patches decreased to 56.3 km
2
. Edge-
expansion growth and inlling growth together exceeded outlying
growth at that time, and the newly generated edge-expansion and
inlling patches were mainly located along north of the main urban
area.
4. Discussion
4.1. Advantages and limitations of S
m
Compared with S, S
m
provides more precise results of multi-temporal
urban growth patterns. Taking the Manlongfeng area in Xishuangbanna
as an example, Fig. 8 further shows the spatial differences in S
m
and S in
identifying multi-temporal urban growth patterns.
The Manlongfeng area experienced signicant urban growth in
recent years. From 2000 to 2005, the newly generated urban patches in
this area were identied as outlying patches using S
m
; however, most
patches were categorized as edge-expansion patches by S. The Man-
longfeng area exhibited signicant outlying growth between 2005 and
2010, while the newly generated outlying patches were still mostly
categorized as edge-expansion patches by S. From 2010 to 2015, the
newly generated urban patches integrated the Manlongfeng area and the
main urban area; thus, S
m
and S showed similar result, and the urban
growth pattern in the Manlongfeng area consequently became edge-
expansion growth and inlling growth. In summary, outlying patches
identied by S
m
were more often classied as edge-expansion patches
using S, because most existing outlying patches were compact with a
smaller size, and the new urban patches would mostly grow along the
edges. Consequently, the misclassication of outlying patches by S
affected the spatial and temporal characteristics of urban growth.
Some researchers argued the necessity of buffer zones, because if a
newly generated urban patch was very close to the old urban patches, it
should not be recognized as the outlying growth. In fact, whether a
newly generated urban patch was resulted from outlying growth should
be considered from both spatial and temporal perspectives. If this patch
was closely connected with the main urban area, it would soon integrate
into the main urban area and then be identied as edge-expansion
growth or inlling growth. In this situation, the occurrence of this
patch would be just a special and rare case due to the selection of the
time phase. If this patch was still isolated from the main urban area after
several time periods, then the patch itself should be considered as
outlying growth. Therefore, the modied urban growth index S
m
, which
focuses on the multi-temporal spatial relationships between the newly
generated urban patch and the main urban area, could properly address
this issue. However, there is an exception when a patch was grown in-
side the main urban area and had no shared boundary. The buffer based
indices, by comparison, could perform better through dening adequate
buffer distance.
In this study, S
m
is calculated based on the main urban area that is
dened as a spatially continuous urban patch including the central
urban area. However, in some cities, there might be several equivalent
urban centers that are disconnected with each other. The performance of
most urban growth indices including S
m
would be affected in this situ-
ation, because the urban growth pattern is decided not only by the
shared edges or buffer area, also by the spatial conguration of urban
areas. For example, if the multi urban areas were very close to each other
and distributed in a ring belt style, it’s difcult to decide the urban
growth pattern when a patch was grown interior to the urban belt.
4.2. Diffusion and coalescence process of urban growth
The results of multi-temporal urban growth pattern identication
reected Dietzel’s hypothetical framework, which describes the
spatially evolving characteristics of urban development as a two-step
process of diffusion and coalescence (Dietzel et al., 2005a).
Many studies have adopted urban growth patterns to test and
delineate the diffusion and coalescence process (Liu et al., 2010; Sun and
Zhao, 2018; Xu et al., 2007; Yu and Ng, 2007). Generally, the dominance
of outlying growth indicates the beginning of the diffusion process,
while the coalescence process can be observed when the urban growth
pattern mainly shows inlling growth and edge-expansion growth (Tian
et al., 2011; Zhao et al., 2014). Nevertheless, since these studies mostly
used S or LEI for the identication of urban growth patterns, which may
improperly identify outlying patches as inlling or edge-expansion
patches, the results could not always reect the diffusion and coales-
cence process very well (Bosch et al., 2020; Chakraborty et al., 2021; Li
et al., 2013). In this study, the modied urban growth index S
m
provides
more precise results of urban growth patterns and is capable of testing
diffusion-coalescence theory. Fig. 9 shows an example of the diffusion-
coalescence process of urban growth in Xishuangbanna. The urban
area of Xishuangbanna in 1976 can be regarded as the “urban seed”. As
the urban area grew, it diffused to new outlying patches. After 1990, the
increasing trend of edge-expansion and inlling growth indicated that
Xishuangbanna began the next stage of diffusion, and urban growth was
“accompanied by organic growth which leads to the outward expansion
of existing urban areas and the inlling of gaps within them” (Dietzel
et al., 2005a). Xishuangbanna gradually reached the so-called “point”,
where the urban patches began to aggregate toward a saturated urban
form. Compared with Xishuangbanna, Phnom Penh experienced much
more edge-expansion and inlling growth. From 1995 to 2000, the
urban growth showed a diffusion process, and outlying growth became
the most signicant urban growth pattern. Subsequently, the urban
landscape experienced a continuous coalescence process with the
dominance of edge-expansion and inlling growth. Ho Chi Min City has
experienced obvious outlying development since 1990, which can be
regarded as the early stage of urban diffusion. The urban growth process
also started to transition from diffusion to coalescence after 2000
because edge-expansion and inlling growth gradually became the
dominant urban growth patterns.
As shown in Fig. 10, the V-shaped trends of the aggregation index in
Xishuangbanna, Phnom Penh and Ho Chi Minh City indicated that these
cities all experienced the process from diffusion to coalescence in recent
decades. In addition, Phnom Penh and Ho Chi Minh City were more
aggregated than Xishuangbanna, since their aggregation indices ach-
ieved higher values than that of Xishuangbanna.
In summary, urban growth in the early years in the case cities can be
regarded as the diffusion process of urban growth, which is always
accompanied by signicant outlying growth and a decrease in the ag-
gregation index. In the transition period from diffusion to coalescence,
the proportion of edge-expansion and inlling growth generally in-
creases, while outlying growth tends to decline continually or in a
uctuating manner. The aggregation indices showed an increasing trend
and reached the highest value by the end of the coalescence process. By
2015, Xishuangbanna, Phnom Penh and Ho Chi Minh City seemed to
remain in the coalescence process. Overall, the results of urban growth
patterns derived from S
m
provided evidence for the hypothesis of
diffusion and coalescence.
5. Highlighting the outlying clusters
Another essential application of multi-temporal urban growth
pattern identication is to characterize the spatial dynamics and struc-
ture of urban expansion. Outlying urban clusters, which usually develop
to be sub-centers, has been commonly discussed in multi-temporal urban
growth studies (He et al., 2018; Jiao et al., 2015). By overlapping
outlying growth patches of all the time periods, outlying clusters were
easily recognized. Fig. 11 shows the outlying clusters and their devel-
opment in different time periods in Xishuangbanna, Phnom Penh and Ho
Chi Minh City.
According to the Master Plan of Xishuangbanna (1999–2020)
H. Cao et al.

Ecological Indicators 142 (2022) 109206
11
(hereinafter referred to as the Master Plan), the spatial structure of the
Xishuangbanna urban area is set as one central urban area and four sub-
centers (Jiangbei area, Gadong area, Gasa area, and Manlongfeng area).
By 2015, the urban spatial structure of Xishuangbanna was mainly in
accordance with the Master Plan except for the Gasa area. Instead, the
airport area became one of the sub-centers after decades of dramatic
expansion. Sitting on the north bank of the Lancang-Mekong River, the
Jiangbei Area showed typical outlying development, and outlying
growth was especially signicant from 1976 to 2000. Although the
outlying growth of the Gadong area lasted for decades, it was the con-
struction of Jinghong Industrial Park in 2005 that marked the fast
development of the Gadong area. The Gadong area was then developed
as a consequence of inlling growth and edge-expansion growth, since
Jinghong Industrial Park merged into the main urban area at the
beginning of its construction. The airport area and Manlongfeng area
rst developed as outlying clusters and then experienced integrated
development with the main urban area. Outlying development around
the airport area started when the Xishuangbanna International Airport
was built in approximately 1990. With the connection of the airport area
and the main urban area through the newly generated urban patches
along the Airport Road, the airport area began to coalesce with the main
urban area after 2005. The Manlongfeng area did not show obvious
urban growth until it was designated as one of the sub-centers by the
Master Plan. The outlying development of the Manlongfeng area mainly
occurred between 2005 and 2010. After 2010, the Manlongfeng area
merged into the main urban area as the result of the coalescence process.
In recent decades, Phnom Penh has launched many satellite city
projects outside the central urban area (Mialhe et al., 2019). Most
outlying clusters, such as the northern peninsula and southern penin-
sula, were developed based on these satellite cities. The northern
peninsula rst experienced fast outlying growth between 1995 and
2000, and with the construction of satellite projects, including the Chroy
Chanvgar (CC) satellite city project and the Ly Yong Phat (LYP) project,
this area has further developed into a sub-center since 2005. Compared
Fig. 9. Diffusion-coalescence process of urban growth (upper) and the case in Xishuangbanna (lower).
Fig. 10. Variation in the aggregation index in the diffusion-coalescence process in Xishuangbanna, Phnom Penh and Ho Chi Minh City.
H. Cao et al.

Ecological Indicators 142 (2022) 109206
12
with the northern peninsula, the southern peninsula experienced
continual outlying development from 1990 to 2015, and the outlying
growth was most signicant between 2010 and 2015. The region north
of the central urban area also exhibited obvious outlying growth from
1995 to 2005. With the construction of Camko city, this region was in-
tegrated into the main urban area in recent years. The Phnom Penh SEZ
was another typical outlying cluster since 2005. Sitting close to the
National 4th Road, it is easier to access the Phnom Penh International
Airport and Sihanoukville port.
The outlying clusters identied in Ho Chi Minh City mainly consisted
of the urban regions in Thu Duc County, Cu Chi County, and Districts 2,
7, and 9. Of these areas, District 7 has always experienced outlying
development in recent decades. Thu Duc County showed the most sig-
nicant outlying growth from 1990 to 1995. With the enhanced
connection between Thu Duc County and the main urban area, Thu Duc
County gradually merged into the main urban area after 1995. Districts
Fig. 11. Outlying clusters and their development. (a) Xishuangbanna; (b) Phnom Penh; (c) Ho Chi Minh City.
H. Cao et al.

Ecological Indicators 142 (2022) 109206
13
2 and 9 experienced the most signicant outlying development between
2000 and 2010. The outlying growth of Districts 2, 7 and 9 most likely
resulted from the activities of foreign property investment (Jung et al.,
2013; Kontgis et al., 2014). Cu Chi County, which used to be regarded as
a rural area, lies northwest of the main urban area. It has shown fast
outlying growth in recent years, especially since 2005.
The outlying clusters in the case cities either retained signicant
outlying growth throughout the past decades or nally integrated into
the main urban area. Consequently, they have been developed as sub-
centers or have shown great potential to be new urban hotspots. Thus,
the identication of multi-temporal urban growth patterns using index
S
m
is able to highlight the outlying clusters and their evolving charac-
teristics. In addition, by describing the relationship between outlying
clusters and the main urban area, the dynamics of urban spatial struc-
ture could also be revealed.
6. Conclusions
In this study, we proposed the modied urban growth index S
m
to
identify multi-temporal urban growth patterns. Taking Xishuangbanna,
Phnom Penh and Ho Chi Minh City as case cities, a 4-step GIS approach
was implemented to identify urban growth patterns using S
m
. In Xish-
uangbanna, outlying growth was always the dominant urban growth
pattern until 2010, while the area and patch number of edge-expansion
patches rst decreased before 2005 and then increased signicantly
thereafter. In Phnom Penh, edge-expansion growth dominated the urban
growth process at most times, and the most signicant outlying growth
occurred during the time periods from 1995 to 2000 and from 2010 to
2015. In Ho Chi Minh City, the newly generated urban patches were
more fragmented with smaller sizes, and outlying growth and the other
two urban growth patterns alternately dominated the urban growth
process. We further compared the results of S
m
and S in identifying
multi-temporal urban growth patterns. Generally, edge-expansion
growth and inlling growth were always the dominant urban growth
patterns according to the results of S because the multi-temporal
outlying growth was classied as edge-expansion growth or inlling
growth. The improper result may further lead to misunderstandings of
the urban growth process and characteristics. Therefore, S
m
performed
better in the identication of multi-temporal urban growth patterns. The
limitations of S
m
were also discussed. When an urban patch was grown
inside the main urban area or a city had multi equivalent and discon-
nected urban areas, the performance of S
m
might be affected.
The results of multi-temporal urban growth pattern identication
using S
m
in the case cities veried the diffusion-coalescence theory. In
the early stage of diffusion, outlying patches always increased signi-
cantly with the dispersion of urban patches. During the transition period
from diffusion to coalescence, outlying growth would show a decreasing
trend in general. When the urban area started to coalesce toward a
saturated landscape, edge-expansion and inlling growth became the
dominant urban growth patterns. By highlighting the outlying clusters
and their multi-temporal urban growth characteristics, the dynamics of
urban spatial structure could always be identied. Thus, with the
identication of multi-temporal urban growth patterns, the modied
urban growth index S
m
is suitable for the applications in delineating the
diffusion-coalescence process and the evolving characteristics of urban
spatial structure.
CRediT authorship contribution statement
Hui Cao: Conceptualization, Data curation, Writing – original draft,
Writing – review & editing, Funding acquisition. Jianglong Chen:
Writing – original draft. Cheng Chen: Data curation. Pingxing Li:
Conceptualization, Writing – original draft, Writing – review & editing,
Funding acquisition.
Declaration of Competing Interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Data availability
Data will be made available on request.
Acknowledgements:
The authors acknowledge funding for this research from the National
Natural Science Foundation of China (Grant No. 41901215, 41871209).
References
Ahrends, A., Hollingsworth, P.M., Ziegler, A.D., Fox, J.M., Chen, H., Su, Y., Xu, J., 2015.
Current trends of rubber plantation expansion may threaten biodiversity and
livelihoods. Global Environ. Change 34, 48–58.
Batty, M., 2002. Thinking about cities as spatial events. Environ. Plann. B Planning &
Design 29, 1–2.
Batty, M., Howes, D., 2001. Predicting temporal patterns in urban development from
remote imagery. In: Donnay, J.P., Barnsley, M.J., Longley, P.A. (Eds.), Remote
Sensing and Urban Analysis. Taylor & Francis, London, pp. 185–204.
Berling-Wolff, S., Wu, J., 2004. Modeling urban landscape dynamics: a case study in
Phoenix, USA. Urban ecosystems 7, 215–240.
Bhatta, B., Saraswati, S., Bandyopadhyay, D., 2010. Urban sprawl measurement from
remote sensing data. Appl. Geogr. 30, 731–740.
Bosch, M., Jaligot, R., Chenal, J., 2020. Spatiotemporal patterns of urbanization in three
Swiss urban agglomerations: insights from landscape metrics, growth modes and
fractal analysis. Landscape Ecol. 35, 879–891.
Burgess, E.W., 1925. The growth of the city: An introduction to a research project. In:
Park, R.E., Burgess, E.W., McKenzie, R.D. (Eds.), The City. The University of Chicago
Press, pp. 47–62.
Camagni, R., Gibelli, M.C., Rigamonti, P., 2002. Urban mobility and urban form: the
social and environmental costs of different patterns of urban expansion. Ecol. Econ.
40, 199–216.
Cao, H., Liu, J., Fu, C., Zhang, W., Wang, G., Yang, G., Luo, L., 2017. Urban Expansion
and Its Impact on the Land Use Pattern in Xishuangbanna since the Reform and
Opening up of China. Remote Sensing 9, 137.
Cao, H., Liu, J., Chen, J., Gao, J., Wang, G., Zhang, W., 2019. Spatiotemporal Patterns of
Urban Land Use Change in Typical Cities in the Greater Mekong Subregion (GMS).
Remote Sensing 11, 801.
Chakraborty, S., Maity, I., Patel, P.P., Dadashpoor, H., Pramanik, S., Follmann, A.,
Novotný, J., Roy, U., 2021. Spatio-temporal patterns of urbanization in the Kolkata
Urban Agglomeration: A dynamic spatial territory-based approach. Sustainable
Cities and Society 67, 102715.
Cui, X., Wang, X., 2015. Urban land use change and its effect on social metabolism: An
empirical study in Shanghai. Habitat International 49, 251–259.
Delphin, S., Escobedo, F.J., Abd-Elrahman, A., Cropper, W.P., 2016. Urbanization as a
land use change driver of forest ecosystem services. Land Use Policy 54, 188–199.
Dietzel, C., Herold, M., Hemphill, J.J., Clarke, K.C., 2005a. Spatio-temporal dynamics in
California’s Central Valley: Empirical links to urban theory. International Journal of
Geographical Information Science 19, 175–195.
Dietzel, C., Oguz, H., Hemphill, J.J., Clarke, K.C., Gazulis, N., 2005b. Diffusion and
coalescence of the Houston Metropolitan Area: evidence supporting a new urban
theory. Environment and Planning B-Planning & Design 32, 231–246.
Dupras, J., Marull, J., Parcerisas, L., Coll, F., Gonzalez, A., Girard, M., Tello, E., 2016.
The impacts of urban sprawl on ecological connectivity in the Montreal Metropolitan
Region. Environ. Sci. Policy 58, 61–73.
Ewing, R., Hamidi, S., 2015. Compactness versus Sprawl. Journal of Planning Literature
30, 413–432.
Ewing, R., Pendall, R., Chen, D., 2002. Measuring sprawl and its impact. Smart Growth
America, Washington, DC.
Fan, X., Zhang, L., Yuan, L., Guo, B., Zhang, Q., Huang, H., 2022. Urbanization and water
quality dynamics and their spatial correlation in coastal margins of mainland China.
Ecol. Ind. 138, 108812.
Fan, J.-S., Zhou, L., 2019. Impact of urbanization and real estate investment on carbon
emissions: Evidence from China’s provincial regions. J. Cleaner Prod. 209, 309–323.
Fei, W., Zhao, S., 2019. Urban land expansion in China’s six megacities from 1978 to
2015. The Science of the total environment 664, 60–71.
Florian, S., Januar, H., 2016. Urban Development and Urbanization in the Greater
Mekong Subregion.
Forman, R.T.T., 1995. Land mosaics: The ecology of landscapes and regions. Cambridge
University Press, Cambridge.
Getu, K., Bhat, H.G., 2021. Analysis of spatio-temporal dynamics of urban sprawl and
growth pattern using geospatial technologies and landscape metrics in Bahir Dar.
Northwest Ethiopia. Land Use Policy 109, 105676.
Goodchild, M.F., 1992. Geographical information science. International journal of
geographical information systems 6, 31–45.
H. Cao et al.

Ecological Indicators 142 (2022) 109206
14
Green, A.M., Barnick, K.A., Pendergast, M.E., S
¸ekercio˘
glu, Ç.H., 2022. Species differences
in temporal response to urbanization alters predator-prey and human overlap in
northern Utah. Global Ecol. Conserv. 36, e02127.
Grimm, N.B., Faeth, S.H., Golubiewski, N.E., Redman, C.L., Wu, J., Bai, X., Briggs, J.M.,
2008. Global Change and the Ecology of Cities. Science 319, 756.
Harris, C.D., Ullman, E.L., 1945. The nature of cities. The Annals of the American
Academy of Political and Social Science 242, 7–17.
He, H.S., DeZonia, B.E., Mladenoff, D.J., 2000. An aggregation index (AI) to quantify
spatial patterns of landscapes. Landscape Ecol. 15, 591–601.
He, Q., He, W., Song, Y., Wu, J., Yin, C., Mou, Y., 2018. The impact of urban growth
patterns on urban vitality in newly built-up areas based on an association rules
analysis using geographical ‘big data’. Land Use Policy 78, 726–738.
Herold, M., Couclelis, H., Clarke, K.C., 2005. The role of spatial metrics in the analysis
and modeling of urban land use change. Comput. Environ. Urban Syst. 29, 369–399.
Hoyt, H., 1939. The structure and growth of residential neighborhoods in American
cities. Federal Housing Administration, Washington D.C.
Huang, X., Xia, J., Xiao, R., He, T., 2017. Urban expansion patterns of 291 Chinese cities,
1990–2015. Int. J. Digital Earth 12, 62–77.
Irwin, E.G., Bockstael, N.E., 2007. The evolution of urban sprawl: evidence of spatial
heterogeneity and increasing land fragmentation. PNAS 104, 20672–20677.
Jia, M., Zhang, H., Yang, Z., 2022. Compactness or sprawl: Multi-dimensional approach
to understanding the urban growth patterns in Beijing-Tianjin-Hebei region. China.
Ecological Indicators 138, 108816.
Jiao, L., Mao, L., Liu, Y., 2015. Multi-order Landscape Expansion Index: Characterizing
urban expansion dynamics. Landscape Urban Plann. 137, 30–39.
Jiao, L., Liu, J., Xu, G., Dong, T., Gu, Y., Zhang, B., Liu, Y., Liu, X., 2018. Proximity
Expansion Index: An improved approach to characterize evolution process of urban
expansion. Comput. Environ. Urban Syst. 70, 102–112.
Jung, S., Huynh, D., Rowe, P.G., 2013. The pattern of foreign property investment in
Vietnam: The apartment market in Ho Chi Minh City. Habitat International 39,
105–113.
Kontgis, C., Schneider, A., Fox, J., Saksena, S., Spencer, J.H., Castrence, M., 2014.
Monitoring peri-urbanization in the greater Ho Chi Minh City metropolitan area.
Appl. Geogr. 53, 377–388.
Lemoine-Rodriguez, R., Inostroza, L., Zepp, H., 2022. Does urban climate follow urban
form? Analysing intraurban LST trajectories versus urban form trends in 3 cities with
different background climates. The Science of the total environment 830, 154570.
Li, C., Li, J., Wu, J., 2013. Quantifying the speed, growth modes, and landscape pattern
changes of urbanization: a hierarchical patch dynamics approach. Landscape Ecol.
28, 1875–1888.
Li, G., Zhang, X., Mirzaei, P.A., Zhang, J., Zhao, Z., 2018. Urban heat island effect of a
typical valley city in China: Responds to the global warming and rapid urbanization.
Sustainable Cities and Society 38, 736–745.
Liu, X., Li, X., Chen, Y., Tan, Z., Li, S., Ai, B., 2010. A new landscape index for quantifying
urban expansion using multi-temporal remotely sensed data. Landscape Ecol. 25,
671–682.
Marquez, L.O., Smith, N.C., 1999. A framework for linking urban form and air quality.
Environ. Modell. Software 14, 541–548.
McGarigal, K., Cushman, S., Ene, E., 2012. FRAGSTATS v4: Spatial Pattern Analysis
Program for Categorical and Continuous Maps. Computer software program
produced by the authors at the University of Massachusetts, Amherst. Available at
the following web site: http://www.umass.edu/landeco/research/fragstats/fragstats
.html.
Mialhe, F., Gunnell, Y., Navratil, O., Choi, D., Sovann, C., Lejot, J., Gaudou, B., Se, B.,
Landon, N., 2019. Spatial growth of Phnom Penh, Cambodia (1973–2015): Patterns,
rates, and socio-ecological consequences. Land Use Policy 87, 104061.
Nguyen, T.B., Samsura, D.A.A., van der Krabben, E., Le, A.-D., 2016. Saigon-Ho Chi Minh
City. Cities 50, 16–27.
Qiu, G., Song, R., He, S., 2019. The aggravation of urban air quality deterioration due to
urbanization, transportation and economic development - Panel models with
marginal effect analyses across China. The Science of the total environment 651,
1114–1125.
Reis, J.P., Silva, E.A., Pinho, P., 2015. Spatial metrics to study urban patterns in growing
and shrinking cities. Urban Geography 37, 246–271.
Shazadeh-Moghadam, H., Minaei, M., Feng, Y., Pontius, R.G., 2019. GlobeLand30 maps
show four times larger gross than net land change from 2000 to 2010 in Asia. Int. J.
Appl. Earth Obs. Geoinf. 78, 240–248.
Shazadeh-Moghadam, H., Minaei, M., Pontius Jr, R.G., Asghari, A., Dadashpoor, H.,
2021. Integrating a Forward Feature Selection algorithm, Random Forest, and
Cellular Automata to extrapolate urban growth in the Tehran-Karaj Region of Iran.
Comput. Environ. Urban Syst. 87, 101595.
Shatkin, G., 1998. ‘Fourth World’cities in the global economy: the case of Phnom Penh,
Cambodia. Int. J. Urban Reg. Res. 22, 378–393.
Shrestha, M.K., York, A.M., Boone, C.G., Zhang, S., 2012. Land fragmentation due to
rapid urbanization in the Phoenix Metropolitan Area: Analyzing the spatiotemporal
patterns and drivers. Appl. Geogr. 32, 522–531.
Sun, Y., Zhao, S., 2018. Spatiotemporal dynamics of urban expansion in 13 cities across
the Jing-Jin-Ji Urban Agglomeration from 1978 to 2015. Ecol. Ind. 87, 302–313.
Tam, V.T., Nga, T.T.V., 2018. Assessment of urbanization impact on groundwater
resources in Hanoi. Vietnam. J Environ Manage 227, 107–116.
Tian, G., Jiang, J., Yang, Z., Zhang, Y., 2011. The urban growth, size distribution and
spatio-temporal dynamic pattern of the Yangtze River Delta megalopolitan region,
China. Ecol. Model. 222, 865–878.
United Nations, 2018. World Urbanization Prospects 2018: Key Facts. United Nations.
Wang, X., Tian, G., Yang, D., Zhang, W., Lu, D., Liu, Z., 2018. Responses of PM2.5
pollution to urbanization in China. Energy Policy 123, 602–610.
Wei, L., Luo, Y., Wang, M., Su, S., Pi, J., Li, G., 2020. Essential fragmentation metrics for
agricultural policies: Linking landscape pattern, ecosystem service and land use
management in urbanizing China. Agric. Syst. 182, 102833.
Wilson, E.H., Hurd, J.D., Civco, D.L., Prisloe, M.P., Arnold, C., 2003. Development of a
geospatial model to quantify, describe and map urban growth. Remote Sens.
Environ. 86, 275–285.
Wu, W., Zhao, S., Zhu, C., Jiang, J., 2015. A comparative study of urban expansion in
Beijing, Tianjin and Shijiazhuang over the past three decades. Landscape Urban
Plann. 134, 93–106.
Xia, C., Zhang, A., Yeh, A.-G.-O., 2020. Shape-weighted landscape evolution index: An
improved approach for simultaneously analyzing urban land expansion and
redevelopment. J. Cleaner Prod. 244, 118836.
Xie, W., Huang, Q., He, C., Zhao, X., 2018. Projecting the impacts of urban expansion on
simultaneous losses of ecosystem services: A case study in Beijing, China. Ecol. Ind.
84, 183–193.
Xu, C., Liu, M., Zhang, C., An, S., Yu, W., Chen, J.M., 2007. The spatiotemporal dynamics
of rapid urban growth in the Nanjing metropolitan region of China. Landscape Ecol.
22, 925–937.
Yang, Y., Wei, C., Xiao, L., Zhong, Z., Li, Q., Wang, H., Wang, W., 2022. Effects of
urbanization on woody plant phylogenetic diversity and its associations with
landscape features in the high latitude northern hemisphere region, Northeast China.
The Science of the total environment 838, 156192.
Yu, X.J., Ng, C.N., 2007. Spatial and temporal dynamics of urban sprawl along two
urban–rural transects: A case study of Guangzhou, China. Landscape Urban Plann.
79, 96–109.
Zhao, S., Zhou, D., Zhu, C., Qu, W., Zhao, J., Sun, Y., Huang, D., Wu, W., Liu, S., 2015.
Rates and patterns of urban expansion in China’s 32 major cities over the past three
decades. Landscape Ecol. 30, 1541–1559.
Zhao, J., Zhu, C., Zhao, S., 2014. Comparing the Spatiotemporal Dynamics of
Urbanization in Moderately Developed Chinese Cities over the Past Three Decades:
Case of Nanjing and Xi’an. J. Urban Plann. Dev. 141, 05014029.
Zhou, X., Chen, H., 2018. Impact of urbanization-related land use land cover changes and
urban morphology changes on the urban heat island phenomenon. Sci. Total
Environ. 635, 1467–1476.
Zou, L., Wang, J., Bai, M., 2022. Assessing spatial–temporal heterogeneity of China’s
landscape fragmentation in 1980–2020. Ecol. Ind. 136, 108654.
H. Cao et al.