Content uploaded by Chao Yang
Author content
All content in this area was uploaded by Chao Yang on Mar 05, 2021
Content may be subject to copyright.
Rapid Urbanization Induced Extensive Forest Loss to Urban Land in
the Guangdong-Hong Kong-Macao Greater Bay Area, China
YANG Chao1, 2, LIU Huizeng1, LI Qingquan1, CUI Aihong3, XIA Rongling4, SHI Tiezhu1, 5, ZHANG Jie1, GAO
Wenxiu5, ZHOU Xiang1, WU Guofeng1, 5
(1. MNR Key Laboratory for Geo-Environmental Monitoring of Great Bay Area & Guangdong Key Laboratory of Urban Informatics &
Shenzhen Key Laboratory of Spatial Smart Sensing and Services, Shenzhen University, Shenzhen 518060, China; 2. College of Informa-
tion Engineering, Shenzhen University, Shenzhen 518060, China; 3. Department of Geography, Hong Kong Baptist University, Hong
Kong 999077, China; 4. School of Land and Resources Engineering, Kunming University of Science and Technology, Kunming 650093,
China; 5. School of Architecture and Urban Planning, Shenzhen University, Shenzhen 518060, China)
Abstract: China has experienced rapid urbanizations with dramatic land cover changes since 1978. Forest loss is one of land cover
changes, and it induces various eco-environmental degradation issues. As one of China’s hotspot regions, the Guangdong-Hong Kong-
Macao Greater Bay Area (GBA) has undergone a dramatic urban expansion. To better understand forest dynamics and protect forest
ecosystem, revealing the processes, patterns and underlying drivers of forest loss is essential. This study focused on the spatiotemporal
evolution and potential driving factors of forest loss in the GBA at regional and city level. The Landsat time-series images from 1987 to
2017 were used to derive forest, and landscape metrics and geographic information system (GIS) were applied to implement further spa-
tial analysis. The results showed that: 1) 14.86% of the total urban growth area of the GBA was obtained from the forest loss in
1987–2017; meanwhile, the forest loss area of the GBA reached 4040.6 km2, of which 25.60% (1034.42 km2) was converted to urban
land; 2) the percentages of forest loss to urban land in Dongguan (19.14%), Guangzhou (18.35%) and Shenzhen (15.81%) were higher
than those in other cities; 3) the forest became increasingly fragmented from 1987–2007, and then the fragmentation decreased from
2007 to 2017); 4) the landscape responses to forest changes varied with the scale; and 5) some forest loss to urban regions moved from
low-elevation and gentle-slope terrains to higher-elevation and steep-slope terrains over time, especially in Shenzhen and Hong Kong.
Urbanization and industrialization greatly drove forest loss and fragmentation, and, notably, hillside urban land expansion may have
contributed to hillside forest loss. The findings will help policy makers in maintaining the stability of forest ecosystems, and provide
some new insights into forest management and conservation.
Keywords: forest loss to urban land; urbanization; spatiotemporal pattern; remote sensing; Guangdong-Hong Kong-Macao Greater Bay
Area (GBA)
Citation: YANG Chao, LIU Huizeng, LI Qingquan, CUI Aihong, XIA Rongling, SHI Tiezhu, ZHANG Jie, GAO Wenxiu, ZHOU Xi-
ang, WU Guofeng, 2021. Rapid Urbanization Induced Extensive Forest Loss to Urban Land in the Guangdong-Hong Kong-Macao
Greater Bay Area, China. Chinese Geographical Science, 31(1): 93−108. https://doi.org/10.1007/s11769-021-1177-9
Received date: 2020-05-21; accepted date: 2020-09-01
Foundation item: Under the auspices of National Natural Science Foundation of China (No. 41890854), Basic Research Program of
Shenzhen Science and Technology Innovation Committee (No. JCYJ20180507182022554), National Key R & D Program of China
(No. 2017YFC0506200), National Natural Science Foundation of China (No. 7181101150), National Natural Science Foundation of
China (No. 41901248), Shenzhen Future Industry Development Funding Program (No. 201507211219247860)
Corresponding author: LI Qingquan. E-mail: liqq@szu.edu.cn; WU Guofeng. E-mail: guofeng.wu@szu.edu.cn
© Science Press, Northeast Institute of Geography and Agroecology, CAS and Springer-Verlag GmbH Germany, part of Springer
Nature 2021
Chin. Geogra. Sci. 2021 Vol. 31 No. 1 pp. 93−108
https://doi.org/10.1007/s11769-021-1177-9
Springer Science Press
www.springerlink.com/content/1002-0063
1 Introduction
Urbanization has experienced a remarkable speed dur-
ing the past half century, the ratio of urban population to
global population increased from 30% in 1950 to 54%
in 2014 (World Bank, 2015), and it will reach 66% in
2050 (Madanian et al., 2018). The urbanization pro-
cessed in Africa and Asia is faster compared with those
in other continents, and it is predicted they will reach
56% and 64% by 2050, respectively (Madanian et al.,
2018). China started to implement its reform and open-
ing up polices in 1978, initiating a socioeconomic trans-
formation process. Over the past 40 years, urbanization
and industrialization have reached an unprecedented
rate. China gradually marched toward a mid- and post-
industrialization period, and the urbanization rate in-
creased by 26% from 1978 to 2009 (NBSC, 2010; Li-
ang et al., 2015). In addition, China experienced an ex-
plosive population growth during this period, the
urban population increased from 210 million in 1982 to
660 million in 2010, and the corresponding urbaniza-
tion rate increased from 20.91% to 49.68% (NBSC,
2013). Continual demands for urban housing, business
and construction land in rapid urbanization period resul-
ted in a dramatic conversion of land use/land cover
types, and the cropland converted into urban land has
become the main source of urban lands (Liu et al., 2010;
Liu et al., 2012; Liu et al., 2014).
Forests are critical natural resources for the survi-
val and development of humankind, and they play an ir-
replaceable role in fighting global climate change (De-
partment, 2010; FAO, 2012; Wu et al., 2019; Garcia et
al., 2020; Seymou, 2020). According to the United Na-
tions Food and Agriculture Organization (FAO), the
global forest area approaches 40 million km2 and ac-
counts for 31% of the Earth’s land surface. Forests
provide socioeconomic benefits and invisible environ-
mental health products for human beings, such as wood
production, oxygen and climate regulation (Gao and
Liu, 2011). However, forest ecosystems are often dis-
turbed by dramatic human activities, especially the urb-
anization and industrialization in recent decades (Lele et
al., 2008; Gao and Liu, 2011; FAO, 2012). The studies
from around the world revealed that the degradation or
even extinction of forest ecosystems could be largely at-
tributed to anthropologic disturbances (Lele et al., 2008,
Lambin and Meyfroidt, 2010; Liu et al., 2013; Liu et al.,
2016a). The deterioration of forest ecosystems is usu-
ally associated with forest loss and fragmentation (Laur-
ance et al., 2000; Miller, 2012), which also have in-
duced many negative eco-environmental consequences,
including species degradation or extinction, soil erosion
and sandy storm attacks (Reddy et al., 2013; Carranza et
al., 2015). Notably, rapid urbanization and industrializa-
tion are currently the largest factor for urban land ex-
pansion in developing and developed countries of the
world (Turner II et al., 2007; Liu et al., 2010b; Yang et
al., 2019a; Xu et al., 2020), affecting ecosystems in loc-
al and global scales (Shen et al., 2008; Yang et al.,
2017a; Girardet, 2020). The demands for commercial
and residential land development exacerbate forest loss
and fragmentation in rapid urbanization periods (Song et
al., 2014). Therefore, investigating forest loss and frag-
mentation under the background of rapid urbanization is
essential for forest ecosystem management and conser-
vation.
Since 1978, China has experienced a transformation
from a socialism planned economy to a market eco-
nomy. Large-scale deforestation gradually occurred in
different cities in China due to commercial timber mar-
ket opening (Liu et al., 2016a). Moreover, local govern-
ments were given the authority to regulate land use
types (e.g., agriculture, building) through land market
reform (Du et al., 2014). As a result, many cities
suffered from persistent forest net loss, because timbers
were harvested for urban construction (Li et al., 2010).
Be conscious of the seriousness of forest loss, the
Chinese government issued a series of policies for forest
recovery and conservation, including the ‘Returning
Farmland to Forest’ program (i.e., increasing forest cov-
ers and preventing soil erosion), and the ‘Grain for
Green’ policy in 1999 (i.e., conversion of farmland to
forest or grassland) (Cao et al., 2009; Deng et al., 2012;
2014; Van Den Hoek et al., 2014). However, the effects
of forest loss prevention programs vary with different
cities and regions, due to the interference of different
local land use planning policies (Mao et al., 2019; Trac
et al, 2013; van Den Hoek et al., 2014). Therefore, re-
vealing the processes and spatiotemporal patterns of
forest changes and forest loss in different cities and re-
gions will be helpful for understanding the dominant
driving forces of forest loss.
The combination of remote sensing images and geo-
graphic information system (GIS) has been widely ap-
94 Chinese Geographical Science 2021 Vol. 31 No. 1
plied in forest dynamic studies, because they can provide
timely and cost-effective information and analyze the
long-time processes and spatiotemporal patterns of
forest changes at multiple scales (Xie et al., 2012; Song
et al., 2014; Jia et al., 2015; Lindquist and D’Annunzio,
2016; Lechner et al., 2020). Landscape metrics provide
new insights in characterizing the detailed patch dynam-
ics of forest changes (Herold et al., 2002; Zengin et al.,
2018; Lv et al., 2019). Landscape indices are employed
to multi-scale or multi-temporal datasets to imply scale
effect and temporal variation. By combination of re-
mote sensing images, GIS and landscape approaches,
the forest loss in various cities and urban agglomera-
tions around the world have been quantified (Li et al.,
2012; Han et al., 2018). Considering various remote
sensing images, high-resolution satellite images show
limitations in geographic coverage and historical archive,
and low-resolution satellite data can not characterize the
detailed changes of forest loss (Setiawan et al., 2014).
Landsat TM (Thematic Mapper) and OLI (Operational
Land Imager) can provide images for over four decades
(Li et al., 2017; Yang et al., 2019a) with a middle spa-
tial resolution, and they have ability in mapping forest
areas at a moderate scale (Kline et al., 2009; Wahyudi et
al., 2018).
Over the past four decades, many researches emphas-
ized the spatiotemporal patterns of forest loss and frag-
mentation at a single scale (city or individual region)
(Song et al., 2014; Jia et al., 2015; Xie et al., 2017; Nav-
arro Cerrillo et al., 2019), and few studies were focused
on the systematic analysis at multiple levels or cross-
city comparisons with spatially consistent datasets. More-
over, the analysis of forest loss to urban land, which is a
general phenomenon in the rapid urbanization regions of
China, is scarce. The Guangdong-Hong Kong-Macao
Greater Bay Area (GBA) is one of the four bay areas in
the world, and it has experienced a rapid urban growth
since 1987 (Yang et al., 2019a). The population of the
GBA accounts for only 5% of China; however, it cre-
ated 11% of China’s total gross domestic product (GDP)
in 2017. To promote sustainable development, the
Chinese government issued the ‘Development Plan for
the GBA’ in 2019. According to this plan, the GBA
planed a world-class urban agglomeration and a high-
quality life circle. However, the urbanization process in
the GBA may greatly affect its forest ecosystem. The
phenomenon of occupying forests in the urbanization
process in the GBA has been exposed from high-resolu-
tion Google Earth images. Therefore, it is particularly
urgent and significant to investigate and understand the
processes of forest loss and fragmentation in the GBA at
different scale perspectives, especially for the forest loss
to urban land. This study focused on revealing the spati-
otemporal evolution and underlying forces of forest loss
in the GBA at the regional and city levels by employing
Landsat time-series images (1987–2017), landscape
metrics and GIS. It is hoped that this study can contrib-
ute to forest management and conservation.
2 Materials and Methods
2.1 Study area
The GBA is located in south China (21°32′N–24°26′N,
111°20′E–115°24′E), and it includes eleven cities: Fo-
shan, Huizhou, Shenzhen, Zhaoqing, Zhuhai, Hong
Kong, Zhongshan, Dongguan, Jiangmen, Guangzhou
and Macao (Fig. 1). The population of the GBA is appro-
ximately 70 million, and it has a total area of 56 000 km2.
The GBA belongs to typical humid subtropical climate
regions, with a large amount of precipitation in summer
(Yu et al., 2019). The urban land of the GBA has expan-
ded from 605.71 km2 in 1987 to 1996.27 km2 in 1997,
4481.96 km2 in 2007 and 7568.19 km2 in 2017 (Yang et
al., 2019a; Fig. 1c). The GBA contributed approxim-
ately 11% of the gross domestic product (GDP) of
China in 2017. The GBA is becoming a world-class bay
area and a well-known urban agglomeration.
2.2 Data source and pre-processing
The satellite data and products used in this study in-
cluded time-series Landsat images, digital elevation
model (DEM) and Google Earth high-resolution images.
Thirty-two cloudless or low-cloud Landsat TM and OLI
images covering the GBA around 1987, 1997, 2007 and
2017 were obtained from the United States Geological
Survey (USGS). The time interval of the satellite data
was 10 years. Most of the images used in this study
were captured in the dry season (October to March),
considering the minimal cloud and low vegetation vari-
ations in this season. Therefore, dry season satellite im-
ages had better capacity in studying land cover change
analysis (Hasan et al., 2019; Yang et al., 2020). Landsat
images have eight (for OLI) or six (for TM) bands at
visible to shortwave wavelengths, with spatial resolu-
YANG Chao et al. Rapid Urbanization Induced Extensive Forest Loss to Urban Land in the Guangdong-Hong ... 95
tion of 30 m. ASTER GDEM products covering the
GBA with a 30 m resolution were also collected from
USGS. The historical high-resolution images of forests
in the GBA were obtained from Google Earth Pro®.
The auxiliary data included urban land dataset, eleva-
tion, slope, socioeconomic data (i.e., GDP and popula-
tion data), and vector data of the GBA administrative di-
vision boundary. The spatial distribution of the urban
land from 1987 to 2017 was also extracted from the
aforementioned Landsat images, which was detailed in
Yang et al. (2019a). Urban land dataset with a spatial
resolution of 30 m provides reliable information on urb-
an expansion dynamics in the GBA (Fig. 1c). The eleva-
tion and slope data were derived from ASTER GDEM
using ArcGIS 10.2 (Figs. 1d and 2e). Socioeconomic
data were collected from Local Statistical Yearbooks in
1987–2017. The projection system used for the vector
and satellite data was WGS_84_UTM_49N. All Land-
sat images and DEM data were clipped with the GBA
boundary dataset. The software ENVI 5.3 was used to
process the Landsat images, including band combina-
tion, FLAASH atmospheric correction, image mosaic,
and image clipping.
2.3 Methods
This study analyzed the changing processes and pat-
terns of forests during a rapid urbanization period and
revealed the spatiotemporal characteristic and driving
forces of forest loss at two scales (the regional level and
city level) with remote sensing, landscape ecology and
spatial analysis method, including three parts: forest ex-
traction and accuracy assessment, forest landscape pat-
tern analysis, and forest loss to urban land dynamic de-
tection.
2.3.1 Forest extraction and accuracy assessment
The forest of the GBA was extracted with an object-ori-
ented support vector machine (O-SVM) in this study. O-
SVM method can combine the high efficiency of bi-
level scale-sets model (BSM) in processing large-scale
images and SVM’s high accuracy in applying small
training samples (Foody and Mathur, 2004; Li et al.,
2010; Hu et al., 2016). Therefore, O-SVM is more effi-
cient in forest extraction than other methods, and can
handle large-scale images while provide high accuracy
(Yu et al., 2017; Yang et al., 2019a). The O-SVM meth-
od was performed through integrating the Scale-Sets-
Image-Analysis-Toolkit (https://github.com/zwhoo/Sca-
Elevation / m
High: 1589
Slop / °
High: 78.98
Low: 0
0 80 160 km40
0 80 km40
0 80 160 km40
0 80 km40
0 1000 2000 km500
(a)
China
GS(2016)1593
N
(e)
(d)(c) SlopElevationUrban Land
Low: −138
Guangdong-Hong Kong-Macao Greater Bay Area
Zhaoqing
Guangzhou
Huizhou
Foshan Dongguan
Zhongshan Shenzhen
Jiangmen
Zhuhai
Macao
Hong Kong
112°E
24°N
23°N
22°N
24°N
23°N
22°N
113°E 114°E 115°E
112°E 113°E 114°E 115°E
1987
1988−1997
Urban land
1998−2007
2008−2017
(b)
Fig. 1 Typical regions of forest loss to urban land in the GBA (Guangdong-Hong Kong-Macao Greater Bay Area ): evidence from
high-resolution Google Earth images
96 Chinese Geographical Science 2021 Vol. 31 No. 1
le-Sets-Image-Analysis-Toolkit) and SVM algorithm in
this study. The parameters of object segmentation scales
in O-SVM were set to 30 (for TM) and 45 (for OLI),
and four kinds of characteristics for each object were
applied for forest information extraction, i.e., texture
(Texture-Variance, Texture-Entropy, Texture-Mean and
Texture-Range), spectral (Spectral-Max, Spectral-Min,
Spectral-Mean and Spectral-STD), spatial (Area, Length,
Compactness, Roundness and Elongation) and normal-
ized differential vegetation index (NDVI) (Yang et al,
2019a). A radial basis function (RBF) was used to con-
struct SVM classifier; the SVM’s Gamma coefficient
was set to 1/n (n refers to the band number for Landsat
images, i.e., n equals six for TM and eight for OLI, re-
spectively), and penalty cost was 100, suggesting high-
er extraction accuracy (Yu et al., 2017). Totally, 692
forest training datasets covering the GBA evenly from
1987–2017 were selected directly from TM/OLI seg-
mented images, 225 samples in 1987, 150 samples in
1997, 147 samples in 2007 and 170 samples in 2017;
while 226 validation samples of forest (81 samples in
1987, 47 samples in 1997, 42 samples in 2007 and 56
samples in 2017) from study area in 1987–2017 were
chosen randomly and evenly from Google Earth Pro®.
The forest extraction accuracy was assessed using four
basic metrics (i.e., Overall Accuracy, Kappa coefficient,
User Accuracyand Producer’s Accuracy) (Congalton,
1991; Yang et al., 2017b; 2019b).
2.3.2 Forest landscape pattern analysis
Four landscape metrics were used to evaluate the com-
plex degree, contiguous level, and fragmented degree of
forest cover (McGarigal, 2015). Landscape shape index
(LSI) was applied to analyze the landscape complexity,
and a larger LSI value indicates a greater complexity
and implies a stronger impact of human activities (Li-
ang et al., 2015). Patch cohesion index (Cohesion) was
applied to assess the contiguous level of forest at the
landscape level. Patch density (PD) and mean patch size
(MPS) were used to quantify forest fragmentation. A
larger PD and a smaller MPS indicate a higher frag-
mentation of forest landscape (Liang et al., 2015). The
software FRAGSTATS 4.2 was employed to calculate
these four metrics with the eight-neighborhood rule
(McGarigal and Marks, 1995). The spatial resolution of
forest raster data were set to 30 m in FRAGSTATS 4.2,
which is consistent with original data source. LSI, MPS
(ha), PD (number/100 ha) and Cohesion (%) were calcu-
lated according to Eqs. (1)–(4), respectively:
LS I =
0.25E
√A(1)
MPS =∑N
i=1ai
N(2)
PD =N
A×10 000 ×100 (3)
Cohension =
1−
n
∑
j=1
pi j
n
∑
j=1
pi j √ai j
/(1−
1
√Z)×100 (4)
where A is the total area of landscapes, E represents the
total length (m) of patch boundary, N is the number of
forest patches, ai is the area of i forest patch, pij means
the perimeter of patch ij in terms of the number of cell
surfaces, aij represents the area of patch ij in terms of
the number of cells, and Z is the total number of cells in
the landscape (McGarigal, 2002; Liang et al., 2015).
2.3.3 Forest loss to urban land dynamic detection
The original forest changed to urban land was defined
as ‘forest loss to urban land’ (Fig. 2). An equation
(Equation (5)) was proposed to quantify and detect the
spatial distribution of forest loss to urban land areas at
the regional and city level from 1987 to 2017 in this
study. In addition, the elevation and slope were em-
ployed to identify the spatiotemporal evolution of forest
loss to urban land on different terrain conditions, be-
cause the suburbs of the GBA are hills and mountains.
F=ft∩ut+1(5)
∩
where F represents the area of forest loss to urban land,
ft is the forest area in t period, ut+1 is the urban land in
t +1 period, and represents intersection operation.
3 Results
3.1 Forest extraction and accuracy assessment
The results of accuracy assessment, quantitative statist-
ics and spatiotemporal evolution of forest in the GBA
are shown in Fig. 3, Table 1, and Figs. 4 and 5, respect-
ively. The accuracy metrics were more than 84%, with
the all accuracy’s average value of each period ap-
proaching 90% in this study (Fig. 3). Forests were ex-
tracted effectively using the O-SVM method (accuracy’s
YANG Chao et al. Rapid Urbanization Induced Extensive Forest Loss to Urban Land in the Guangdong-Hong ... 97
average value = 88% in 1987, 90% in 1997, 89% in
2007 and 90% in 2017), and the accuracy of all years
satisfies the requirement of land cover change analysis
(Foody, 2002; Zhang et al., 2010). Generally, the forest
experienced a dynamic change at the regional and city
level from 1987 to 2017, and the spatial distribution of
forest was concentrated in urban suburbs, hills and
mountains (Fig. 4 and Fig. 5). At the regional level, the
forest of the GBA increased first then decreased
(Fig. 4a), the area decreased from 30 230.26 km2 in
1987 to 26 189.66 km2 in 2017, and the total loss area
was 4040.6 km2 (Table 1 and Fig. 4b). At the city level,
the trends of forest changes of all cities were consistent
with the GBA, except for Zhaoqing, Huizhou and Ma-
cao (Fig. 4a). Among the eleven cities, Zhaoqing had
the largest forest area, and the average area exceeded
10 000 km2 in 1987–2017 (Fig. 4a). In contrast, Macao
had the smallest forest area, with an average forest area
of only 3.52 km2 (Fig. 4a). Notably, the total change of
forest areas in Huizhou, Dongguan and Guangzhou
showed gain, while the forest areas of other cities
showed loss during 1987–2017 (Fig. 4b and Table 1).
3.2 Forest landscape pattern analysis
Fig. 6 shows the features and trends of four landscape
metrics for the forest area changes of the GBA and elev-
en cities in 1987–2017. Generally, a disparity of land-
scape responses to forest area changes was observed
during the study period. For the GBA level, the frag-
mentation degree of forest cover showed a trend
of first increasing (1987–2007) and then decreasing
(2007–2017); the increased PD and decreased MPS of
forest cover indicated that the forests were becoming
scattered patches in 1987–2007 (Figs. 6a and 6b).
Moreover, the forest cover of the GBA had the largest
fragmentation degree in 2007, which can be inferred
from the maximum PD value (0.067/100 ha) and minim-
um MPS value (724.57 ha) (Figs. 6a and 6b). Notably,
the first decreasing trend (1987–2007) and then increas-
ing trend (2007–2017) of LSI in the GBA showed that
(a)
(b)
Forest loss to urban
land
Forest loss to urban land
Urban Land
Urban Land
2018
2016
22°27'14.63″ N, 113°18'56.76″ E
22°27'28.21″ N, 113°24'05.90″ E
2006
2007
Forest
Forest
Fig. 2 Typical forest loss to urban area : Evidence from high-resolution Google Earth images
Accuracy value / %
90
85
80
1987 1997
Year
20172007
75
Overall accuracy Kappa coefficient
Producer’s accuracy User accuracy
Fig. 3 Accuracy metrics of forest extraction from 1987 to 2017
in the GBA (Guangdong-Hong Kong-Macao Greater Bay Area)
of China
98 Chinese Geographical Science 2021 Vol. 31 No. 1
the human disturbance to forests increased first and then
decreased (Fig. 6c). The decreasing Cohesion of the
GBA also indicated that the connections among forest
patches became increasingly weaker in 1987–2007
(Fig. 6d). On the contrary, the increasing Cohesion of
the GBA indicated the strong connections among forest
patches in 2007–2017 (Fig. 6d). The Cohesion changes
also demonstrated the fragmentation variations of forest
covers. The Cohesion decreased by 0.023% from 1987
to 2007 in the GBA (Fig. 6d), suggesting that forest
patches’ spatial distribution tended to be scattered and
decentralized. The exploitation of construction environ-
ment during 1987–2007 likely contributed to this trans-
formation to a certain extent.
From the perspective of city level, the PD of forests
in different cities presented first increasing and then de-
creasing trend during 1987–2017, except for Zhuhai,
Huizhou and Jiangmen (Fig. 6a). The trends of MPS
were contrary to those of PD in all cities (Fig. 6b). The
trends of PD and MPS in Zhaoqing, Zhongshan, Guang-
1
10
100
1000
10000
100000
1987 1997 2007 2017
−4500
−3500
−2500
−1500
−217.11
Foshan
−2507.49
Jiangmen
−1629.7
Zhaoqing
−213.52
Zhuhai
−158.77
HongKong
Macao
−2.55
−500
500
1500
LgArea / km2
GBA
Foshan
Huizhou
Jiangmen
Zhaoqing
Zhuhai
Zhongshan
HongKong
Macao
Dongguan
Shenzhen
Guangzhou
Area / km2
Zhaoqing
Huizhou
Macao
Huizhou
+665.53
+5.88
Dongguan
Guangzhou
+300.01
Zhongshan
−99.59
+Increase
−
Loss
−4040.6
GBA
(b)
Year
(a)
Shenzhen
−181.53
Fig. 4 Forest dynamics in the GBA (Guangdong-Hong Kong-Macao Greater Bay Area) from 1987 to 2017: (a) change trend of forest
area; (b) total change of forest area
Table 1 Forest area and forest area changes in the GBA (Guangdong-Hong Kong-Macao Greater Bay Area) and eleven cities from
1987 to 2017 (km2)
Study areas
Forest area Forest area changes
1987 1997 2007 2017 1987–1997 1997–2007 2007–2017 1987–2017
GBA 30230.26 35176.35 26975.70 26189.66 4946.09 –8200.65 –786.04 –4040.60
Foshan 752.17 1151.59 787.62 535.05 399.42 –363.97 –252.56 –217.11
Huizhou 6594.42 9360.21 6376.80 7259.95 2765.79 –2983.41 883.16 665.53
Jiangmen 4996.75 6832.38 4395.81 2489.26 1835.63 –2436.57 –1906.54 –2507.49
Zhaoqing 12406.50 9720.70 9056.87 10776.80 –2685.80 –663.83 1719.93 –1629.70
Zhuhai 479.59 635.02 368.60 266.06 155.43 –266.41 –102.54 –213.52
Zhongshan 344.26 373.70 352.31 244.67 29.44 –21.38 –107.64 –99.59
Hong Kong 629.09 657.67 676.29 470.32 28.57 18.62 –205.97 –158.77
Macao 5.33 4.78 1.19 2.78 –0.55 –3.59 1.59 –2.55
Dongguan 445.82 990.46 595.32 451.69 544.64 –395.14 –143.63 5.88
Shenzhen 908.83 921.11 728.55 727.30 12.28 –192.56 –1.25 –181.53
Guangzhou 2656.73 4517.96 3627.60 2956.75 1861.23 –890.36 –670.85 300.01
YANG Chao et al. Rapid Urbanization Induced Extensive Forest Loss to Urban Land in the Guangdong-Hong ... 99
zhou and Shenzhen were consistent with those of the
GBA during 1987–2017, but the values of PD and MPS
were different (Figs. 6a and 6b). The PD values of
Zhaoqing were lower than those of the GBA, and the
MPS values of Huizhou were higher than those of the
GBA (Figs. 6a and 6b). It is noteworthy that Macao had
the lowest MPS, LSI and Cohesion values in all cities in
1987–2017 (Figs. 6b, 6c and 6d), suggesting that the
patch areas, patch numbers and complexity of forests in
Macao were smaller than those in other cities.
3.3 Forest loss to urban land dynamics from 1987
to 2017
Fig. 7 and Table 2 show the spatiotemporal characterist-
ics of forest loss to urban land at the regional and city
level. Generally, the forest area loss to urban land was a
dynamic process during study period, and the spatial
feature of forest loss to urban land was mainly concen-
trated in the regions with low elevations (< 80 m) and
gentle slopes (< 5°) (Fig. 8, Tables 3 and 4). However,
some patches of forest loss to urban land were tran-
sitioned from lower-elevation and gentle-slope terrains
to higher-elevation (80–400 m) and steep-slope (5°–30°)
terrains over time, especially for Jiangmen, Zhaoqing,
Shenzhen and Hong Kong (Fig. 8, Tables 3 and 4). The
changes of forest loss to urban land at different ele-
vations and slopes may result from the urbanization and
industrialization in suburbs. At the regional level, the
forest loss to urban land experienced a first increase
then declining trend during 1987–2017 (Fig. 9a), and
reached the maximum value (585.27 km2) in 1997–2007.
In addition, the total area of forest loss to urban land in
the GBA was 1034.42 km2 during 1987–2017 (Table 2).
At the city level, the trends of forest loss to urban
land were also first increasing then decreasing during
1987–2017, except for Guangzhou, Foshan and Zhong-
shan (Fig. 9b). Shenzhen had the largest transformation
area of forest to urban land (reaching 47.05 km2, almost
50% of the GBA during 1987–1997 (Table 2), resulting
from rapid urbanization. During 1997–2017, the maxim-
um transformation areas of forest to urban land were ob-
served in Dongguan (1997–2007) and Guangzhou
(2007–2017), and reached 145.41 km2 and 92.56 km2,
respectively. Moreover, the total areas of forest loss to
(a) 1987
111°25′E
23°45′N
22°40′N
21°35′N
23°45′N
22°40′N
21°35′N
Legend
Urban
Forest
112°30′E 113°35′E 114°40′E
111°25′E 112°30′E 113°35′E 114°40′E
0 40 80 160 km
(b) 1997
111°25′E
23°45′N
22°40′N
21°35′N
23°45′N
22°40′N
21°35′N
Legend
Urban
Forest
112°30′E 113°35′E 114°40′E
111°25′E 112°30′E 113°35′E 114°40′E
Other land
Other land 0 40 80 160 km
(c) 2007
111°25′E
23°45′N
22°40′N
21°35′N
23°45′N
22°40′N
21°35′N
Legend
Urban
Forest
112°30′E 113°35′E 114°40′E
111°25′E 112°30′E 113°35′E 114°40′E
Other land
0 40 80 160 km
(d) 2017
111°25′E
23°45′N
22°40′N
21°35′N
23°45′N
22°40′N
21°35′N
Legend
Urban
Forest
112°30′E 113°35′E 114°40′E
111°25′E 112°30′E 113°35′E 114°40′E
Other land
0 40 80 160 km
Fig. 5 Spatiotemporal characteristics of forest during 1987 and 2017 in the GBA (Guangdong-Hong Kong-Macao Greater Bay Area)
100 Chinese Geographical Science 2021 Vol. 31 No. 1
urban land in Dongguan, Guangzhou and Shenzhen
ranked as the top three of the GBA in 1987–2017, and
their percentages accounted for 19.14%, 18.35%, and
15.81%, respectively (Table 2). In contrast, this percent-
GBA
Foshan
Huizhou
Jiangmen
Zhaoqing
Zhuhai
Zhongshan
Hong Kong
Macao
Dongguan
Shenzhen
Guangzhou
0
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18
0.20
1
10
100
1000
10000
0
20
40
60
80
100
120
140
160
95.5
96.0
96.5
97.0
97.5
98.0
98.5
99.0
99.5
100.0
1987 1997 2007 2017
Cohesion / %
PD / (Number/100 ha)
2017200719971987 2017200719971987
2017
20071997
1987
LgMPS / ha
(b)(a)
(c)
LSI
(d)
Year
Year
Year
Year
Fig. 6 The landscape patterns of forest cover changes from 1987 to 2017: (a) Patch density (PD), (b) mean patch size (MPS), (c) land-
scape shape index (LSI) and (d) patch cohesion (Cohesion)
111°25′E
23°45′N
22°40′N
21°35′N
23°45′N
22°40′N
21°35′N
112°30′E 113°35′E 114°40′E
111°25′E 112°30′E 113°35′E 114°40′E
0 40 80 160 km
Forest loss
Dongguan
Shenzhen 0 20 km
0 20 km
1987−1997
1997−2007
2007−2017
Background
Fig. 7 The spatiotemporal characteristics of forest loss to urban land in the GBA from 1987 to 2017
YANG Chao et al. Rapid Urbanization Induced Extensive Forest Loss to Urban Land in the Guangdong-Hong ... 101
age was less than 0.1% in Macao, which was the lowest
among all eleven cities (Table 2). It is noteworthy that
the spatial distribution of patches of forest loss to urban
land in Shenzhen, Dongguan, Zhongshan and Foshan
were distributed evenly over time, and the process of
urban growth occupied many suburban forest (Fig. 7),
suggesting that the urbanization and industrialization of
suburbs were extensive and decentralized.
Forest loss to urban in 1987−1997 Forest loss to urban in 1997−2007 Forest loss to urban in 2007−2017
<80
80
−200
200−400
400−600
>600
Elevation / m
<5
5
−10
10−400
400−600
>600
Slope / °
(a) (b)
0
0 20 km
40
0 20 km
80 160 km 0 40 80 160 km
Fig. 8 The spatiotemporal characteristics of forest area loss to urban land at different elevations and slopes from 1987 to 2017 in GBA
Table 2 Forest loss to urban land at regional and city level in GBA from 1987 to 2017 / km2
Region 1987–1997 1997–2007 2007–2017 Total loss
GBA 92.97 585.27 356.18 1034.42 (100.00)
Foshan 2.34 32.40 58.77 93.51 (9.04)
Huizhou 2.87 90.33 20.15 113.35 (10.96)
Jiangmen 2.67 64.43 43.04 110.14 (10.65)
Zhaoqing 11.29 17.80 17.04 46.13 (4.46)
Zhuhai 4.80 18.85 6.09 29.74 (2.88)
Zhongshan 3.42 18.67 25.71 47.8 (4.62)
Hong Kong 3.31 25.28 12.66 41.25(3.99)
Macao 0.09 0.76 0 0.85 (0.08)
Dongguan 9.04 145.41 43.57 198.02 (19.14)
Shenzhen 47.05 79.93 36.56 163.54 (15.81)
Guangzhou 6.10 91.19 92.56 189.85 (18.35)
Note: The figures in brackets indicate the proportion of the total loss of each city to the total loss of GBA / %
102 Chinese Geographical Science 2021 Vol. 31 No. 1
0
20
40
60
80
100
120
140
160
0
100
200
300
400
500
600
700
Forest loss to urban area unit / km
2
Forest loss to urban area unit / km
2
1987−1997 1997−2007 2007−2017 1987−1997 1997−2007 2007−2017
Guangzhou
Foshan
Zhongshan
(a) (b)
GBA Foshan
Huizhou
Jiangmen
Zhaoqing
Zhuhai
Zhongshan
Hong Kong
Macao
Dongguan
Shenzhen
Guangzhou
Year Year
Fig. 9 The trends of forest loss to urban land at regional and city level from 1987 to 2017: (a) GBA and (b) 11 cities within GBA
Table 3 Forest loss to urban land at different elevations from 1987 to 2017 in the GBA / km2
Region
Forest loss to urban land during 1987–1997 Forest loss to urban land during 1997–2007 Forest loss to urban land during 2007–2017
< 80 80–200 200–400 400–600 > 600 < 80 80–200 200–400 400–600 > 600 < 80 80–200 200–400 400–600 > 600
GBA 78.10 12.64 1.35 0.68 0 550.66 23.72 9.69 1.51 0 332.30 19.47 2.53 1.40 0.43
Foshan 2.25 0.03 0 0 0 32.17 0 0 0 0 58.22 0.52 0.03 0 0
Huizhou 2.46 0.38 0.03 0 0 81.91 7.27 1.05 0 0 16.74 2.75 0.64 0 0
Jiangmen 2.44 0 0.22 0 0 62.49 1.53 0.06 0 0 39.58 1.98 0.38 0.68 0.43
Zhaoqing 9.49 0.15 0.93 0.68 0 16.38 0.97 0.31 0.01 0 11.40 4.40 0.61 0.63 0
Zhuhai 4.78 0 0 0 0 18.71 0.16 0 0 0 5.38 0.71 0.01 0 0
Zhongshan 3.46 0 0 0 0 18.72 0.06 0 0 0 25.50 0.20 0 0 0
Hong Kong 2.90 0.38 0 0 0 9.95 7.06 7.19 1.35 0 11.39 0.66 0.50 0.10 0
Macao 0.09 0 0 0 0 0.52 0.23 0 0 0 0 0 0 0 0
Dongguan 8.36 0.64 0 0 0 145.34 0.40 0 0 0 43.33 0.23 0 0 0
Shenzhen 36.50 10.38 0.14 0 0 74.39 5.33 0.25 0 0 30.63 5.66 0.28 0 0
Guangzhou 5.36 0.68 0.04 0 0 90.07 0.71 0.82 0.15 0 90.12 2.34 0.10 0 0
Table 4 Forest loss to urban land at different slopes from 1987 to 2017 in the GBA / km2
Region
Forest loss to urban land in 1987–1997 Forest loss to urban land in 1997–2007 Forest loss to urban land in 2007–2017
< 5 5–10 10–20 20–30 > 30 < 5 5–10 10–20 20–30 > 30 < 5 5–10 10–20 20–30 > 30
GBA 63.04 18.40 9.07 1.90 0.36 468.55 69.03 36.99 10.09 0.92 255.33 65.17 29.62 5.31 0.69
Foshan 1.42 0.49 0.29 0.06 0.01 26.14 5.02 0.90 0.09 0.01 48.31 7.59 2.55 0.27 0.05
Huizhou 1.58 0.64 0.56 0.09 0 67.65 10.98 9.84 1.69 0.07 13.74 3.09 2.65 0.63 0.02
Jiangmen 1.98 0.40 0.23 0.05 0 51.90 7.95 3.29 0.83 0.12 27.06 9.67 5.04 0.10 0.28
Zhaoqing 7.88 1.66 1.26 0.31 0.14 11.01 3.60 2.19 0.67 0.20 6.11 5.14 4.52 1.18 0.10
Zhuhai 3.93 0.60 0.22 0.03 0 15.80 1.67 1.12 0.27 0.01 3.18 1.13 1.29 0.45 0.03
Zhongshan 2.30 0.89 0.26 0.01 0 16.37 1.84 0.53 0.04 0 20.95 3.36 1.22 0.15 0.03
Hong Kong 1.36 0.87 0.88 0.14 0.01 5.35 4.62 9.95 5.25 0.38 9.44 1.69 1.01 0.40 0.11
Macao 0.05 0.03 0.01 0 0 0.17 0.16 0.30 0.14 0 0 0 0 0 0
Dongguan 6.08 1.74 1.00 0.16 0.03 135.39 8.47 1.64 0.19 0.06 37.14 5.09 1.20 0.14 0
Shenzhen 33.58 9.53 3.20 0.63 0.09 64.55 11.44 3.54 0.39 0.05 22.86 9.14 4.05 0.49 0.02
Guangzhou 2.87 1.56 1.17 0.41 0.08 74.23 13.26 3.70 0.54 0.03 66.55 19.26 6.08 0.62 0.05
YANG Chao et al. Rapid Urbanization Induced Extensive Forest Loss to Urban Land in the Guangdong-Hong ... 103
4 Discussion
4.1 Driving forces of forest loss and fragmentation
China has experienced remarkable urbanizations since
1978, especially in the GBA (Zhang and Weng, 2016;
Zhang et al., 2016; Yang et al., 2019a). The urban land
in the GBA expanded from 605.71 km2 to 7568.19 km2
(a total expansion of 6962.48 km2) in 1987–2017, and
the GBA is experiencing a rapid transition period from
urbanization to suburbanization (Yang et al., 2019a).
The forest loss to urban land in the GBA was 1034.42 km2
during 1987–2017, and 14.86% of total urban growth
area in the GBA was obtained from forest loss. During
this period, urbanization and rural industrialization have
induced large-scale transformation of land use types
(Jordan et al., 2007; Liang et al., 2015), one of which
was the transformation of low-altitude forests to built-up
lands (Li et al., 2010). In the early stage of the reform
and opening-up, the eco-environmental problems caused
by forest loss were not taken into account in economic
development. Low-altitude flat forests also became the
primary choice for urban construction, due to their ideal
terrain. Extensive infrastructure constructions occupied
a large area of forests in Shenzhen, which led to the
maximum transformation area of forest to urban land in
1987–1997. It is worth noting that local governments
developed many satellite towns and industrial parks to
control urban sprawl and evacuate overcrowded popula-
tion and industries in the original urban core (NDRC,
2014; Zhang et al., 2016; Yang and Li et al., 2019), and
thus industrial parks and satellite towns arose in the sub-
urbs of Guangzhou, Shenzhen, Foshan, Zhongshan,
Dongguan and Zhuhai in 1997–2017. However, satel-
lite towns and industrial parks inevitably occupied some
forests, resulting in a large forest loss to urban land in
the suburbs in the GBA during 1997–2017 (585.27 km2
for 1997–2007 and 356.18 km2 for 2007–2017). Some
studies prove that the growing urban transportation net-
works could divide the landscape formed by land cover
into countless small patches, resulting in landscape frag-
mentation (Gobattoni et al., 2011; Liang et al., 2015).
Therefore, forest fragmentation in the GBA can be at-
tributed to rapid urban sprawl and industrialization to a
certain extent.
Urban expansion always prefers to choose farmland
for development because generally the terrain of farm-
land is plains and the development costs are low. China
has converted extensive farmlands into urban lands
since reform and opening-up (Liu et al., 2014; Liu et al.,
2015; Liu et al., 2016b; Hu et al., 2018). Being aware of
the seriousness of farmland loss, the Chinese govern-
ment issued a series of policies, such as China’s Nation-
al General Land Use Plan (1997–2010 and 2006–2020),
to prevent the loss of farmland (Zhong et al., 2014; Xu
et al., 2015). The high-quality cultivated land (Class I
and II types) in China is distributed in flat regions with
slopes of 0°–6°, and most of them belong to the basic
farmland, which can not be developed. The mandatory
measures and requirements of strict farmland conserva-
tion policies push local governments to develop urban
lands on hillsides or mountains with low elevations and
gentle slopes. A large amount of high quality farmlands
in the plain areas were converted into urban lands in the
early stage of the reform and opening up (1987–1997).
Therefore, it is not surprising that the areas of forest loss
to urban land gradually moved towards hillsides in the
GBA during 1997–2017, especially for hilly cities with
less farmland, such as Shenzhen and Hong Kong.
It is worth noting that urban development and GDP
growth are closely associated with political achieve-
ments in China (Liu et al., 2014c). The local govern-
ments within the GBA tended to develop industrial
clusters in pursuit of high GDP growth to reach politic-
al achievements; however, the basic farmland protec-
tion regions were forbidden for development, which res-
ulted in a large number of industrial parks in suburbs
and hillsides being developed during 1997–2017, espe-
cially for Dongguan, Foshan, Zhongshan and Shenzhen.
The development of industrial clusters usually occupies
some forests, resulting in eco-environment degradation.
These results suggest that more attentions to forest con-
servation are needed.
4.2 Landscape responses to forest dynamics
Our results revealed that the effects of forest changes on
the landscape varied at regional and city level. For the
regional level, we found that forest fragmentation and
complexity increased in the early stage (1987–2007),
which confirmed the general observation that urbaniza-
tion leads to increasing landscape fragmentation and
complexity (Collinge, 1996; Chen et al., 2007). However,
a decreasing trend of forest landscape fragmentation
was observed in 2007–2017, suggesting that a reduction
of human disturbance and a growth of environmental
104 Chinese Geographical Science 2021 Vol. 31 No. 1
protection awareness can adjust forest distributions to
avoid the acceleration of landscape fragmentation and
complexity. In addition, the landscape responses to
forest changes are not always monotonic, but vary with
spatial and temporal scale. For the city level, forest dis-
tributions in most cities presented a decreasing suburb-
to-central urban areas gradient (i.e., the farther away
from the city centers, the more forests were distributed),
which is similar to the general observation in the GBA.
Moreover, we found that a higher fragmentation and
lower contiguous degree of forests appeared in the
farther mountainous regions, confirming that the dis-
turbance of anthropologic activities in forests is mainly
concentrated in highly urbanized areas. The fragmenta-
tion trends of Zhaoqing, Zhongshan, Guangzhou and
Shenzhen were consistent with those of the GBA dur-
ing 1987–2017, while the fragmentation trends of
Zhuhai, Huizhou and Jiangmen were different from
those of the GBA, moreover their fragmentation de-
grees were different. These results suggested that the
urbanization levels and volumes of forest resources for
these cities were different in 1987–2017. Therefore, the
landscape responses to forest changes at the city level
were also not monotonic, but varied according to space
and time scale.
4.3 Limitations and future works
There are some limitations in our study, which need to
be further explored. The GBA has a wide geographical
coverage and heavy rainfall in summer, which makes it
difficult to obtain enough cloudless Landsat images in a
same season. The resolution of Landsat images is
30 m, which makes it difficult to obtain more precise
forest boundaries; thus, high-resolution images may
hold potential in improving forest loss to urban land
studies. This study was focused on the forest loss and
forest loss to urban land of the GBA in 1987–2017, and
the forest recovery was not analysed in depth. Forest
change is a dynamic process, and the forest area in the
GBA had a recovery period in 1987–1997, with a restor-
ation area of 4946.09 km2. Among the eleven cities,
Huizhou, Jiangmen and Guangzhou had more restora-
tion areas than other cities, which could be attributed to
the reforestation project ‘Greening Guangdong in 10
Years’ initiated in 1985 and the mountainous terrain of
these three cities (Trac et al., 2013; Hasan et al., 2019).
Therefore, the forest recovery in the GBA needs to be
discussed in depth in the future. In order to balance the
contradiction of urban growth and eco-environment pro-
tection, studying the ecological effects of forest loss and
the correlation between urbanization and forest land-
scape changes is urgently required by local governments.
5 Conclusions
This study first extracted forest boundaries of GBA
from 1987 to 2017 using Landsat time-series images by
object-oriented support vector machine method, and
then revealed the spatiotemporal features of forest loss
to urban land, landscape patterns of forest dynamics at
regional and city level by combining landscape metrics,
and GIS techniques. The main conclusions were as fol-
lows: 1) The spatial distribution of forest was concen-
trated in urban suburbs, hills and mountains of the GBA.
Forest of the GBA increased first then decreased, and
the trends of forest changes of all cities were consistent
with the GBA, except for Zhaoqing, Huizhou and Ma-
cao. 2) Landscape responses to forest change varied
with spatial and temporal scale. Forests became increas-
ingly fragmented in 1987–2007, and then fragmentation
decreased in 2007–2017 at the regional level, which was
consistent with that in Zhaoqing, Zhongshan, Guang-
zhou and Shenzhen. 3) The total urban growth area in
the GBA was 6962.48 km2, of which 14.86% was ob-
tained from forest loss. The total area of forest loss in
the GBA reached 4040.6 km2 in 1987–2017, of which
25.60% was converted to urban lands. The percentages
of forest loss to urban land in Dongguan (19.14%),
Guangzhou (18.35%) and Shenzhen (15.81%) were
higher than those in other cities. 4) Urbanization and in-
dustrialization drove forest loss to urban land. Hillside
urban land expansion contributed to mountain or hill-
side forest loss in the GBA. These findings will be help-
ful to policy makers for maintaining the stability of
forest ecosystem, and provide some new insights into
forest management and conservation. Our results also
suggest that urban lands on hillsides are at risk, because
the changes of regional topographic features at higher-
elevation and steeper-slope terrains may cause surface
subsidence and deterioration of ecological quality.
References
Cao S X, Chen L, Liu Z D, 2009. An investigation of Chinese at-
YANG Chao et al. Rapid Urbanization Induced Extensive Forest Loss to Urban Land in the Guangdong-Hong ... 105
titudes toward the environment: case study using the grain for
green project. AMBIO: A Journal of the Human Environment,
38(1): 55–64. doi: 10.1579/0044-7447-38.1.55
Carranza M L, Hoyos L, Frate L et al., 2015. Measuring forest
fragmentation using multitemporal forest cover maps: forest
loss and spatial pattern analysis in the Gran Chaco, central Ar-
gentina. Landscape and Urban Planning, 143: 238–247. doi:
10.1016/j.landurbplan.2015.08.006
Cerrillo R M N, Vieira D J E, Ochoa-Gaona S et al., 2019. Land
cover changes and fragmentation in mountain neotropical eco-
systems of Oaxaca, Mexico under community forest manage-
ment. Journal of Forestry Research, 30: 143–155. doi:
10.1007/s11676-017-0568-3
Chen M W, Zong Y G, Ma Q et al., 2007. Quantitative study on
spatio-temporal change of urban landscape pattern based on
RS/GIS: a case of Xi ’an metropolitan area in China. In: Pro-
ceedings of SPIE 6752, Geoinformatics 2007: Remotely Sensed
Data and Infor-mation. Nanjing: SPIE. doi: 10.1117/12.
760778
Collinge S K, 1996. Ecological consequences of habitat frag-
mentation: implications for landscape architecture and plan-
ning. Landscape and Urban Planning, 36(1): 59–77. doi:
10.1016/S0169-2046(96)00341-6
Congalton R G, 1991. A review of assessing the accuracy of clas-
sifications of remotely sensed data. Remote Sensing of Envir-
onment, 37(1): 35–46. doi: 10.1016/0034-4257(91)90048-B
Deng L, Shangguan Z P, Li R, 2012. Effects of the grain-for-
green program on soil erosion in China. International Journal
of Sediment Research, 27(1): 120–127. doi: 10.1016/S1001-
6279(12)60021-3
Deng L, Liu G B, Shangguan Z P, 2014. Land-use conversion and
changing soil carbon stocks in China’s ‘Grain-for-Green’ pro-
gram: a synthesis. Global Change Biology, 20(11): 3544–3556.
doi: 10.1111/gcb.12508
Department of Forest Resources Management (DFRM), SFA,
2010. The 7th national forest inventory and status of forest re-
sources. Forest Resources Management, (1): 1–8. (in Chinese)
Du J F, Thill J C, Peiser R B et al., 2014. Urban land market and
land-use changes in post-reform China: a case study of Beijing.
Land-scape and Urban Planning, 124: 118–128. doi:
10.1016/j.landurbplan.2014.01.012
FAO-Food and Agricultural Organization of the United Nations,
2012. State of the World’S Forests, 2012. Rome, Italy: FAO.
http://www.fao.org/docrep/016/i3010e/i3010e00.htm.
Foody G M, 2002. Status of land cover classification accuracy as-
sessment. Remote Sensing of Environment, 80(1): 185–201.
doi: 10.1016/S0034-4257(01)00295-4
Foody G M, Mathur A, 2004. A relative evaluation of multiclass
image classification by support vector machines. IEEE Trans-
actions on Geoscience and Remote Sensing, 42(6): 1335–1343.
doi: 10.1109/TGRS.2004.827257
Gao J, Liu Y S, 2011. Deforestation in Heilongjiang Province of
China, 1896-2000: severity, spatiotemporal patterns and
causes. Applied Geography, 35(1-2): 345–352. doi: 10.1016/
j.apgeog.2012.08.001
Garcia C A, Savilaakso S, Verburg R W et al., 2020. The global
forest transition as a human affair. One Earth, 2(5): 417–428.
doi: 10.1016/j.oneear.2020.05.002
Girardet H, 2020. People and nature in an urban world. One
Earth, 2(2): 135–137. doi: 10.1016/j.oneear.2020.02.005
Gobattoni F, Pelorosso R, Lauro G et al., 2011. A procedure for
mathematical analysis of landscape evolution and equilibrium
scenarios assessment. Landscape and Urban Planning, 103(3-
4): 289–302. doi: 10.1016/j.landurbplan.2011.08.011
Han R, Cui N X, Guo L, 2018. Monitoring the effects of the grain
to green projects using landscape metrics: a case study in east-
ern Si-chuan, China. IOP Conference Series: Earth and Envir-
onmental Science, 189: 052009. doi: 10.1088/1755-1315/
189/5/052009
Hasan S, Shi W Z, Zhu X L et al., 2019. Monitoring of land
use/land cover and socioeconomic changes in south China over
the last three decades using Landsat and Nighttime light data.
Remote Sensing, 11(4): 1658. doi: 10.3390/rs11141658
Herold M, Scepan J, Clarke K C, 2002. The use of remote sens-
ing and landscape metrics to describe structures and changes in
urban land uses. Environment and Planning A: Economy and
Space, 34(8): 1443–1458. doi: 10.1068/a3496
Hu Y J, Kong X B, Zheng J et al., 2018. Urban expansion and
farmland loss in Beijing during 1980-2015. Sustainability,
10(11): 3927. doi: 10.3390/su10113927
Hu Z W, Li Q G, Zou Q et al., 2016. A bilevel scale-sets model
for hierarchical representation of large remote sensing images.
IEEE Transactions on Geoscience and Remote Sensing,
54(12): 7366–7377. doi: 10.1109/TGRS.2016.2600636
Jia M M, Wang Z M, Zhang Y Z et al., 2015. Landsat-based es-
timation of mangrove forest loss and restoration in Guangxi
Province, China, influenced by human and natural factors.
IEEE Journal of Selected Topics in Applied Earth Observa-
tions and Remote Sensing, 8(1): 311–323. doi: 10.1109/
JSTARS.2014.2333527
Jordan N, Boody G, Broussard W et al., 2007. Environment: sus-
tainable development of the agricultural bio-economy. Science,
316(5831): 1570–1571. doi: 10.1126/science.1141700
Kline J D, Moses A, Azuma D et al., 2009. Evaluating satellite
imagery-based land use data for describing forestland develop-
ment in western Washington. Western Journal of Applied
Forestry, 24(4): 214–222. doi: 10.1093/wjaf/24.4.214
Lambin E F, Meyfroidt P, 2010. Land use transitions: socio-eco-
logical feedback versus socio-economic change. Land Use
Policy, 27(2): 108–118. doi: 10.1016/j.landusepol.2009.09.003
Laurance W F, Vasconcelos H L, Lovejoy T E, 2000. Forest loss
and fragmentation in the Amazon: implications for wildlife
conservation. Oryx, 34(1): 39–45. doi: 10.1046/j.1365-3008.
2000.00094.x
Lechner A M, Foody G M, Boyd D S, 2020. Applications in re-
mote sensing to forest ecology and management. One Earth,
2(5): 405–412. doi: 10.1016/j.oneear.2020.05.001
Lele N, Joshi P K, Agrawal S P, 2008. Assessing forest fragment-
ation in northeastern region (NER) of India using landscape
matrices. Ecological Indicators, 8(5): 657–663. doi:
106 Chinese Geographical Science 2021 Vol. 31 No. 1
10.1016/j.ecolind.2007.10.002
Li M S, Mao L J, Zhou C G et al., 2010. Comparing forest frag-
mentation and its drivers in China and the USA with Globcov-
er v2.2. Journal of Environmental Management, 91(12):
2572–2580. doi: 10.1016/j.jenvman.2010.07.010
Li M S, Mao L J, Lu C et al., 2012. Temporal change in frag-
mentation in China’s primary forest ecoregions. Advanced Ma-
terials Research, 433-440: 1286–1296. doi: 10.4028/www.sci-
entific.net/AMR.433-440.1286
Li X, Zhou Y, Asrar G R et al., 2017. Characterizing spatiotem-
poral dynamics in phenology of urban ecosystems based on
Landsat data. Science of the Total Environment, 605-606:
721–734. doi: 10.1016/j.scitotenv.2017.06.245
Liang C, Jiang P H, Chen W et al., 2015. Farmland protection
policies and rapid urbanization in China: a case study for
Changzhou city. Land Use Policy, 48: 552–566. doi:
10.1016/j.landusepol.2015.06.014
Lindquist E J, D’Annunzio R, 2016. Assessing global forest land-
use change by object-based image analysis. Remote Sensing,
8(8): 678. doi: 10.3390/rs8080678
Liu J Y, Zhang Z X, Xu X L et al., 2010. Spatial patterns and
driving forces of land use change in china during the early 21st
century. Journal of Geographical Sciences, 20(4): 483–494.
doi: 10.1007/s11442-010-0483-4
Liu Jiyuan, Zhang Qian, Hu Yunfeng, 2012. Regional differ-
ences of China’s urban expansion from late 20th to Early 21st
Century based on remote sensing information. Chinese Geo-
graphical Science, 22: 1–14. doi: 10.1007/s11769-012-0510-8
Liu J Y, Kuang W H, Zhang Z X et al., 2014. Spatiotemporal
characteristics, patterns, and causes of land-use changes in
China since the late 1980s. Journal of Geographical Sciences,
24(2): 195–210. doi: 10.1007/s11442-014-1082-6
Liu T, Liu H, Qi Y J, 2015. Construction land expansion and cul-
tivated land protection in urbanizing China: insights from na-
tional land surveys: 1996-2006. Habitat International, 46:
13–22. doi: 10.1016/j.habitatint.2014.10.019
Liu Tingxiang, Zhang Shuwen, Tang Junmei et al., 2013. Com-
parison and analysis of agricultural and forest land changes in
typical agricultural regions of northern mid-latitudes. Chinese
Geographical Science, 23(2): 163–172. doi: 10.1007/s11769-
012-0576-3
Liu Y L, Feng Y H, Zhao Z et al., 2016a. Socioeconomic drivers
of forest loss and fragmentation: a comparison between differ-
ent land use planning schemes and policy implications. Land
Use Policy, 54: 58–68. doi: 10.1016/j.landusepol.2016.01.016
Liu Fang, Zhang Zengxiang, Shi Lifeng et al., 2016b. Urban ex-
pansion in China and its spatial-temporal differences over the
past four decades. Journal of Geographical Sciences, 26(10):
1477–1496. doi: 10.1007/s11442-016-1339-3
Liu Y S, Wang J Y, Long H L, 2010b. Analysis of arable land
loss and its impact on rural sustainability in Southern Jiangsu
Province of China. Journal of Environmental Management,
91(3): 646–653. doi: 10.1016/j.jenvman.2009.09.028
Liu Y S, Fang F, Li Y H, 2014b. Key issues of land use in China
and implications for policy making. Land Use Policy, 40:
6–12. doi: 10.1016/j.landusepol.2013.03.013
Liu Z, Geng Y, Zhang P et al., 2014c. Emergy-based comparat-
ive analysis on industrial clusters: economic and technological
development zone of Shenyang area, China. Environmental
Science and Pollution Research, 21(17): 10243–10253. doi:
10.1007/s11356-014-2854-3
Lv H L, Yang Y B, Zhang D et al., 2019. Perimeter-area ratio ef-
fects of urbanization intensity on forest characteristics, land-
scape patterns and their associations in Harbin city, northeast
China. Urban Ecosystems, 22(4): 631–642. doi: 10.1007/
s11252-019-00850-0
Madanian M, Soffianian A R, Koupai S S et al., 2018. Analyzing
the effects of urban expansion on land surface temperature pat-
terns by landscape metrics: a case study of Isfahan city, Iran.
Environmental Monitoring and Assessment, 190(4): 189. doi:
10.1007/s10661-018-6564-z
Mao D H, He X Y, Wang Z M et al., 2019. Diverse policies lead-
ing to contrasting impacts on land cover and ecosystem ser-
vices in Northeast China. Journal of Cleaner Production, 240:
117961. doi: https://doi.org/10.1016/j.jclepro.2019.117961
McGarigal K, Marks B J, 1995. FRAGSTATS: Spatial Pattern
Analysis Program for Quantifying Landscape Structure. Port-
land: USDA Forest Service, 122.
McGarigal K, 2002. Landscape pattern metrics. In: El-Shaarawi
A H and Piegorsch W W (eds.). Encyclopedia of Environmet-
rics. Chichester: John Wiley & Sons, 1135–1142.
McGarigal K, 2015. Fragstats Help. Amherst: University of Mas-
sachusetts.
Miller M D, 2012. The impacts of Atlanta ’s urban sprawl on
forest cover and fragmentation. Applied Geography, 34:
171–179. doi: 10.1016/j.apgeog.2011.11.010
NBSC (National Bureau of Statistics of China), 2010. China
Compendium of Statistics. Beijing: China Statistics Press. (in
Chinese)
NBSC (National Bureau of Statistics of China), 2013. China
Compendium of Statistics. Beijing: China Statistics Press. (in
Chinese)
NDRC (National Development and Reform Commission of
China), 2014. National New-type Urbanization Plan. Avail-
able at: http://www.gov.cn/zhengce/2014-03/16/content_
2640075.htm (accessed on 16 March 2014)
Reddy C S, Sreelekshmi S, Jha C S et al., 2013. National assess-
ment of forest fragmentation in India: landscape indices as
measures of the effects of fragmentation and forest cover
change. Ecological Engineering, 60: 453–464. doi: 10.1016/
j.ecoleng.2013.09.064
Setiawan Y, Yoshino K, Prasetyo L B, 2014. Characterizing the
dynamics change of vegetation cover on tropical forestlands
using 250 m multi-temporal MODIS EVI. International Journ-
al of Applied Earth Observation and Geoinformation, 26:
132–144. doi: 10.1016/j.jag.2013.06.008
Seymour F, 2020. Seeing the forests as well as the (Trillion) trees
in corporate climate strategies. One Earth, 2(5): 390–393. doi:
10.1016/j.oneear.2020.05.006
Shen W J, Wu J G, Grimm N B et al., 2008. Effects of urbaniza-
YANG Chao et al. Rapid Urbanization Induced Extensive Forest Loss to Urban Land in the Guangdong-Hong ... 107
tion-induced environmental changes on ecosystem functioning
in the phoenix metropolitan region, USA. Ecosystems, 11:
138–155. doi: 10.1007/s10021-007-9085-0
Song D X, Huang C, Sexton J O et al., 2014. Size and frequency
of forest loss and gain in China during 2000–2005. In: Americ-
an Geo-physical Union, Fall Meeting 2014. San Francisco:
AGU, Tao R, Su F B, Liu M X et al., 2010. Land leasing and
local public finance in China ’s regional development: evid-
ence from prefecture-level cities. Urban Studies, 47(10):
2217–2236. doi: 10.1177/0042098009357961
Trac C J, Schmidt A H, Harrell S et al., 2013. Environmental re-
views and case studies: is the returning farmland to forest pro-
gram a success? three case studies from Sichuan. Environment-
al Practice, 15(3): 350–366. doi: 10.1017/S146604661300
0355
Turner II B L, Lambin E F, Reenberg A, 2007. The emergence of
land change science for global environmental change and sus-
tainability. Proceedings of the National Academy of Sciences of
the United States of America, 104(52): 20666–20671. doi:
10.1073/pnas.0704119104
Van Den Hoek J, Ozdogan M, Burnicki A et al., 2014. Evaluat-
ing forest policy implementation effectiveness with a cross-
scale remote sensing analysis in a priority conservation area of
Southwest China. Applied Geography, 47: 177–189. doi:
10.1016/j.apgeog.2013.12.010
Wahyudi A, Liu Y, Corcoran J, 2018. Combining landsat and
landscape metrics to analyse large-scale urban land cover
change: a case study in the Jakarta Metropolitan Area. Journal
of Spatial Science, 64(3): 515–534. doi: 10.1080/14498596.
2018.1443849
World Bank, 2015. World bank database. http://data.worldbank.
org. Cited 13 May 2015.
Wu Z, Dai E, Wu Z F et al., 2019. Future forest dynamics under
climate change, land use change, and harvest in subtropical
forests in southern China. Landscape Ecology, 34(4): 843–863.
doi: 10.1007/s10980-019-00809-8
Xie H L, Kung C C, Zhao Y L, 2012. Spatial disparities of re-
gional forest land change based on ESDA and GIS at the
county level in Beijing-Tianjin-Hebei area. Frontiers of Earth
Science, 6(4): 445–452. doi: 10.1007/s11707-012-0338-7
Xie H L, He Y F, Zhang N et al., 2017. Spatiotemporal changes
and fragmentation of forest land in Jiangxi province, China.
Journal of Forest Economics, 29(1): 4–13. doi: 10.1016/j.
jfe.2017.08.004
Xu G L, Huang X J, Zhong T Y et al., 2015. Assessment on the
effect of city arable land protection under the implementation
of China’s national general land use plan (2006-2020). Habitat
International, 49: 466–473. doi: 10.1016/j.habitatint.2015.06.
017
Yang C, Wu G F, Ding K et al., 2017b. Improving land use/land
cover classification by integrating pixel unmixing and decision
tree methods. Remote Sensing, 9(12): 1222. doi: 10.3390/
rs9121222
Yang C, Li Q Q, Hu Z W et al., 2019a. Spatiotemporal evolution
of urban agglomerations in four major bay areas of US, China
and Japan from 1987 to 2017: Evidence from remote sensing
images. Science of the Total Environment, 671: 232–247. doi:
10.1016/j.scitotenv.2019.03.154
Yang C, Wu G F, Chen J Y et al., 2019b. Simulating and fore-
casting spatio-temporal characteristic of land-use/cover change
with numerical model and remote sensing: a case study in
Fuxian Lake Basin, China. European Journal of Remote Sens-
ing, 52(1): 374–384. doi: 10.1080/22797254.2019.1611387
Xu G, Zhou Z, Jiao L et al., 2020. Compact urban form and ex-
pansion pattern slow down the decline in urban densities: a
global perspective. Land Use Policy, 94: 104563. doi: 10.1016/
j.landusepol.2020.104563
Yang C, Zhang C C, Li Q Q et al., 2020. Rapid urbanization and
policy variation greatly drive ecological quality evolution in
Guang-dong-Hong Kong-Macau greater bay area of China: a
remote sensing perspective. Ecological Indicators, 115:
106373. doi: 10.1016/j.ecolind.2020.106373
Yang J, Sun J, Ge Q S et al., 2017a. Assessing the impacts of urb-
anization-associated green space on urban land surface temper-
ature: a case study of Dalian, China. Urban Forestry & Urban
Greening, 22: 1–10. doi: 10.1016/j.ufug.2017.01.002
Yu Sisi, Sun Zhongchang, Guo Huadong et al., 2017. Monitoring
and analyzing the spatial dynamics and patterns of megacities
along the maritime Silk Road. Journal of Remote Sensing,
21(2): 169–181. (in Chinese)
Yu Z W, Yao Y W, Yang G Y et al., 2019. Strong contribution of
rapid urbanization and urban agglomeration development to re-
gional thermal environment dynamics and evolution. Forest
Ecology and Management, 446: 214–225. doi: 10.1016/j.
foreco.2019.05.046
Zengin H, Değermenci A S, Bettinger P, 2018. Analysis of tem-
poral changes in land cover and landscape metrics of a man-
aged forest in the west black sea region of northern Turkey:
1970-2010. Journal of Forestry Research, 29: 139–150. doi:
10.1007/s11676-017-0423-6
Zhang J, Zhou Y K, LI R Q et al., 2010. Accuracy assessments
and uncertainty analysis of spatially explicit modeling for land
use/cover change and urbanization: a case in Beijing metropol-
itan area. Science China Earth Science, 53: 173–180. doi:
10.1007/s11430-009-0199-4
Zhang L, Weng Q H, 2016. Assessment of urban growth in the
Pearl River Delta, China, using time series Landsat imagery.
In: Weng Q H (ed). Remote Sensing for Sustainability. London:
CRC Press.
Zhang Z X, Wen Q K, Liu F et al., 2016. Urban expansion in
china and its effect on cultivated land before and after initiat-
ing “reform and open policy. Science China Earth Sciences,
59(10): 1930–1945. doi: 10.1007/s11430-015-0160-2
Zhong T Y, Mitchell B, Huang X J, 2014. Success or failure:
evaluating the implementation of China’s national general land
use plan (1997-2010). Habitat International, 44: 93–101. doi:
10.1016/j.habitatint.2014.05.003
108 Chinese Geographical Science 2021 Vol. 31 No. 1
