Available via license: CC BY-NC-ND 3.0
Content may be subject to copyright.
DOI: 10.2478/eces-2021-0008 ECOL CHEM ENG S. 2021;28(1):87-100
Xuan ZHAO
1
and Jianjun LIU
1*
QUANTIFYING URBAN VEGETATION COVERAGE CHANGE
WITH A LINEAR SPECTRAL MIXING MODEL:
A CASE STUDY IN XI’AN, CHINA
Abstract: With the rapid development of urban area of Xi’an in recent years, the contradiction between ecological
environmental protection and urban development has become prominent. The traditional remote sensing
classification method has been unable to meet the accuracy requirements of urban vegetation monitoring.
Therefore, how to quickly and accurately conduct dynamic monitoring of urban vegetation based on the spectral
component characteristics of vegetation is urgent. This study used the data of Landsat 5 TM and Landsat 8 OLI in
2011, 2014 and 2017 as main information source and LSMM, region of variation grid analysis and other methods
to analyse the law of spatial-temporal change of vegetation components in Xi’an urban area and its influencing
factors. The result shows that: (1) The average vegetation coverage of the study area from 2011 to 2017 reached
more than 50 %, meeting the standard of National Garden City (great than 40 %). The overall vegetation coverage
grade was high, but it had a decreasing trend during this period. (2) The vegetation in urban area of Xi’an
experienced a significant change. From 2011 to 2017, only 30 % of the low-covered vegetation, 24.39 % of the
medium-covered vegetation and 20.15 % of the high-covered vegetation remained unchanged, while the
vegetation in the northwest, northeast, southwest and southeast of the edge of the city’s third ring changed
significantly. (3) The vegetation quality in urban area of Xi’an has decreased from 2011 to 2014 with 6.9 % of
vegetation coverage reduced; while from 2014 to 2017, the overall vegetation quality of this area has improved
with 2.1 % of the vegetation coverage increased, which was mainly attributed to urban construction and Urban
Green Projects. This study not only can obtain the dynamic change information of urban vegetation quickly, but
also can provide suggestions and data support for urban planning of ecological environmental protection.
Keywords: urban vegetation changes, vegetation components partition, Linear Spectral Mixing Model (LSMM),
vegetation quality, remote sensing
Introduction
With the formulation and implementation of the national plan for building city clusters
and metropolitan areas, China’s urbanization rate has reached 59.58 % in 2018 [1]. As more
and more rural population is moving to cities, the ecological space in urban areas are
threatened and there are prominent contradictions between ecological environmental
protection and economic development [2]. In this case, the spatial-temporal change
monitoring of urban environment will become more and more important [3]. Urban
vegetation is the sum of all kinds of plants such as forests, thickets, hedges, flower beds
also situated in green roofs, grasslands, trees and crops owned by parks, campuses, temples,
squares, courts, courtyards, streets, farmlands and idle spaces in the city [4-6].
As an important element of urban ecosystem, urban vegetation plays an irreplaceable role
1
College of Landscape Architecture and Art, Northwest A&F University, Xianyang 712100, China
*
Corresponding author: ljj@nwafu.edu.cn
Xuan Zhao and Jianjun Liu
88
in promoting the development of urban production, enriching citizens’ life and maintaining
ecological equilibrium [7]. For example, it can alleviate “Urban Heat Island”, improve
regional climate and so on [8].
With the rapid development of urbanization, the change of vegetation coverage has
become an important basis of measuring urban ecological status and promoting urban
ecological landscape planning [9, 10]. At present, researches on urban vegetation coverage
dynamics that scholars at home and abroad carried out are mainly as follows. (1) Image
monitoring based on time series [11]. Du et al. [12] used MODIS NDVI data and
meteorological observation data during 2000 to 2016 and adopted maximum synthesis,
trend analysis and other related methods to study the distribution characteristics of NDVI in
Qinghai-Tibet Plateau at different periods and its relationship with precipitation and
temperature. Li et al. [13] studied the law of ecological environment evolution in the
Yangtze River Delta based on MODIS influence. But this method generally has a low
resolution ratio, so that it is difficult to reveal the subtle change of vegetation.
(2) Vegetation change monitoring based on medium-resolution land use classification data.
Liu et al. [14] used multi-source remote sensing data to clarify the relationship between
land use change and urban ecological environment in urban area of Beijing. Hussein et al.
[15] extracted land use classification data by Landsat images and monitored urban
vegetation coverage. This method usually depends on land use classification standards, so it
is difficult to reflect the dynamic conversion between vegetation in different areas with
different coverage grades. (3) Urban land monitoring based on Unmanned Aerial Vehicles
(UAV) and high-resolution images [16]. Liang et al. [17] pointed out that the UAV remote
sensing in the aspect of urban green space information extraction is very mature, which can
effectively evaluate urban green spaces and ecological services. Hashim et al. [18] studied
urban green space extraction and vegetation index thresholds using high-resolution images.
However, there are some limitations in application and promotion due to high-resolution
images’ large volume of data and its high cost. Monitoring on urban vegetation change
mostly focuses on land use classification, but it is time-consuming and labour-intensive.
And the computer classification and recognition algorithms are usually used as statistical
regression or training methods, most of which are supervisory modelling analysis [19, 20].
In the current researches based on land use classification, the "majority" method is often
used to classify mixed pixels, and vegetation components in non-vegetation pixels are often
ignored, resulting in statistical errors. How to eliminate these statistical errors from the
perspective of remote sensing mechanism and spectral decomposition and how to find out
vegetation components contained in non-vegetation units are worth studying.
Xi’an is the only megacity in northwest China and a key city along the Belt and Road
Initiative [21]. In 2014, the establishment of Xixian New Area was approved by the State
Council; in 2018, the National Development and Reform Commission (NDRC) issued
a document to approve the Guanzhong Plain City Cluster Development Plan, clearly
proposing to build Xi’an into the ninth National Central City [22]. This development trend
indicates that with the urban construction speed of Xi’an, its urban environment is
changing. Therefore, it is of great significance to carry out research on vegetation coverage
change in Xi’an. This study takes the urban area of Xi’an within the third ring as the study
area, using the TM remote sensing data in 2011, 2014 and 2017 as the information source
and taking the high spatial resolution Google images of the same period as references to
analyse three points as follows. (1) The change of vegetation coverage and vegetation
components in Xi’an urban area during the past decade. (2) The law of vegetation spatial
Quantifying urban vegetation coverage change with a linear spectral mixing model: A case study …
89
variation and mutual conversion in urban area of Xi’an. (3) The influence factors of
vegetation change in urban area of Xi’an. The results of this research will provide
references for urban vegetation construction and green space system planning in the study
area.
Materials and methods
Study area
Xi’an, the study area, is located between 107°40’~109°49’E and 33°39’~34°45’N
(Fig. 1). The south and southeast of this city is bounded by the main ridge of the Qinling
Mountains, which is adjacent to Hanzhong City and Shangluo City; the west of this city is
bounded by Taibai Mountain and Qinghua Taiyuan, bordering Baoji City; the northwest is
bounded by the Weihe River, facing Xianyang City across the river; and the east is bounded
by the Linghe River and Ruyuan Mountains, bordering Weinan City. The climate of this
area belongs to the Semi-humid Monsoon Climate of Warm Temperature Zone with four
distinct seasons.
Fig. 1. Study area
Data source and pre-processing
The remote sensing data in this study comes from the Geospatial Data Cloud of
Chinese Academy of Sciences and the administrative boundary data of the study area
comes from Xi’an Database of Forest Resources Second Inventory. Considering that the
cloud cover on the day when the remote sensing data was obtained was less than 10 % and
the influences of vegetation growth season and other factors, 6 bands of the Landsat 5 TM
sensor on June 4, 2011 and 7 bands of Landsat 8 OLI sensor on May 11, 2014 and April 17,
2017 were finally selected, with a spatial resolution of 30 m. The orbit number is 127/36.
For eliminating the effects of different data source and time, we used relative radiometric
calibration method, while treated the Landsat 5 TM images as reference and other images
conduct relative calibration. All the calibrations including FLAASH atmospheric correction
and geometric correction for the original images was conducted with ENVI 5.3 [23]
software.
Xuan Zhao and Jianjun Liu
90
Extraction of vegetation coverage grade
Remote sensing inversion method establishes correlation with vegetation coverage
grade through establishing vegetation remote sensing spectral information of the study area,
and then obtains vegetation coverage grade [24-26]. When the interaction between incident
radiation and the land-cover material in the scene involves a single endmember, then the
pixel’s spectral response can be satisfactorily modelled by the weighted sum of the
endmembers’ spectral responses, with the weights estimated by the fraction of the pixel
covered by the endmember. The modelling process is then known as Linear Spectral
Mixture Model (LSMM) [27]. The Linear Spectral Mixing Model (LSMM) decomposition
adopted in this study does not require the measurement of ground vegetation coverage data
for modelling which is easy to popularize [28]. Generally, LSMM’s definition is that the
reflectance of pixels in a certain spectral band is a linear combination of the reflectance of
the terminal units constituting pixels and its proportion in the pixel area as the weight
coefficient [29-31]. The formula is as follows:
1
N
i ki k i
k
R f C
λ λ λ
ε
=
= +
(1)
1
1
N
ki
k
f
=
=
(2)
1/2
2
1
( ) /
i
i
RMS N
λ
λ
ε
=
=
(3)
where R
i
λ
is the spectral reflection of the i pixel in the λ band; N is the number of basic
components; in N ≤ m + 1, m is the number of bands; f is the value of component occupied
by the k basic component in its corresponding i pixel;
C
kλ
is the spectral reflection of the k
basic component in the λ band; ε
iλ
is the error value of the i pixel in the λ band;
RMS
is the
error image of root-mean-square.
Endmembers collection
According to the characteristics of land cover in this study area, the endmembers can
be defined as four categories: bare soil, vegetation, impervious layer and water body.
Generally, there are many methods to collect endmembers, such as endmembers extraction
based on geometric vertex, Pixel Purity Index (PPI) or Sequential Maximum Angle Convex
Cone (SMACC), etc. [32] In this study, the endmembers were selected by establishing
Region of Interest (ROI). Generally, ROI is defined as selecting and applying information
about the area you want to capture in remote sensing image processing. First, it used
Minimum Noise Fraction (MNF) transformation to remove the noise of mask data and
confirm the intrinsic dimensions of data, so as to reduce the amount of calculation in
subsequent study [33-35]. In this way, the images in three stages were changed into six
band components through MNF transformation. This study synthesized the bands of the
first three components to obtain false colour images. Then, based on the Google Earth
historical images of the same period as a reference, ROI was established. At last, the
original range of the endmembers that influence the spectrum was collected [36] (Fig. 2).
Quantifying urban vegetation coverage change with a linear spectral mixing model: A case study …
91
Fig. 2. Spectral reflectance of each band: a) in 2011, b) in 2014, c) in 2017
LSMM decomposition and validation
Using the collected spectral profile of endmembers, the vegetation coverage of this
study area was obtained through LSMM decomposition with the restrictive Least Squares.
According to the RMS images of data’s root-mean-square in three stages obtained through
LSMM decomposition, the mean error was 1.69, which met the analysis accuracy
requirements of vegetation coverage change.
Changes of vegetation coverage grade
Vegetation components extracted through LSMM were defined as vegetation coverage
grades. According to the realities of the study area, the vegetation coverage in this area was
divided into five grades (Table 1) using the classification method proposed in the literature
[37] and was denoted by F
g
. Then the spatial distribution of vegetation coverage grades is
obtained. Through transition matrix and hierarchical encoding (Table 1), changing trends of
vegetation coverage grades from 2011 to 2014 and from 2014 to 2019 are calculated.
The formula is as follows [38]:
( 1)
10
i j ti j t i j
Y X X
× × + ×
= × + (4)
where i, j represent the line and column code number of raster graphics; Y
i×j
represents
vegetation coverage change of pixel in the line i, column j from t to t+1 period; X
ki×j
represents the code number of the previous vegetation coverage; and
X
(k+1)i×j
represents the
code number of the latter vegetation coverage [39].
Table 1
Thresholds of different vegetation coverage grades (F
g
: fraction of vegetation coverage)
Vegetation coverage No-covered Low-covered
Medium-covered
High-covered Full-covered
Grade 1 2 3 4 5
Coverage degree [%] 0 ≤
F
g
< 20 20 ≤
F
g
< 40 40 ≤
F
g
< 60 60 ≤
F
g
< 80 80 ≤
F
g
Xuan Zhao and Jianjun Liu
92
Analysis of vegetation quality grade change based on geospatial grid method
This study used grid analysis to describe the spatial distribution of vegetation change
[40]. Considering the area and the reasonable form of the vegetation distribution of the
study area comprehensively, 480 × 480 m grids were divided. The attributes of vegetation
evolution type in the grid are defined according to the weight of the vegetation change type
in the grid. If the weight of vegetation evolution type in grid is more than 50 %, then define
this type as grid transition. In this way, the main spatial distribution of vegetation evolution
types can be obtained.
The vegetation change types were divided based on the transition matrix and the
spatial distribution of vegetation coverage grade. The increase of vegetation coverage grade
is defined as high-transition type, an unchanged vegetation coverage grade is defined as
invariant-transition type, and a decline in vegetation coverage grade is defined
as low-transition type. The formula is presented as follows:
( 1)
i j t i j ti j
TY X X
× + × ×
= − (5)
where TY
i×j
represents vegetation coverage change type from t to t+1 period; X
(t+1)×j
represents the code number of the previous vegetation coverage; X
ki×j
represents the code
number of the latter vegetation coverage (Table 1). If TY
i×j
> 0, the vegetation type is
defined as high-transition type; if TY
i×j
< 0, the vegetation type is defined as low-transition
type; and if TY
i×j
= 0, it is defined as invariant-transition type.
Results and discussion
Vegetation components characteristics in Xi’an urban area
The average vegetation coverage of the study area in three stages reached 50 %, and
the coverage rate was relatively high (Fig. 3, Table 2), which exceeded the evaluation
standard of National Forest City (great than 40 %). The annual average vegetation coverage
decreased by 11 % from 2011 to 2014, 3 % from 2014 to 2017, and 13 % from 2011 to
2017 respectively. The average reduction of standard deviation of vegetation coverage
grade in study area during these three stages was 1.73, 6.17, 7.90 % respectively, where
mainly occurred in the surrounding of the urban city. From the area of different vegetation
coverage grades, it can be seen that there is an increasing trend of no vegetation coverage
during all these three stages where mainly occurred in the north and northwest of urban
area; in the variation trend of low, medium, high, and full vegetation coverage, except that
low coverage in 2011 to 2014 and full coverage in 2014 to 2017 increased slightly, the rest
of them showed different degrees of downward trend.
Table 2
Statistics of average vegetation coverage and its grade composition area between three stages of the study area
Year Area of different vegetation coverage [km
2
]
F
g
No coverage Low coverage
Medium coverage High coverage
Full coverage
2011 58.00 190.28 94.72 75.33 50.72 46.70
2014 48.00 238.58 99.20 58.24 32.87 28.86
2017 45.00 251.80 76.79 51.27 32.72 45.15
∆14-11 –11.00 48.30 4.47 –17.09 –17.85 –17.84
∆17-14 –3.00 13.22 –22.40 –6.97 –0.14 16.30
∆17-11 –13.00 61.52 –17.93 –24.06 –17.99 –1.54
Quantifying urban vegetation coverage change with a linear spectral mixing model: A case study …
93
Fig. 3. Vegetation coverage grade: a) in 2011, b) in 2014, c) in 2017
Transitions of vegetation coverage grades in urban area of Xi’an
During 2011 to 2017, areas without vegetation coverage were located in impervious
areas (located in the core of urban city) such as urban buildings, water bodies, etc., among
which 83.12 % remained unchanged (Table 5, Fig. 4c). The rest was mainly transferred to
low and medium coverage grades. The transition to low coverage areas were located at the
north, south and southwest to central part of study area; and the transition to medium
coverage areas were located at the north and south of study area. Areas that transferred to
no coverage grade include all kinds of coverage grade components. The comprehensive
transition effect led to a 7.90 km
2
reduction of no coverage area and the proportion of low
coverage area remained unchanged. The proportion of the rest areas that mainly transferred
to no coverage, medium coverage, high coverage and full coverage accounted for 50.10,
13.28, 3.91 and 2.71 % respectively.
Table 3
The transition matrix of different vegetation coverage grade area from 2011 to 2014
Transition rate [%]
2014
Grade No coverage Low coverage
Medium
coverage High coverage
Full coverage
2011
No coverage 86.93 10.22 2.04 0.58 0.24
Low coverage
44.61 42.97 9.41 1.98 1.02
Medium
coverage 20.37 35.70 31.06 8.68 4.19
High coverage
16.76 15.55 31.58 24.99 11.12
Full coverage
15.73 8.63 12.82 22.87 39.96
Xuan Zhao and Jianjun Liu
94
Fig. 4. a) Transfer distribution maps of area proportion of vegetation coverage grade from 2011 to 2014,
b) transfer distribution maps of area proportion of vegetation coverage grade from 2014 to 2017,
c) transfer distribution maps of area proportion of vegetation coverage grade from 2011 to 2017,
note: N means no coverage, L means low coverage, M means medium coverage, H means high
coverage, F means full coverage
Table 4
The transition matrix of different vegetation coverage grade area from 2014 to 2017
Transition rate [%]
2017
Grade No coverage Low coverage
M
edium
coverage High coverage
Full coverage
2014
No coverage 84.80 11.11 2.56 0.84 0.68
Low coverage
33.16 36.66 21.98 5.69 2.51
Medium
coverage 15.59 17.05 29.35 24.04 13.97
High coverage
12.79 7.17 13.36 23.81 42.87
Full coverage
12.65 4.65 6.18 11.25 65.28
Table 5
The transition matrix of different vegetation coverage grade area from 2011 to 2017
Transition rate [%]
2017
Grade No coverage Low coverage
Medium
coverage High coverage
Full coverage
2011
No coverage 83.12 10.66 3.73 1.47 1.02
Low coverage
50.10 30.00 13.28 3.91 2.71
Medium
coverage 29.98 22.71 24.39 13.25 9.67
High coverage
25.14 12.77 17.22 20.15 24.72
Full coverage
24.07 9.01 9.35 12.89 44.69
Quantifying urban vegetation coverage change with a linear spectral mixing model: A case study …
95
During 2011 to 2014, areas without vegetation coverage were widely located in
impervious areas (located in the core of urban city) such as urban buildings, water bodies,
etc. (Table 3, Fig. 4a), among which 86.93 % remained unchanged. The other 13.07 %
mainly transferred to low and medium coverage grades which were mainly located at the
central of Weiyang District, the south of Baqiao District and Yanta District. Areas that
transferred to no coverage grade included all kinds of coverage grade components.
The comprehensive transition effect led to a 48.30 km
2
increase of no coverage area and the
proportion of low coverage area remained unchanged (42.97 %). The proportion of the rest
areas that mainly transferred to no coverage and medium coverage accounted for 44.61 %
and 9.41 % respectively; areas transferred to no coverage were distributed all over the study
area; areas that transferred to no coverage, medium coverage, high coverage and full
coverage had a comprehensive impact that resulted in the increase of the low coverage area
by 4.47 km
2
.
During 2014 to 2017, areas without vegetation coverage were widely located in
impervious areas such as urban buildings, water bodies, etc., among which 84.80 %
remained unchanged. The rest 15.20 % mainly transferred to low and medium coverage
grades which were mainly located at the north, southwest, southeast and east of the study
area (Table 4, Fig. 4b). Areas that transferred to no coverage grade included all kinds of
coverage grade components. The comprehensive transition effect led to a 13.22 km
2
increase of no coverage area and the proportion of low coverage area remained unchanged
(36.66 %). The proportion of the rest areas that mainly transferred to no coverage, medium
coverage, high coverage and full coverage accounted for 33.16, 21.98, 5.69, 2.51 %
respectively; areas transferred to no coverage and medium coverage were distributed all
over the study area and areas transferred to high coverage and full coverage were mainly
distributed at the fringe of study area; areas that transferred to no coverage, medium
coverage, high coverage and full coverage had a comprehensive impact that resulted in the
decrease of the low coverage area by 22.40 km
2
.
Spatial dynamics of quality grade of vegetation coverage
There were differences in vegetation coverage grade changes among 2011, 2014 and
2017. The rise of vegetation coverage grade is considered to have a trend of vegetation
improvement; the invariable vegetation coverage grade is considered to have an unchanged
trend; and the decline of vegetation coverage grade is considered to have a trend of
vegetation degradation [41-43]. From 2011 to 2014 (Fig. 5a), 6.9 % of vegetation coverage
reduced in the study area. The dynamic changes of vegetation coverage in study area were
mainly located at the west of Weiyang District, the east of Xincheng District, the north of
Baqiao District and the west of Yanta District. The number of grids with declining
vegetation coverage grade was greater than that of grids with increasing vegetation
coverage grade, indicating that these areas had a trend of vegetation degradation (Fig. 5).
From 2014 to 2017 (Fig. 5b), 2.1 % of the vegetation coverage increased in the study area.
The dynamic changes of vegetation coverage in study area were mainly located at the west
of Weiyang District, the north of Baqiao District and the southeast of Yanta District.
The number of grids with increasing vegetation coverage grade was greater than that
of grids with declining vegetation coverage grade, indicating that these areas had a trend of
vegetation improvement. From 2011 to 2017 (Fig. 5c), the location of dynamic change
of vegetation coverage in study area was similar to that from 2014 to 2017. The number of
grids with increasing vegetation coverage grade was slightly less than that of grids with
Xuan Zhao and Jianjun Liu
96
declining vegetation coverage grade and the grids density was relatively large, indicating
that the overall trend of vegetation improvement was similar to that of vegetation
degradation during the past six years.
Fig. 5. Distribution of vegetation coverage grade dynamic: a) from 2011 to 2014, b) from 2014 to 2017,
c) from 2011 to 2017
Through the results analysis of this study, it can be seen that the vegetation coverage
within the third ring of the main urban area of Xi’an has undergone different degrees of
quantitative changes and spatial changes in the past six years. This change was in line with
the law of general urban development. Generally, the changes in city’s central area were
relatively small, while the changes in the urban fringe were relatively large, which was the
results that influenced by combined factors such as nature, man-made factors, urban
socio-economic development and so on. But it was mainly affected by the following
factors:
(1) Urban infrastructure had an impact on vegetation coverage change. With the
development of this study area and the introduction of new urban planning, the urban
construction of Xi’an was strengthened constantly. According to the survey data of
Xi’an Bureau of Planning and Natural Resources [44], during the three stages from
2011 to 2017, Xi’an completed a batch of comprehensive urban plans, such as
Research on the overall Planning and Development Strategy of Xi’an International
Metropolis, Overall Urban Planning of Greater Xi’an and so on. These construction
measures affected the change of land using types for urban construction and thus
indirectly affected the change of vegetation coverage. According to the Xi’an
Statistical Yearbooks, the areas of urban construction land using and road areas in
2011, 2014, and 2017 increased by 36.45 and 24.26 % on average. With the demand of
Quantifying urban vegetation coverage change with a linear spectral mixing model: A case study …
97
urban construction, some areas that used to be low and medium vegetation coverage
transferred to construction land, resulting in a reducing trend of vegetation coverage in
a short-term. However, this trend will be improved and enhanced with the construction
of the Urban Green Space System.
(2) Landscape engineering had an impact on vegetation coverage change. There were
92 newly built or renovated green space squares in Xi'an in 2011, 60 in 2014, and 62 in
2017 [43]. At the same time, the afforestation area in the study area was 6.66 km
2
in
2011, 7.06 km
2
in 2014, 1.24 km
2
in 2017. We can see that the increase of green square
areas and afforestation areas obviously affected the improvement of vegetation
coverage grade and effectively promotes the change of it.
(3) To some extent, the improvement of vegetation coverage was affected by the climate
change, because of air temperate and precipitation increased in recent decades, beside
the effects of urban heat island could not be neglected. As it is well known, the
decision of government had the largest effects on the coverage changes, because the
effects of tree species and tree ages selection were greater than the growth of trees.
Therefore, selecting the trees that could adapt to local soil and climate conditions is
crucial for sustaining a health urban ecosystem.
Conclusion
Research on the change of urban vegetation coverage plays a significant role in
formulating and adjusting the urban vegetation construction planning and improving he
urban human settlement environment. Meanwhile, it can provide a basis for the fine
management of urban vegetation. The conclusions of this study are as follows:
(1) During 2011 to 2017, the vegetation construction within the third ring of the main
urban area of Xi'an has achieved remarkable results. The results of the study showed
that although it fluctuated in different periods, the vegetation coverage declined in
2011 compared with that in 2014 or 2017, with an average decline of 12 %. However,
the average vegetation coverage in these six years was above 50 %, which exceeded
the evaluation standard (40 %) of National Forest City. The fluctuation of vegetation
coverage affected by the development of urban infrastructure in a short-term is
a normal phenomenon which does not affect the overall increase trend of urban
vegetation coverage.
(2) The spatial distribution of vegetation coverage in the main urban area of Xi’an is
a dynamic process. In addition to the relatively stable changes in the central area, the
northwest, northeast, southwest, and southeast fringe areas of study area had different
degrees of change, which was consistent with the urban development progress of Xi’an
in the past six years.
(3) The geospatial grid analysis method of urban vegetation changes based on LSMM can
intuitively express changes of vegetation quality. From 2011 to 2014, three stages
vegetation coverage quality of Xi’an had a trend of degradation in the west of Weiyang
District, the east of Xincheng District, the north of Baqiao District and the west of
Yanta District; from 2014 to 2017, the quality of vegetation coverage had a trend of
improvement in the west of Weiyang District, the north of Baqiao District and the
southeast of Yanta District; from 2011 to 2017, the density of vegetation coverage in
study area increased and the overall quality improved as well.
Xuan Zhao and Jianjun Liu
98
Acknowledgements
This work was supported by Research on Vegetation Restoration Techniques on Steep
Loess Slope in Qianyang, Shaanxi (K303021613).
References
[1] Li X, Li T, Li H, Qi J, Hu L. Research on the online consumption effect of China’s urbanization under
population aging background. Sustainability. 2019;11:4349. DOI: 10.3390/su11164349.
[2] Elhoseny H, Elhoseny M, Riad AM, Hassanien AE. A framework for big data analysis in smart cities. Int
Conf Adv Machine Learning Technol Appl. 2018;723:405-14. DOI:
10.1007/978-3-319-74690-6_40.
[3] Lv Z, Hu B, Lv H. Infrastructure monitoring and operation for smart cities based on IoT system. IEEE Trans
Industrial Informatics. 2019;16:1957-62. DOI: 10.1109/TII.2019.2913535.
[4] Richards DR, Belcher RN. Global changes in urban vegetation cover. Remote Sens. 2020;12(1):23. DOI:
10.3390/rs12010023.
[5] Deska I, Mrowiec M, Ociepa E, Michniewski M. Impact of the hydrogel amendment and the dry period
duration on the green roof retention capacity. Ecol Chem Eng S. 2020;27:357-71. DOI:
10.2478/eces-2020-0023.
[6] Gdela M, Widomski MK, Musz-Pomorska A. Hydraulic efficency of selected intensive green roof
substrates. Ecol Chem Eng A. 2019;26(1-2):37-45. DOI: 10.2428/ecea.2019.26(1-2)4.
[7] Tsai SB. Wang K. Using a novel method to evaluate the performance of human resources in green logistics
enterprises. Ecol Chem Eng S. 2019;26(4):629-40. DOI: 10.1515/eces-2019-0045.
[8] Oh RRY, Richards DR, Yee ATK. Community-driven skyrise greenery in a dense tropical city provides
biodiversity and ecosystem service benefits. Landscape Urban Planning. 2018;169:115-23. DOI:
10.1016/j.landurbplan.2017.08.014.
[9] Meng XY, Gao X, Li SY, Lei JQ. Spatial and temporal characteristics of vegetation NDVI changes and the
driving forces in Mongolia during 1982-2015. Remote Sens. 2020;12:603-28. DOI: 10.3390/rs12040603.
[10] Voorde TVD. Spatially explicit urban green indicators for characterizing vegetation cover and public green
space proximity: a case study on Brussels, Belgium. Int J Digital Earth. 2017;10:798-813.
DOI:10.1080/17538947.2016.1252434.
[11] Elhoseny M. Multi-object detection and tracking (MODT) machine learning model for real-time
video surveillance systems. Circuits Systems Signal Processing. 2020;39:611-30. DOI:
10.1007/s00034-019-01234-7.
[12] Du JQ, Zhao CX, Shu JM, Jiaerheng A, Yuan XJ, Yin JQ, et al. Spatiotemporal changes of vegetation on the
Tibetan Plateau and relationship to climatic variables during multiyear periods from
1982-2012. Environ Earth Sci. 2016;75:77. DOI:10.1007/s12665-015-4818-4.
[13] Li ZH, Gao W, Gao ZQ, Shi RH, Liu CH. A study on assessment of urbanization and ecosystem changes
based on MODIS time series in Shanghai municipality from 2000 to 2009. Proc SPIE. 2010;7809:78090R.
DOI: 10.1117/12.858605.
[14] Liu T, Yang XJ. Monitoring Urban Growth and Land Changes in Beijing, China’s Capital City by Remote
Sensing: Progress and Challenges: An Interdisciplinary Perspective. In: Challenges Towards Ecological
Sustainability in China. 2019:55-67. DOI: 10.1007/978-3-030-03484-9_4.
[15] Hussein SO. Monitoring urban greenness evolution using multitemporal Landsat imagery in the city of Erbil
(Iraq). Central European Geol. 2018;62:1-12. DOI: 10.1556/24.61.2018.10.
[16] Lv Z. The security of Internet of drones. Computer Commun. 2019;148:208-14. DOI:
10.1016/j.comcom.2019.09.018.
[17] Liang HL, Li WZ, Zhang QP, Zhu W, Chen D, Liu J, et al. Using unmanned aerial vehicle data to assess the
three-dimension green quantity of urban green space: A case study in Shanghai, China. Landscape Urban
Planning. 2017;164:81-90. DOI: 10.1016/j.landurbplan.2017.04.006.
[18] Hashim H, Abd Latif Z, Adnan NA. Urban vegetation classification with NDVI thresold value method with
very high resolution (VHR) PLEIADES Imagery. Int Arch Photogramm. Remote Sens Spat Inf Sci.
2019;237-40. DOI: 10.5194/isprs-archives-XLII-4-W16-237-2019.
[19] Rwanga S, Ndambuki J. Accuracy assessment of land use/land cover classification using remote sensing and
GIS. Int J Geosci. 2017;8:611-22. DOI: 10.4236/ijg.2017.84033.
[20] Kussul N, Lavreniuk M, Skakun S, Shelestov A. Deep learning classification of land cover and crop types
using remote sensing data. IEEE Geosci Remote Sensing Lett. 2017;14:778-82. DOI:
10.1109/LGRS.2017.2681128.
Quantifying urban vegetation coverage change with a linear spectral mixing model: A case study …
99
[21] Ramasamy B, Yeung MCH. China's one belt one road initiative: The impact of trade facilitation versus
physical infrastructure on exports. World Econ. 2019;42:1673-94. DOI: 10.1111/twec.12808.
[22] National Development and Reform Commission of the People’s Republic of China. Notice of the National
Development and Reform Commission and the Ministry of Housing and Urban-Rural Development on
Printing and Distributing the Development Plan of the Guanzhong Plain Urban Agglomeration. 2018.
No. 220. Available from: http://www.ndrc.gov.cn/zcfb/zcfbtz/201802/t20180207_876904.html.
[23] John AR. Remote Sensing Digital Image Analysis. 2013. ISBN: 9783642300615. DOI:
10.1007/978-3-642-30062-2.
[24] Xie YC, Sha ZY, Yu M. Remote sensing imagery in vegetation mapping: a review. J Plant Ecol.
2008;1:6240-52. DOI: 10.1093/jpe/rtm005.
[25] Quast R, Albergel C, Calvet JC, Wagner W. A generic first-order radiative transfer modelling approach for
the inversion of soil and vegetation parameters from scatterometer observations. Remote Sens. 2019;11:285.
DOI: 10.3390/rs11030285.
[26] Wang J, Du Y, Liu Z, Cheng H. Determining diagnostic indicators for fine-scale short vegetation
aboveground biomass inversion using a HVRU-based analysis approach. Ecol Indicators. 2020;111:106033.
DOI: 10.1016/j.ecolind.2019.106033.
[27] Zanotta DC, Haertel V, Shimabukuro YE, Renno CD. Linear spectral mixing model for identifying potential
missing endmembers in spectral mixture analysis. IEEE Trans Geosci Remote Sens. 2013;52:3005-12. DOI:
10.1109/TGRS.2013.2268539.
[28] Li H, Lei J, Wu J. Evolution analysis of vegetation cover under the disturbance of rare earth mining: a case
in Lingbei mining area. J Appl Sci Eng. 2017;20:393-400. DOI: 10.6180/jase.2017.20.3.14.
[29] Adami M, Bernardes S, Arai E, Freitas RM, Shimabukuro YE, Espírito-Santo FDB, et al. Seasonality of
vegetation types of South America depicted by moderate resolution imaging spectroradiometer (MODIS)
time series. Int J Appl Earth Observation Geoinform. 2018;69:148-63. DOI: 10.1016/j.jag.2018.02.010.
[30] Fassoni-Andrade AC, Zanotta DC, Guasselli LA, Andrade AM. Linear spectral mixing model for estimating
optically active components in estuarine waters of Patos Lagoon, Brazil. Int J Remote Sens.
2017;38:4767-81. DOI: 10.1080/01431161.2017.1323281.
[31] Li C, Liu P, Zou C, Sun F, Cioffi JM. Yang L. Spectral-efficient cellular communications with coexistent
one-and two-hop transmissions. IEEE Trans Vehicular Technol. 2015;65:6765-72. DOI:
10.1109/TVT.2015.2472456.
[32] Dong Di, Wang Di. Comparisons of ERDAS and ENVI in thematic mapping. 2011 IEEE 3rd Int Conf
Communication Software Networks. 2011. DOI: 10.1109/ICCSN.2011.6014623.
[33] Ran P, Eyal BD. Assessing the detection limit of petroleum hydrocarbon in soils using hyperspectral
remote-sensing. Remote Sensing Environ. 2019;224:145-53. DOI: 10.1016/j.rse.2019.01.026.
[34] Kumar C, Chatterjee S, Oommen T. Mapping hydrothermal alteration minerals using high-resolution
AVIRIS-NG hyperspectral data in the Hutti-Maski gold deposit area, India. Int J Remote Sens.
2020;41:794-812. DOI: 10.1080/01431161.2019.1648906.
[35] Zhao Y, Yang C. Information fusion robust guaranteed cost Kalman estimators with uncertain noise
variances and missing measurements. Int J Systems Sci. 2019;50:2853-69. DOI:
10.1080/00207721.2019.1690719.
[36] Bian JH, Li AN, Zhang ZJ, Zhao W, Lei GB, Yin GF, et al. Monitoring fractional green vegetation cover
dynamics over a seasonally inundated alpine wetland using dense time series HJ-1A/B constellation images
and an adaptive endmember selection LSMM model. Remote Sensing Environ. 2017;197:98-114. DOI:
10.1016/j.rse.2017.05.031.
[37] Zhang J, Liu Y, Zhou DM, Zhang Q, Chen WN. Spatial-temporal character of vegetation cover and its
influence factors in the Shule River Basin China, during 1985-2011. Human Ecol Risk Assess: Int J.
2020;26:608-20. DOI: 10.1080/10807039.2018.1528437.
[38] Peng J, Liu YH, Shen H, Han YN, Pan YJ. Vegetation coverage change and associated driving forces in
mountain areas of Northwestern Yunnan, China using RS and GIS. Environ Monit Assess.
2012;184:4787-98. DOI: 10.1007/s10661-011-2302-5.
[39] Liu X, Zhou W, Bai Z. Vegetation coverage change and stability in large open-pit coal mine dumps in China
during 1990-2015. Ecol Eng. 2016;95:447-51. DOI: 10.1016/J.ECOLENG.2016.06.051.
[40] Zhao K, Jin BX, Fan H, Song WW, Zhou SY, Jiang YY. High-performance overlay analysis of massive
geographic polygons that considers shape complexity in a cloud environment. Int J Geo-Information.
2019;8:290. DOI: 10.3390/ijgi8070290.
[41] Yan LB, He RX, Kašanin-Grubin M, Luo G, Peng H, Qiu JX. The dynamic change of vegetation cover and
associated driving forces in Nanxiong Basin, China. Sustainability. 2017;9:443. DOI: 10.3390/su9030443.
Xuan Zhao and Jianjun Liu
100
[42] Liu SY, Huang SZ, Xie YY, Wang H, Huang Q, Leng GY, et al. Spatial-temporal changes in vegetation
cover in a typical semi-humid and semi-arid region in China: Changing patterns, causes and implications.
Ecol Indicators. 2019;98:462-75. DOI: 10.1016/j.ecolind.2018.11.037.
[43] Du JQ, Fu Q, Fang SF, Wu JH, He P, Quan ZJ. Effects of rapid urbanization on vegetation cover in the
metropolises of China over the last four decades. Ecol Indicators. 2019;107:105458. DOI:
10.1016/j.ecolind.2019.105458.
[44] Yang Z, Song J, Cheng D, Xia J, Li Q, Ahamad MI. Comprehensive evaluation and scenario simulation for
the water resources carrying capacity in Xi'an city, China. J Environ Manage. 2019;230:221-33. DOI:
10.1016/j.jenvman.2018.09.085.
