International Journal of Applied Earth Observation and Geoinformation 38 (2015) 216–228

Contents lists available at ScienceDirect

International Journal of Applied Earth Observation and

Geoinformation
journal homepage: www.elsevier.com/locate/jag

Analysis on spatio-temporal trends and drivers in vegetation growth

during recent decades in Xinjiang, China
Jiaqiang Du a,b,∗ , Jianmin Shu a , Junqi Yin b , Xinjie Yuan b , Ahati Jiaerheng b ,
Shanshan Xiong c , Ping He a , Weiling Liu a
a
State Key Laboratory of Environmental Criteria and Risk Assessment, Chinese Research Academy of Environmental Sciences, Beijing 100012, China
b
Xinjiang Academy of Environmental Protection Science, Xinjiang 830011, China
c
Bureau of Hydrology, The Ministry of Water Resources of the People’s Republic of China, Beijing 100053, China

a r t i c l e i n f o a b s t r a c t

Article history: Vegetation plays an important role in regulating the terrestrial carbon balance and the climate system, and
Received 9 June 2014 also overwhelmingly dominates the provisioning of ecosystem services. In this study, a non-stationary
Received in revised form 6 January 2015 1982–2012 AVHRR NDVI3g time series, the newest dataset, were used to evaluate spatio-temporal pat-
Accepted 9 January 2015
terns of seasonal vegetation changes in Xinjiang province of China at regional, biome and pixel scales
over progressively longer periods from 18 to 31 years, starting in 1982, and their linkages to climatic
Keywords:
factors and human activities were analyzed. At regional scale, the increases were statistically significant
AVHRR NDVI3g
for autumn NDVI during fourteen periods, for growing season and summer NDVI during the most periods,
Hydrothermal conditions
Planting structure
and for spring only during the first four periods. The rates of NDVI increase in growing season and all sea-
Correlation sons significantly decreased over fourteen periods. At pixel scale, areas with significant browning rapidly
Regional scale increased over fourteen periods for growing season and all seasons, and these areas were mainly concen-
Pixel scale trated in northern desert of Xinjiang. Vegetation growth in Xinjiang was regulated by both moisture and
thermal conditions: the response of NDVI in spring and autumn was more sensitive to thermal factors,
such as temperature and potential evapotranspiration, and correlations between NDVI and precipitation
and between NDVI and humidity index were stronger in summer and growing season. Extensive use
of fertilizers and expanded farmland irrigated area increased vegetation growth for cropland. However,
the rapid increase in the proportion of cotton cultivation and use of drip irrigation may reduce spring
NDVI in the part of farmlands. Trend analysis during the multiple nested time series may contribute to a
better and deep understanding of NDVI dynamic and foreseeing changes in the future. Accordingly, NDVI
in Xinjiang will continuously increase at regional scale and the areas showing significant browning will
also furthermore grow.
© 2015 Elsevier B.V. All rights reserved.

Introduction
Vegetation is the most critical element of the terrestrial ecosys-
tem, serving as a medium in water circulation, material cycling,
energy flow, and information transfer. Vegetation plays a crucial
role in regulating the terrestrial carbon balance and climate sys-
tem (Peng et al., 2012, 2011; Piao et al., 2011). Therefore, it is
important to monitor vegetation changes. Changes in the spatial
and temporal pattern of vegetation alter the structure and function
of landscapes, consequently influencing biodiversity and ecologi-
cal processes (Li et al., 2012). Vegetation growth can be strongly
∗ Corresponding author at: Chinese Research Academy of Environmental Science,
8 Dayangfang, Chaoyang District, Beijing 100012, China. Tel.: +86 10 84933564.
affected by climate change and human activity (Goetz et al., 2005;
Myneni et al., 1997; Nemani et al., 2003; Peng et al., 2012). The
study of vegetation dynamics has become one of the key issues in
global change (Peng et al., 2012). There are large inter-annual vari-
ations and intra-annual dynamics in vegetation growth because
precipitation typically has large inter-annual variations in amount
and distribution within a year (Buyantuyev and Wu 2009; Li et al.,
2012; Qi et al., 2009; Wessels et al., 2007). The influence of cli-
mate change on vegetation growth is therefore more evident (IPCC,
2007). Due to high spatial coverage and a long time series, remotely
sensed data have become the most important source for moni-
toring vegetation dynamics at large scales (Fensholt et al., 2012;
Pettorelli et al., 2012; Wessels et al., 2007). The normalized differ-
ence vegetation index (NDVI) dataset, which was often used as a
proxy for vegetation productivity, is one of the key metrics of land

http://dx.doi.org/10.1016/j.jag.2015.01.006
0303-2434/© 2015 Elsevier B.V. All rights reserved.
 J. Du et al. / International Journal of Applied Earth Observation and Geoinformation 38 (2015) 216–228
Fig. 1. Distribution of vegetation types and climate stations in Xinjiang, China.
degradation, and it has been suggested that this index can be used
as an integral measure of ecosystem services (Bai and Dent, 2009;
Fensholt et al., 2012; Fensholt and Rasmussen, 2011). The GIMMS
NDVI dataset is considered to be one of the best datasets available
for long-term NDVI trend analysis (Alcaraz-Segura et al., 2010; Beck
et al., 2011; Beck and Goetz, 2011; Fensholt et al., 2012). The long-
term NDVI time series data provides a powerful tool to understand
vegetation growth history, monitor current conditions and prepare
for future changes (Piao et al., 2011; van Leeuwen et al., 2006). In
2014, released the newest NDVI dataset, calling AVHRR NDVI3g ,
and the quality inspection of the dataset displayed the AVHRR
NDVI3g show good quality (Pinzon and Tucker, 2014; Wang et al.,
2014; Zeng et al., 2013). The linear trend analysis over an overall
long-term series may obscure significant trend changes appearing
within shorter duration (de Jong et al., 2012). The comparing anal-
ysis between results of multi-periods, changes in vegetation NDVI
over the most recent several years and dynamics of NDVI trend
yet have not been well researched in the arid area of northwestern
China.
The Xinjiang province of northwest China is deep in the inte-
rior of Eurasia far from the ocean. It is the main part of China’s arid
regions and represents typical arid regions of Eurasia. Its complex
topography, arid continental climate, desert soil and vegetation act
together to form diverse biomes. In recent decades, the effect of
climate change has become evident (Wang et al., 2013; Wu et al.,
2010; Zhao et al., 2011b), especially in the terrestrial ecosystem of
Xinjiang (Mohammat et al., 2013; Piao et al., 2003; Ren et al., 2007;
Wang et al., 2011; Zhao et al., 2011a,b). A number of studies have
also indicated that NDVI data showed greening of Xinjiang during
recent decades, with increased precipitation as the main driving
factor (Mohammat et al., 2013; Piao et al., 2011; Zhao et al., 2011b).
However, other researchers state that the simple linear trend could
not accurately show temporal patterns of changes in vegetation
growth (Beck et al., 2011; de Jong et al., 2012; Piao et al., 2011;
Wessels et al., 2007, 2012). As the studies mentioned above were
conducted on larger spatial scales than Xinjiang, their conclusions
may not be applicable to the Xinjiang region. In most cases the time
period of these studies spanned the duration of the study, therefore
217
the continuity and robustness of changing processes and trends are
inadequate. Therefore, it is necessary to better understand the NDVI
changes and their climate controls during longer time scales in dif-
ferent seasons with the newest dataset. The objectives of this study
were to analyze vegetation growth dynamics relative to climatic
and human activity at regional, biome and pixel scales during grow-
ing season in Xinjiang. To analyze temporal patterns and dynamic
processes, we estimated NDVI trends and correlations between
NDVI and climatic factors over fourteen time periods: 1982–1999,
1982–2000, 1982–2001, . . ., 1982–2011 and 1982–2012.
Materials and methods
Study area
The Xinjiang Uygur Autonomous Region covers a total area
of 1665,000 km2 and spans 34.25 ◦ N–49.17 ◦ N in latitude and
73.33 ◦ E–96.42 ◦ E in longitude (Fig. 1). It is a typical mountain–basin
system including three east–west mountain ranges and two major
basins between these three mountain ranges. These mountains
consist of the Altai Mountains, Tianshan Mountains, and the Kun-
lun Mountains from north to south, and the basins consist of the
northern Junggar basin and the southern Tarim basin.
This area has a typical inner-continental climate, featured by a
wide range of temperatures, low precipitation, strong wind, and
low humidity. The annual precipitation and annual average tem-
perature in North Xinjiang ranges from 100 to 500 mm, and from
4 ◦ C to 8 ◦ C, respectively. In South Xinjiang, annual precipitation
ranges from 20 to 100 mm, and average temperature from 10 ◦ C to
13 ◦ C, respectively. Vegetation type ranges from forest to meadow,
steppe and desert in a vertical direction along mountains. This is
especially evident in the Altai Mountains and Tianshan Mountains.
Data sources and processing
The GIMMS NDVI3g dataset was derived from the NOAA/AVHRR
dataset at a spatial resolution of 0.083◦ × 0.083◦ over a 15-d interval
from 1982 to 2012. The GIMMS NDVI3g dataset has been corrected
218 J. Du et al. / International Journal of Applied Earth Observation and Geoinformation 38 (2015) 216–228

0.20 Growing Season 0.15 Spring

0.19
NDVI

NDVI
0.18 0.13

y = 0.0003 x - 0.4292 y = 0.0001 x - 0.0978

0.17 R = 0.1902 R = 0.0257

0.16 0.11
1982 1987 1992 1997 2002 2007 2012 1982 1987 1992 1997 2002 2007 2012
Year Year

0.30 Summer 0.18 Autumn

y = 0.0003 x - 0.5232
R = 0.1756
0.28
NDVI
NDVI

0.16

0.26

y = 0.0005 x - 0.6665
R = 0.1883
0.24 0.14
1982 1987 1992 1997 2002 2007 2012 1982 1987 1992 1997 2002 2007 2012
Year Year

Fig. 2. Interannual variations in NDVI over Xinjiang during 1982–2012.

to minimize the effects of volcanic eruptions, sensor design differ-
ences between the AVHRR/2 and AVHRR/3 instruments, and thus
can be used to evaluate the long-term trends in vegetation activity
(Pinzon and Tucker, 2014). The long-term monthly mean temper-
ature and monthly precipitation data were acquired from China
Meteorological Data Sharing Service System. The locations of the
53 weather stations across Xinjiang are shown in Fig. 1. The data
on vegetation type were digitized from the 1:1,000,000 vegetation
map (Editorial Committee for Vegetation Map of China, 2001). Veg-
etation types were classified as follows: forest (1.52% area ratio, the
same as below), shrub (0.52%), grass (32.28%), crop (3.48%), desert
(37.87%) and non-vegetation (24.33%). The distribution of different
types of vegetation is shown in Fig. 1. Data on the consumption of
chemical fertilizer, effective irrigated area and cotton acreage were
collected from the Xinjiang Statistical Yearbook (Statistics Bureau
of Xinjiang Uygur Autonomous Region, 1989–2012, 2005).
The monthly GIMMS NDVI dataset was obtained using the Max-
imum Value Composite (MVC) method. In order to reduce the
influence of sparsely vegetated pixels on the NDVI trend, and
according to other researchers in this field (Mohammat et al., 2013;
Myneni et al., 1997; Zhao et al., 2011a,b), pixels with a mean grow-
ing season NDVI <0.05 were excluded in this study.
Methods
In order to gain insight into the contribution of each sea-
son to the change in growing season NDVI, the growing season
was divided into three seasons: spring (March–May), summer
(June–August) and autumn (September–November). NDVI of the
growing season and different seasons were calculated as the aver-
age of NDVI during the corresponding months.
To detect trends in NDVI and climate, a least-squares linear
regression was applied. In order to characterize the moisture con-
ditions in Xinjiang, the reference crop evapotranspiration (ET0 ) and
a humidity index (HI, defined as the ratio of precipitation to ET0 )
were calculated. Evapotranspiration is the only term that appears
in both a water balance equation and a land surface energy flow
(Gao et al., 2007; Xu and Singh, 2005), and is closely related to
the water use efficiency of ecosystems, particularly in arid regions
(Gao et al., 2007; Zhao et al., 2011b). The ET0 was calculated using
Thornthwaite method according to Zhao et al. (2011b).
To further explore the climatic factors driving NDVI change,
inter-annual correlations between NDVI and climatic variables
(temperature, precipitation, ET0 and HI) were calculated using the
Pearson correlation coefficient. Similar to Zhao et al., (2011b), we
used the ordinary Kriging method to generate a temperature and
precipitation map with the same resolution and geographic coor-
dinate system as those of the GIMMS NDVI3g dataset. We also
calculated the percentage area that showed positive or negative
trends in the fourteen nested time series. Significance tests of
trends in vegetation dynamics and correlations between NDVI and
climate were conducted using the correlation coefficient of the lin-
ear regression. Based on the results of the significance tests, trends
or correlations were classified according to three ranks: promi-
nent or highly significant (P < 0.01), significant (P < 0.05), and not
significant (P > 0.05).
 J. Du et al. / International Journal of Applied Earth Observation and Geoinformation 38 (2015) 216–228 219

16 Spatial patterns of NDVI trend

Growing
During 1982–2012, areas characterized by significantly
Spring
decreased NDVI mainly concentrated in Junggar basin, east and
12 Summer south of Tarim basin, and northeast of Xinjiang; those by sig-
NDVI 1000 / year

Autumn nificantly increased NDVI mainly distributed in the central and

solid symbol for P < 0.05
8 NDVI change during other periods were similar to that during
4 The size of areas with positive trends were commonly larger
0
1 3 5 7 9 11 13 15
Periods
Fig. 3. Variations of seasonal NDVI interannual slope during 14 period.
Results
NDVI changes
Changes in NDVI at a regional scale
In growing season and three seasons, there was a positive
increase in NDVI with fluctuations during fourteen periods (Fig. 2).
In particular, this increase was significant in autumn for fourteen
periods, in summer and growing season for most periods, and in
spring only for the first four periods (Fig. 3). The rates of increase in
summer was largest during most periods, followed by those usu-
ally in autumn. As the NDVI record grew in length, the rates of
increase showed a highly significantly decrease (growing season:
R2 = 0.90, spring: R2 = 0.87, summer: R2 = 0.80, autumn: R2 = 0.93,
P < 0.01, n = 14, the same order as below) in growing season and
three seasons.
Scale of vegetation type
In growing season and autumn, NDVI showed significant
increase for forest, shrub, grassland, and cropland during most
periods, and for desert during the first seven periods. In spring,
significant NDVI increases were found only for shrub and grass-
land during most periods, and decreases for desert during the last
five periods. In summer, NDVI displayed significant increase for
cropland and shrub during all periods, and for forest, grassland
and desert only during part periods (Table 1). The amount of NDVI
change in five biomes decreased high-significantly as the lengths
of the periods increased.
northern Xinjiang, and west of Tarim basin. Spatial patterns of
1982–2012, but areas characterized by significantly change in
NDVI were general small (Fig. 4).
than those with negative trends (Table 2) for growing season and
three seasons across all periods, especially during the first few peri-
ods. At the 0.05 significance level, areas that showed increased
NDVI were also markedly larger than those with decreased NDVI
for growing season and three seasons during the first nine periods.
Areas that showed positive trends of NDVI highly significant
decreased over fourteen periods in growing season and all sea-
sons (R2 = 0.91, R2 = 0.84, R2 = 0.81, R2 = 0.94, n = 14). Areas with
significant increases in NDVI decreased over fourteen periods for
growing season and three seasons, but significantly decrease only
for growing season (R2 = 0.34, n = 14). However, areas characterized
by significant decrease in NDVI in growing season and three sea-
sons significantly increased (R2 = 0.96, R2 = 0.95, R2 = 0.97, R2 = 0.93,
n = 14) over fourteen periods.
Correlations between NDVI and climatic drivers
Effects of climate change on NDVI at a regional scale
The differences in correlations between regional average NDVI
and contemporary temperature, precipitation, ET0 and HI were
obvious across seasons. NDVI in the growing season were signifi-
cantly correlated with precipitation in all fourteen periods analyzed
here (Table 3), and with temperature, ET and HI only in part peri-
ods. In spring, positive correlations between NDVI and temperature
and between NDVI and ET were significant in fourteen periods, and
those between NDVI and precipitation and between NDVI and HI
were negatively non-significant. NDVI in the summer were signif-
icantly correlated with precipitation and HI, and non-significantly
correlated with temperature and ET in all periods. Positive correla-
tions between NDVI and temperature in autumn were significant
only in part periods, and those between NDVI and other climatic
factors were relatively weak in all periods.
In growing season, NDVI were significantly positively cor-
related with four climatic factors (Table 4) for grassland and
cropland, with precipitation and HI for desert, with temperature
and ET for forest, with temperature, precipitation and ET for shrub

Table 1
Statistics of trends in growing and seasonal NDVI for different biomes in Xinjiang (10−4 /yr).

Growing season Spring

1982–1999 1982–2003 1982–2008 1982–2012 1982–1999 1982–2003 1982–2008 1982–2012

Desert 6.42* 4.20* 0.56 0.45 3.95 3.00 −0.27 −0.82

Grassland 12.52** 8.88** 4.64* 4.70** 8.37* 6.84* 4.71* 3.80*
Forest 15.64** 10.63** 6.41** 6.54** 12.57 7.43 6.53 5.04
Cropland 30.39** 20.10** 12.84** 14.36** 13.12 7.10 1.32 0.19
Shrub 11.80** 8.34** 4.66** 5.01** 7.90* 6.02 2.46** 0.60**

Summer Autumn

1982–1999 1982–2003 1982–2008 1982–2012 1982–1999 1982–2003 1982–2008 1982–2012

Desert 9.36* 4.79 0.55 1.07 5.94* 4.80** 1.41 1.10

Grassland 15.09** 9.66** 4.37 6.15** 14.11** 10.14** 4.84* 4.16*
Forest 12.30* 7.23 2.60 6.03* 22.06** 17.22** 10.09* 8.54*
Cropland 49.48** 32.13** 21.55** 26.16** 28.56** 21.07** 15.66** 16.72**
Shrub 12.65** 7.91* 4.57** 6.85** 14.84** 11.1** 6.95** 7.57**
*
P < 0.05 (statistical significance of linear regression of Pearson correlation).
**
P < 0.01 (statistical significance of linear regression of Pearson correlation).
220 J. Du et al. / International Journal of Applied Earth Observation and Geoinformation 38 (2015) 216–228

Fig. 4. Spatial distribution of NDVI trends over Xinjiang during the period 1982–1999, 1982–2006 and 1982–2012.

Table 2
Area fraction of the NDVI displaying different statistical significance level with progressively longer time series since 1982(%). N, P and S stand for R < 0, R > 0, and statistical
significance of linear regression of Pearson correlation less 0.05, respectively.

Growing season Spring Summer Autumn

P SN SP P SN SP P SN SP P SN

1982–1999 83.44 0.95 31.07 72.02 1.17 12.37 77.33 0.96 23.46 81.77 0.91
1982–2000 84.28 0.86 32.13 80.42 0.64 18.07 75.49 1.03 23.63 80.90 0.90
1982–2001 83.61 1.03 33.21 83.35 0.59 22.21 72.61 1.26 20.68 79.34 1.21
1982–2002 80.59 1.80 33.41 80.48 1.26 24.11 68.45 2.89 20.63 76.77 1.71
1982–2003 75.91 3.31 30.16 72.36 2.57 19.74 65.33 4.22 20.04 73.72 2.38
1982–2004 70.40 5.83 27.10 66.31 4.20 18.48 61.10 6.72 19.32 67.63 3.62
1982–2005 66.21 9.02 27.34 60.59 7.08 17.38 57.03 9.04 20.09 64.77 4.90
1982–2006 61.91 12.42 25.05 56.47 10.64 17.03 53.83 11.84 19.56 59.81 6.80
1982–2007 60.10 14.98 25.08 54.07 13.83 17.48 53.06 13.74 17.96 58.09 9.45
1982–2008 56.17 18.04 23.36 51.75 17.93 18.07 50.94 17.51 17.26 52.93 12.80
1982–2009 55.60 19.36 24.57 51.03 19.62 19.21 51.30 19.24 18.20 51.11 14.38
1982–2010 57.89 19.33 27.56 50.08 21.55 18.64 54.90 18.23 21.82 54.16 14.27
1982–2011 57.18 21.55 29.09 50.76 23.98 20.63 55.32 20.93 24.67 51.15 18.37
1982–2012 55.63 23.80 29.61 49.17 26.39 20.76 53.02 24.04 25.25 51.63 20.66
 J. Du et al. / International Journal of Applied Earth Observation and Geoinformation 38 (2015) 216–228 221

Table 3
Correlation coefficients between NDVI and climatic factors during fourteen periods.

Growing Season Spring Summer Autumn

RNDVI-T RNDVI-P RNDVI-ET RNDVI-HI RNDVI-T RNDVI-P RNDVI-ET RNDVI-HI RNDVI-T RNDVI-P RNDVI-ET RNDVI-HI RNDVI-T RNDVI-P RNDVI-ET RNDVI-HI

1982–1999 0.39 0.48* 0.34 0.40 0.52* 0.10 0.51* 0.00 0.28 0.60** 0.27 0.55* 0.29 0.10 0.20 0.08
1982–2000 0.42 0.47* 0.38 0.38 0.61** −0.02 0.63** −0.13 0.31 0.60** 0.3 0.55* 0.3 0.08 0.22 0.06
1982–2001 0.44 0.45* 0.41 0.36 0.63** −0.06 0.64** −0.17 0.28 0.60** 0.26 0.55* 0.37 0.13 0.26 0.09
1982–2002 0.45* 0.46* 0.42 0.37 0.63** −0.05 0.64** −0.16 0.27 0.59** 0.25 0.54* 0.42 0.12 0.32 0.06
1982–2003 0.46* 0.43* 0.43* 0.34 0.64** −0.11 0.66** −0.21 0.27 0.59** 0.25 0.54* 0.42* 0.13 0.32 0.07
1982–2004 0.41* 0.42* 0.39 0.34 0.56** −0.12 0.58** −0.21 0.26 0.59** 0.24 0.54** 0.42* 0.13 0.32 0.07
1982–2005 0.40 0.42* 0.38 0.34 0.53** −0.13 0.56** −0.21 0.25 0.57** 0.24 0.52** 0.43* 0.12 0.34 0.05
1982–2006 0.31 0.44* 0.30 0.36 0.50* −0.12 0.53** -0.20 0.21 0.57** 0.2 0.53** 0.30 0.14 0.23 0.08
1982–2007 0.31 0.446 0.31 0.36 0.48* −0.12 0.51** −0.20 0.2 0.55** 0.19 0.51** 0.31 0.12 0.23 0.06
1982–2008 0.22 0.46* 0.20 0.39* 0.42* −0.11 0.45* −0.19 0.12 0.56** 0.11 0.53** 0.26 0.13 0.19 0.07
1982–2009 0.22 0.45* 0.21 0.38* 0.42* −0.11 0.45* −0.19 0.12 0.51** 0.1 0.48* 0.26 0.14 0.19 0.09
1982–2010 0.22 0.50** 0.22 0.43* 0.44* −0.16 0.46* −0.24 0.18 0.49** 0.17 0.466 0.31 0.26 0.25 0.18
1982–2011 0.24 0.49** 0.25 0.42* 0.44* −0.16 0.47** −0.24 0.25 0.49** 0.24 0.45* 0.29 0.26 0.23 0.18
1982–2012 0.24 0.49** 0.25 0.41* 0.43* −0.15 0.46** −0.24 0.24 0.49** 0.24 0.456 0.28 0.26 0.23 0.18

RNDVI-T : correlation coefficients between NDVI and temperature, RNDVI-P : NDVI and precipitation, RNDVI-ET : NDVI and ET, RNDVI-HI : NDVI, the same as below.
*
P < 0.05 (statistical significance of linear regression of Pearson correlation).
**
P < 0.01 (statistical significance of linear regression of Pearson correlation).

during 1982–2012. In spring, positive correlations between NDVI
and thermal factors (temperature and ET) were found for all veg-
etation types, but the significant correlations only for grassland,
forest and shrub; significant negative correlations between NDVI
and precipitation were exhibited for forest and between NDVI and
HI for forest and grassland. In summer, only cropland NDVI and all
climatic factors were significantly positively correlated, significant
positive correlations between NDVI and precipitation and between
NDVI and HI were found for grassland and shrub, but for forest
only correlation between NDVI and precipitation were significant
positive. In autumn, significant positive correlations between NDVI
and temperature and between NDVI and ET were found for for-
est, cropland and shrub, but for grassland only between NDVI and
temperature.
Spatial patterns of correlation between NDVI and climate
Areas with positive correlation between NDVI and tempera-
ture for growing season, spring and autumn were larger than 50%
across all periods, and significant positive correlations between
NDVI and temperature were found in more pixels for spring
than for growing season, summer and autumn during all peri-
ods (Table 5). These pixels are mainly distributed in the Tianshan
Mountains, Altai Mountains and the western and northern of
Tarim basin (Fig. 5). For growing season and all seasons, pix-
els with significant negative correlations between NDVI and
temperature increased high-significantly (R2 = 0.91, R2 = 0.72,
R2 = 0.94, R2 = 0.89, n = 14), and those with positive correla-
tion decreased high-significantly (R2 = 0.89, R2 = 0.67, R2 = 0.85,
R2 = 0.86, n = 14) as the study period grows in length. Pixels with
significant negative correlations between NDVI and temperature
in growing season and all seasons are mainly concentrated in the
Junggar basin, the eastern and southern of Tarim basin and the
northeastern Xinjiang. Comparing with the first periods, areas with
significant positive correlations between NDVI and temperature for
growing season and all seasons expanded somewhat, then those
with significant negative correlations rapid expanded.
The NDVI in growing season and summer is positively correlated
with precipitation in most of the study area (more than 70% of the
pixels), and significant positive correlations is observed in more
than 15% (Table 6) of the pixels during all periods, mostly in north-
ern and western Xinjiang (Fig. 6). For spring, NDVI significantly
and negatively correlated with precipitation in Altai Mountains and
Tianshan Mountains (Fig. 6).
Positive correlations between NDVI and ET in spring and autumn
were found in more than 70% of the study area during over-
whelming majority of periods, and areas with significantly positive
correlations between NDVI and ET for spring are more than
that for other seasons (Table 7). Similar to correlations between
NDVI and temperature, pixels with significant negative correlation
between NDVI and ET for growing season and all seasons increased

Table 4
Correlation coefficients between growing and seasonal NDVI and climatic factors for different biomes during 1982–2012.

Growing season Spring

RNDVI-T RNDVI-P RNDVI-ET RNDVI-HI RNDVI-T RNDVI-P RNDVI-ET RNDVI-HI

Desert −0.04 0.43* −0.02 0.41* 0.00 0.14 0.00 0.14

Grassland 0.38* 0.48** 0.38* 0.37* 0.67** −0.35 0.71** −0.48**
Forest 0.47** 0.28 0.466 0.15 0.71** −0.47** 0.78** −0.62**
Cropland 0.36* 0.476 0.36* 0.37* 0.24 0.03 0.24 −0.03
Shrub 0.43* 0.42* 0.46** 0.31 0.45* −0.26 0.49** −0.35

Summer Autumn

RNDVI-T RNDVI-P RNDVI-ET RNDVI-HI RNDVI-T RNDVI-P RNDVI-ET RNDVI-HI

Desert 0.05 0.33 0.04 0.32 −0.03 0.32 −0.04 0.31

Grassland 0.29 0.52** 0.29 0.47** 0.36* 0.21 0.30 0.12
Forest 0.23 0.36* 0.26 0.31 0.53** 0.03 0.45* −0.14
Cropland 0.38* 0.51** 0.36* 0.45* 0.49** 0.21 0.41* 0.08
Shrub 0.31 0.51** 0.31 0.46* 0.54** 0.26 0.51** 0.15
*
P < 0.05 (statistical significance of linear regression of Pearson correlation).
**
P < 0.01 (statistical significance of linear regression of Pearson correlation).
222 J. Du et al. / International Journal of Applied Earth Observation and Geoinformation 38 (2015) 216–228

Table 5
Area fraction (%) displaying different statistical significance levels of the correlation between NDVI and temperature with progressively longer time series since 1982.

Growing season Spring Summer Autumn

P SN SP P SN SP P SN SP P SN SP

1982–1999 68.66 0.97 10.92 67.43 1.90 14.66 59.85 2.09 5.40 63.10 2.29 8.80
1982–2000 69.45 0.98 12.64 76.53 0.49 19.82 59.96 2.47 6.56 62.40 2.32 9.42
1982–2001 69.15 1.32 15.36 79.78 0.48 23.20 56.71 3.23 5.78 63.88 3.06 10.80
1982–2002 66.86 2.33 16.23 79.71 0.51 25.21 52.90 4.40 5.94 62.93 3.03 12.98
1982–2003 66.78 1.86 16.48 82.16 0.36 25.60 52.57 4.44 5.99 62.25 3.35 13.42
1982–2004 62.19 2.81 15.13 75.64 0.77 23.46 50.53 5.00 5.51 60.70 3.12 12.74
1982–2005 58.96 4.67 15.63 70.82 1.89 22.24 48.38 5.74 5.67 59.86 4.24 13.93
1982–2006 54.56 8.95 13.11 67.68 2.78 21.49 45.23 7.36 5.53 53.39 7.43 11.70
1982–2007 53.51 12.40 13.04 62.89 5.31 20.57 43.71 8.80 5.49 53.59 8.43 12.53
1982–2008 49.67 17.54 12.30 58.14 10.15 21.24 41.03 12.29 5.77 51.25 10.50 12.52
1982–2009 49.74 18.22 12.84 57.95 11.88 21.44 40.95 11.47 4.91 51.14 10.89 12.08
1982–2010 50.13 18.04 12.82 58.39 9.05 21.22 42.87 10.56 5.91 53.45 11.30 14.20
1982–2011 51.22 19.35 13.51 58.61 7.40 20.61 44.63 13.30 7.66 51.37 14.69 14.28
1982–2012 50.99 20.10 13.13 58.09 9.34 20.46 43.40 16.14 8.45 50.97 11.30 11.90

Fig. 5. Spatial distributions of correlations between NDVI and temperature.

 J. Du et al. / International Journal of Applied Earth Observation and Geoinformation 38 (2015) 216–228 223

Table 6
Area fraction (%) displaying different statistical significance levels of the correlation between NDVI and precipitation with progressively longer time series since 1982.

Growing season Spring Summer Autumn

P SN SP P SN SP P SN SP P SN SP

1982–1999 75.98 0.6 17.19 61.26 2.95 4.83 72.18 1.13 17.08 44.35 3.45 2.18
1982–2000 75.67 0.65 17.04 55.49 3.45 2.93 72.83 1.22 17.7 43.89 3.41 2.04
1982–2001 75.19 0.8 17.25 50.66 4.17 2.35 73.71 1.35 17.51 45.75 4.04 2.48
1982–2002 74.61 0.93 18.59 48.11 4.82 2.12 72.98 1.47 17.29 45.36 4.07 2.44
1982–2003 71.62 1.74 18.16 45.12 5.48 1.81 72.45 1.72 17.26 44.97 4.99 2.69
1982–2004 72.55 1.4 17.92 44.22 5.26 1.31 73.21 1.11 17.66 46.21 3.9 2.86
1982–2005 70.56 1.92 17.68 43.48 6.35 1.41 71.49 1.35 17.46 45.25 3.46 2.6
1982–2006 72.94 1.34 18.28 43.96 5.25 1.42 73.36 1.02 17.94 47.55 3.11 3.03
1982–2007 71.09 1.34 17.71 44.15 4.08 1.41 70.9 1.11 17.66 45.08 3.44 2.98
1982–2008 72.81 1.05 18.57 45.56 3.95 1.51 73.21 0.74 17.2 46.23 3.39 3.24
1982–2009 73.1 0.96 17.72 43.65 4.17 1.49 73.3 0.6 16.16 45.06 3.92 3.63
1982–2010 74.08 0.99 21.45 41.86 5.54 1.88 72.11 0.85 15.86 52.21 2.99 6.7
1982–2011 74.88 0.74 21.03 41.7 5.25 1.67 72.9 0.82 15.5 54.32 2.46 6.32
1982–2012 74.43 0.59 19.76 44.62 6.11 1.31 71.43 0.9 15.1 52.72 2.31 5.1

Fig. 6. Spatial distributions of correlations between NDVI and precipitation.

224 J. Du et al. / International Journal of Applied Earth Observation and Geoinformation 38 (2015) 216–228

Table 7
Area fraction (%) displaying different statistical significance levels of the correlation between NDVI and ET with progressively longer time series since 1982.

Growing season Spring Summer Autumn

P SN SP P SN SP P SN SP P SN SP

1982–1999 63.99 1.52 8.48 69.84 0.63 14.46 60.12 2.08 5.53 61.77 2.65 7.54
1982–2000 66.62 1.54 10.77 80.24 0.1 20.9 60.28 2.37 6.97 60.76 2.52 7.64
1982–2001 66.24 2.04 12.93 82.29 0.21 27.12 56.29 3.06 5.88 60.47 3.19 8.07
1982–2002 64.21 2.93 12.52 81.45 0.27 28.38 52.44 4.24 5.95 60.75 2.98 9.4
1982–2003 64.03 2.69 12.55 84.38 0.1 28.49 52.02 4.19 5.87 60.07 3.41 9.03
1982–2004 59.57 3.64 10.81 77.7 0.38 25.84 49.89 4.82 5.49 59.35 3 8.86
1982–2005 56.56 5.29 10.92 74.51 0.73 25.02 47.76 5.6 5.65 57.78 3.87 10.31
1982–2006 52.55 7.78 9.63 71.23 1.08 24.21 45.15 7.13 5.43 51.53 6.49 8.93
1982–2007 51.2 10.65 10.1 64.74 3 22.56 43.67 8.58 5.47 52.1 6.5 8.63
1982–2008 46.47 16.01 9.74 59.34 6.89 22.18 40.64 12.08 5.62 50.39 7.65 8.95
1982–2009 46.63 16.77 10.14 59.25 8.23 22.5 40.74 11.11 4.45 50.25 8.56 8.63
1982–2010 47.02 16.95 10.35 59.7 6.47 22.46 42.47 10.26 5.42 53.07 9.16 10.52
1982–2011 48.47 19.5 11.22 59.64 7.41 22.46 44.13 12.66 7.07 50.93 12.37 10.84
1982–2012 47.93 21.08 11.52 59.11 10.03 22.49 42.92 15.43 7.92 50.61 10.64 8.95

Fig. 7. Spatial distributions of correlations between NDVI and ET.

 J. Du et al. / International Journal of Applied Earth Observation and Geoinformation 38 (2015) 216–228 225

Table 8
Area fraction (%) displaying different statistical significance levels of the correlation between NDVI and HI with progressively longer time series since 1982.

Growing season Spring Summer Autumn

P SN SP P SN SP P SN SP P SN SP

1982–1999 73.8 0.71 14.34 58.66 4.13 4.7 70.52 1.25 15.82 42.36 3.32 2.78
1982–2000 73.32 0.78 14.26 51.67 5.28 2.57 71.4 1.35 16.3 41.9 3.34 2.61
1982–2001 72.65 1.2 14.64 46.39 6.21 1.97 72.27 1.57 16.32 43.65 3.7 2.83
1982–2002 72.44 1.28 15.64 44.22 6.8 1.94 71.91 1.62 16.44 42.66 3.62 2.6
1982–2003 69.77 1.93 15.06 40.95 7.75 1.51 71.47 1.79 16.45 42.64 4.51 2.69
1982–2004 71.1 1.43 15.08 41.3 7.61 1.4 72.37 1.17 16.83 43.8 3.51 2.84
1982–2005 69.97 1.92 15.07 41.38 8.3 1.4 71.1 1.38 16.97 43.14 3.48 2.83
1982–2006 72.62 1.29 15.82 42.76 7.43 1.43 73.15 1.03 17.22 45.75 3.01 3.41
1982–2007 71.81 1.03 15.06 44.29 6.51 1.31 71.17 1.03 16.99 43.63 3.31 3.41
1982–2008 73.73 0.76 16.46 46.72 6.87 1.6 73.3 0.68 16.81 45.15 3.41 3.78
1982–2009 73.98 0.75 16.22 45.47 6.7 1.34 73.59 0.55 16.02 45.07 3.84 3.9
1982–2010 75.17 0.73 18.73 50.73 1.31 2.57 72.26 0.83 15.27 51.59 3.16 6.32
1982–2011 76.27 0.67 18.4 43.68 8.16 1.29 73.33 0.78 14.99 53.96 2.61 6.24
1982–2012 75.77 0.64 17.75 46.61 9.11 1.33 72.11 0.86 14.5 52.8 2.46 5.08

high-significantly (R2 = 0.92, R2 = 0.80, R2 = 0.93, R2 = 0.89, n = 14),
and those with positive correlation decreased high-significantly
(R2 = 0.88, R2 = 0.69, R2 = 0.85, R2 = 0.83, n = 14) as the study period
grows in length. Spatial patterns of correlations between NDVI and
ET are very similar to that between NDVI and temperature during
all periods (Fig. 7).
For growing season and summer, more than 70% and 15% of the
whole pixels were characterized by positive correlation between
NDVI and HI during all periods and significantly positive correlation
between NDVI and HI during most periods (Table 8), respec-
tively, whereas the relationships between NDVI and HI for spring
and autumn are relatively weak. Spatial patterns of correlations
between NDVI and HI are very similar to that between NDVI and
precipitation during all periods (Fig. 8).
Discussion
Analysis of the NDVI trend
Overall, the findings presented in this study indicated that NDVI
significantly increased in most seasons at the regional scale. This is
in agreement with the conclusions of NDVI turning green, resulting
from studies conducted in arid regions worldwide (Fensholt et al.,
2012), in Eurasia (Piao et al., 2011) and other areas (Dai et al., 2011;
Peng et al., 2012; Wang et al., 2011; Zhao et al., 2011a,b).
The growing season NDVI and NDVI in different seasons
increased significantly and continuously, suggesting that on the
time scale of 30 years vegetation growth in Xinjiang showed an
overall increasing tendency. This was especially true in summer
and autumn, when the rate of NDVI increase was relatively large.
The increase in NDVI during the growing season was mostly caused
by an increase in NDVI during summer and autumn.
Although the growing season NDVI increased highly signifi-
cantly for all five vegetation types except desert over the most
periods, the season for which the increase in NDVI occurred was
different for different vegetation types and for different periods.
Increase in cropland and grassland NDVI mainly occurred in sum-
mer and autumn, and the increase in summer were relatively large.
Increases in forest and shrub NDVI mainly occurred in autumn.
At a regional scale, the rates at which NDVI changed were grad-
ually diminished and tended to stabilize over the fourteen periods,
but at pixel scale, area coverage with significant negative trends
was significantly increased in growing season and all seasons,
which led to a significant decrease in the amount of NDVI change at
the regional scale. In addition, the phenomenon of the modifiable
temporal unit problem may partly contribute to reduction of NDVI
increment over periods at regional scale (Cheng and Adepeju, 2014;
de Jong and de Bruin, 2012). Although strictly speaking, changes
in NDVI in these regions with positive trend were not statisti-
cally significant, they also reflected the slowdown of the regional
NDVI increase. Thus, the regional average may mask or obscure the
phenomenon that at pixel scale. While there was an increase in
the size of areas that showed significant decrease in NDVI, there
was a significant NDVI increase on a regional scale as the time
period extended. Area that experienced severe NDVI change is the
object where realistic management needs to focus. The long-term
and consistent NDVI datasets offer a cheap, verifiable, viable way
to quickly identify areas concerning dramatic vegetation change,
supporting managers in their effort to design and apply adaptive
management strategies.
The spatial patterns of areas with significant NDVI changes in
the growing season, summer and autumn were mostly similar, and
the differences were mainly in the size of the range and the extent
of the change. In general, among all seasons areas with significant
increase in NDVI in growing season were the largest, and areas with
significant decrease in summer or spring were the largest. Growing
season, summer and autumn NDVI increased significantly mainly
in cropland and the spatial extent of cropland area with significant
NDVI increase was notably larger than the size of areas with crops
in the vegetation type map (Figs. 1 and 4). Areas in which vege-
tation coverage improved to a relatively large degree were mainly
located in agricultural areas in oases and surrounding regions and
to a lesser extent this also occurred in a few scattered areas in
some mountainous regions, consistent with results of related stud-
ies (Wang et al., 2011; Zhao et al., 2011b). Meanwhile, as the time
period extended the range of areas showing significant increases
in NDVI expanded notably to surrounding areas with oases at the
centre. This suggests that artificial oases formed by agricultural
development contributed to this expansion (Piao et al., 2003).
The response of NDVI to climate change and human activity
Although studies have suggested that NDVI change in arid and
semi-arid regions are mainly affected by growing season precip-
itation and ET0 and that correlation of NDVI with temperature is
relatively weak (Nemani et al., 2003; Zhao et al., 2011a,b), the
results of the present study indicated that vegetation growth in
Xinjiang was limited by the combination of moisture and ther-
mal conditions. This was verified by the significant correlations
between seasonal NDVI and temperature, precipitation, ET0 , and
HI at regional and pixel scale. The difference between our results
and the conclusions of related studies was mainly caused by the
difference in the time period studied. Trend analyses based on a
finite time series often leads to different outcomes due to differ-
ences in starting points, ending points and inflection points (de Jong
et al., 2012; Wessels et al., 2007, 2012). Extending the time series as
226 J. Du et al. / International Journal of Applied Earth Observation and Geoinformation 38 (2015) 216–228
Fig. 8. Spatial distributions of correlations between NDVI and HI.
much as possible and focusing on the process of change are partic-
ularly important in studies that monitor vegetation dynamics and
its relationship with climate change.
The responses of vegetation to climate varied seasonally; this
was broadly consistent with results from studies conducted in
Eurasia (Piao et al., 2011) and central Asia (Mohammat et al., 2013).
In spring and autumn, NDVI was more highly correlated with tem-
perature and ET0 than with precipitation and HI, whereas the
opposite was observed in summer. This reflected the sensitivity
of vegetation growth to heat and indicated that temperature was
the main climatic constraint to vegetation growth at beginning and
end of the growing season, while precipitation was most limiting
in the middle of growing season.
The findings that NDVI was significantly positively correlated
with temperature in spring is consistent with results obtained in
studies conducted in Asian inland areas (Mohammat et al., 2013).
Areas with significantly positive correlations between NDVI and
temperature were localized in the Altai Mountains, the Tianshan
Mountains and the Kunlun Mountains. This suggests that at high
elevation, the effect of temperature on vegetation growth was
greater than at low elevation. The spatial distribution of areas with
positive correlations between NDVI and temperature for autumn
was similar to that for spring, but was not as clear.
The vegetation type where the NDVI in spring significantly
decreased was mainly cropland, which was distributed in the
downstream region of the Manas river (area 1), the Aksu river, and
the intersection of the Tarim and Hotan rivers (area 2). Except for
spring which NDVI displayed a significant decrease, NDVI in these
areas exhibited significant increases in other seasons. Consider-
ing that NDVI in spring surrounding these areas mostly showed
a significant increase, human activities may be one of the driving
factors of a decrease in NDVI. Changes in crop planting structure
may cause a significant decrease in NDVI in some farmlands over
these periods. In recent years, cotton acreage in Xinjiang has rapidly
expanded (Fan et al., 2013). Maximum cotton growth occurs in
July, which is later than maize, wheat and other crops (Cao et al.,
2008). Therefore, NDVI in cotton farmland in spring was signifi-
cantly lower than that in farmland planted other crops (Hao et al.,
 J. Du et al. / International Journal of Applied Earth Observation and Geoinformation 38 (2015) 216–228
2012). During 1982–2012, the proportion of cotton acreage to total
crop acreage showed significant negative correlations with farm-
land NDVI in spring in above two areas (R2 = 0.20, P < 0.05, n = 31;
R2 = 0.32, P < 0.01, n = 30, respectively). In addition, after 2000, drip
irrigation in oasis agricultural areas of Xinjiang largely increased,
and areas that previously utilized flood and furrow irrigation were
changed to the current drip irrigation under film mulch (Fan et al.,
2013). After film mulching in spring, seedlings of other weeds can
hardly penetrate the mulch and sprout successfully. At this point
the leaves of the crop seedlings have not fully covered the ground,
and as a result, pixels that mix the mulch, bare land and crop
seedlings show small NDVI values.
Although cropland NDVI is closely related to climatic fac-
tors in the growing season, summer and autumn, the impact of
human activity cannot be ignored. In 1982–2012, cropland NDVI
showed a highly significant positive correlations with both the
amount of chemical fertilizers used and the effective irrigation
area (R2 = 0.49, P < 0.01, n = 31; R2 = 0.42, P < 0.01, n = 31) and these
correlations were stronger than that between NDVI and climatic
factors. This suggests that NDVI in irrigated agricultural areas in
an arid environment strongly depends on human agricultural pro-
duction activities. The significant increase in NDVI in the Manas
River basin of northern Xinjiang was related to the continuous
increase in the spatial extent of irrigated areas in this region (Fan
et al., 2013). Over the past 31 years, annual precipitation obvi-
ously decreased in the Junggar basin, for example, in Beitashan
climate station with a rate of −4.28 mm/10 a and in Qijiaojing cli-
mate station with −8.82 mm/10 a. There was also slightly reduced
precipitation in Qitai, Fuhai, Tulufan climate stations, and so on.
The most prominent warming trend is found in the eastern Junggar
basin, which along with the concurrent decrease in precipitation,
may strengthen the drought stress already imposed in this region.
Consequently, NDVI decrease in these regions due to drought stress
strengthened by reduced precipitation and warming.
The long-term NDVI time series data provides a powerful tool to
understand vegetation growth history, monitor current condition
and prepare for the future change (Piao et al., 2011; van Leeuwen
et al., 2006). Although the method of linear regression character-
ized by simple, robust, and its results are easy to interpret (Fensholt
et al., 2009; Peng et al., 2012), the overall linear trend cannot accu-
rately show actual temporal patterns of vegetation growth change
during the whole study period (de Jong et al., 2011, 2012; Piao et al.,
2011). For instance, summer NDVI in study area exhibited com-
bined greening and browning trends during 1982–2012, which was
averaged out by simple linear trend analysis during whole period.
Because of time lag effects of vegetation growth to climate change
(Yang and Piao, 2006; Yu et al., 2012), it was difficult to simu-
late future vegetation change using general ecological models (Yu
et al., 2012). Investigation spatio-temporal changes of vegetation
and its relationship with climatic factors during multiple nested
periods is likely becoming one of methods effectively addressed
this problems in the study on vegetation change and projection.
Over progressively longer periods since beginning year of the study,
detecting the changes of vegetation growth and exploring the trend
in the annual variation rates of NDVI during each period maybe
provide the future trajectory of vegetation change. It can thus be
predict that the magnitude of NDVI increase for growing season
and all seasons will further reduce, particularly for spring. At pixel
scale, area with significant decrease in NDVI mostly will succes-
sively expand in the future.
Conclusions
Over three decades, vegetation activity in Xinjiang showed
an overall increasing tendency. The rate of the NDVI increase
227
gradually decreased over fourteen periods for growing season and
all seasons, particularly for spring. The slowing of the NDVI increase
across fourteen periods at a regional scale resulted from a rapid
increase in areas where there was a significant NDVI reduction.
Investigations into changes in NDVI need to be limited to specific
time periods, and the results may vary greatly or even reverse dur-
ing different time periods. Hence, in this type of study the length of
the study period should be expanded as much as possible and the
course of the change should be explored in depth.
Vegetation growth was regulated by hydrothermal conditions,
and there were obvious differences in the sensitivity of vegetation
growth to water and heat across different seasons. In spring and
autumn, temperature is the dominant limiting factor for vegeta-
tion growth in most regions of Xinjiang, and in summer, vegetation
was mainly constrained by water. NDVI decrease in eastern Junggar
basin due to drought stress strengthened by reduced precipitation
and warming. Meanwhile, change in NDVI is significantly affected
by human activities. The extensive use of fertilizers and increased
spatial scale of farmland irrigation resulted in increased vegetation
cover of cropland. In addition to climatic factors, the rapid increase
in the proportion of cotton cultivation and the gradual promotion
of drip irrigation under film mulch is an important driver in the
reduction of NDVI observed in some of the farmlands in spring.
In the researches focusing on long time sequence of NDVI, trend
analysis during the multiple nested time series may contribute to a
better and deep understanding of NDVI changes and foreseeing the
future trajectory. So, the magnitude of NDVI increase for growing
season and all seasons will further reduce in the future, but area
with significantly NDVI decrease mostly will significantly increase
for growing season and three seasons.
Acknowledgements
This study was supported by the National Natural Science Foun-
dation of China (grant 41001055), the Special Fund of Environment
Protection Research in the Public Interest (grant 201209027-5) and
the Fundamental Research Funds for the Central Public-interest
Scientific Institution (grant 2009KYYW10 and 2012-YSKY-13). The
authors thank the NASA Global Inventory Modeling and Map-
ping Studies (GIMMS) group for producing and sharing the AVHRR
GIMMS NDVI datasets. NASA/MODIS Land Discipline Group is
thanked for sharing the MODIS LAND data.
References
IPCC, 2007. Climate Change 2007. The Physical Science Basis. Cambridge University
Press, Cambridge, UK.
Alcaraz-Segura, D., Liras, E., Tabik, S., et al., 2010. 1–1999 NDVI trends in the
Iberian Peninsula across four time-series derived from the AVHRR sensor:
LTDR, GIMMS, FASIR, and PAL-II. Sensor 10, 1291–1314.
Bai, Z., Dent, D., 2009. Recent land degradation and improvement in China. AMBIO:
J. Hum. Environ. 38, 150–156.
Beck, H.E., McVicar, T.R., van Dijk, A.I.J.M., et al., 2011. Global evaluation of four
AVHRR–NDVI data sets: Intercomparison and assessment against Landsat
imagery. Remote Sens. Environ. 115, 2547–2563.
Beck, P.S.A., Goetz, S.J., 2011. Satellite observations of high northern latitude
vegetation productivity changes between 1982 and 2008: ecological variability
and regional differences. Environ. Res. Lett. 6, 5501–5511.
Buyantuyev, A., Wu, J., 2009. Urbanization alters spatiotemporal patterns of
ecosystem primary production: a case study of the Phoenix metropolitan
region USA. J. Arid. Environ. 73, 512–520.
Cao, W., Liu, J., Ma, R., 2008. Regional planning of Xinjiang cotton growing areas for
monitoring and recognition using remote sensi. Trans. Chin. Soc. Agric. Eng. 24,
172–176.
Cheng, T., Adepeju, M., 2014. Modifiable temporal unit problem (MTUP) and its
effect on space-time cluster detection. PLoS ONE 9, e100465.
Dai, S., Zhang b, Wang, H., et al., 2011. Vegetation cover change and the driving
factors over northwest China. J. Arid. Land 3, 25–33.
de Jong, R., de Bruin, S., 2012. Linear trends in seasonal vegetation time series and
the modifiable temporal unit problem. Biogeosciences 9, 71–77.
de Jong, R., de Bruin, S., Schaepman, M., et al., 2011. Quantitative mapping of global
land degradation using Earth observations. Int. J. Remote Sens. 32, 6823–6853.
228 J. Du et al. / International Journal of Applied Earth Observation and Geoinformation 38 (2015) 216–228
de Jong, R., Verbesselt, J., Schaepman, M.E., et al., 2012. Trend changes in global
greening and browning: contribution of short-term trends to longer-term
change. Global Change Biol. 18, 642–655.
Editorial Committee for Vegetation Map of China, 2001. Vegetation Atlas of China.
Science Press, Beijing, China.
Fan, Z., Wu, S., Wu, Y., et al., 2013. The land reclamation in Xinjiang since the
founding of new China. J. Nat. Res. 28, 713–720.
Fensholt, R., Langanke, T., Rasmussen, K., et al., 2012. Greenness in semi-arid areas
across the globe 1981–2007 — an earth observing satellite based analysis of
trends and drivers. Remote Sens. Environ 121, 144–158.
Fensholt, R., Rasmussen, K., 2011. Analysis of trends in the Sahelian ‘rain-use
efficiency’ using GIMMS NDVI: RFE and GPCP rainfall data. Remote Sens.
Environ. 115, 438–451.
Fensholt, R., Rasmussen, K., Nielsen, T.T., et al., 2009. Evaluation of earth
observation based long term vegetation trends — intercomparing NDVI time
series trend analysis consistency of Sahel from AVHRR GIMMS, Terra MODIS
and SPOT VGT data. Remote Sens. Environ. 113, 1886–1898.
Gao, G., Chen, D., Xu C-y, et al., 2007. Trend of estimated actual evapotranspiration
over China during 1960–2002. J. Geophys. Res.: Atmos. 112, D11120.
Goetz, S.J., Bunn, A.G., Fiske, G.J., et al., 2005. Satellite-observed photosynthetic
trends across boreal North America associated with climate and fire
disturbance. Proc. Natl. Acad. Sci. 102, 13521–13525.
Hao, P., Niu, Z., Wang, L., et al., 2012. Multi-source automatic crop pattern mapping
based on historical vegetation index profiles. Trans. Chin. Soc. Agric. Eng. 28,
123–131.
Li, A., Wu, J., Huang, J., 2012. Distinguishing between human-induced and
climate-driven vegetation changes: a critical application of RESTREND in inner
Mongolia. Landsc. Ecol. 27, 969–982.
Mohammat, A., Wang, X., Xu, X., et al., 2013. Drought and spring cooling induced
recent decrease in vegetation growth in inner Asia. Agric. For. Meteorol.
178-179, 21–30.
Myneni, R.B., Keeling, C.D., Tucker, C.J., et al., 1997. Increased plant growth in the
northern high latitudes from 1981 to 1991. Nature 386, 698–702.
Nemani, R.R., Keeling, C.D., Hashimoto, H., et al., 2003. Climate-driven increases in
global terrestrial net primary production from 1982 to 1999. Science 300,
1560–1563.
Peng, J., Liu, Z., Liu, Y., et al., 2012. Trend analysis of vegetation dynamics in
Qinghai–Tibet plateau using Hurst exponent. Ecol. Indic. 14, 28–39.
Peng, S., Chen, A., Xu, L., et al., 2011. Recent change of vegetation growth trend in
China. Environ. Res. Lett. 6, 4027–4029.
Pettorelli, N., Chauvenet, A.L.M., Duffy, J.P., et al., 2012. Tracking the effect of
climate change on ecosystem functioning using protected areas: Africa as a
case study. Ecol. Indic. 20, 269–276.
Piao, S., Fang, J., Zhou, L., et al., 2003. Interannual variations of monthly and
seasonal normalized difference vegetation index (NDVI) in China from 1982 to
1999. J. Geophys. Res. Atmos. 108, 4401.
Piao, S., Wang, X., Ciais, P., et al., 2006. Changes in satellite-derived vegetation
growth trend in temperate and boreal Eurasia from 1982 to 2006. Global
Change Biol. 17, 3228–3239.
Pinzon, J., Tucker, C., 2014. A non-stationary 1981–2012 AVHRR NDVI3g time
series. Remote Sens. 6, 6929–6960.
Qi, Q., Wang, T., Kou, X., et al., 2009. Temporal and spatial changes of vegetation
cover and the relationship with precipitation in Jinghe watershed of China.
Chin. J. Plant Ecol. 33, 253–264.
Statistics Bureau of Xinjiang Uygur Autonomous Region, 1989–2012. Xinjiang
Statistical Yearbook. China Statistics Press, Beijing, China.
Statistics Bureau of Xinjiang Uygur Autonomous Region, 2005. Xinjiang Five
Decades Yearbook. China Statistics Press, Beijing, China.
Ren, J., Liu, H., Yin, Y., et al., 2007. Drivers of greening trend across vertically
distributed biomes in temperate arid Asia. Geophys. Res. Lett. 34, L07707.
van Leeuwen, W.J.D., Orr, B.J., Marsh, S.E., et al., 2006. Multi-sensor NDVI data
continuity: uncertainties and implications for vegetation monitoring
applications. Remote Sens. Environ. 100, 67–81.
Wang, H., Chen, Y., Xun, S., et al., 2013. Changes in daily climate extremes in the
arid area of northwestern China. Theor. Appl. Climatol. 112, 15–28.
Wang, J., Dong, J., Liu, J., et al., 2014. Comparison of gross primary productivity
derived from GIMMS NDVI3g, GIMMS, and MODIS in Southeast Asia. Remote
Sens. 6, 2108–2133.
Wang, Z., Shi, Q., Wang, T., et al., 2006. Spatial-temporal characteristics of
vegetation cover change in mountain-oasis-desert system of Xinjiang from
1982 to 2006. J. Nat. Res. 26, 609–618.
Wessels, K.J., Prince, S.D., Malherbe, J., et al., 2007. Can human-induced land
degradation be distinguished from the effects of rainfall variability? A case
study in South Africa. J. Arid Environ. 68, 271–297.
Wessels, K.J., van den Bergh, F., Scholes, R.J., 2012. Limits to detectability of land
degradation by trend analysis of vegetation index data. Remote Sens. Environ.
125, 10–22.
Wu, Z., Zhang, H., Krause, C., et al., 2010. Climate change and human activities: a
case study in Xinjiang, China. Clim. Change 99, 457–472.
Xu, C.Y., Singh, V.P., 2005. Evaluation of three complementary relationship
evapotranspiration models by water balance approach to estimate actual
regional evapotranspiration in different climatic regions. J. Hydrol. 308,
105–121.
Yang, Y., Piao, S., 2006. Variations in grassland vegetation cover in relation to
climatic factors on the Tibetan Plateau. J. Plant Ecol. 30, 1–8.
Yu, H., Xu, J., Okuto, E., et al, 2012. Seasonal Response of Grasslands to Climate
Change on the Tibetan Plateau. PLoS ONE 7, e49230.
Zeng, F.-W., Collatz, G., Pinzon, J., et al., 2013. Evaluating and quantifying the
climate-driven interannual variability in global inventory modeling and
mapping studies (GIMMS) normalized difference vegetation index (NDVI3g) at
global scales. Remote Sens. 5, 3918–3950.
Zhao, X., Tan, K., Fang, J., 2011a. NDVI-based international and seasonal variations
of vegetation activity in Xinjiang during the period of 1982–2006. Arid. Zone
Res. 28, 10–16.
Zhao, X., Tan, K., Zhao, S., et al., 2011b. Changing climate affects vegetation growth
in the arid region of the northwestern China. J. Arid. Environ. 75, 946–952.