See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/309921339

Drought trends based on the VCI and its correlation with climate factors in the
agricultural areas of China from 1982 to 2010

Article in Environmental Monitoring and Assessment · October 2016

DOI: 10.1007/s10661-016-5657-9

CITATIONS READS

46 612

8 authors, including:

Xiaojin Qian Liang Liang

Nanjing University of Posts and Telecommunications Jiangsu University
14 PUBLICATIONS 124 CITATIONS 36 PUBLICATIONS 662 CITATIONS

SEE PROFILE SEE PROFILE

Qiu Shen Zhixiao Liu

Beijing Normal University Jishou University
18 PUBLICATIONS 245 CITATIONS 7 PUBLICATIONS 189 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Estimation of Crop Residue Cover using Multisource Remote Sensing Data View project

Climate change and its impact on food security in China View project

All content following this page was uploaded by Xiaojin Qian on 06 January 2019.

The user has requested enhancement of the downloaded file.

Environ Monit Assess (2016) 188:639
DOI 10.1007/s10661-016-5657-9

Drought trends based on the VCI and its correlation

with climate factors in the agricultural areas of China
from 1982 to 2010
Xiaojin Qian & Liang Liang & Qiu Shen & Qin Sun &
Lianpeng Zhang & Zhixiao Liu & Shuhe Zhao &
Zhihao Qin

Received: 19 April 2016 / Accepted: 20 October 2016

# Springer International Publishing Switzerland 2016

Abstract Drought is a type of natural disaster that has the VCI increased in the majority of agricultural areas,
the most significant impacts on agriculture. Regional indicating that the drought frequency decreased over
drought monitoring based on remote sensing has be- time, and the decreasing trend in the southern region
come popular due to the development of remote sensing was more notable than that in the northern region. A
technology. In this study, vegetation condition index correlation analysis showed that temperature and wind
(VCI) data recorded from 1982 to 2010 in agricultural velocity were the main factors that influenced drought in
areas of China were obtained from advanced very high the agricultural areas of China. From a regional perspec-
resolution radiometer (AVHRR) data, and the temporal tive, excluding precipitation, the climate factors had
and spatial variations in each drought were analyzed. various effects on drought in different regions.
The relationships between drought and climate factors However, the correlation between the VCI and precipi-
were also analyzed. The results showed that from 1982 tation was low, possibly due to the widespread use of
to 2010, the agricultural areas that experienced frequent artificial irrigation technology, which reduces the reli-
and severe droughts were mainly concentrated in the ance of agricultural areas on precipitation.
northwestern areas and Huang-Huai Plain. Moreover,
Keywords Agricultural area . Drought . VCI . Climate
Electronic supplementary material The online version of this factor . Correlation coefficient
article (doi:10.1007/s10661-016-5657-9) contains supplementary
material, which is available to authorized users.

X. Qian : L. Liang (*) : Q. Shen : Q. Sun : L. Zhang Introduction

School of Geodesy and Geomatics, Jiangsu Normal University,
Xuzhou, China Drought is one of the most costly types of disasters and
e-mail: liangliang198119@163.com
can affect natural habitats, ecosystems, agricultural sys-
L. Liang : S. Zhao : Z. Qin tems, and urban water supplies (Jiao et al. 2016). Nearly
School of Geographic and Oceanographic Sciences, Nanjing half of the countries in the world have suffered from
University, Nanjing, China drought (Field et al. 2012; Bokal et al. 2014; Masud
X. Qian et al. 2015). Therefore, it is critical to monitor and issue
Key Laboratory of Digital Earth Science, Institute of Remote warnings about droughts. Agricultural drought refers to
Sensing and Digital Earth, Chinese Academy of Science, Beijing, a lack of moisture for crops due to external environmen-
China tal factors. Droughts impact normal crop growth and can
Z. Liu cause crop failure (Wilhite 2000). Agriculture is the
College of Biology and Environment Science, Jishou University, foundation of China’s economy, and climate change is
Jishou, China projected to directly influence food security and
639 Page 2 of 13 Environ Monit Assess (2016) 188:639

sustainable development. Moreover, agriculture is se- using the TVDI based on moderate-resolution imaging
verely constrained by the climate and weather (Dai spectroradiometer (MODIS) data and indicated that a
2013). Therefore, studies regarding agriculture drought large portion of the study area was humid from May to
and its relationship with climate factors have become September 2005, drought occurred in the western part
popular among governments and researchers of the study area in August, and the eastern part of the
worldwide. study area was humid in June. Additionally, the
Currently, the common methods of agriculture approach accurately predicted the agriculture drought
drought monitoring can be divided into site-based and conditions in the study area during the crop growth
remote sensing-based indices. Site-based indices in- season in 2005. Dutta et al. (2015) attempted to iden-
clude the standardized precipitation index (SPI), tify the spatiotemporal extent of agricultural drought
Palmer drought severity index (PDSI), and crop mois- over Rajasthan using VCI and assessed the
ture index (CMI). Remote sensing-based indices are performance of VCI by comparing the estimates to
divided into bare surface indices (e.g., thermal inertia meteorological drought indicators. They found that
and microwave moisture inversion) and vegetation cov- VCI was useful for monitoring the onset, duration,
er indices (Liu et al. 2016). Site-based indices can be and spatiotemporal extent of agricultural drought. In
established based on precipitation, temperature, evapo- addition, many scholars have performed annual
ration, runoff, and other meteorological and hydrologi- analyses of regional drought in recent years. For
cal data from observation stations. However, it is diffi- example, Ma and Fu (2006) analyzed annual drought
cult to continuously and dynamically monitor a wide in northern China using the surface humidity index,
range of drought areas using this method due to the the PDSI, and inverted soil moisture and revealed the
limited spatial density of observation stations (Wang temporal and spatial variations of drought in the region
et al. 2012a; Zhou et al. 2014; Guo and Wang 2015; over the past 54 years. He et al. (2011) analyzed the
Hao et al. 2015). temporal and spatial characteristics of the frequency of
With the development of remote sensing technolo- extreme drought on inter-annual and inter-decadal
gy, it is possible to access extensive, multi-scale, and scales and during summer and winter monsoons
multi-temporal integrated surface information in real based on daily data from 108 meteorological stations
time and monitor a wide range of drought conditions in southwestern China from 1960 to 2009. Liang et al.
(Wang et al. 2012b; Zhao 2013). A series of drought (2014) analyzed the annual drought trends in China
monitoring indices based on remote sensing data have using the TVDI, which was calculated using MODIS
been proposed, including the normalized difference data from 2001 to 2010, and drew conclusions regard-
vegetation index (NDVI) (Rouse et al. 1974), the ing its relationship to climate factors. These studies
temperature vegetation drought index (TVDI) provide a framework for agricultural drought monitor-
(Sandholt et al. 2002), and the vegetation condition ing and annual analyses of drought. However, the
index (VCI) (Kogan 1995). Scholars have conducted relationships between drought and climate factors must
numerous studies of the temporal and spatial be investigated further and better understood, especial-
distributions of regional droughts that affect crops ly in agricultural areas.
based on these drought indices. For example, In the indices based on vegetation cover, NDVI can
Ozelkan et al. (2016) estimated drought, which was only reflect the climate, soil, and thermal impacts on
expressed by SPI in a rain-fed and irrigated agricultur- vegetation in an isolated and qualitative manner. From a
al area in the heart of the Southeast Anatolia, Turkey, macroscopic perspective, it is difficult to establish a
using Landsat 8 Operational Land Imager (OLI) NDVI unified and quantitative standard for assessing the
data. The results showed positive correlations between growth conditions of crops; thus, VCI was proposed
SPI and NDVI over rain-fed areas and negative by Kogan. It includes the real-time NDVI information
correlations over irrigated agricultural areas, and the and historical information regarding NDVI change.
best correlations between spring SPI and summer Moreover, this method can decrease or eliminate the
NDVI reflected the critical influence of spring impacts on NDVI associated with different geographical
watering on crop growth and summer yield in the locations, ecological systems, soil conditions, etc., as
region. Liu et al. (2014a) monitored the drought con- well as reflect crop water shortages (Kogan 1995; Liu
ditions in Jilin, China, during the plant growth period and Kogan 1996). Based on the surface vegetation
Environ Monit Assess (2016) 188:639 Page 3 of 13 639

growth conditions identified by satellite monitoring, the Materials and methods

VCI evaluates drought based on changes in vegetation
conditions, and its precision has been validated in many General description of the study area
studies and applications. Furthermore, it is easier to
calculate the VCI than the TVDI or other indices be- China is a vast territory that can be divided into three
cause the VCI can be calculated using data from the natural regions based on differences in geographical
near-infrared and red bands. In addition, the sensitivity location and natural and human geography: the eastern
of the VCI for monitoring drought is significantly higher monsoon region, the northwestern arid and semi-arid
than that of the NDVI (Sha et al. 2013); therefore, the region, and the Qinghai-Tibet alpine region. The eastern
VCI has been widely used in drought monitoring and monsoon region is divided into northern and southern
analysis (Cai et al. 1995; Domenikiotis et al. 2004; regions that are bounded by the Qinling Mountains and
Bajgiran et al. 2008; Guan et al. 2008; Quiring and Huaihe River (January 0 °C isothermal and 800 mm
Ganesh 2010). The advanced very high resolution radi- isohyet). Thus, China can be divided into the following
ometer (AVHRR) is attached to National Oceanic and four geographic areas: the southern region (103°–123°
Atmospheric Administration (NOAA) satellites. The E, 22°–34° N), northern region (103°–135° E, 33°–53°
AVHRR can dynamically monitor large areas of vege- N), northwestern region (73°–123° E, 37°–50° N), and
tation because its data are characterized by long tempo- Tibetan Plateau (73°–104° E, 27°–40° N) (Liang et al.
ral series, high temporal resolution, and moderate spatial 2014). To better analyze the agricultural areas, they are
resolution. divided into four regions according to the boundaries of
Therefore, this study calculated the VCI using the four geographic areas in China because it is difficult
NOAA/AVHRR data and performed annual analyses to describe the boundaries of agricultural areas in detail.
of the temporal and spatial distributions of drought in The study evaluates the correlations between drought
agricultural areas in China from 1982 to 2010. and climate factors; therefore, the climate conditions in
Moreover, the correlations between the VCI and annual these agricultural areas of China should be introduced
mean precipitation, temperature, sunshine duration, briefly. The annual mean precipitation decreases from
wind velocity, and atmospheric pressure were examined the southeast coast to the inland northwest. The annual
to understand the climate factors that influence droughts mean precipitation is more than 1600 mm in southeast
in China. These correlations provide a scientific basis coast areas and less than 50 mm in the majority of
for mitigating the impacts of climate change on agricul- northwest inland areas. The latitude range of agricultural
tural production. The flowchart of the study is shown in areas is extensive, and the temperature varies accord-
Fig. 1. ingly. In winter, the lowest temperature is lower than

Fig. 1 Flowchart of the method

639 Page 4 of 13 Environ Monit Assess (2016) 188:639

minus 30 °C, while the highest temperature is above indexes: VCI, TCI, and VHI. The VCI, which spans
20 °C. The annual mean sunshine duration exhibits the 52 periods in a year, was obtained using IDL program-
opposite trend as that of the precipitation distribution. In ming, and its spatial resolution is 16 km and temporal
the northwestern region and Tibetan Plateau, the annual resolution is 7 days. Then, monthly data were synthe-
mean sunshine duration is generally more than 3000 h, sized using the maximum value composite method, and
but in most southern region, it is less than 2000 h. The annual data were obtained using the average value com-
annual mean wind velocity is higher in northern and posite method. The VCI was calculated using the fol-
coastal areas than that in southern and inland areas. lowing formula (Mao et al. 2005; Mao et al. 2007):
Notably, the annual mean wind velocity in coastal areas
and three northern regions (northeast, northwest and
northern China) is the largest in agricultural areas. NDVIi −NDVImin
VCI ¼ 100 
China has a typical continental monsoon climate. In NDVImax −NDVImin
plain areas, winter is characterized by cold temperatures
and high atmospheric pressure, while summer is char-
acterized by high temperatures and low pressure. where NDVIi is the NDVI of the pixels in a specific
Therefore, the highest atmospheric pressure generally period, and NDVImax and NDVImin are the maximum
appears in January, and the lowest pressure occurs in and minimum NDVI values in a specific period, respec-
July and August. However, the months with the highest tively. The maximum and minimum values of the de-
and lowest atmospheric pressures are reversed in areas nominator reflect the best and worst conditions for veg-
with elevations of 4000–5000 m compared (http://www. etation growth, respectively, and the difference between
cma.gov.cn). them reflects the living conditions of the local vegeta-
According to the second national land survey, there is tion. The numerator is the difference between the known
approximately 135.385 million hectares of cultivated and minimum values of the NDVI for a specific period,
land, including paddy fields and dry land, in China, as which reflects the vegetation growth. A smaller value
shown in Fig. 2. The eastern region, which includes indicates that less vegetation grew. The VCI is calculat-
26.297 million hectares of cultivated land, accounts for ed using this formula, which results in a number ranging
19.4 % of the arable land in China, and the cultivated from 0 to 100. If the VCI ranges from 70 to 100, it
land primarily consists of paddy fields, which are gen- indicates normal vegetation conditions and no drought.
erally used to plant rice, rapeseed, sugar cane, and other A value between 50 and 70 indicates moderate vegeta-
crops. The central area contains 30.715 million hectares tion conditions and mild drought, whereas a value from
of arable land, which accounts for 22.7 % of the arable 30 to 50 indicates poor vegetation growth and moderate
land in China and is mainly used to grow wheat, rice, drought, and below 30 indicates extremely poor growth
and cotton. The western cultivated land area covers conditions and severe drought. Larger values of the VCI
50.435 million hectares or 37.3 % of the arable land in indicate good vegetation growth and low degrees of
China. This area is in an arid or extremely arid climate drought, whereas smaller values of the VCI indicate
zone with an annual precipitation of less than 200 mm. poor vegetation growth and higher degrees of drought
However, the region has numerous inland rivers, which (Sha et al. 2013; Dutta et al. 2015).
make it a commodity grain and cotton production base The National Meteorological Information Center in
in China (Guan et al. 2008). The northeastern region, China provided 557 effective site observations of
which contains 27.938 million hectares of cultivated ground-recorded meteorological data from 1982 to
land, accounts for 20.6 % of the arable land in China 2010, as shown in Fig. 2. The monthly mean precipita-
and is mainly dry land that is typically planted with tion, air temperature, sunshine duration, wind velocity,
wheat, beans, peanuts, and sugar beets (http://www. and atmospheric pressure were calculated using the
mlr.gov.cn/). daily observation data from each site. The annual mean
meteorological values were then determined based on
Acquisition of the VCI and meteorological data the monthly mean values. Kriging interpolation was
used to generate a raster map of the meteorological data
The AVHRR data are provided by VHP vegetation with a spatial resolution of 1000 m. Then, the data in the
health products from NOAA, which include three agricultural areas were obtained by masking.
Environ Monit Assess (2016) 188:639 Page 5 of 13 639

Fig. 2 Spatial distributions of agricultural areas and meteorological sites in 2010

Analysis of trends in the drought distribution period (27 years; there are no data from 1994 and 2000).
and climate factors in agricultural areas from 1982 In the following study, the total drought frequency and
to 2010 different levels of drought frequency will be analyzed.
However, the total drought frequency is the sum of the
To analyze changes in the trends of the VCI and the mild drought, moderate drought, and severe drought fre-
climate factors in agricultural areas from 1982 to 2010, quencies. Therefore, in the total drought frequency analy-
the droughts were divided into four levels—normal sis, a frequency up to 0.8 is considered a high frequency
(VCI ≥ 70), mild (50 ≤ VCI < 70), moderate and 0.4 is considered a low frequency. However, in dif-
(30 < VCI < 50), and severe (VCI ≤ 30)—based on ferent levels of drought frequency analysis, a frequency of
literature concerning aridity classification standards up to 0.6 is considered a high frequency and 0.2 is
(Guan et al. 2008; Sha et al. 2013; Dutta et al. 2015). considered a low frequency.
The corresponding grades of the VCI values are shown In addition, the trends in the VCI and climate factors
in Table 1. Then, different levels of drought frequency were calculated using the following formula to analyze the
(f) were calculated using the following formula: associated changes in agricultural areas (Zhang et al.
2011b):
n
f ¼ Xn X n X n 
N
n i¼1
x i t i − x
i¼1 i
t
i¼1 i
Slope ¼ Xn  Xn 2
where f is the drought frequency, n is the number of
n t 2−
i¼1 i
t
i¼1 i
different levels of drought, and N is the length of the study
Table 1 Classification of the VCI where xi is the VCI or the value of a climate factor in
Level VCI value
year i, n is the length of the time series, and ti is the
number of the year from 1982 to 2010 (1–27). A posi-
Normal ≥70 tive slope indicates that the VCI or climate factor in-
Mild drought 50 ≤ VCI < 70 creased during the study period. Otherwise, the value
Moderate drought 30 < VCI < 50 decreased. The increasing trend in the VIC reflects
Severe drought ≤30 improved vegetation growth and drought abatement.
Thus, the VCI trend contrasts the drought trend. The
639 Page 6 of 13 Environ Monit Assess (2016) 188:639

legends of the figures of drought and climate factor of a climate factor in year i, respectively, and x and y are
trends were mainly based on positive-negative proper- the mean VCI and the mean value of a climate factor,
ties or spatial distributions. respectively, from 1982 to 2010.
Based on the linear correlation coefficient calcula-
Correlation analysis of the VCI and climate factors tions, the partial correlation coefficient was calculated
using the following formula:
The Pearson correlation coefficient can reflect the de- Rxy −Rxz Ryz
gree and direction of the correlation between two vari- Rxyz ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  ffi
ables. A larger absolute value indicates a stronger cor- 2 2
1−Ryz 1−Rxz
relation between the two variables. The partial correla-
tion coefficient is an indicator that measures the degree Bwhere Rxyz is the partial correlation coefficient of
of linear correlation between two variables among sev- variables x and y after fixing variable z, Rxy is the
eral variables when the impacts of multiple other vari- correlation coefficient of variables x and y, Rxz is the
ables are controlled. The internal linear relationships correlation coefficient of variables x and z, and Ryz is the
between two variables can be more accurate and reliable correlation coefficient of variables y and z.
by comparing the partial correlation coefficient and the
correlation coefficient (Yan 2003). The Pearson correla-
tion coefficients and partial correlation coefficients were
calculated using SPSS 19.0 (IBM SPSS Inc., Chicago, Results and analysis
USA) to analyze the relationships between drought and
various climate factors. Correlation coefficients were Analysis of the spatial and temporal distributions
calculated using the following formula: of drought in agricultural areas from 1982 to 2010
Xn   
i¼1
x i −x y i −y Analysis of the distribution of drought frequency
Rxy ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 rX ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Xn  n  2
i¼1
x i −x i¼1
y i −y The spatial distribution of the drought frequency in
agricultural areas of China was determined using
where Rxy is the correlation coefficient, n is the length NOAA/AVHRR data from 1982 to 2010. The results
of the time series, and i is the number of the year from are shown in Fig. 3. The drought frequency in the vast
1982 to 2010 (1–27). xi and yi are the VCI and the value majority of agricultural areas was greater than 0.6, and it

Fig. 3 Spatial distribution of the

drought frequency
Environ Monit Assess (2016) 188:639 Page 7 of 13 639

was greater than 0.8 in the northern and northwestern severity (Zhang et al. 2011a). However, severe drought
regions. Drought on the Northeastern Plain is caused by in the northwest is largely caused by its specific geo-
the excessive reclamation of cultivated land, which de- graphical environment. This region is located inland,
creases the wetland area and reduces the ability of the away from the ocean, and lacks precipitation; therefore,
soil to supply and store water, and by the destruction of the drought frequency is comparatively high. The
the vegetation environment via the mining of mineral drought frequency in the southern agricultural areas
deposits, construction of reservoirs and dams, and was relatively low, but in some areas, such as Yunnan,
overgrazing. In addition, less rain has led to serious it remained high, possibly due to the climate, the land-
desertification. Such high drought frequencies mainly form, and other natural factors, as well as poor cultiva-
occur because the Huang-Huai Plain is located in a tion practices, inadequate water conservancy strategies,
humid and semi-humid climate zone; therefore, the tem- and other human-induced factors (Yin and Li 2013). In
perature in this region increases rapidly before rain general, the drought frequency was generally high in the
events occur, which causes substantial evaporation of northern agricultural areas and low in the southern areas.
soil moisture. Moreover, high population and industry Additionally, the agricultural areas that experienced fre-
densities in districts increase the demand for water and quent droughts were mainly concentrated in the north-
decrease surface runoff, which increases the drought ern and northwestern areas.

Fig. 4 Spatial distribution of drought frequency at different levels (a normal, b mild drought, c moderate drought, d severe drought)
639 Page 8 of 13 Environ Monit Assess (2016) 188:639

Analysis of the distribution of the frequency of drought variations, and other factors. In general, the majority of the
at different levels national agricultural areas were affected by various de-
grees of drought. The Northeastern Plain mainly experi-
As shown in Fig. 4, the frequency of drought at each of the enced mild droughts, whereas the Huang-Huai Plain and
four drought levels, normal, mild, moderate, and severe, central region experienced moderate droughts, and the
was calculated. Then, the significant differences between northwestern region experienced moderate and severe
the frequencies of drought at these levels were identified. droughts. However, most agricultural areas were affected
The regions that had high normal frequencies were mainly by mild and moderate droughts.
concentrated in the southern regions, especially in the
Sichuan Basin, which experienced no drought for approx- Analysis of the distribution of the drought trends
imately 40 % of the 27 years. The frequency in the
northeastern and northwestern regions was zero, which The spatial distribution of drought trends is shown in
indicates that drought persisted in these regions. The Fig. 5. Overall, the VCI increased in the vast majority of
frequency of mild drought exhibited a certain variation agricultural areas, especially in the Sichuan Basin, sug-
with latitude and tended to decrease gradually from north gesting that the drought trend decreased over time.
to south. The mild drought frequency in most of the However, in a few agricultural areas, such as parts of
Northeastern Plain was at least 0.6, and the frequency in the Middle-Lower Yangtze Plain, the Northeastern
the Huang-Huai Plain and the southern region was lower Plain, Ningxia, Gansu, northern Xinjiang, Mongolia,
than 0.4. In addition to the few extreme areas in the and so on, the VCI decreased, suggesting that the
northeastern and northwestern regions, the frequency of drought severity gradually increased from 1982 to
moderate drought was concentrated in the 0.4 to 0.6 range. 2010. In recent years, the urbanization process has ac-
In most agricultural areas, the frequency of severe drought celerated, and wetland areas have been significantly
was zero. The frequency was low, i.e., less than 0.2, in the reduced. These may be the dominant factors behind
northwestern and central regions and on the Huang-Huai drought occurrence in the Yangtze River Delta.
Plain. The analysis of the four levels of drought revealed a
considerable difference in the drought frequency in the Climate factor trends in the agricultural areas of China
Sichuan Basin. The levels were typically normal and from 1982 to 2010
moderate, which were attributed to the unbalanced spatio-
temporal distribution of precipitation because of the ter- The trends in the annual mean precipitation, tempera-
rain, abnormal atmospheric circulation, regional climate ture, sunshine duration, wind velocity, and atmospheric

Fig. 5 Spatial distribution of the

drought trend
Environ Monit Assess (2016) 188:639 Page 9 of 13 639

Fig. 6 Spatial distribution of trends in climate factors (a annual mean precipitation, b annual mean temperature, c annual mean sunshine
duration, d annual wind velocity, e annual atmospheric pressure)
639 Page 10 of 13 Environ Monit Assess (2016) 188:639

pressure are shown in Fig. 6. The mean precipitation the precipitation and atmospheric pressure were relative-
exhibited a significant increasing trend in the Huang- ly apparent. These changes may be related to changes in
Huai Plain area, especially in the central part of atmospheric circulation and ocean currents.
Shandong. This trend occurred because regional terrain
lifting causes coastal water vapor to condense, which Analysis of the correlation between the VCI
increases the amount of rain. In the Northeastern Plain, and the climate factors in the agricultural areas of China
the Middle-Lower Yangtze Plain, and many areas in the
central region, the mean precipitation exhibited a clear To analyze the correlations between the VCI and climate
decreasing trend. In the agricultural areas of China, the factors, this study included calculations of the correlation
mean temperature generally increased, which was con- coefficients between the VCI and the annual mean pre-
sistent with the global warming phenomenon, especially cipitation, temperature, sunshine duration, wind velocity,
in the Middle-Lower Yangtze Plain and Loess Plateau. and atmospheric pressure, as shown in Table 2. The
However, the change was not apparent in the results showed that the VCI in agricultural areas had a
Northeastern Plain, the middle of the Huang-Huai significant positive correlation with the mean tempera-
Plain, the western Sichuan Basin, and parts of Guangxi ture, indicating that an increase in the mean temperature
and Yunnan. The overall trend in the mean sunshine caused an upward trend in the VCI, which reflected a
duration was small. The trend decreased in the Huang- decline in drought. A statistical analysis showed that the
Huai Plain, the Middle-Lower Yangtze Plain, and the maximum mean temperature was approximately 25 °C in
southern part of the Northeastern Plain, whereas the southern agricultural areas, approximately 15 °C in the
trend increased in mid-western China and the northern northern region, and approximately 11 and 12 °C in the
part of the Northeastern Plain. Regarding the mean wind northwestern region and Tibetan Plateau, respectively.
velocity, an increasing trend was mainly concentrated in These values did not exceed the most suitable tempera-
the Sichuan Basin and parts of Shan-Gan-Ning, and the ture for growing the vast majority of crops (Liu et al.
trend decreased in the majority of the other regions, 2014b). In addition, this increase in temperature promot-
especially the eastern coastal areas, Shandong peninsu- ed vegetation growth to a certain extent, which led to
la, and the middle of the Northeastern Plain. The region- higher values of the VCI. Conversely, the correlation
al differences in the mean atmospheric pressure were between the VCI and the mean wind velocity was highly
more significant. At the junction of Anhui, Henan, and negative because a high wind velocity leads to greater
mid-western Shandong, a decreasing trend was notable, evaporation and more rapid loss of surface moisture,
whereas significant increasing trends appeared in which intensify drought conditions. The relationship be-
Gansu, Shanxi, and several southern areas. The changes tween the VCI and mean sunshine duration was not
in other regions were small, but there was a decrease in significant in the agricultural areas of China, except for
the majority of these areas. In summary, the trends of the in the northern region. The VCI was slightly correlated
climate factors exhibited temporal and spatial differ- with mean precipitation, partly because artificial irriga-
ences from 1982 to 2010. The annual mean sunshine tion has become popular in China, and a relatively com-
duration generally remained the same, and changes in plete water conservancy system that has been established

Table 2 Correlation coefficients of the VCI and climate factors in agricultural areas (time scale 1 year)

Region Mean precipitation Mean temperature Mean sunshine Mean wind Mean atmospheric
duration velocity pressure

Southern −0.174 0.802** 0.028 −0.712** −0.299

Northern 0.069 0.632** −0.426* −0.745** −0.441*
Northwestern 0.059 0.650** −0.198 −0.679** −0.294
Tibetan Plateau 0.111 0.789** 0.171 −0.415* −0.416*
Total −0.052 0.766** −0.228 −0.761** −0.382*

*The correlation is significant at the 0.05 level (two-tailed test)

**The correlation is significant at the 0.01 level (two-tailed test)
Environ Monit Assess (2016) 188:639 Page 11 of 13 639

Table 3 Partial correlation coefficients between the VCI and climate factors in agricultural areas (time scale 1 year)

Region Mean precipitation Mean temperature Mean sunshine duration Mean wind velocity Mean atmospheric pressure

Southern −0.185 0.578 0.052 −0.341 −0.331

Northern 0.044 0.505 −0.136 −0.478 −0.453
Northwestern 0.355 0.353 0.236 −0.560 −0.603
Tibetan Plateau 0.346 0.508 0.488 −0.220 −0.648
Total 0.103 0.442 0.139 −0.442 −0.382

in recent years. Therefore, most cultivated areas rely 1982 to 2010. Additionally, the agricultural areas
mainly on artificial irrigation, and their dependence on that suffered from frequent, severe droughts were
precipitation has decreased, as has the influence of pre- mainly concentrated in the northwestern area and
cipitation on drought. the Huang-Huai Plain. However, most agricultural
As shown in Table 3, to more accurately represent the areas were influenced by mild and moderate
relationship between the VCI and climate factors in droughts. Moreover, the VCI increased in the over-
agricultural areas, the partial correlation coefficients whelming majority of agricultural areas, illustrating
were calculated. Overall, temperature and wind velocity that drought exhibited a decreasing tendency over
were the main factors that influenced drought in the time, and the decreasing trend in the southern region
agricultural areas of China. From a regional perspective, was larger than that in the northern region.
temperature was the main factor the influenced drought 2. The analysis showed that the climate factor trends in
in the southern region. Temperature, wind velocity, and the agricultural areas of China were inconsistent
atmospheric pressure were the main driving factors in from 1982 to 2010. The annual mean temperatures
the northern region, and those in the northwestern region in the four regions increased, and this phenomenon
were wind velocity and atmospheric pressure, whereas was consistent with global warming. However, dif-
those in the Tibetan Plateau region were temperature, ferent areas and regions exhibited different in-
sunshine duration, and atmospheric pressure. This anal- creases. The trends in the other four types of climate
ysis illustrates that the relationship between the climate factors differed between the geographical regions,
factors is relatively complex in the agricultural areas of and the mean atmospheric pressure varied
China. The VCI is affected by temperature, sunshine significantly.
duration, wind velocity, atmospheric pressure, human 3. The analysis of the correlations between the climate
activities, and other factors, and no single factor funda- factors and the VCI showed that temperature and
mentally determines the level of drought in agricultural wind velocity were the main factors that influenced
areas. drought in the agricultural areas of China, and there
were several geographical differences in the factors
that affected drought. Excluding precipitation, the
Discussion and conclusion climate factors affected drought differently in dif-
ferent regions. In addition, the VCI was slightly
This study presented an analysis of the main factors that correlated with mean precipitation, likely due to
influence drought and provided a scientific basis for the popularization of artificial irrigation systems
mitigating the impacts of climate change on agricultural and relatively complete water conservancy. This
production by analyzing drought trends and the relation- conclusion is in agreement with that of Liang et al.
ships between the VCI and the climate factors in the (2014). Furthermore, the comparison between the
agricultural areas of China from 1982 to 2010. We correlations and the partial correlation analysis of
arrived at the following conclusions: the climate factors and VCI showed that the corre-
lations between the climate factors are relatively
1. The drought frequency was high in the northern complicated, and no single factor can radically de-
agricultural areas and low in the southern areas from termine the level of regional drought.
639 Page 12 of 13 Environ Monit Assess (2016) 188:639

Remote sensing data and the choice of the vegetation Guo, N., & Wang, X. P. (2015). Advances and developing oppor-
tunities in remote sensing of drought. Journal of Arid
index are critical factors in analyses of the characteristics
Meteorology, 33, 1–18.
of the temporal and spatial distributions of drought in Hao, C., Zhang, J., & Yao, F. (2015). Combination of multi-sensor
agricultural areas over long periods. With the rapid devel- remote sensing data for drought monitoring over Southwest
opment of remote sensing technology and increasingly China. International Journal of Applied Earth Observation
precise drought monitoring and research, more compre- and Geoinformation, 35, 270–283. doi:10.1016/j.
jag.2014.09.011.
hensive and detailed analyses of drought scenarios will be He, J. Y., Zhang, M. J., Wang, P., Wang, S. J., & Wang, X. M.
performed by combining remote sensing data from mul- (2011). Climate characteristics of the extreme drought events
tiple sources and a variety of drought indices in the in Southwest China during recent 50 years. Acta
agricultural areas of China. Geographica Sinca, 66, 1179–1190.
Jiao, W., Zhang, L., Chang, Q., Fu, D., Cen, Y., & Tong, Q. (2016).
Evaluating an enhanced vegetation condition index (VCI)
Acknowledgments This research is supported by the National
based on VIUPD for drought monitoring in the continental
Science Foundation of China (nos. 41401473 and 31560130); the
United States. Remote Sensing, 8(3), 224. doi:10.3390
Postdoctoral Science Foundation of China (2013M531329); the
/rs8030224.
National Basic Research Program of China (no. 2010CB951503);
and the National Innovation and Entrepreneurship Training Pro- Kogan, F. N. (1995). Droughts of the late 1980s in the United
gram for Undergraduates (nos. 201610320004Z and States as derived from NOAA polar-orbiting satellite data.
201310320048Z). Bulletin of the American Meteorological Society, 76, 655–
668. doi:10.1175/1520-0477(1995)076<0655:DOTLIT>2.0.
CO;2.
Liang, L., Zhao, S., Qin, Z., He, K., Chen, C., Luo, Y., et al.
References (2014). Drought change trend using MODIS TVDI and its
relationship with climate factors in China from 2001 to 2010.
Journal of Integrative Agriculture, 13(7), 1501–1508.
Bajgiran, P., Darvishsefat, A. A., Khalili, A., & Makhdoum, M. F.
doi:10.1016/S2095-3119(14)60813-3.
(2008). Using AVHRR-based vegetation indices for drought
monitoring in the northwest of Iran. Journal of Arid Liu, W. T., & Kogan, F. N. (1996). Monitoring regional drought
Environments, 72(6), 1086–1096. doi:10.1016/j. using the vegetation condition index. International Journal
jaridenv.2007.12.004. of Remote Sensing, 17(14), 2761–2782. doi:10.1080
/01431169608949106.
Bokal, S., Grobicki, A., Kindler, J., & Thalmeinerova, D. (2014).
From national to regional plans—the integrated drought Liu, L. W., Zhang, W. P., Duan, Y. H., Xing, L. X., Chen, X. B., &
management Programme of the global water Partnership for Zhao, M. (2014a). Terrain corrected TVDI for agricultural
Central and Eastern Europe. Weather and Climate Extremes, drought monitoring using MODIS data. Acta Geographica
3, 37–46. doi:10.1016/j.wace.2014.03.006. Sinca, 34, 3704–3711.
Cai, B., Lu, W. J., & Zheng, X. J. (1995). Using meteorological Liu, Z. J., Liu, J. Y., & Shao, Q. Q. (2014b). Optimum temperature
satellite’s VCI to monitor soil state. Remote Sensing for Land of vegetation growth for various land cover types in China.
& Resource, 4, 45–50. Journal of Geo-Information Science, 16, 1–7.
Dai, A. (2013). Increasing drought under global warming in ob- Liu, X., Zhu, X., Pan, Y., Li, S., Liu, Y., & Ma, Y. (2016).
servations and models. Nature Climate Change, 3, 52–58. Agricultural drought monitoring: progress, challenges, and
Domenikiotis, C., Spiliotopoulos, M., Tsiros, E., & Dalezios, N. R. prospects. Journal of Geographical Sciences, 26(6), 750–
(2004). Early cotton yield assessment by the use of the 767. doi:10.1007/s11442-016-1297-9.
NOAA/AVHRR derived vegetation condition index (VCI) Ma, Z., & Fu, C. (2006). Some evidence of drying trend over
in Greece. International Journal of Remote Sensing, 25(14), northern China from 1951 to 2004. Chinese Science Bulletin,
2807–2819. doi:10.1080/01431160310001632729. 51(23), 2913–2925. doi:10.1007/s11434-006-2159-0.
Dutta, D., Kundu, A., Patel, N. R., Saha, S. K., & Siddiqui, A. R. Mao, K., Qin, Z., Shi, J., & Gong, P. (2005). A practical split-window
(2015). Assessment of agricultural drought in Rajasthan algorithm for retrieving land-surface temperature from MODIS
(India) using remote sensing derived vegetation condition data. International Journal of Remote Sensing, 26(15), 3181–
index (VCI) and standardized precipitation index (SPI). 3204. doi:10.1080/01431160500044713.
Egyptian Journal of Remote Sensing and Space Science, Mao, K. B., Shi, J. C., Li, Z. L., & Tang, H. J. (2007). An RM-NN
18(1), 53–63. doi:10.1016/j.ejrs.2015.03.006. algorithm for retrieving land surface temperature and emis-
Field, C. B., Barros, V., Stocker, T. F., Qin, D., Dokken, D. J., Ebi, sivity from EOS/MODIS data. Journal of Geophysical
K. L., et al. (2012). Managing the risks of extreme events and Research: Atmospheres (1984–2012), 112, 1–17.
disasters to advance climate change adaptation. A special Masud, M. B., Khaliq, M. N., & Wheater, H. S. (2015). Analysis of
report of working groups I and II of the intergovernmental meteorological droughts for the Saskatchewan River Basin
panel on climate change. Cambridge: Cambridge University using univariate and bivariate approaches. Journal of
Press. Hydrology, 522, 452–466. doi:10.1016/j.jhydrol.2014.12.058.
Guan, X. D., Guo, N., Huang, J. P., Ge, J. M., & Zhang, Z. H. Ozelkan, E., Chen, G., & Ustundag, B. B. (2016). Multiscale object-
(2008). Applicability analysis of VCI to monitoring north- based drought monitoring and comparison in rainfed and
west China drought. Plateau Meteorology, 27, 1046–1053. irrigated agriculture from Landsat 8 OLI imagery.
Environ Monit Assess (2016) 188:639 Page 13 of 13 639

International Journal of Applied Earth Observation and MODIS. Bulletin of Soil and Water Conservation, 32, 139–
Geoinformation, 44, 159–170. doi:10.1016/j.jag.2015.08.003. 211.
Quiring, S. M., & Ganesh, S. (2010). Evaluating the utility of the Wilhite, D. A. (2000). Drought as a natural hazard: concepts and
vegetation condition index (VCI) for monitoring meteorolog- definitions. Drought a Global Assessment, 1, 3–18.
ical drought in Texas. Agricultural and Forest Meteorology, Yan, L. K. (2003). Application of correlation coefficient and
150(3), 330–339. doi:10.1016/j.agrformet.2009.11.015. biased correlation coefficient in related analysis. Journal of
Rouse, J. W. J., Haas, R. H., Schell, J. A., & Deering, D. W. Yunnan University of Finance and Economics, 19, 78–80.
(1974). Monitoring vegetation systems in the Great Plains Yin, H., & Li, Y. H. (2013). Summary of advance on drought study
with ERTS. NASA Special Publication, 351, 309. in Southwest China. Journal of Arid Meteorology, 31, 182–
Sandholt, I., Rasmussen, K., & Andersen, J. (2002). A simple 193.
interpretation of the surface temperature/vegetation index Zhang, Q., Zhang, L., Cui, X. C., & Zeng, J. (2011a). Progresses
space for assessment of surface moisture status. Remote and challenges in drought assessment and monitoring.
Sensing of Environment, 79(2–3), 213–224. doi:10.1016 Progresses and Challenges in Drought Assessment and
/S0034-4257(01)00274-7. Monitoring, 26, 763–778.
Sha, S., Guo, N., Li, Y. H., Ren, Y. L., & Li, Y. P. (2013). Zhang, Y. D., Zhang, X. H., & Liu, S. R. (2011b). Correlation
Comparison of the vegetation condition index with meteoro- analysis on normalized difference vegetation index (NDVI)
logical drought indices: a case study in Henan province. of different vegetations and climatic factors in Southwest
Journal of Glaciology and Geocryology, 35, 990–998. China. Chinese Journal of Applied Ecology, 22(2), 323–330.
Zhao, Y. S. (2013). Analysis principle and method of remote
Wang, J. S., Li, Y. H., Wang, R. Y., Feng, J. Y., & Zhao, Y. X.
sensing application. Beijing: Science Press.
(2012a). Preliminary analysis on the demand and review of
Zhou, D., Zhang, B., Gui, R. P., Zhang, C. L., Yang, S. W., & Ji, D.
progress in the field of meteorological drought research.
M. (2014). Analysis of drought characteristics of Shaanxi
Journal of Arid Meteorology, 30, 497–508.
province in recent 50 years based on standardized precipita-
Wang, Z. M., Yang, H. B., Ji, Q. C., & Li, X. T. (2012b). Remote tion evapotranspiration index. Journal of Natural Resources,
sensing monitoring for drought in Henan province based on 29, 677–688.

View publication stats