Accepted Manuscript

Characterization of droughts during 2001–2014 based on remote

sensing: A case study of Northeast China

DianMin Cong, ShuHe Zhao, Cheng Chen, Zheng Duan

PII: S1574-9541(16)30100-5
DOI: doi: 10.1016/j.ecoinf.2017.03.005
Reference: ECOINF 753
To appear in: Ecological Informatics
Received date: 24 July 2016
Revised date: 9 March 2017
Accepted date: 29 March 2017

Please cite this article as: DianMin Cong, ShuHe Zhao, Cheng Chen, Zheng Duan ,
Characterization of droughts during 2001–2014 based on remote sensing: A case study of
Northeast China. The address for the corresponding author was captured as affiliation for
all authors. Please check if appropriate. Ecoinf(2017), doi: 10.1016/j.ecoinf.2017.03.005

This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.
 ACCEPTED MANUSCRIPT

Characterization of Droughts During 2001-2014

Based on Remote Sensing: A Case Study of Northeast
China
DianMin Cong1,2,3, ShuHe Zhao1,2,3*, Cheng Chen4, Zheng Duan5
1
School of Geographic and Oceanographic Sciences, Nanjing University, Nanjing 210023, China;
2
Jiangsu Provincial Key Laboratory of Geographic Information Science and Technology, Nanjing University

PT
210023, Nanjing, China;
3
Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application,

RI
Nanjing 210023, China;
4
Nanjing Hydraulic Research Institute, Nanjing 210029, China;
5

SC
Chair of Hydrology and River Basin Management, Technical University of Munich, Munich 80333, Germany;

*
Corresponding author
NU
E-mail: zhaosh@nju.edu.cn

Abstract: Northeast China, the most important region for commercial grain
MA

farming in China, is vulnerable to drought due to high fluctuation in monthly rainfall.

Timely, accurate and effective drought monitoring is very essential for securing the
D

output of grain farming. In this study, three widely used drought indices were
E

compared using satellite soil moisture data and their capability for drought monitoring
PT

were evaluated in Northeast China. Three indices are Normalized Monthly

CE

Precipitation Anomaly Percentage (NPA), Vegetation Health Index (VHI) and

AC

Normalized Vegetation Supply Water Index (NVSWI). In order to know the

relationship between rainfall and drought especially over different local climate zones

and give a more detail strategy for crop irrigation, a time-lag relationship is

investigated based on the TRMM 3B43 monthly precipitation and NVSWI. Tendency

rate (slope) was used to characterize the change of drought events in the region for the

period 2001-2014. Results showed that: (1) the three selected indices were suitable for
 ACCEPTED MANUSCRIPT

drought monitoring, with NVSWI performing the best in terms of the highest

correlation with soil moisture, (2) an obvious time-lag was observed between rainfall

and drought with the time lag being one month for all three climate zones, (3) drought

occurred more frequently in spring and winter, while in summer drought occurred

more easily in the west than in the east in Northeast China, (4) overall the frequency

PT
of drought was decreasing from 2001 to 2014 in Northeast China.

RI
Keywords: drought indices; time lag; change trend of drought; Northeast China

SC
1. Introduction NU
Drought not only affects the growth of crops, influences global food prices, but also
MA

contributes to political unrest [1-3]. Northeast China has been more vulnerable to

drought since the 1990s, and more than 25% of the total cropping areas have been
D

affected by drought over the past two decades [4], causing a serious threat to economy
E

development in the region. Thus it is very essential for obtaining a timely, accurate
PT

drought monitoring so that the output grain production can be secured under the
CE

drought impact of global changes [5].

AC

Traditional methodologies of drought monitoring consist of field measurement

and meteorological assessment. The field measurement of soil moisture is generally

considered as the most accurate method for drought monitoring. However, this

method faces challenges due to its burden work in obtaining sufficient sampling

points for drought monitoring on a regional scale [6]. Meteorological assessment by

climatic variables such as temperature and precipitation also encounters the problem
 ACCEPTED MANUSCRIPT

of spare stations data inaccuracy, and non-timely acquisition. For example, the

Monthly Precipitation Anomaly Percentage (PA) [7-8] and the Standard Precipitation

Index (SPI) [9] were always used for drought monitoring in the past.

Satellite remote sensing has been proved to be an efficient and reliable tool for

drought monitoring [10-12]. It is particularly effective in characterizing the temporal

PT
and spatial evolution of drought [13]. Generally, the satellite-based methods for

RI
drought monitoring can be divided into four classes. The first class comprises

SC
Vegetation Index (VI) based methods which uses vegetation indices reflecting the
NU
growth condition of the vegetation to link with drought. The widely used vegetation

indices include the NDVI (Normalized Difference Vegetation Index) [14], VCI
MA

(Vegetation Condition Index) [15] and EVI (Enhanced Vegetation Index) [16]. These

remote sensing indices can indicate the reduction of leaf area index and chlorophyll
D

after vegetation was attacked by drought. However, the change of vegetation

E
PT

greenness is not synchronous with drought event, therefore there is usually a certain

time lag between this kind of vegetation indices and the real drought occurrence.
CE

The second class is the combined usage of VI and Land Surface Temperature
AC

(LST). By combining the VCI and Temperature Condition Index (TCI) [17], Kogan

and Unganai et al. proposed a new method called the Vegetation Health Index (VHI)

[18-20]. Carlson et al. proposed a new index called the Vegetation Supply Water

Index (VSWI) based on the division of NDVI and LST [21]. An index named

Temperature Vegetation Dryness Index (TVDI), which was based on the NDVI-LST

triangular space, was proposed by Sandholt et al. [22]. These drought indices,
 ACCEPTED MANUSCRIPT

combing the advantages of VI and LST in drought monitoring, have been widely used

in the past years [5, 21, 23]. However, TVDI is more suitable for small regions

covered by full vegetation [22]. Also, these indices are based on statistical analysis,

which are difficult to express the drought situation accurately.

The third class method for drought monitoring is based on the theory of surface

PT
energy balance with soil as the observation object. Watsonet al. first proposed the

RI
thermal inertia model [24], and Price used the surface energy balance theory to

SC
improve the model [25]. Idso et al. used the canopy temperature minus the air
NU
temperature as an index of crop water status and developed the Crop Water Stress

Index (CWSI) [26]. Moran et al. considered that most of the theoretical parameters in
MA

the CSWI and the vegetation coverage have nearly linear relationship, then a Water

Deficit Index (WDI) model was proposed in order to extend its applicability [27].
D

However, the methods based on the theory of surface energy balance are only
E
PT

applicable to bare soil or periods with low vegetation cover. Additionally, the

temperature derived from remote sensing image may suffer from large uncertainties.
CE

Active and passive microwave remote sensing is the fourth way of drought
AC

monitoring. It has been widely used to retrieve soil moisture, with a particular

advantage that data can be obtained regardless of atmospheric conditions [28].Passive

microwave remote sensing plays an important role in monitoring drought of the large

scale area [29], with soil moisture data provided by a series of airborne and

space-borne passive microwave radiometers, such as the Tropical Rainfall

Measurement Mission (TRMM) [30-31] and the Advanced Microwave Scanning

 ACCEPTED MANUSCRIPT

Radiometer (AMSR-E) [32-33].The method of active microwave remote sensing

retrieves soil moisture mainly by the radar reflection, such as the ESA

ERS-1/2C-Band SAR. Though microwave remote sensing is very sensitive to the soil

moisture, its spatial resolution is always coarser. For example, the spatial resolution of

AMSR-E soil moisture data is only 0.25°. Table 1 summarizes the calculation of all

PT
drought indices mentioned above.

RI
Table 1. Summary of drought indices mentioned above.

Drought Index Calculation formula Drought Index Calculation formula

SC
Temperature
Crop Water Stress
Vegetation Dryness
Index (CWSI)
NU
Index (TVDI)
Monthly
Vegetation
Precipitation
Condition Index
MA

Anomaly
(VCI)
Percentage (PA)
Normalized
Vegetation Supply Vegetation Health
D

Water Index Index (VHI)

(NVSWI)
E

Normalized
PT

Monthly Vegetation Supply

Precipitation Water Index
Anomaly (VSWI)
CE

Percentage (NPA)
Standard
Water Deficit Index
Precipitation Index
AC

(WDI)
(SPI)
Temperature
Condition Index —— ——
(TCI)
Where represents rainfall, represents the mean rainfall, represents the amount of evapotranspiration,
represents the amount of potential evapotranspiration, K is calculated based on the precipitation probability
distribution, and and are all constants.

The capability of aforementioned drought indices varies greatly in different areas,

and it is very essential for evaluating the performance of a certain drought index
 ACCEPTED MANUSCRIPT

before using it. However, to our knowledge, the performances of these indices in

different areas have been seldom validated in previous studies. In addition, there is

always a time lag between rainfall occurrences and response of drought, and the lag

time may be different in different climate zones as a result of the different main

vegetation types. Accurate information on this time lag is meaningful for the irrigation

PT
guide in different climate zones, which needs investigation. In this study, the AMSR-E

RI
soil moisture data at lower spatial resolution was firstly introduced to compare the

SC
performance of the three frequently used drought indices, which were calculated

based on Moderate Resolution Imaging Spectroradiometer (MODIS) data at relatively

NU
higher spatial resolution, to retrieve key drought indicators for Northeast China. In
MA

addition, the time lag between rainfall and moisture conditions derived from the best

drought index for different climate zones was further investigated.

D

This study has three main objectives: (1) to evaluate the three widely used
E
PT

drought indices (NPA, VHI and NVSWI) and determine the best one for drought

monitoring in Northeast China, (2) to identify the time lag between rainfall and
CE

drought based on the determined best drought index in different climate zones in
AC

Northeast China, (3) to explore the temporal-spatial patterns and the change trend of

drought in Northeast China during the period 2001-2014.

2. Data and Methods

2.1 Study area

Northeast China is a large geographical region with a longitude extending from 115 to
 ACCEPTED MANUSCRIPT

135ºN, and latitude spanning from 38 to 53ºE. The region comprises approximately

1.267 million square kilometers, accounting for 13% of the land area in China.

Administratively, the region is composed of Heilongjiang Province, Liaoning

Province, Jilin Province and the eastern part of Inner Mongolia Autonomous Region.

Northeast China has a complex landform with the Changbai Mountain to the east,

PT
Lesser Khingan Mountains to the north and Great Khingan Mountains to the west.

RI
The average elevation of the Northeast Plain is about 200 meters. The climate in the

SC
region is a typical temperate monsoon pattern with four clearly distinct seasons. It is

characterized by hot and rainy summers and cold and dry winters. The climate zones
NU
change from a humid zone to a semiarid zone from the east to the west (Fig.1). As the
MA

most important grain producing area in China, it is mainly covered by crops, which

cover 45.46% of the total study area. The main crops include wheat, corn, rice, and
D

soybean. As a result of the region with substantial forest stock volume in China, the
E

percentage of the study area covered by forest reaches up to 28.51%.

PT
CE
AC

Fig.1. Geographical location, land cover types, meteorological stations and three climate
 ACCEPTED MANUSCRIPT

regions of Northeast China.

The annual average precipitation of Northeast China varies from years to years.

As shown in Fig.2, the annual average precipitation was 452 mm/year in 2001 while

was 749 mm/year in 2013. The rainfall in 2009 (582 mm/year) is the closet year to the

average of the annual average precipitation (580 mm/year) from 2001 to 2014. It was

PT
recorded that Northeast China was attacked by drought seriously in 2001 and the

RI
drought area reached 3.75 million hectares in Heilongjiang Province, and 2.7 million

SC
hectares in Jilin Province of which the area attacked seriously by drought was 2.04
NU
million hectares, causing a great economical loss.
MA
E D
PT
CE
AC

Fig.2. Averagely annual rainfall during 2001-2014 of Northeast China.

2.2 Data acquisition and pre-processing

2.2.1 MODIS NDVI and LST products

The NDVI and LST products from the MODIS products (MOD13C2 and MOD11C3,

respectively) for the period 2001-2014 were used in this study. They were obtained
 ACCEPTED MANUSCRIPT

from the NASA’s Earth Observing System Data and Information System

(http://reverb.echo.nasa.gov/). Both products are at 30-day time intervals and at 0.05°

spatial resolution. The monthly LST and monthly NDVI were extracted from above

two products and re-projected using IDL/ENVI software. Then the above data were

resampled to spatial resolution of 0.25° by pixel averaging to maintain the same

resolution with the AMSR-E soil moisture data used in this paper.

PT
2.2.2 AMSR-E soil moisture data

RI
The AMSR-E soil moisture data, with 1-day intervals and 0.25° spatial resolution,

SC
were used to verify which index among the selected three indices (detailed in Section
NU
2.3) can represent the soil moisture status best. The AMSR-E soil moisture data for

the year 2009 were obtained from the NASA’s Earth Observing System Data and
MA

Information System (http://reverb.echo.nasa.gov/). The typical year 2009 was selected

because it received annual precipitation that was closest to the average rainfall during
D

the period of 2001-2014. The monthly averaged soil moisture data for the year 2009
E

were calculated using the IDL/ENVI software.

PT

2.2.3 TRMM 3B43 precipitation product

CE

The monthly TRMM 3B43 precipitation from January to December of 2001-2014

AC

were obtained from the NASA’s Earth Observing System Data and Information

System（http://reverb.echo.nasa.gov/）to calculate the monthly average rainfall in

Northeast China. This product extends from 50°N to 50°S at the spatial resolution of

0.25°. The TRMM 3B43 product was used to calculate the index NPA and in the

further analysis of the time lag between drought indices and precipitation. Its

effectiveness was verified with in-situ measurements from rain gauge stations.
 ACCEPTED MANUSCRIPT

2.2.4 In-situ rain gauge station data

The measured monthly rainfall data from 2001 to 2014 were obtained from the China

Meteorological Data Sharing Service System (http://cdc.cma.gov.cn/). In order to

validate the TRMM 3B43 monthly precipitation data in Northeast China, the monthly

average in situ precipitation data, collected from the 78 meteorological stations (Fig.1)

PT
in Northeast China and its surrounding areas, were averaged to obtain the monthly

precipitation data from 2001 to 2014. Due to the limited distribution range of TRMM

RI
3B43 products, the selected meteorological stations are all between 50°N and 50°S.

SC
2.2.5 GLC2000 land cover product
NU
The Land Cover (LC) data of the study area were obtained from the GLC2000

product from the Geospatial Data Cloud (http://www.giscloud.cn/). It is at the spatial

MA

resolution of 1km. The GLC2000 land cover product was used to obtain the main

vegetation cover type in each of the three climate zones.

E D

2.3 Methodology
PT

The procedures in this study are illustrated in the flowchart (Fig.3). The whole
CE

procedures include the following four steps: (1) remote sensing data were

pre-processed to obtain two drought indices (VHI and NVSWI), and the TRMM 3B43
AC

gridded precipitation data were first validated with in-situ rainfall data and then were

used to obtain the third drought index NPA, (2) the applicability of the three drought

indices over different climate zones in Northeast China was investigated based on the

correlation analysis between the above three drought indices and the AMSR-E soil

moisture data, (3) the drought index with best performance was chosen to analyze the

time lag relationship between precipitation occurrences and drought over different
 ACCEPTED MANUSCRIPT

climate zones, (4) the temporal and spatial distribution and the change trend of

drought from 2001 to 2014 were analyzed. Details about the procedures are described

in the following subsections (Fig.3).

PT
RI
SC
NU
MA
E D

Fig.3. Flowchart used for drought monitoring in this study, in which boxes correspond to the
PT

procedures from (1) to (4) detailed in Section 2.3.

CE

2.3.1 NPA

The Monthly Precipitation Anomaly Percentage (PA) refers to the rainfall deficit
AC

compared with the mean value for the same period. It is calculated with the following

formula [7].

(1)

where represents the current rainfall, is the mean rainfall during the same period.

The PA can be used to monitor drought by the degree of deviation compared to the
 ACCEPTED MANUSCRIPT

average rainfall. The values of the PA can be positive or negative. A Monthly

Precipitation Anomaly Percentage (NPA) with values from 0-1 could be derived using

Eq. (2).

(2)

where and are the minimum and maximum values of the NPA,

PT
respectively.

RI
2.3.2 VHI

SC
Soil moisture is closely related with vegetation growth and land surface temperature.
NU
Therefore, the NDVI and LST data could be used to construct indices for drought

monitoring. VCI suggests a normalization of NDVI values relative to the absolute

MA

maximum and minimum of the NDVI. TCI used the brightness temperature values to

assess drought instead of the vegetation coverage level. The VHI were calculated by
E D

combining the VCI and TCI. The VCI, TCI and VHI were calculated with the
PT

following formulas [18].

CE

(3)
AC

(4)

(5)

where and are the minimum and maximum value of NDVI

respectively, and are the minimum and maximum value of LST

respectively, and are the weight coefficients of VCI and TCI. Since the

contribution of moisture and temperature during the vegetation cycle is currently not
 ACCEPTED MANUSCRIPT

known, we assume that the share of VCI and TCI is equal ( ). The VHI of

the study area from January to December of the entire period 2001-2014 was

calculated.

2.3.3 NVSWI

When vegetation suffers from drought, with the lack of water, leaf stomata will close

PT
in order to sustain water in the canopy. Therefore, evapotranspiration decreases and

RI
LST increases. Based on the assumption that with enough soil water supplies, the LST

SC
observed over vegetation will be lower as a result of evapotranspiration than that

observed over bare land. The NVSWI is constructed and its formula is as follow [21]:
NU
(6)
MA

Because the VSWI can only represent the relative spatial location and is not

comparable on the time series, the NVSWI was put forward and it was calculated with
D

the following formula [23]:

E
PT

(7)

where and are minimum and maximum value of VSWI of the

CE

pixel during the period of study, respectively. The NVSWI from January to December
AC

of the entire period 2001-2014 in Northeast China was calculated.

2.3.4 Validation of TRMM 3B43 Precipitation Data

To validate the accuracy of the TRMM 3B43 precipitation data, the average rainfall

derived from in-situ measurements of the study area from 2001-2014 was compared

with the TRMM monthly 3B43 precipitation product using linear regression analysis

on 168 observations (12 observations each year from 2001-2014).

 ACCEPTED MANUSCRIPT

The correlation coefficient (R) and the Root Mean Square Error (RMSE) were

calculated to measure the consistency and deviation between TRMM rainfall data and

the in situ rainfall measurements.

2.3.5 Tendency Rate of Drought

To analyze whether the drought was increasing or decreasing from 2001-2014 in

PT
Northeast China, the tendency rate of drought was calculated based on the monthly

NVSWI data. It was calculated with the following formula.

RI
SC
(8)
NU
where is the series number of the year, n is the length of time series, represents

the frequency of drought occurring in the year. If the slope is positive, it

MA

represents that the frequency of drought occurring is increasing, and the degree of

drought becomes worse during the time series. On the contrary, the negative slope
D

indicates that drought degree is increasing during the time series.

E
PT

3. Results
CE

3.1 Validation of TRMM 3B43 monthly precipitation

product with in situ rainfall data
AC

In this study, the TRMM 3B43 precipitation product shows good consistency in

magnitude and temporal pattern with in-situ rainfall data, although it is overestimated

in the winter and sometimes underestimated in the summer (Fig.4). Fig.5 shows the

scatterplot of TRMM 3B43 monthly precipitation against the measured rainfall data

for the whole period. Comparison of TRMM 3B43 and in situ data showed that

TRMM 3B43 had a good and significant correlation with a high correlation
 ACCEPTED MANUSCRIPT

coefficient of 0.99 at p < 0.001. The RMSE value of 6.47 mm/month indicates that

there is a small deviation between the in situ rainfall measurement data and the

TRMM rainfall product on the whole. Therefore, TRMM 3B43 product can be used as

an alternative to in situ rainfall measurements for a better characterization of spatial

patterns of precipitation in the Northeast China in this study.

PT
RI
SC
NU
MA
E D

Fig.4. Comparison of TRMM rainfall product with in situ rainfall data at monthly timescale.
PT
CE
AC

Fig.5. Scatterplot of TRMM 3B43 monthly precipitation against the measured rainfall data
 ACCEPTED MANUSCRIPT

for the whole period from 2001-2014.

3.2 Evaluation of three drought indices in Northeast China

To select the most appropriate drought index in Northeast China, the performances of

three frequently used indices (NPA, VHI and NVSWI) were evaluated based on the

correlation analysis using the AMSR-E soil moisture data. It should be noted that the

PT
NPA beyond 50°N have no data (Fig.6) because TRMM 3B43 product only covers the

areas within 50°N to 50°S. In order to characterize the drought degree in detail, all

RI
three indices NVSWI, VHI and NPA were divided into five levels: severe drought,

SC
moderate drought, slight drought, normal condition and wetness. These levels

correspond to numerical ranges of 0-0.2, 0.2-0.4, 0.4-0.6, 0.6-0.8 and 0.8–1

NU
respectively. Moreover, the AMSR-E soil moisture data of three typical months (May,
MA

July and September) in 2009 were selected in order to determine the ability of drought

monitoring quantitatively. The representative months selected above are all during the
D

spring wheat growing season.

E

Fig.6 shows the spatial patterns of different drought indices in Northeast China.
PT

It was shown that the spatial patterns of three different indices vary considerably. The
CE

Lesser Khingan Mountains and their surroundings were identified as wet by the

NVSWI, but the opposite was true by both the VHI and NPA in May, which may be as
AC

a result of the low vegetation coverage for VHI and low precipitation for NPA.

NVSWI showed that the drought in south of Greater Hinggan Mountains and Songnen

Plain had aggravated from July to September while the VHI showed exactly the

opposite, which may be as a result of the high LST for VHI. The spatial distribution

pattern of NPA was variable because of the uneven distribution of rainfall, and the

spatial distribution of VHI is not spatially comparable. Thus, NVSWI performs best
 ACCEPTED MANUSCRIPT

out of the three drought indices considering the spatial distribution (Fig.6).

PT
RI
SC
NU
MA
E D
PT
CE

Fig.6. Spatial distribution of three different drought indices in Northeast China for the three

typical months in 2009.

AC

In order to evaluate the ability for drought monitoring of the three drought

indices, the correlation analyses were performed between the drought indices and

AMSR-E soil moisture data. Fig.7 shows the fitting results of NVSWI, VHI and

NPA with AMSR-E soil moisture data in the three typical months (May, July and

September) of 2009 respectively. In May, the best drought index was NVSWI with

the highest value of correlation coefficient being 0.68. Both VHI and NPA showed
 ACCEPTED MANUSCRIPT

negative correlations with soil moisture data, which could be due to the low

vegetation coverage for VHI and low precipitation for NPA. The correlation

coefficient between VHI and the soil moisture data was only 0.09. In July and

September, NVSWI and the AMSR-E soil moisture data registered highest

correlation coefficients with values of 0.56 and 0.53. In conclusion, NVSWI had a

significantly higher correlation with the soil moisture data than VHI and NPA,

PT
indicating its highest ability to reflect the soil moisture. Therefore, NVSWI

RI
possessed the best performance and could be effectively used in the drought

SC
monitoring of Northeast China.
NU
MA
E D
PT
CE
AC

Fig.7. Correlations between monthly AMSR-E soil moisture data and the three drought

indices (NVSWI, VHI and NPA) for the three typical months in 2009.
 ACCEPTED MANUSCRIPT

3.3 Time lags between mean monthly NVSWI and rainfall in

different climate zones
Rainfall is the key factor affecting the amount of soil moisture, and the chance of

drought occurrence is proportional to the rainfall. Information on the lag time between

PT
rainfall and soil moisture implied by drought index especially over locally different

RI
climate zones enables a more detailed strategy for crop irrigation. Therefore, it is of

high importance to study the relationship between drought and rainfall. In this study,

SC
the annual total rainfall amount was calculated based on the monthly TRMM 3B43.
NU
The relationship between mean monthly NVSWI of different climate zones and

rainfall was analyzed in the three typical years of 2001 (dry), 2009 (average) and
MA

2013 (wet) (as specified in Fig.2).

Fig.8 shows the relation between mean monthly NVSWI and mean monthly
D

rainfall in three different climate zones during the three typical years (2001, 2009 and
E
PT

2013). It was shown that the monthly rainfall was usually very small in January and

February. Then it increased and reached a maximum in June or July. Afterwards, it

CE

decreased gradually from August to December. The monthly rainfall time-series

showed a trend of periodic change and the rainfall reached peaks in July 2013 among
AC

the three years. The temporal pattern of monthly averaged NVSWI of different

climate zone was in phase with the monthly rainfall time-series. Fluctuation of

monthly averaged NVSWI in humid areas and semi-humid areas was more valiant

while the amplitude of NVSWI in semi-arid areas was relatively small. It may

potentially be because that the semi-arid area, that was covered by grass mostly

(85.59%) (Table 2), had relatively lower vegetation coverage than humid areas and
 ACCEPTED MANUSCRIPT

semi-humid areas, causing reduced ability of adjusting temperature. Thus, the value of

NDVI was low in summer while the LST was relatively high, leading to a low value

of NVSWI when it was calculated by Eq. (6).

Table 2. Distribution of land cover classes in the three different climate zones derived from

GLC2000 land cover product.

PT
Land Cover Class Humid areas (%) Semi-humid areas (%) Semi-arid areas (%)

Bare Land 0.04 0.13 0.09

RI
Build 0.64 1.11 0.38

Crop 33.89 67.52 10.79

SC
Forest 60.68 14.7 2.29

Grass 4.6 15.83 85.59

NU
Water 0.16 0.7 0.86
MA
E D
PT
CE
AC

Fig.8. Relation between mean monthly NVSWI and mean monthly rainfall in three different

climate zones during the three typical years (2001, 2009 and 2013). Mean monthly rainfall is

plotted on the secondary Y-axis.

Due to the diverse growing conditions, climate zones are usually covered by
 ACCEPTED MANUSCRIPT

differently main vegetation. For example, the humid area in this study is mostly

covered by forest (60.68%) (Table 2). Due to the different water storage capability of

vegetation, it is expected to be a different lag time between the rain and soil moisture

[23]. In order to analyze the lag period between the rain and the soil moisture in

different climate zones, the monthly averaged NVSWI and TRMM rainfall data of

different climate zones from 2001 to 2014 were calculated. Then the correlation

PT
analyses between monthly mean NVSWI and rainfall time series were carried out for

RI
the three different climate zones by using a different time lag from 0 to 3 months.

SC
Table 3 showes that correlation coefficients reached the highest for all climate zones

when the time lag is 1 month. All the correlation analyses passed the significance test
NU
at P<0.05.

Table 3. Square of correlation coefficient (R2) between monthly mean NVSWI and monthly
MA

rainfall time series for the different climate zones under different time lag condition (0-3

months).
D

Time lag (month) Humid areas Semi-humid areas Semi-arid areas

E

0 0.66 0.68 0.70

PT

1 0.67 0.72 0.76

CE

2 0.30 0.33 0.34

3 0.01 0.02 0.02

AC

3.4 Temporal and spatial distribution of drought in

Northeast China
The temporal and spatial distribution of drought in Northeast China was calculated

based on NVSWI. Fig.9 shows the spatial distribution of NVSWI in six months

(January, March, May, July, September and November) of the above three typical

years. It was shown that most of Northeast China was affected by drought in January,
 ACCEPTED MANUSCRIPT

except in the southeastern of Lesser Khingan Mountains and some areas in the east of

the Jilin Province. As time went on, the degree of drought in southern Northeast China

abated and the focus of drought shifted gradually to the north in March, although the

north of Northeast China and the eastern part of Inner Mongolia Plateau were still

seriously affected by drought. In May, with increased rainfall amount, most of the

PT
areas were wet. But as the major grain-producing areas, the Songnen Plain, Liaohe

RI
Plainand Sanjiang Plain were still affected by slight drought. Additionally, the eastern

SC
part of the Inner Mongolia Plateau was suffering from moderate drought. In July and

September when highest rainfall amount occurred, the drought on the above three
NU
plains diminished, and most of Northeast China was wet. However, drought returned
MA

in November, and the Songnen Plain, Liaohe Plain and Sanjiang Plain were attacked

by severe drought. The year most affected by drought was 2001, as shown by a
D

massive areas of severe drought in January and May, which was consistent with the
E

lowest rainfall totals in 2001 (Fig.2).

PT
CE
AC
 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
Fig.9. Spatial distribution of drought in Northeast China in the three typical years (2001,
MA

2009 and 2013).

D

3.5 Drought change trend in Northeast China during

E

2001-2014
PT

In order to characterize the change trend of drought from 2001-2014 in details, the
CE

occurrence frequency of drought, including severe drought, moderate drought, slight

drought, were totaled monthly during 2001-2014. Then, the rate of drought
AC

occurrence in every year and the drought change trend from 2001 to 2014 were

calculated.

Fig. 10 shows the occurrence frequency of drought during 2001-2014 in

Northeast China, which displays no obvious regularity. The highest frequency of

drought was in 2001(35.81%), which was consistent with the least amount of rainfall
 ACCEPTED MANUSCRIPT

(Fig.2) and the most serious drought in 2001 (Fig.9). The smallest occurrence

frequency of drought was in 2010 (24.63%) which could be due to the higher amount

of rainfall than normal years (Fig.2). The slope of the fitting line is negative, with a

value of -0.003, indicating that overall the frequency of drought was decreasing from

2001 to 2014. Fig.11 shows the distribution of drought slope in Northeast China. The

PT
drought degree has somewhat decreased in the Songnen Plain, Liaohe Plain and in the

RI
eastern part of Inner Mongolia Autonomous Region over the past 14 years. However,

SC
the central part of the Greater Khingan Mountains, most areas of the Lesser Khingan

Mountains and most areas in southwestern Northeast China were affected by drought
NU
more frequently from 2001 to 2014.
MA
E D
PT
CE
AC

Fig.10. Frequency of drought occurrence in Northeast China from 2001 to 2014.

 ACCEPTED MANUSCRIPT

PT
RI
SC
NU
Fig.11. Distribution of the drought change trend in Northeast China from 2001-2014.
MA

4. Discussion
In this study, the advantage of using the NVSWI was justified by correlation analysis
D

between drought indices and AMSR-E soil moisture data. Generally, the in-situ
E
PT

measurement of soil moisture is considered as the most accurate method for drought

monitoring. However, in-situ measurements are often available for only small scales
CE

and limited areas. Given the spatial resolution of the NVSWI depending on the used
AC

satellite data (0.05° in this study), the AMSR-E soil moisture data are more

compatible with the gridded data like NVSWI than in-situ measurements when

performing the linear regression analysis to retrieve the best index. In addition, the

gridded soil moisture data have been used to evaluate the performance of drought

indices in previous studies [34]. The good validity of the AMSR-E soil moisture data

also has been evaluated by a lot of existing studies in other areas, indicating its
 ACCEPTED MANUSCRIPT

reliability in reflecting the actual soil moisture for large scale [35-37].

NVSWI is identified as the most adequate index among the three investigated

indices to measure drought in Northeast China. In fact, many studies have used VSWI

to monitor drought [23, 38]. The NVSWI, normalized based on the VSWI, has more

objective indicator than the original VSWI [23]. In general, NPA monitors drought

PT
following precipitation only, which may have very small difference in some months

RI
with low precipitation in Northeast China [7, 39]. Also, the degree of drought cannot

SC
be monitored. The spatial distribution of VHI is not spatially comparable because the

values of NDVI and LST vary greatly in different months [18]. NVSWI characterizes
NU
the drought by the combination of the growth condition of vegetation and land surface
MA

temperature, thereby leading to the best performance among the three drought indices

(Fig.6 and Fig.7).

D

There were also a lot of studies focused on the regional drought in Northeast.
E

Song et al. found that the Songnen Plain suffered the most severe drought in spring
PT

[39], Huang et al. found that drought in the western area is more severe than in the
CE

eastern area in Inner Mongolia [40], which are all consistent with the results shown in
AC

Fig.9. Zhou et al. found that the trend of drought was decreasing during 2001-2013 in

Northeast China [41], which is in accordance with the results shown in Fig.10. In

conclusion, the temporal-spatial distribution and change trend of drought in Northeast

China is consistent with the previous study focused on the national or regional

drought monitoring, although there may be slight difference in detail due to the

differently selected drought indices.

 ACCEPTED MANUSCRIPT

It can be found from the Table 3 that the difference in the R2 between time lag of

0 month and 1 month is very small in all cases, which may be because the Northeast

China is not an extremely arid region and most areas belong to the climate zones

ranging from semi-arid areas to humid areas. Thus, the soil is not extremely short of

moisture and the effect of rainfall is not particularly evident to the drought relief.

PT
However, the difference in R2 between time lag of 0 month and 1 month became

RI
larger from humid areas to semi-arid areas (Table 3). It may be due to the different

SC
ability of water conservation for different vegetation cover types in each climate zone.

The strong ability of water conservation of forests makes a more moist soil
NU
environment, thereby leading to the smallest difference in R2 in humid area between
MA

time lag of 0 month and 1 month. Crops and grasses have relatively lower ability of

water conservation and the soil is shortage of water to some extent. Thus, the effect of
D

rainfall is more evident to the drought relief than the forest, and the difference in R2 in
E

semi-humid areas and semi-arid areas is more lager than humid areas between time
PT

lag of 0 month and 1 month.

CE

The time lag is expected to be different in each climate zones due to its uniquely
AC

main vegetation types with different capacity of water storage [23]. The humid areas

covered mostly by tall and dense forests are expected to have a longest time lag

because forests possess the best capacity of water retention with deepest roots to tab

groundwater compared with crops and grasses [42]. Similarly, semi-arid areas covered

mostly by grasses should have a shorter time lag due to the lower capacity of water

retention for grasses compared with forests. As for semi-humid areas covered mostly
 ACCEPTED MANUSCRIPT

by crops, although the water storage capacity of crops may be similar to or even lower

than that of grasses, artificial irrigation has a great effect on the time lag [43]. Thus,

semi-arid areas should have a time lag similar to or longer than arid areas. However,

because the data used in this study were all at the monthly timescale, the time lag for

all the different climate zones were found to be 1 month. Thus, the MODIS data at

PT
higher time scales (8 or 16 days) should be used in further studies to obtain a better

RI
understanding of the time lag.

SC
5. Conclusions NU
The applicability of three widely used drought indices (NPA, VHI and NVSWI) was

evaluated based on a correlation analysis between AMSR-E soil moisture data and the
MA

drought indices over the three climate zones. The results showed that the NVSWI

registered highest correlation coefficients with the AMSR-E soil moisture data with
D

the values of 0.68 in May, 0.56 in July and 0.53 in September, indicating its best
E
PT

applicability in monitoring drought in Northeast China.

An obvious time lag was observed between rain occurrences and drought in
CE

different climate zones with the time lag being one month for all three climate zones.
AC

However, the time lag is expected to be different in each climate zones due to its

uniquely main vegetation types with different capacity of water storage. Thus, the

MODIS data at higher time scales (8 or 16 days) will be used in further studies to

obtain a better understanding of the time lag.

Most areas in Northeast China were attacked by drought in winter. In summer

the spatial distribution of drought showed that drought occurs more easily in the west
 ACCEPTED MANUSCRIPT

than in the east of Northeast China. The temporal patterns of the drought were

apparently influenced by seasonal variations, with drought occurring more frequently

in spring and winter in Northeast China. The change trend of drought from 2001 to

2014 was also characterized. It is shown that the variation of drought trend was

apparent among different parts of the region, and overall the frequency of drought was

PT
decreasing from 2001 to 2014.

RI
Acknowledgements

SC
This study were supported by the Natural Science Foundation of China (Grant No.
NU
41671429), National Basic Research Program of China (Grant No. 2010CB951503)

and the Project funded by Nanjing Hydraulic Research Institute (Grant No. Y916032).
MA

The authors would like to thank NASA’s Earth Observing System Data and the

Information System for providing free MODIS data, the China Meteorological Data
D

Sharing Service System for supporting station precipitation data, and the Geospatial
E
PT

Data Cloud for supporting Land Cover data. We also appreciate the editor and

anonymous reviewers for their helpful comments.

CE

References
AC

1. Kelley, C. P., Mohtadi, S., Cane, M. A., Seager, R., & Kushnir, Y. (2015). Climate

change in the Fertile Crescent and implications of the recent Syrian drought.

Proceedings of the National Academy of Sciences, 112(11), 3241-3246.

2. Wu, J., Zhou, L., Mo, X., Zhou, H., Zhang, J., & Jia, R. (2015). Drought

monitoring and analysis in China based on the Integrated Surface Drought Index

(ISDI). International Journal of Applied Earth Observation and Geoinformation,

 ACCEPTED MANUSCRIPT

41, 23-33.

3. Zhang, J., Mu, Q., & Huang, J. (2016). Assessing the remotely sensed Drought

Severity Index for agricultural drought monitoring and impact analysis in North

China. Ecological Indicators, 63, 296-309.

4. Qin, Z., Tang, H., Li, W., Zhang, H., Zhao, S., & Wang, Q. (2014). Modelling

PT
impact of agro-drought on grain production in China. International Journal of

RI
Disaster Risk Reduction, 7, 109-121.

SC
5. Dhorde, A. G., & Patel, N. R. (2016). Spatio-temporal variation in terminal

drought over western India using dryness index derived from long-term MODIS
NU
data. Ecological Informatics, 32, 28-38.
MA

6. Rahimzadeh-Bajgiran, P., Berg, A. A., Champagne, C., & Omasa, K. (2013).

Estimation of soil moisture using optical/thermal infrared remote sensing in the

D

Canadian Prairies. ISPRS Journal of Photogrammetry and Remote Sensing, 83,

E

94-103.
PT

7. Shao-E, Y., & Bing-Fang, W. (2010, March). Calculation of monthly precipitation

CE

anomaly percentage using web-serviced remote sensing data. In Advanced

AC

Computer Control (ICACC), 2010 2nd International Conference on (Vol. 5, pp.

621-625). IEEE.

8. Chen, C., & Zhao, S. (2016). Drought monitoring and analysis of Huanghuai Hai

plain based on TRMM precipitation data. remote sensing for Land and Resources,

28(1),122－129.

9. McKee, T. B., Doesken, N. J., & Kleist, J. (1993, January). The relationship of
 ACCEPTED MANUSCRIPT

drought frequency and duration to time scales. In Proceedings of the 8th

Conference on Applied Climatology (Vol. 17, No. 22, pp. 179-183). Boston, MA,

USA: American Meteorological Society.

10. Carlson, T. N., Gillies, R. R., & Perry, E. M. (1994). A method to make use of

thermal infrared temperature and NDVI measurements to infer surface soil water

PT
content and fractional vegetation cover. Remote sensing reviews, 9(1-2), 161-173.

RI
11. Tucker, C. J., Justice, C. O., & Prince, S. D. (1986). Monitoring the grasslands of

SC
the Sahel 1984-1985. International Journal of Remote Sensing, 7(11), 1571-1581.

12. Wang, H., Lin, H., & Liu, D. (2014). Remotely sensed drought index and its
NU
responses to meteorological drought in Southwest China. Remote Sensing Letters,
MA

5(5), 413-422.

13. Wan, Z., Wang, P., & Li, X. (2004). Using MODIS land surface temperature and
D

normalized difference vegetation index products for monitoring drought in the

E

southern Great Plains, USA. International Journal of Remote Sensing,25(1),

PT

14. Tucker, C. J., Justice, C. O., & Prince, S. D. (1986). Monitoring the grasslands of
CE

the Sahel 1984-1985. International Journal of Remote Sensing, 7(11), 1571-1581.

AC

15. Kogan, F. N. (1990). Remote sensing of weather impacts on vegetation in

non-homogeneous areas. International Journal of Remote Sensing, 11(8),

1405-1419.

16. Qing Liu, H., & Huete, A. (1995). A feedback based modification of the NDVI to

minimize canopy background and atmospheric noise. Geoscience and Remote

Sensing, IEEE Transactions on, 33(2), 457-465.

 ACCEPTED MANUSCRIPT

17. Kogan, F. N. (1995). Application of vegetation index and brightness temperature

for drought detection. Advances in Space Research, 15(11), 91-100.

18. Kogan, F. N. (1998). Global drought and flood-watch from NOAA polar-orbitting

satellites. Advances in Space Research, 21(3), 477-480.

19. Unganai, L. S., & Kogan, F. N. (1998). Drought monitoring and corn yield

PT
estimation in Southern Africa from AVHRR data. Remote Sensing of

RI
Environment, 63(3), 219-232.

SC
20. Wang, H., Lin, H., & Liu, D. (2014). Remotely sensed drought index and its

responses to meteorological drought in Southwest China. Remote Sensing Letters,

NU
5(5), 413-422.
MA

21. Carlson, T. N., Gillies, R. R., & Perry, E. M. (1994). A method to make use of

thermal infrared temperature and NDVI measurements to infer surface soil water
D

content and fractional vegetation cover. Remote sensing reviews, 9(1-2), 161-173.
E

22. Sandholt, I., Rasmussen, K., & Andersen, J. (2002). A simple interpretation of the
PT

surface temperature/vegetation index space for assessment of surface moisture

CE

status. Remote Sensing of environment, 79(2), 213-224.

AC

23. Abbas, S., Nichol, J. E., Qamer, F. M., & Xu, J. (2014). Characterization of

drought development through remote sensing: A case study in central Yunnan,

China. Remote Sensing, 6(6), 4998-5018.

24. Watson, K., Rowan, L. C., & Offield, T. W. (1971). Application of thermal

modeling in the geologic interpretation of IR images (Thermal modeling for IR

images geologic interpretation, discussing physical parameters role in materials

 ACCEPTED MANUSCRIPT

natural environmental diurnal temperature behavior). In International Symposium

on Remote Sensing of Environment, 7th, University of Michigan, Ann Arbor,

Mich (pp. 2017-2041).

25. Price, J. C. (1985). On the analysis of thermal infrared imagery: The limited

utility of apparent thermal inertia. Remote sensing of Environment, 18(1), 59-73.

PT
26. Idso, S. B., Jackson, R. D., Pinter, P. J., Reginato, R. J., & Hatfield, J. L. (1981).

RI
Normalizing the stress-degree-day parameter for environmental variability.

SC
Agricultural Meteorology, 24, 45-55.

27. Moran, M. S., Clarke, T. R., Inoue, Y., & Vidal, A. (1994). Estimating crop water
NU
deficit using the relation between surface-air temperature and spectral vegetation
MA

index. Remote sensing of environment, 49(3), 246-263.

28. Peng, J., Loew, A., Zhang, S., Wang, J., & Niesel, J. (2016). Spatial downscaling
D

of satellite soil moisture data using a vegetation temperature condition index.

E

Geoscience and Remote Sensing, IEEE Transactions on, 54(1), 558-566.

PT

29. Schmugge, T. J., Kustas, W. P., Ritchie, J. C., Jackson, T. J., & Rango, A. (2002).
CE

Remote sensing in hydrology. Advances in water resources, 25(8), 1367-1385.

AC

30. Duan, Z., Bastiaanssen, W. G. M., & Liu, J. (2012). Monthly and annual

validation of TRMM Mulitisatellite Precipitation Analysis (TMPA) products in

the Caspian Sea Region for the period 1999–2003. In Geoscience and Remote

Sensing Symposium (IGARSS), 2012 IEEE International (pp. 3696-3699). IEEE.

31. Chen, C., Zhao, S., Duan, Z., & Qin, Z. (2015). An Improved Spatial

Downscaling Procedure for TRMM 3B43 Precipitation Product Using

 ACCEPTED MANUSCRIPT

Geographically Weighted Regression. Selected Topics in Applied Earth

Observations and Remote Sensing, IEEE Journal of, 8(9), 4592-4604.

32. Grant, J. P., Wigneron, J. P., De Jeu, R. A. M., Lawrence, H., Mialon, A.,

Richaume, P., & Kerr, Y. (2016). Comparison of SMOS and AMSR-E vegetation

optical depth to four MODIS-based vegetation indices. Remote Sensing of

PT
Environment, 172, 87-100.

RI
33. Zhao, J., Xu, J., Xie, X., & Lu, H. (2016). Drought monitoring based on TIGGE

SC
and distributed hydrological model in Huaihe River Basin, China. Science of the

Total Environment, 553, 358-365.

NU
34. Zhang, J., Mu, Q., & Huang, J. (2016). Assessing the remotely sensed Drought
MA

Severity Index for agricultural drought monitoring and impact analysis in North

China. Ecological Indicators, 63, 296-309.

D

35. Crow, W. T., Miralles, D. G., & Cosh, M. H. (2010). A quasi-global evaluation
E

system for satellite-based surface soil moisture retrievals. IEEE Transactions on

PT

Geoscience and Remote Sensing, 48(6), 2516-2527.

CE

36. Draper, C. S., Walker, J. P., Steinle, P. J., De Jeu, R. A., & Holmes, T. R. (2009).
AC

An evaluation of AMSR–E derived soil moisture over Australia. Remote Sensing

of Environment, 113(4), 703-710.

37. Jackson, T. J., Cosh, M. H., Bindlish, R., Starks, P. J., Bosch, D. D., Seyfried, M.,

& Du, J. (2010). Validation of advanced microwave scanning radiometer soil

moisture products. IEEE Transactions on Geoscience and Remote Sensing,

48(12), 4256-4272.
 ACCEPTED MANUSCRIPT

38. Jain, S. K., Keshri, R., Goswami, A., & Sarkar, A. (2010). Application of

meteorological and vegetation indices for evaluation of drought impact: a case

study for rajasthan, india. Natural Hazards, 54(3), 643-656.

39. Song, X., Li, L., Fu, G., Li, J., Zhang, A., Liu, W., & Zhang, K. (2014).

Spatial–temporal variations of spring drought based on spring-composite index

PT
values for the Songnen Plain, Northeast China. Theoretical and applied

RI
climatology, 116(3-4), 371-384.

SC
40. Huang, J., Xue, Y., Sun, S., & Zhang, J. (2015). Spatial and temporal variability

of drought during 1960–2012 in Inner Mongolia, north China. Quaternary

NU
International, 355, 134-144.
MA

41. Zhou L, Wu J, Mo X, Zhou H., & Diao C. (2017). Quantitative and detailed

spatiotemporal patterns of drought in China during 2001–2013. Science of The

D

Total Environment, 589(1), 136-145.

E

42. Allen, C. D. (2008). Mechanisms of plant survival and mortality during drought:
PT

why do some plants survive while others succumb to drought? New Phytologist,
CE

178(4), 719.
AC

43. J. M. Grünzweig, Valentine, D. W., & F. S. Chapin III. (2015). Successional

changes in carbon stocks after logging and deforestation for agriculture in interior

alaska: implications for boreal climate feedbacks. Ecosystems, 18(1), 132-145.

 ACCEPTED MANUSCRIPT

Highlights
1. NVSWI was the best drought index for the drought monitoring in Northeast

China compared with NPA and VHI.

2. An obvious time lag was observed between rain occurrences and drought in

different climate zones with the time lag being one month for all three climate

PT
zones.

RI
3. The temporal and spatial patterns of drought in Northeast China were explored.

SC
4. Overall the frequency of drought was decreasing from 2001 to 2014 in Northeast

China.
NU
MA
E D
PT
CE
AC