Accepted Manuscript
Analysis and prediction of reference evapotranspiration with climate change in
Xiangjiang River Basin, China
Xin-e Tao, Hua Chen, Chong-yu Xu, Yu-kun Hou, Meng-xuan Jie
PII:
S1674-2370(15)00084-8
DOI:
10.1016/j.wse.2015.11.002
Reference:
WSE 38
To appear in:
Water Science and Engineering
Please cite this article as: Tao, X.-e, Chen, H., Xu, C.-y., Hou, Y.-k., Jie, M.-x., Analysis and prediction of
reference evapotranspiration with climate change in Xiangjiang River Basin, China, Water Science and
Engineering (2015), doi: 10.1016/j.wse.2015.11.002.
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
Analysis and prediction of reference evapotranspiration with
climate change in Xiangjiang River Basin, China1
Xin-e Taoa, Hua Chena,*, Chong-yu Xua, b, Yu-kun Houa, Meng-xuan Jiea
State Key Laboratory of Water Resources and Hydropower Engineering Science, Wuhan
University, Wuhan 430072, China
b
Department of Geosciences, University of Oslo, Oslo N-0316, Norway
Received 6 April 2015; accepted 9 September 2015
RI
PT
a
Abstract
1. Introduction
TE
D
M
AN
U
SC
Reference evapotranspiration ( ET0 ) is often used to estimate actual evapotranspiration in water balance
studies. In this study, the present and future spatial distributions and temporal trends of ET0 in the Xiangjiang
River Basin (XJRB) in China were analyzed. ET0 during the period from 1961 to 2010 was calculated with
historical meteorological data using the FAO Penman-Monteith (FAO P-M) method, while ET0 during the period
from 2011 to 2100 was downscaled from the Coupled Model Intercomparison Project Phase 5 (CMIP5) outputs
under two emission scenarios, representative concentration pathway 4.5 and representative concentration pathway
8.5 (RCP45 and RCP85), using the statistical downscaling model (SDSM). The spatial distribution and temporal
trend of ET0 were interpreted with the inverse distance weighted (IDW) method and Mann-Kendall test method,
respectively. Results show that: (1) the mean annual ET0 of the XJRB is 1 006.3 mm during the period from 1961
to 2010, and the lowest and highest values are found in the northeast and northwest parts due to the high latitude
and spatial distribution of climatic factors, respectively; (2) the SDSM performs well in simulating the present
ET0 and can be used to predict the future ET0 in the XJRB; and (3) CMIP5 predicts upward trends in annual
ET0 under the RCP45 and RCP85 scenarios during the period from 2011 to 2100. Compared with the reference
period (1961 to 1990), ET0 increases by 9.8%, 12.6%, and 15.6% under the RCP45 scenario and 10.2%, 19.1%,
and 27.3% under the RCP85 scenario during the periods from 2011 to 2040, from 2041 to 2070, and from 2071 to
2100, respectively. The predicted increasing ET0 under the RCP85 scenario is greater than that under the RCP45
scenario during the period from 2011 to 2100.
Key words: Reference evapotranspiration ( ET0 ); Spatial-temporal variation; Climate change; Statistical
downscaling; Xiangjiang River Basin
AC
C
EP
According to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change
(IPCC), the global warming trend, which is mainly caused by the increasing amount of greenhouse
gas emissions, will continue (IPCC, 2014). China has also experienced a pronounced warming
over the last few decades (Piao et al., 2010). Rising temperature is expected to strengthen the
hydrological cycle because of the ability of warm air to hold and redistribute more moisture,
which causes the change in atmospheric circulation (Allen and Ingram, 2002; Wentz et al., 2007;
Bates et al., 2008; Durack et al., 2012; Ye et al., 2013). Despite responding to climate change in
different ways, changes in precipitation, runoff, groundwater flow, evapotranspiration, and soil
moisture indicate that the hydrological cycle has been intensified along with the rising temperature
around the world during the 20th century (Allen and Ingram, 2002; Alan et al., 2003; Wentz et al.,
2007; Gao et al., 2012; Mitchell et al., 2012; Wang et al., 2012; Allan et al., 2013).
Evapotranspiration is an important component of the hydrological cycle and water balance in
addition to precipitation and runoff, which also play key roles in the energy budget in the
earth-atmosphere system (Xu et al., 2006; Wang et al., 2007). As the only connection linking the
water balance and land surface energy balance, evapotranspiration is considered the most
significant indicator for climate change and the water cycle (Xu et al., 2005; Xu and Singh, 2005;
Wang et al., 2013). The process of evapotranspiration governs the moisture transfer between soil
This work was supported by the National Natural Science Foundation of China (Grants No. 51190094,
51339004, and 51279138).
* Corresponding author
Email address: Chua@whu.edu.cn (Hua Chen)
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
and atmosphere in a watershed and is heavily influenced by climate change (Fisher et al., 2011).
Evapotranspiration, together with precipitation, determinates the humidity of a region.
Of all the components in the hydrological cycle, evapotranspiration is the most difficult to
estimate. There are few direct measures for actual evapotranspiration ( ETa ) over global land areas.
Because of the lack of observation data, reference evapotranspiration ( ET0 ) is often used to
estimate ETa in most cases, such as scheduling irrigation, managing water resources, analyzing
regional climate humidity conditions, and providing information for hydrological, agricultural,
and environmental models. ET0 is defined as the rate of evapotranspiration from a hypothetical
crop with an assumed crop height of 0.12 m, a fixed surface resistance of 70 s/m, and an albedo of
0.23, which would closely resemble the evapotranspiration from an extensive surface of green
grass of uniform height, actively growing, well-watered, and completely shading the ground
(Allen et al., 1998). With abundant water available at the reference evapotranspiring surface, the
main factors affecting ET0 are climatic factors (Allen et al., 1998). Therefore, ET0 is a climatic
variable and can be computed using meteorological data (Xu et al., 2006). The FAO
Penman-Monteith (FAO P-M) method is recommended as the sole standard method for the
calculation of ET0 , and the required data are available. The FAO P-M method is relatively
complete in theory, with comprehensive consideration of various climatic factors affecting
evapotranspiration (Allen et al., 1998).
There have been many related studies on hydro-climatic changes at the regional scale in
recent years, mainly focusing on analyzing the spatial distribution and temporal trend of ET0
based on historical data and identifying the main climatic factors affecting ET0 (Bandyopadhyay
et al., 2009; Zhang et al., 2011; Liu et al., 2012; Zuo et al., 2012; Ye et al., 2013). It is also
necessary to use climate models for the reliable prediction of ET0 . General circulation models
(GCMs) are used to predict future changes in ET0 . Due to the coarse spatial resolution of GCM
outputs, downscaling techniques are used to obtain the weather and climate information at a local
scale from relatively coarse-resolution GCMs (Wilby and Dawson, 2007).
Downscaling methods are divided into two categories, the physically based dynamic
downscaling method and the empirically based statistical downscaling method. The dynamic
downscaling models, in particular the regional climate model (RCM), have a clear physical
meaning. Their definite weaknesses, however, are higher computational cost and strong
dependence on the boundary conditions induced by GCMs (Wang et al., 2013). The statistical
downscaling method establishes a statistical relationship between observations and large-scale
variables, and the relationship is then used based on the GCM data to obtain the local scale
variables. The statistical downscaling method has been widely used because of its convenience in
implementation and low computation requirements (Chu et al, 2010; Wang et al., 2013). As the
first tool of the downscaling method, which is freely offered to researchers on climate change
impacts, the statistical downscaling model (SDSM) is widely used, due to its simplicity and
superior capability (Fowler and Wilby, 2007; Wilby et al., 2002).
Previous work in China has mainly focused on temperature and precipitation downscaling;
ET0 downscaling began to appear in recent years (Li et al., 2012; Wang et al., 2013). Li et al.
(2012) examined the present and future spatiotemporal characteristics of ET0 on the Loess
Plateau in China, found that ET0 significantly increased during the period from 1961 to 2009 due
to a downward trend in relative humidity and an upward trend in temperature, and demonstrated a
continuous increasing trend in the 21st century with a more pronounced upward trend after 2050.
Li et al. (2012) suggested that the increasing trend in ET0 could possibly influence water
resources on the Loess Plateau in the 21st century and some countermeasures should be taken to
reduce the adverse impacts. Wang et al. (2013) investigated the spatial and seasonal changes in
ET0 on the Tibetan Plateau and determined that, considering its spatial-temporal variation, an
average ET0 series presented a zigzag pattern (increasing–decreasing–increasing) with two joint
points in 1973 and 1993, and predicted a continuous increasing trend in ET0 in the 21st century.
In the Yangtze River Basin, there have already been studies related to ET0 . For example,
ACCEPTED MANUSCRIPT
2. Materials and methods
2.1 Study area and data
M
AN
U
SC
RI
PT
Gao et al. (2012) detected a decreasing trend in ET0 , caused mainly by a significant decrease of
the net total radiation and secondarily by a significant decrease of the wind speed over the basin.
Ye et al. (2013) analyzed the variation of ET0 and its affecting factors in the Poyang Lake
Catchment. However, there has been no prediction of ET0 in the Yangtze River Basin until now,
despite the importance of the Yangtze River Basin in China. It is necessary to mention that human
activities, such as the construction of hydraulic engineering like the Three-Gorges Dam, contribute
most to the variation of evaporation in the Yangtze River Basin. In this study, the Xiangjiang
River Basin (XJRB), which is a sub-basin of the Yangtze River Basin, was chosen for its natural
ecological environment and the fact that it is not seriously affected by human activities. The latest
outputs of the Coupled Model Intercomparison Project Phase 5 (CMIP5) recommended by the
Fifth Assessment Report of IPCC were used in this study for ET0 prediction, specifically the new
generation of two emission scenarios, representative concentration pathway 4.5 and representative
concentration pathway 8.5 (RCP45 and RCP85), instead of the old climate scenarios.
The primary objectives of this study were (1) to analyze the spatial distribution and temporal
trend of ET0 in the XJRB during the period from 1961 to 2010, (2) to investigate the adaptability
of SDSM for ET0 downscaling in the study area, and (3) to predict future ET0 during the period
from 2011 to 2100 with the SDSM under the RCP45 and RCP85 scenarios, in order to study how
ET0 will change in the XJRB in the 21st century.
AC
C
EP
TE
D
As an important tributary of the Yangtze River, the Xiangjiang River is the largest river in
Hunan Province, in southern China. The Xiangjiang River originates in the Haiyang Mountains of
Guangxi Province, flows through Yongzhou, Hengyang, Zhuzhou, Xiangtan, Loudi, Changsha,
and Yueyang cities, and discharges into Dongting Lake, with a length of 856 km and a total basin
area of nearly 94 660 km2. In Hunan Province, the river has a length of 670 km and a basin area of
nearly 85 383 km2. The XJRB covers 40% of the total area, and 60% of the population of Hunan
Province, and is responsible for 63.5% of the gross domestic product (Zhang et al., 2014).
Ranging from 110°E to 115°E and 24°N to 29°N, the XJRB has a subtropical humid
monsoon climate with four distinct seasons, hot and humid summers, cold and dry winters, and a
mean annual temperature of 16 to 19°C. In addition, the basin has an average annual rainfall of 1
200 to 1 700 mm, most of which falls from April to July, and a large variability in runoff, with a
range of 100 to 20 800 m3/s in the main channel.
Three sets of data, historical meteorological data, National Centers for Environmental
Prediction (NCEP)/National Center for Atmospheric Research (NCAR) reanalysis data, and future
GCM grid outputs, were used in this study to calculate and analyze ET0 in the XJRB for the
periods from 1961 to 2010 and from 2011 to 2100, respectively.
Historical meteorological data from 14 national meteorological stations were collected from
the China Meteorological Data Sharing Service System, including daily observations of air
temperature (T), relative humidity (RH), sunshine duration (SD), and wind speed (WS), for the
period from 1960 to 2010. Daily records of the all climate variables were verified by the China
Meteorological Administration before delivery. Of the 14 meteorological stations, seven are inside
the basin and seven are close to its boundary. The location and distribution of these meteorological
stations are shown in Fig. 1. All of these meteorological stations are less than 300 meters above
sea level, except for the Nanyue Station near the central XJRB, a station located near the famous
and scenic Hengshan Mountain, which has an elevation of 1 265.9 meters above sea level. Of
these meteorological stations, 11 are located in Hunan Province, two in Jiangxi Province, and one
in Guangdong Province.
The gridded NCEP/NCAR reanalysis data from 1961 to 2010 were interpolated into
station-scale predictors with the inverse distance weighted (IDW) method. Meanwhile, the
ACCEPTED MANUSCRIPT
TE
D
M
AN
U
SC
RI
PT
corresponding GCMs were the second generation of the Canadian Earth system model (CanESM2)
from the CMIP5. Two emission scenarios were considered in this study: the RCP45 and RCP85
scenarios. The CMIP5 models are forced by gradually increased radiative force, finally stabilizing
at 4.5 W/m2 in the RCP45 scenario, which is as effective as a level of 850 ppm CO2-equivalent
concentration by the year 2100, and 8.5 W/m2 in the RCP85 scenario, which is as effective as a
level of more than 1 370 ppm CO2-equivalent concentration by the year 2100. The RCP45
scenario represents medium-low RCP, while the RCP85 scenario represents high RCP. The XJRB
has played an important role in the economic development of Hunan Province, and its
environmental condition is severely concerning. In the past few years, Hunan Province committed
itself to creating more balanced economic growth and reducing the influence of economic
activities on the environment in the XJRB. Thus, the prediction of future CO2 levels in the XJRB
could be based on a coordinated development of the economy and the environment.
Fig. 1 Location of meteorological stations in XJRB
2.2 FAO P-M method
AC
C
EP
The FAO P-M method is recommended as the sole standard method for ET0 calculation
(Allen et al., 1998). The method is physically based and explicitly incorporates physiological and
aerodynamic parameters (Xu et al., 2006). Its accuracy and reliability have been widely verified
under different climatologic conditions in numerous regions around the world, including China.
Following the procedures in Allen et al., (1998), the FAO P-M method was used to calculate daily
ET0 in this study, and monthly and annual ET0 were the accumulation of daily ET0 :
ET0 =
0.408∆ ( Rn − G ) + γ 900 (Tmean + 273)  u2 ( es − ea )
∆ + γ (1 + 0.34u2 )
(1)
where ET0 is reference evapotranspiration (mm), ∆ is the slope of the saturated vapour pressure
(kPa/°C), Rn is the net radiation at the crop surface (MJ/(m2·d)), G is the soil heat flux density
(MJ/(m2·d)), Tmean is the mean daily air temperature at a height of 2 m (°C), u2 is wind speed at a
height of 2 m (m/s), es–ea is the saturation vapour pressure deficit (kPa), and γ is the psychrometric
constant (kPa/°C).
2.3 Statistical downscaling model
The SDSM is a coupling of the stochastic weather generator and regression algorithm
promoted by Wilby et al. (2002). It first helps users select the most representative large-scale
ACCEPTED MANUSCRIPT
2.4 Mann-Kendall test method
M
AN
U
SC
RI
PT
climatic variables (the predictors) by testing the variability in the climate (the predictand) at a
particular site using statistical models. Then, the climate variables are utilized to generate
statistical relationships with daily observed data. This relationship can be applied to obtaining
daily weather data for a future time period with the GCM-derived predictors.
Wilby and Dawson (2007) suggested that identifying empirical relationships between
predictors, such as mean sea level pressure, and single site predictands, such as air temperature at
stations, is central to all statistical downscaling methods and is often the most time-consuming
step in the process (Wang et al., 2013). In this study, the screen variable function in SDSM 4.2
was used when appropriate predictors were being chosen for model calibration of ET0
downscaling. As the atmospheric circulation features were not remarkably different in the XJRB,
the predictors were chosen according to integrated consideration of correlation analysis results of
three stations randomly picked from all the 14 stations in the XJRB.
NCEP reanalysis data were used as the inputs to the SDSM to downscale ET0 , the
predictand in this study, in the calibration period from 1961 to 1990 to develop the regression
model, and in the validation period from 1991 to 2003 to validate the model. Multiple regression
models for each station were derived from the selected station-scale predictors and station-scale
predictands (i.e., ET0 in this study). After calibration with data from 1961 to 1990 and validation
with data from 1991 to 2003, the relationships between selected predictors and ET0 at each
station were built.
TE
D
The rank-based nonparametric Mann-Kendall test method (Mann, 1945; Kendall, 1975;
Hamed and Rao, 1998), which has been widely used for trend detecting in climatic and hydrologic
time series because it does not need to assume any distribution form for the data and has a similar
effect to the parametric method (Burn and Elnur, 2002; Bandyopadhyay et al., 2009; Gao et al.,
2007; Ye et al., 2013), was applied to evaluate the significance of the trends in ET0 and other
climatic factors. The equations of the Mann-Kendall test method are given below (Mann, 1945;
Kendall, 1975) and the statistic S is calculated as
n −1
S =∑
n
∑ sgn( x
EP
i =1 j =i +1
i
− xj )
1

sgn( xi − x j ) =  0
 −1

(2)
xi > x j
xi = x j
xi < x j
(3)
AC
C
where xi and xj are the sequential data values, and n is the length of the data set. The statistic S
tends to be normally distributed for large n, with mathematical expectation and variance given by
Hamed and Rao (1998):
(4)
E (S ) = 0
Var ( S ) =
q

1
n ( n − 1)( 2n + 5) − ∑t p ( t p − 1)( 2t p + 5) 
18 
p −1

(5)
where t p is the number of ties for the pth value, and q is the number of tried values. Then, the
standard statistic Z, following the standard normal distribution, is formulated as




Z =



The null hypothesis of no trend is rejected if
( S − 1)
Var ( S )
0
( S + 1)
Var ( S )
S >0
S =0
S <0
(6)
ACCEPTED MANUSCRIPT
Z > Z1−α
(7)
2
where Z is the standard normal variate, and α is the significance level for the test, taken to be 0.05
in this study. A positive value of Z indicates an increasing trend, whereas a negative value
indicates a decreasing trend. Z1−α 2 is the critical value of Z from the standard normal table, and,
for the significance level of 5%, the value of Z1−α 2 is 1.96.
RI
PT
2.5 Performance assessment
SC
Three statistics, i.e., the relative error ( Re ), coefficient of determination ( R2 ), and
Nash-Suttcliffe efficiency coefficient (NSE) (Nash and Sutcliffe, 1970), were used to evaluate the
performance of SDSM for ET0 downscaling. R2 represents the strength of the relationship
between the observed value and simulated value, and NSE reflects how well the volume and
timing of the simulated value fit with the observed value. The closer the value of Re is to 0 and
the values of R2 and NSE are to 1, the more successfully the model performs.
3. Results and discussion
M
AN
U
3.1 Spatial distribution of mean annual ET0 during period from 1961 to 2010
AC
C
EP
TE
D
The spatial distribution of mean annual ET0 in the XJRB during the period from 1961 to
2010 is presented in Fig. 2, interpolated by the IDW method using data from all 14 meteorological
stations. The mean annual ET0 of the XJRB during the period from 1961 to 2010 is 1 060.3 mm,
and the variation coefficient for the mean annual ET0 of the 14 stations is 0.09, indicating a low
spatial variation of ET0 in the XJRB. The lowest mean annual ET0 appears at the Nanyue
Station, located at the center of the basin, with a value of 768.1 mm, while the highest mean
annual ET0 appears at the Lingling Station, located in the southwest part of the basin, with a
value of 1 087.1 mm. From Fig. 2, it can also be seen that the mean annual ET0 in the upper
XJRB was higher than it was in the lower XJRB. In the upper XJRB, ET0 increases from the
southwest part of the basin (1 050 mm) to the southern part of the basin (1 080 mm) and then
decreases from the southern part of the basin (1 080 mm) to the southeast part of the basin (1 020
mm).
Fig. 2 Spatial distribution of mean annual ET0 in XJRB during period from 1961 to 2010 (units: mm)
3.2 Temporal trend in annual ET0 during period from 1961 to 2010
The spatial distribution of the temporal trend in annual ET0 at each station in the XJRB
during the period from 1961 to 2010 is shown in Fig. 3, while Fig. 4(a) presents the linear trend of
ACCEPTED MANUSCRIPT
TE
D
M
AN
U
SC
RI
PT
the mean annual ET0 of the XJRB during the period from 1961 to 2010. Downward trends in
annual ET0 are detected at nine of the 14 meteorological stations, of which four show significant
downward trends with α < 0.05. Five stations present upward trends, and only one station shows a
significant upward trend with α < 0.05. Stations with significant downward trends are mainly
situated in the middle and eastern parts of the basin, with two such stations inside the XJRB, i.e.,
the Shuangfeng Station with an annual ET0 change rate of –1.24 mm/year and the Chenzhou
Station with an annual ET0 change rate of –1.36 mm/year, and two such stations outside the
XJRB, i.e., the Yichun Station with an annual ET0 change rate of –1.55 mm/year and the
Lianxian Station with an annual ET0 change rate of –0.90 mm/year. A significant upward trend
in annual ET0 occurs at the Shaoyang Station, located at the boundary of the XJRB, with an
annual ET0 change rate of 0.89 mm/year. As for the mean annual ET0 of the XJRB, an
insignificant downward trend with a change rate of –0.33 mm/year is detected during the period
from 1961 to 2010.
Fig. 3 Temporal trend in annual ET0 at each station in XJRB during period from 1961 to 2010
AC
C
EP
In order to explore the reasons for change in ET0 , the linear regression method was used to
analyze the temporal trends in the related climatic factors (Figs. 4(b) through (e)), including mean
air temperature ( Tmean ), RH, SD, and WS. As shown in Figs. 4(b) and (c), a significant upward
trend in Tmean , with a change rate of 0.018°C/year, and a significant downward trend in RH, with a
change rate of –0.085%/year, are factors contributing to the upward trend in annual ET0 , while a
significant downward trend in SD, with a change rate of –0.011 h/year, and a significant
downward trend in WS, with a change rate of –0.015 m/(s·year), are factors contributing to the
downward trend in annual ET0 .
SC
RI
PT
ACCEPTED MANUSCRIPT
Fig. 4 Linear trend of mean annual ET0 of XJRB and climatic factors during period from 1961 to 2010
M
AN
U
As shown in Fig. 4, significant changes in the mean annual ET0 of the XJRB and RH begin
in 2002. RH decreases approximately by 10% (from 81.5% to 71.3%) from 2002 to 2004 and the
mean annual ET0 of the XJRB increases by approximately 120 mm (from 946.6 mm to 1 065.4
mm) from 2002 to 2004. The leap of the mean annual ET0 of the XJRB in the period from 2002
to 2010 is affected by the increase of Tmean and WS and the decrease of RH.
Linear regression models were also established for annual ET0 and the climatic factors, and
the values of R2 between annual ET0 and different climatic factors are listed in Table 1.
Moreover, the significance test, the t-test, was performed on linear regression relationships
between annual ET0 and the climatic factors, and the values of the significance level, α, are also
listed in Table 1.
TE
D
Table 1 Correlation between annual ET0 and climatic factors during period from 1961 to 2010
Climatic factors
Tmean
RH
SD
WS
R2
0.20
α
<0.50
0.51
0.75
0.06
>0.50
<0.02
<0.50
AC
C
EP
The correlation between annual ET0 and SD is significant with α < 0.02. Meanwhile, the
value of R2 between annual ET0 and SD is the highest one of the four R2 values. Though the
R2 between annual ET0 and RH is the second highest, the correlation between annual ET0 and
RH is insignificant, with α > 0.50. As for Tmean and WS, the significance levels of correlations
between annual ET0 and Tmean and annual ET0 and WS are both less than 0.50. However, the
values of R2 between annual ET0 and Tmean and between annual ET0 and WS are quite low,
indicating that ET0 is more sensitive to SD in the XJRB for the period from 1961 to 2010.
3.3 Prediction of ET0 during period from 2011 to 2100
3.3.1 Calibration and validation of SDSM for ET0 downscaling
In this study, the NCEP reanalysis data for the period from 1961 to 2010 were divided into
two periods, and the data of the first 30 years (1961 to 1990), which was determined to be the
calibration period, were used to construct downscaling models. Based on the multiple regression
equations between the selected predictors and predictand ( ET0 ) for each station, the data of the
later 20 years (1991 to 2010), which was determined to be the validation period, were used to
validate the regression models. The performance of SDSM for ET0 downscaling could be
evaluated through comparison with the ET0 series calculated by the FAO P-M method.
Six of the 26 predictors of the NCEP reanalysis data were selected as the inputs for the
SDSM model; they are listed in Table 2 and are similar to the predictors used in previous studies
ACCEPTED MANUSCRIPT
(Li et al., 2012; Wang et al., 2013). Taking the Yueyang Station as an example, the correlations
between ET0 and the six station-scale NCEP predictors are presented in Table 2. It can be seen
that the absolute value of correlation coefficients ranges from 0.56 to 0.75, indicating strong
correlation between the six NCEP predictors and ET0 . Thus, it is feasible to downscale ET0 with
these predictors.
Selected predictor
Abbreviation
1
850 hPa specific
humidity
Mean sea level
pressure
500 hPa zonal
velocity
500 hPa geopotential
height
Air temperature at
850 hPa
Air temperature at 2
m height
nceph_8s
2
3
4
5
6
ncepmslp
ncepp5_u
ncepp500
ncept850
nceptemp
Correlation
coefficient
0.56
–0.68
–0.70
0.61
0.75
SC
No.
RI
PT
Table 2 Correlations between predictors and ET0 taking Yueyang Station as an example
0.74
M
AN
U
Using predictors listed in Table 2, daily ET0 during the period from 1961 to 2010 was
downscaled. The performance of SDSM in daily ET0 downscaling was evaluated, and it is
presented in Table 3. The comparison of monthly ET0 series calculated by the FAO P-M method
( ET0 -PM) and downscaled with the SDSM model ( ET0 -SDSM) in the validation period is shown
in Fig. 5. According to Table 3, Re values in both calibration and validation periods were within
the acceptable range, and R2 and NSE are all above 0.84, demonstrating the strong applicability
of the SDSM in ET0 downscaling in the study area. Thus, the established SDSM can be used to
predict future ET0 in the XJRB.
Table 3 Performance of SDSM in daily ET0 downscaling
Period
Re (%)
0.10
7.16
R2
0.87
0.86
NSE
0.87
0.84
EP
TE
D
Calibration
Validation
AC
C
Fig. 5 Comparison of ET0 -PM with ET0 -SDSM in validation period
3.3.2 Prediction of ET0 during period from 2011 to 2100
Based on the calibration and validation of the SDSM model described above, ET0 during
the period from 2011 to 2100 in the XJRB was predicted. As shown in Fig. 6, significant upward
trends in the mean annual ET0 of the XJRB are detected under both the RCP45 and the RCP85
scenarios.
RI
PT
ACCEPTED MANUSCRIPT
Fig. 6 Prediction of mean annual ET0 of XJRB under RCP45 and RCP85 scenarios during period from 2011 to
2100
M
AN
U
SC
Taking the calibration period (1961 to 1990) of SDSM model as a reference period, changes
in the mean annual ET0 in three future periods, 2011 to 2040, 2041 to 2070, and 2071 to 2100,
under the RCP45 and RCP85 scenarios are shown in Table 4. The mean annual ET0 of the XJRB
in the reference period is 1 040.5 mm, and ET0 will increase by 9.8%, 12.6%, and 15.6% under
the RCP45 scenario, and 10.2%, 19.1%, and 27.3% under the RCP85 scenario in the periods of
2011 to 2040, 2041 to 2070, and 2071 to 2100, respectively. For the mean annual ET0 of the
XJRB, the increasing of predicted value under the RCP85 scenario is greater than that under the
RCP45 scenario.
Table 4. Change rates of future mean annual ET0 of XJRB under RCP45 and RCP85 scenarios compared with
reference period
Period
2011–2040
2041–2070
2071–2100
Change rate (%)
RCP45 scenario
9.8
12.6
15.6
RCP85 scenario
10.2
19.1
27.3
AC
C
EP
TE
D
Furthermore, the predicted variation ranges of the mean annual ET0 of the XJRB in three
future periods are presented in Fig. 7, where the dotted line in Fig. 7 refers to the mean annual
ET0 of the XJRB in the reference period. In the period from 2011 to 2040, the mean annual ET0
of the XJRB under the RCP45 scenario is similar to that under the RCP85 scenario. A difference
occurs in the period from 2041 to 2070, when the mean annual ET0 of the XJRB under the
RCP85 scenario, with a change rate of 4.9% to 37.6% as compared with the reference period, is
clearly higher than that under the RCP45 scenario, with a change rate of 1.8% to 22.6% as
compared with the reference period. In the period from 2071 to 2100, the gap is further widened,
because the RCP45 scenario represents medium-low RCP, and the RCP85 scenario represents high
RCP.
M
AN
U
SC
RI
PT
ACCEPTED MANUSCRIPT
Fig. 7 Variation ranges of predicted mean annual ET0 for three periods under RCP45 and RCP85 scenarios
4. Conclusions
AC
C
EP
TE
D
In this study, the spatial distribution and temporal trend of ET0 during the period from 1961
to 2010 were examined, and future ET0 during the period from 2011 to 2100 was downscaled
from CMIP5 outputs under two emission scenarios (the RCP45 and RCP85 scenarios) using the
SDSM. The major conclusions are as follows:
(1) The mean annual ET0 of the XJRB is 1 060.3 mm during the period from 1961 to 2010,
the lowest mean annual ET0 (768.1 mm) is found near the center of the basin, and the highest
mean annual ET0 (1 087.1 mm) appears in the southwest part of the basin. Of the 14
meteorological stations selected in this study, downward trends in annual ET0 are detected at
nine of them, four stations of which present significant downward trends. Five stations present
upward trends, but only one of them shows a significant upward trend. The spatial variation of
ET0 is caused by the combined effect of the related climatic factors. As for the mean annual ET0
of the XJRB, an insignificant downward trend with a change rate of –0.33 mm/year was detected
during the period from 1961 to 2010. The temporal trend in annual ET0 is mainly caused by the
significant downward trend in SD during the period from 1991 to 2010.
(2) In both the calibration and validation periods, the SDSM performs well in ET0
downscaling in the study area. The comparison of monthly ET0 series downscaled by the SDSM
with that calculated by the FAO P-M method shows Re , R2 , and NSE values of 0.10%, 0.87, and
0.87 in the calibration period and 7.16%, 0.86, and 0.84 in the validation period, respectively. In
general, the SDSM can be used to predict future ET0 in the XJRB.
(3) Using CMIP5 outputs, upward trends in annual ET0 were predicted under the RCP45
and RCP85 scenarios during the period from 2011 to 2100. The increase under the RCP85
scenario will be greater than that under the RCP45 scenario. Compared with the reference period
(1961 to 1990), ET0 will increase by 9.8%, 12.6%, and 15.6% under the RCP45 scenario and
10.2%, 19.1%, and 27.3% under the RCP85 scenario in the periods from 2011 to 2040, 2041 to
2070, and 2071 to 2100, respectively.
ACCEPTED MANUSCRIPT
References
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
Alan, D.Z., Justin, S., Edwin, P.M., Bart, N., Eric, F.W., Dennis, P.L., 2003. Detection of intensification in
global-and continental-scale hydrological cycles: Temporal scale of evaluation. Journal of Climate 16(3),
535–547. http://dx.doi.org/ 10.1175/1520-0442(2003)016<0535:DOIIGA>2.0.CO;2
Allan, C., Xia J., Pahl-Wostl, C., 2013. Climate change and water security: challenges for adaptive water
management. Current Opinion in Environmental Sustainability 5(6), 625–632. http://dx.doi.org/
10.1016/j.cosust.2013.09.004
Allen, R.G. Pereira, L.S., Raes. D., 1998. Crop evapotranspiration: Guidelines for computing crop water
requirements. In: FAO Irrigation and Drainage Paper No. 56. FAO, Rome.
Allen, M.R., Ingram, W.J., 2002. Constraints on future changes in climate and the hydrologic cycle. Nature
419(6903), 224–232. http://dx.doi.org/ 10.1038/nature01092.
Bandyopadhyay, A., Bhadra, A., Raghuwanshi, N.S., Singh, R., 2009. Temporal trends in estimates of reference
evapotranspiration over India. Journal of Hydrologic Engineering 14(5), 508–515. http://dx.doi.org/
10.1061/(ASCE)HE.1943–5584.0000006.
Bates, B., Kundzewicz, Z.W., Wu, S., Palutikof, J., 2008. Climate Change and Water. Intergovernmental Panel on
Climate Change (IPCC). Geneva.
Burn, D.H., Elnur, M.A.H., 2002. Detection of hydrologic trends and variability. Journal of hydrology 255(1),
107–122. http://dx.doi.org/10.1016/S0022-1694(01)00514-5.
Chu, J.T., Xia, J., Xu, C.Y., Singh, V.P., 2010. Statistical downscaling of daily mean temperature, pan evaporation
and precipitation for climate change scenarios in Haihe River, China. Theoretical and Applied Climatology
99(1-2), 149–161. http://dx.doi.org/10.1007/s00704-009-0129-6
Durack, P.J., Wijffels, S.E., Matear, R.J., 2012. Ocean salinities reveal strong global water cycle intensification
during 1950 to 2000. Science 336 (6080), 455–458.
Fisher, J.B., Whittaker, R.J., Malhi, Y., 2011. ET come home: potential evapotranspiration in geographical
ecology. Global Ecology and Biogeography 20(1), 1–18. http://dx.doi.org/10.1111/j.1466-8238.2010.00578.x.
Fowler, H.J., Wilby. R.L., 2007. Beyond the downscaling comparison study. International Journal of Climatology
27(12), 1543–1545. http://dx.doi.org/10.1002/joc.1616.
Gao, G., Chen, D., Xu, C.Y., Simelton, E., 2007. Trend of estimated actual evapotranspiration over China during
1960–2002. Journal of Geophysical Research 112(D11). http://dx.doi.org/10.1029/2006JD008010.
Gao, G., Xu, C.Y., Chen, D.L., Singh, V.P., 2012. Spatial and temporal characteristics of actual evapotranspiration
over Haihe River basin in China estimated by the complementary relationship and the Thornthwaite water
balance model. Stochastic Environmental Research and Risk Assessment 26(5), 655–669. http://dx.doi.org/
10.1007/s00477-011-0525-1.
Hamed, K.H., Rao, A.R., 1998. A modified Mann-Kendall trend test for autocorrelated data. Journal of Hydrology
204(1-4), 182–196. http://dx.doi.org/10.1016/S0022-1694(97)00125-X.
IPCC, 2014. Climate Change 2014: Impacts, Adaptation, and Vulnerability. Cambridge University Press, New
York.
Kendall, M.G., 1975. Rank Correlation Methods. Charles Griffin, London.
Li, Z., Zheng, F.L. Liu, W.Z., 2012. Spatiotemporal characteristics of reference evapotranspiration during
1961-2009 and its projected changes during 2011-2099 on the Loess Plateau of China. Agricultural and Forest
Meteorology 154–155(6), 147–155. http://dx.doi.org/10.1016/j.agrformet.2011.10.019.
Liu, W., Peng, S., Wang, W., 2012. Reference evapotranspiration changes in the Haihe River basin during past 50
years. Procedia Engineering 28(5), 258–263. http://dx.doi.org/10.1016/j.proeng.2012.01.716.
Mann, H.B., 1945. Non-parametric tests against trend. Econometrica 13, 245–259.
Mitchell, M., Curtis, A., Sharp, E., Mendham, E., 2012. Directions for social research to underpin improved
groundwater management. Journal of Hydrology 448(15), 223–231. http://dx.doi.org/10.1016/j.jhydrol.
2012.04.056.
Nash, J., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models, part I: A discussion of principles.
Journal of Hydrology 10(3), 282–290. http://dx.doi.org/10.1016/0022-1694(70)90255-6.
Piao, S., Ciais, P., Huang, Y., Shen, Z., Peng, S., Li, J., Fang, J., 2010. The impacts of climate change on water
resources and agriculture in China. Nature 467(7311), 43–51. http://dx.doi.org/10.1038/nature09364.
Wang, W.G., Shao, Q.X., Peng, S.Z., Xing, W.Q., Yang, T., Luo, Y.F., Yong, B., Xu, J.Z., 2012. Reference
evapotranspiration change and the causes across the Yellow River Basin during 1957–2008 and their spatial
and seasonal differences. Water Resources Research 48(5), W05530. http://dx.doi.org/10.1029/2011W
R010724.
Wang, W.G., Xing, W.Q., Shao, Q.X., Yu, Z.B., Peng, S.Z., Yang, T., Yong, B., Taylor, J., Singh, V.P., 2013.
Changes in reference evapotranspiration across the Tibetan Plateau: Observations and future projections based
ACCEPTED MANUSCRIPT
AC
C
EP
TE
D
M
AN
U
SC
RI
PT
on statistical downscaling. Journal of Geophysical Research: Atmospheres 118(10), 4049-4068.
http://dx.doi.org/10.1002/jgrd.50393.
Wang, Y., Jiang, T., Bothe, O., Fraedrich, K., 2007. Changes of pan evaporation and reference evapotranspiration
in the Yangtze River basin. Theoretical and Applied Climatology 90(1), 13–23. http://dx.doi.org/
10.1007/s00704-006-0276-y.
Wentz, F.J., Ricciardulli, L., Hilburn, K., Mears, C., 2007. How much more rain will global warming bring?
Science. 317(5835), 233–235.
Wilby, R.L., Dawson, C.W., Barrow, E.M., 2002. SDSM: A decision support tool for the assessment of regional
climate
change
impacts.
Environmental
Modelling
and
Software.
17(2),
145–157.
http://dx.doi.org/10.1016/S13 64-8152(01)00060-3.
Wilby, R.L., Dawson, C.W., 2007. SDSM 4.2: A Decision Support Tool for the Assessment of Regional Climate
Change Impacts, User Manual. Department of Geography, Lancaster University, Lancaster.
Xu, C.Y. Singh, V.P., 2005. Evaluation of three complementary relationship evapotranspiration models by water
balance approach to estimate actual regional evapotranspiration in different climatic regions. Journal of
Hydrology 308, 105–121. http://dx.doi.org/10.1016/j.jhydrol.2004.10.024.
Xu, C.Y., Widén E., Halldin. S., 2005. Modelling hydrological consequences of climate change-progress and
challenges. Advances in Atmospheric Sciences 22(6), 789–797. http://dx.doi.org/10.1007/BF02918679.
Xu, C.Y., Gong, L.B., Jiang, T., Chen, D.L., Singh, V.P., 2006. Analysis of spatial distribution and temporal trend
of reference evapotranspiration and pan evaporation in Changjiang (Yangtze River) catchment. Journal of
Hydrology 327(1-2), 81–93. http://dx.doi.org/10.1016/j.jhydrol.2005.11.029.
Ye, X.C., Li, X.H., Liu, J., Xu, C.Y., Zhang, Q., 2013. Variation of reference evapotranspiration and its
contributing climatic factors in the Poyang Lake catchment, China. Hydrological Processes 28(25), 6151–6162.
http://dx.doi.org/10.1002/hyp.10117.
Zhang, Q., Xu, C.Y., Chen, X.H., 2011. Reference evapotranspiration changes in China: Natural processes or
human influences? Theoretical and Applied Climatology 103(3-4), 479–488. http://dx.doi.org/ 10.1007/s00704
-010-0315-6.
Zhang, Z., Chen, Y., Wang, P., Shuai, J.B., Tao, F.L., Shi, P.J., 2014. River discharge, land use change, and
surface water quality in the Xiangjiang River, China. Hydrological Processes 28(13), 4130–4140.
http://dx.doi.org/10.1002/hyp.9938.
Zuo, D.P., Xu, Z.X., Yang, H., Liu, X.C., 2012. Spatiotemporal variations and abrupt changes of potential
evapotranspiration and its sensitivity to key meteorological variables in the Wei River basin, China.
Hydrological Processes 26(8), 1149–1160. http://dx.doi.org/10.1002/hyp.8206.