Agricultural and Forest Meteorology 233 (2017) 183–194

Contents lists available at ScienceDirect

Agricultural and Forest Meteorology

journal homepage: www.elsevier.com/locate/agrformet

Spatiotemporal change and trend analysis of potential

evapotranspiration over the Loess Plateau of China during 2011–2100
Shouzhang Peng a,b , Yongxia Ding c , Zhongming Wen a,b , Yunming Chen a,b,∗ , Yang Cao a,b ,
Jingyu Ren d
a
State Key Laboratory of Soil Erosion and Dryland Farming on the Loess Plateau, Northwest A&F University, Yangling 712100, China
b
Institute of Soil and Water Conservation, Chinese Academy of Sciences and Ministry of Water Resources, Yangling 712100, China
c
College of Earth Environmental Sciences, Lanzhou University, Lanzhou 730000, China
d
Institute of Soil and Water Conservation, Northwest A&F University, Yangling 712100, China

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

Article history: The spatiotemporal change and trend of annual potential evapotranspiration (PET) over the Loess Plateau
Received 13 September 2016 of China from 2011 to 2100 are assessed in this work. PET is calculated using the Hargreaves model
Received in revised form with monthly mean, maximum, and minimum temperatures and a 1 km spatial resolution, which are
19 November 2016
generated using the Delta downscaling method and general circulation models (GCMs) with four rep-
Accepted 24 November 2016
resentative concentration pathway (RCP) scenarios. The PET trend is detected via Mann–Kendall and
Available online 30 November 2016
Sen’s slope estimator tests. The following results are drawn: (1) the Delta downscaling method shows
a favorable performance in detecting GCM monthly temperatures based on the mean absolute error
Keywords:
Potential evapotranspiration
and regression analysis between downscaled data and independent surface observations. Among the 28
Delta downscaling GCMs, the NorESM1-M and GISS-E2-R models show the best performance in reproducing the monthly
Hargreaves model mean/maximum and minimum temperatures over the Loess Plateau, respectively; (2) the average
Mann–Kendall trend test annual PET over this region will increase by 12.7%–23.9% from 1961 to 1990 to the end of this century
Sen’s slope estimator (2071–2100). However, these increments show strong spatial variations; (3) the annual PET during the
Loess plateau 2011–2100 period at each grid of the region demonstrates a significantly increasing trend under each RCP
scenario, while RCP2.6, RCP4.5, RCP6.0, and RCP8.5 have average magnitudes of trend of 10.4 mm/10yr,
17.7 mm/10yr, 21 mm/10yr, and 29.7 mm/10yr, respectively; (4) the annual PET with significant trends
during the other three periods (2011–2040, 2041–2070, and 2071–2100) present various spatial distribu-
tions in their magnitudes of trend under the aforementioned RCP scenarios. RCP2.6 showed a significant
decrease during 2041–2070 and 2071–2100, although such trends are only observed at 0.3% and 1.2% of
the Loess Plateau, respectively; and (5) the spatial results provide some information, such as locations and
area ratios, which are valuable in assessing future PET changes and trends. These spatiotemporal results
represent the PET changes and trends in detail and provide insights for developing flexible adaptation
and mitigation strategies to combat the effects of global warming in this region.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction annual precipitation and air temperature over the LP decreased and
increased, respectively (Bi et al., 2009; Miao et al., 2011; Sun et al.,
The Loess Plateau (LP) of China is known for its severe 2015; Wang et al., 2012), thereby exacerbating the water shortage
soil erosion (Sun et al., 2015). Given that most areas in this and threatening the development of this region (Li et al., 2012a).
region have semiarid and sub-humid climates with an annual Under the effects of global warming, the temperature over the LP
mean precipitation (1961–2009) of 143.6 mm–811.8 mm along the will continuously increase throughout the century (Li et al., 2012b)
southeast–northwest direction, the LP is also threatened by severe and subsequently accelerate the water loss in this region (Aouissi
water shortage (Li et al., 2012a). Over the past 50 years, the et al., 2016). Potential evapotranspiration (PET) is a key hydrologi-
cal variable for quantifying water loss in a region; this variable can
be used to calculate actual evapotranspiration, schedule irrigation
∗ Corresponding author at: No. 26, Xinong Road, Institute of Soil and Water Con- projects, and prepare input data for hydrological models (Aouissi
servation, Yangling, Shaanxi 712100, China. et al., 2016; Li et al., 2012a). Therefore, the future PET in the LP
E-mail address: [email protected] (Y. Chen).

http://dx.doi.org/10.1016/j.agrformet.2016.11.129
0168-1923/© 2016 Elsevier B.V. All rights reserved.
184 S. Peng et al. / Agricultural and Forest Meteorology 233 (2017) 183–194

must be investigated to develop suitable adaptation and mitigation amount of information on the detailed spatial variations of PET
strategies for combating the effects of global warming. change and trend over the LP during future periods hinders the
Studies on PET have a long history and have proposed sev- development of flexible adaptation and mitigation strategies to
eral calculation methods, including the Penman–Monteith (Allen combat the effects of global warming in this region.
et al., 1998), Hargreaves (Hargreaves and Samani, 1985), and This study aims (1) to generate future monthly mean, maxi-
Priestly–Taylor methods (Priestley and Taylor, 1972), which can mum, and minimum temperature data with high spatial resolution
indirectly calculate PET using meteorological data (Aouissi et al., using the Delta downscaling method and GCMs data, and (2) to
2016). The Penman–Monteith method has been considered as a calculate and assess future annual PET changes, trends, and their
universal standard for estimating PET (Allen et al., 1998; Aouissi spatial variations based on the downscaled temperature data dur-
et al., 2016). However, this method requires five daily weather ing 2011–2100.
parameters that influence evaporation, namely, solar radiation,
minimum and maximum air temperature, relative humidity, and
2. Data and methods
wind speed. The other methods also provide acceptable results
(Arnold et al., 1998). Considered the best method for estimating PET
2.1. Study area
in the LP (Zhao et al., 2004), the Hargreaves method only depends
on three monthly climate parameters, namely, mean, maximum,
The LP region is situated in North China (33.7◦ –41.3◦ N,
and minimum temperatures (Hargreaves and Samani, 1985).
100.8◦ –114.6◦ E) and traversed by the upper–middle reaches of the
General circulation models (GCMs) provide valuable infor-
Yellow River (Fig. 1). The region starts from the Taihang Mountains
mation on long-term temperature projections at a global or
in the east, reaches the Riyue Mountains in the west, and borders on
sub-continental scale (IPCC, 2013). The Coupled Model Intercom-
the Qinling and Yin Mountains in the south and north, respectively
parison Project Phase5 (CMIP5) of the World Climate Research
(Liu et al., 2016). The LP has a warm or temperate continental mon-
Program provides the state-of-the-art multi-model dataset used
soon climate with extensive monsoonal influence. Approximately
by the Intergovernmental Panel on Climate Change for its fifth
60%–70% annual precipitation occurs in this region from June to
assessment report. Most of the extant GCMs have horizontal reso-
September in the form of high-intensity storms (Wang et al., 2012).
lutions of a few hundred kilometers (Meehl et al., 2007), and several
The annual mean temperature ranges from 3.6 ◦ C in the north-
researchers have attempted to downscale CMIP5 model datasets to
west to 14.3 ◦ C in the southeast. The region demonstrates a large
a 0.5◦ (approximately 55 km) (Brekke et al., 2013). However, spatial
diurnal temperature range throughout the year and faces dry and
resolution limits the ability of GCM datasets to represent complex
cold winters, hot and humid summers, rapid temperature reduc-
topography, land surface characteristics, and other processes in the
tion during autumn, and rapid temperature increase during spring.
climate system (Xu et al., 2016). Therefore, GCM datasets cannot
The annual potential evaporation in this area (865 mm–1274 mm)
draw realistic and reliable temperature change data at fine scales,
is estimated to exceed the precipitation level (Li et al., 2012a). The
which are required when developing suitable adaptation and mit-
region spans arid, semiarid, and semi-humid zones and is consid-
igation strategies at the regional-to-local scale (Giorgi et al., 2009).
ered as a semiarid-to-semihumid transitional zone that is sensitive
Dynamic or statistical downscaling models for GCM data are neces-
to climate change (Liu and Sang, 2013).
sary for downscaling a temperature dataset because these models
add values to the data (Mosier et al., 2014; Xu et al., 2016). Dynamic
downscaling requires numerous inputs and computational require- 2.2. Data collection
ments (Brekke et al., 2013; Xu et al., 2016) and sometimes cannot
represent reliable temperature change data at fine scales compared GCM monthly mean, maximum, and minimum tempera-
with statistical downscaling (Ahmed et al., 2013; Dosio et al., 2015; ture data were obtained from the downscaled CMIP5 dataset
Xu et al., 2016). Therefore, statistical downscaling can generate a with a 0.5◦ resolution (http://gdo-dcp.ucllnl.org/downscaled cmip
high spatial resolution temperature dataset for drawing PET infor- projections/). This dataset was processed from raw 28 GCM data
mation of the LP at a fine scale. using the bias-correction spatial disaggregation (BCSD) method
The statistical downscaling framework often involves regres- described in Brekke et al. (2013). Table 1 lists the basic information
sion and Delta downscaling methods. The former constructs of the 28 GCM models and the associated institutions. This dataset
multiple linear regression relationships between raw GCM data covers the period from January 1950 to December 2100; the data
and station observations in the historical period, and then applies from 1950 to 2005 are historical data under the historical emis-
these relationships to future GCM grid outputs (Li et al., 2012b; sions scenario, while those from 2006 to 2100 are future data under
Timm et al., 2015). This framework generates substantial climate future representative concentration pathway (RCP) scenarios (i.e.,
element data at target stations in future periods, but cannot con- RCP2.6, RCP4.5, RCP6.0, and RCP8.5).
struct a high spatial resolution future climate dataset that prevents The high-resolution reference climatologies were presented as
drawing detailed spatial variations at fine scales. The latter uses a grid data with a 1 km (approximately 30 ) resolution as suggested
low-resolution monthly time series and high-resolution reference by the Chinese Ecosystem Research Network (CERN, www.cnern.
climatology as inputs; the high-resolution climatology input must org.cn). These climatologies were processed using spatial interpola-
present a physically representative, fine-scale distribution of the tion and geographic information systems that covered 740 national
meteorological variable over the landscape (Brekke et al., 2013; weather stations in China and a 1 km digital elevation model. These
Mosier et al., 2014). Unlike the direct interpolation of low spatial instruments covered each month from 1961 to 2000.
resolution sources to a higher spatial resolution, the Delta down- The surface observation data for the monthly mean, maximum,
scaling process incorporates high-resolution orographic effects and minimum temperatures were obtained from 113 national
that are not represented in low-resolution input grids (Mosier et al., weather stations in and around the LP (Fig. 1) (http://data.cma.
2014). cn/); these data were used to evaluate the downscaled monthly
Previous studies have assessed the future PET change and trend temperature grids from January 1991 to December 2005 (period of
over the LP (Li et al., 2012a). However, these studies were conducted the available temperature data). The surface observation data for
at the station scale, used the regression downscaling framework to monthly pan evaporation were obtained from 26 evaporation pans
obtain future climate data, and adopted a geo-statistical interpola- belonging to the national weather stations in and around the LP
tion method to describe the spatial variations of PET. The limited (Fig. 1); these data were used to evaluate the calculated monthly
 S. Peng et al. / Agricultural and Forest Meteorology 233 (2017) 183–194 185

Table 1
Summary of 28 general circulation models from CMIP5.

Model acronym Institution References

1 ACCESS1.0 Commonwealth Scientific and Industrial Research Organization and Marsland et al. (2013)
Bureau of Meteorology, Australia
2 BCC-CSM1.1 Beijing Climate Center, China Meteorological Administration, China Xin et al. (2013)
3 BCC-CSM1.1(m) Ren et al. (2016)
4 BNU-ESM Beijing Normal University, China Ji et al. (2014)
5 CanESM2 Canadian Centre for Climate Modelling and Analysis, Canada Chylek et al. (2011)
6 CESM1-BGC NSF/DOE NCAR, USA Long et al. (2013)
7 CESM1-CAM5 Neale et al. (2013)
8 CMCC-CM Centro Euro-Mediterraneo per I Cambiamenti Climatici, Italy Scoccimarro et al. (2011)
9 CNRM-CM5 Centre National de Recherches Meteorologiques, Meteo-France, France Voldoire et al. (2013)
10 CSIRO-MK-3.6.0 Australian Commonwealth Scientific and Industrial Research Rotstayn et al. (2010)
Organization, Australia
11 EC-EARTH EC-EARTH consortium, Europe Hazeleger et al. (2012)
12 FGOALS-g2 Institute of Atmospheric Physics, Chinese Academy of Sciences, China Zhou et al. (2013)
13 FIO-ESM The First Institution of Oceanography, SOA, China Qiao et al. (2013)
14 GFDL-CM3 NOAA Geophysical Fluid Dynamics Laboratory, USA Donner et al. (2011)
15 GFDL-ESM2G Dunne et al. (2012)
16 GFDL-ESM2M Dunne et al. (2012)
17 GISS-E2-H-CC NASA Goddard Institute for Space Studies, USA Wang et al. (2015)
18 GISS-E2-R Schmidt et al. (2006)
19 GISS-E2-R-CC Wang et al. (2015)
20 HadCM3 Met Office Hadley Centre, UK Collins et al. (2001)
21 INMCM4.0 Institute for Numerical Mathematics, Russia Volodin et al. (2010)
22 IPSL-CM5A-LR Institut Pierre-Simon Laplace, France Dufresne et al. (2013)
23 MIROC4 h Japan Agency for Marine-Earth Science and Technology, University of Zhang et al. (2015)
24 MIROC5 Tokyo, and National Institute for Environmental Studies (Japan) Watanabe et al. (2010)
25 MIROC-ESM Watanabe et al. (2011)
26 MIROC-ESM-CHEM Watanabe et al. (2011)
27 MRI-CGCM3 Meteorological Research Institute, Japan Yukimoto et al. (2012)
28 NorESM1-M Norwegian Climate Centre, Norway Bentsen et al. (2013)

Fig. 1. Location and digital elevation model of Loess Plateau region in China, 131 national weather stations and 26 evaporation pans distributed in and around the region.
186 S. Peng et al. / Agricultural and Forest Meteorology 233 (2017) 183–194

Fig. 2. Schematic illustration of spatial downscaling process. The downscaled mean temperature (MeanT) field is constructed from GCM data in July 2005.

PET grids from January 1985 to December 2005 (period of the avail- interpolation methods, including nearest-neighbor, bilinear, cubic
able pan evaporation data). spline, and bicubic interpolations. The nearest-neighbor interpo-
lation uses the values of nearby grids to estimate the value of
the original grid. Bilinear interpolation fits a linear function over
2.3. Delta downscaling procedure each interval on the original grid in one dimension and in another.
Cubic spline interpolation fits a “natural” spline (i.e., without a
The Delta downscaling method was applied on the afore men- tension parameter) to each interval. Bicubic interpolation is an
tioned data to produce monthly mean, maximum, and minimum extension of cubic interpolation that interpolates the value on a
temperature grids with a 1 km resolution from January 1950 to two-dimensional regular grid. The interpolated surface is smoother
December 2100. A part of the BCSD method, the Delta downscal- than the corresponding surfaces obtained via bilinear or nearest-
ing process has been described in many studies (Maurer et al., neighbor interpolation. These four interpolation methods were
2002; Ning et al., 2015; Wood et al., 2002). Fig. 2 shows a rect- conducted using the interp2 function of MATLAB.
angular region (red zone in Fig. 1) that covers the LP region and
illustrates the components and steps of the Delta downscaling pro-
cess for determining mean temperature using the GCM 0.5◦ time 2.4. Evaluation of the Delta downscaling method
series and CERN 30 climatology datasets. The first step (Fig. 2(a))
constructs a 0.5◦ climatology for each month from the 0.5◦ time- The spatial downscaling process for temperature was conducted
series dataset. The low-resolution climatology is produced using over the red zone in Fig. 1. Surface observation records from 113
the 1961–2000 period, which is also used by CERN to construct its national weather stations across the red zone were used to verify
climatology data. A 0.5◦ anomaly (Fig. 2(b)) is then calculated. The the accuracy of each downscaled grid. The observation records from
anomaly for temperature is computed as the difference between January 1991 to December 2005 were used to verify the downscaled
the time series element and climatology. The anomaly is then inter- results.
polated to the 30 CERN grid using the spatial interpolation method The average model-performance error was quantified using
(Fig. 2(c)). The final step (Fig. 2(d)) transforms the high-resolution mean absolute error (MAE), which is a more natural measure of
anomaly back to an absolute surface by scaling this anomaly using average error compared with the root-mean-square error (Wang
the CERN climatology for the corresponding month. This transfor- and Chen, 2014; Willmott and Matsuura, 2005). This accuracy esti-
mation undoes the creation of the anomaly; therefore, addition is mator can be mathematically expressed as follows:
used for temperature.
The anomaly grid from the original to high-resolution coordi- 1 n
MAE = × × |Pi − Oi |, (1)
nates, as illustrated in Fig. 2(c), can be interpolated using several n i=1
 S. Peng et al. / Agricultural and Forest Meteorology 233 (2017) 183–194 187

Fig. 3. Comparisons of the observed (x) and downscaled (y) values of monthly mean temperature (MeanT, ◦ C), maximum temperature (MaxT, ◦ C), and minimum temperature
(MinT, ◦ C) at the four seasons (spring: from March to May; summer: from June to August; autumn: from September to November; and winter: from December to February).
The MeanT/MaxT and MinT values during 1/1991–12/2005 are downscaled from the NorESM1-M and GISS-E2-R models, respectively, using the bilinear interpolation method.

where Pi and Oi are the downscaled and observed values, respec- pendent data (Gocic and Trajkovic, 2013). Two non-parametric
tively, and n is the number of records of all validated stations. methods (Mann–Kendall and Sen’s slope estimator tests) were used
to detect the PET trends over the LP in each grid from 2011 to
2.5. Calculation of PET 2100. The Mann–Kendall test provides a measure (ZS ) that indi-
cates whether the long-term change of a variable is significant or
Monthly PET (mm/month) is calculated via the Hargreaves not (Atta-ur-Rahman and Dawood, 2016). A comparison analysis
method and by using monthly mean, maximum, and minimum was conducted at the 95% confidence level. The time series trend
temperatures. Several improvements were applied to the original was significant at the 95% confidence level when |ZS | > 1.96. ZS > 1.96
equation (Hargreaves and Samani, 1982, 1985). The form used in indicates a significant increase, while ZS < −1.96 indicates a signif-
this study was published in 1985 (Hargreaves and Samani, 1985) icant decrease. The magnitude of the PET trend was calculated via
and expressed as follows: Sen’s slope estimator test (Atta-ur-Rahman and Dawood, 2016).

PET = 0.0023 × S0 × (MaxT − MinT)0.5 × (MeanT + 17.8) , (2)

3. Results
where MaxT is the maximum air temperature for a specific month
(◦ C), MinT is the minimum air temperature for a specific month (◦ C), 3.1. Evaluation of downscaled temperature and calculated PET
MeanT is the mean air temperature for a specific month (◦ C), and S0
is the extra-terrestrial solar radiation (mm/month) at the top of the Table 2 shows the MAE between the downscaled and observed
Earth’s atmosphere on a horizontal surface (Allen et al., 1998). S0 monthly MeanT, MaxT, and MinT under the four interpolation
can be calculated according to the solar constant, number of days methods. The MAE under bilinear interpolation is the smallest
in a year, latitude, and solar decimation. The detailed calculation for each GCM although such error slightly differs across the four
procedure has been described in Allen et al. (1998) and Zhao et al. interpolation methods. Similarly, the MAE under the bilinear inter-
(2004). polation ranges from 1.342 ◦ C to 1.599 ◦ C for monthly MeanT, from
1.715 ◦ C to 1.978 ◦ C for monthly MaxT, and from 1.336 ◦ C to 5.527 ◦ C
2.6. Trend analysis methods for monthly MinT. The NorESM1-M model demonstrates the best
performance in reproducing the monthly MeanT and MaxT over the
The significant trends in climatologic time series can be LP, while the GISS-E2-R model demonstrates the best performance
tested using parametric and non-parametric methods. Paramet- in reproducing the monthly MinT in the LP. The regression analy-
ric trend tests require the data to be independent and normally sis reveals that the monthly temperature values downscaled by the
distributed, while non-parametric trend tests only require inde- most suitable GCMs data are close to the observed values at the four
188 S. Peng et al. / Agricultural and Forest Meteorology 233 (2017) 183–194

Table 2
MAE between the observed and downscaled monthly MeanT, MaxT, and MinT from 113 national weather stations during 1/1991–12/2005.

Monthly MeanT (◦ C) Monthly MaxT (◦ C) Monthly MinT (◦ C)

Bicubic Bilinear Nearest Spline Bicubic Bilinear Nearest Spline Bicubic Bilinear Nearest Spline

ACCESS1.0 1.460 1.458 1.462 1.460 1.824 1.822 1.824 1.824 1.442 1.439 1.444 1.442
BCC-CSM1.1 1.419 1.418 1.419 1.419 1.780 1.779 1.781 1.780 1.415 1.414 1.416 1.415
BCC-CSM1.1(m) 1.449 1.447 1.450 1.449 1.819 1.818 1.820 1.819 1.458 1.457 1.460 1.458
BNU-ESM 1.429 1.429 1.430 1.429 1.822 1.821 1.822 1.822 1.392 1.391 1.394 1.392
CanESM2 1.363 1.361 1.364 1.363 1.795 1.794 1.796 1.795 1.339 1.337 1.341 1.339
CESM1-BGC 1.434 1.432 1.435 1.434 1.753 1.751 1.754 1.753 1.482 1.480 1.483 1.482
CESM1-CAM5 1.392 1.390 1.392 1.392 1.718 1.716 1.718 1.718 1.415 1.412 1.416 1.415
CMCC-CM 1.484 1.482 1.484 1.484 1.891 1.889 1.891 1.891 1.455 1.453 1.456 1.455
CNRM-CM5 1.451 1.449 1.451 1.451 1.852 1.850 1.852 1.852 1.395 1.393 1.395 1.395
CSIRO-MK-3.6.0 1.424 1.422 1.425 1.424 1.803 1.801 1.803 1.803 1.372 1.370 1.373 1.372
EC-EARTH 1.444 1.442 1.444 1.444 1.865 1.864 1.866 1.866 1.411 1.409 1.412 1.411
FGOALS-g2 1.470 1.469 1.470 1.470 1.833 1.832 1.833 1.833 1.446 1.444 1.447 1.446
FIO-ESM 1.534 1.533 1.535 1.534 1.909 1.908 1.910 1.909 1.512 1.511 1.513 1.512
GFDL-CM3 1.383 1.382 1.384 1.384 1.772 1.770 1.772 1.772 1.389 1.387 1.389 1.389
GFDL-ESM2G 1.415 1.413 1.416 1.415 1.727 1.725 1.729 1.728 1.476 1.474 1.478 1.476
GFDL-ESM2M 1.425 1.423 1.426 1.425 1.809 1.807 1.809 1.809 1.405 1.403 1.406 1.405
GISS-E2-H-CC 1.499 1.498 1.500 1.499 1.921 1.919 1.922 1.921 1.423 1.421 1.424 1.423
GISS-E2-R 1.347 1.346 1.347 1.347 1.729 1.727 1.730 1.729 1.338 1.336 1.338 1.338
GISS-E2-R-CC 1.407 1.406 1.407 1.407 1.802 1.800 1.802 1.802 1.409 1.407 1.410 1.409
HadCM3 1.505 1.503 1.507 1.505 1.860 1.858 1.862 1.860 1.499 1.497 1.501 1.499
INMCM4.0 1.564 1.562 1.565 1.564 1.940 1.939 1.942 1.941 1.536 1.532 1.538 1.536
IPSL-CM5A-LR 1.549 1.548 1.550 1.549 1.725 1.723 1.725 1.725 1.557 1.555 1.559 1.557
MIROC4h 1.471 1.469 1.472 1.471 1.828 1.825 1.830 1.828 1.461 1.459 1.462 1.461
MIROC5 1.446 1.445 1.447 1.446 1.847 1.845 1.848 1.847 5.528 5.527 5.527 5.528
MIROC-ESM 1.379 1.377 1.380 1.379 1.788 1.786 1.788 1.788 1.388 1.386 1.391 1.388
MIROC-ESM-CHEM 1.600 1.599 1.600 1.600 1.980 1.978 1.980 1.980 1.598 1.597 1.599 1.598
MRI-CGCM3 1.571 1.570 1.571 1.571 1.949 1.947 1.950 1.949 1.553 1.551 1.553 1.553
NorESM1-M 1.343 1.342 1.344 1.343 1.717 1.715 1.718 1.717 1.341 1.339 1.341 1.341

Bicubic, Bilinear, Nearest, and Spline represent the bicubic, bilinear, nearest-neighbor, and cubic spline interpolations, respectively.

was divided into three periods (2011–2040, 2041–2070, and

2071–2100) to assess the spatiotemporal change and trend of
annual PET in the future period.
Fig. 5 shows the spatial patterns of mean annual PET changes in
the future period (2011–2100) under the four RCP scenarios relative
to the reference period (1961–1990). Under each RCP, the zones
that showed the largest changes were located in high-elevation
regions, such as Wutai Mountain in the northeast, Hua Mountain
in the south, and Qilian Mountains in the west. Table 3 shows the
corresponding Min, Max, Mean, and coefficient of variation (CV)
for each spatial distribution in Fig. 5. The CV indicates that RCP8.5
shows the largest spatial variation (10.9%), while RCP2.6 shows the
smallest spatial variation (9.2%). The difference between Max and
Min indicates that RCP8.5 shows the most extreme spatial variation
(20.6%; ranging from 11.9% to 32.6%), while RCP2.6 shows the most
moderate spatial variation (10.9%; ranging from 7.6% to 18.5%). The
Mean indicates that the PET over the LP changes by 9.5% for RCP2.6,
10.8% for RCP4.5, 10.1% for RCP6.0, and 14.1% for RCP8.5.
Fig. 6 shows the spatial patterns of the mean annual PET changes
in 2011–2040, 2041–2070, and 2071–2100 under the four RCP sce-
Fig. 4. Comparisons of the observed (x) and calculated (y) values of potential evap-
otranspiration (PET, mm) during 1/1985–12/2005. narios relative to the reference period (1961–1990). During each
period, the zones with the largest PET change were located in
some high-elevation regions under all RCP scenarios as shown
seasons (Fig. 3). Although the monthly PET values calculated by the
in the 2011–2100 period. Table 3 shows the corresponding Min,
most accurate MeanT, MaxT, and MinT values are slightly smaller
Max, Mean, and CV for each spatial distribution in Fig. 6. The
than the observed values, the statistical indicators between these
CV indicates that RCP4.5 shows the largest spatial variation in
values present a high evaluation for calculating the monthly PET
2011–2040 (21.8%), while RCP6.0 shows the smallest spatial vari-
(Fig. 4).
ation in 2071–2100 (8.3%). The difference between Max and Min
indicates that RCP8.5 shows the most extreme spatial variation
3.2. Spatiotemporal changes in PET in 2071–2100 (31.9%; ranging from 20.6% to 52.5%), while RCP6.0
shows the most moderate spatial variation in 2011–2040 (9.7%;
As mentioned in Section 3.1, the monthly PET during the ranging from 1.8% to 11.5%). The Mean indicates that the averaged
1950–2100 period was calculated by the MeanT, MaxT, and MinT, PET change over the LP increases with time under each RCP scenario
and the temperature values were downscaled by the most suit- in the future periods.
able GCM data using the bilinear interpolation method. A reference
period (1961–1990) was set, and the future period (2011–2100)
 S. Peng et al. / Agricultural and Forest Meteorology 233 (2017) 183–194 189

Fig. 5. Geographic distributions of the changes in the mean annual PET (%) from the reference period (1961–1990) to the future period (2011–2100) under four emission
scenarios.

Table 3
Spatial distribution characteristics of mean annual PET change (%) in the future periods relative to the reference period (1961–1990).

2011–2040 2041–2070 2071–2100 2011–2100

RCP2.6 RCP4.5 RCP6.0 RCP8.5 RCP2.6 RCP4.5 RCP6.0 RCP8.5 RCP2.6 RCP4.5 RCP6.0 RCP8.5 RCP2.6 RCP4.5 RCP6.0 RCP8.5

Min 3.9 2.3 1.8 3.5 7.8 9.3 5.7 10.3 9.8 12.7 13.3 20.6 7.6 8.7 7.3 11.9
Max 14.8 14.1 11.5 14.2 19.2 25.3 19.5 31.2 23.2 32.6 33.2 52.5 18.5 24.0 21.4 32.6
Mean 6.2 5.0 4.3 5.4 9.7 11.7 9.0 12.9 12.7 15.8 17.0 23.9 9.5 10.8 10.1 14.1
CV 15.3% 21.8% 19.7% 17.5% 9.1% 9.9% 12.7% 12.9% 9.5% 9.2% 8.3% 9.2% 9.2% 10.5% 10.4% 10.9%

Min, Max, Mean, and CV are the minimum, maximum, mean, and coefficient of variation, respectively.

3.3. Trend analysis in PET scenario during 2011–2040. Specifically, the significantly increas-
ing trend ranges from 10.9 mm/10yr to 41.9 mm/10yr with a mean
Mann–Kendall and Sen’s slope estimator tests were conducted of 26.5 mm/10yr for RCP2.6, from 9.8 mm/10yr to 48.7 mm/10yr
at each grid over the LP. Fig. 7 maps the spatial patterns of with a mean of 29.1 mm/10yr for RCP4.5, from 7.7 mm/10yr to
annual PET trend during 2011–2100 (i.e., 90-year span) under 29.3 mm/10yr with a mean of 18.5 mm/10yr for RCP6.0, and from
the four RCP scenarios, while Table 4 lists the corresponding sta- 7.4 mm/10yr to 30.9 mm/10yr with a mean of 20.2 mm/10yr for
tistical indexes (Min, Max, Mean, and CV) and area ratios (AR) RCP8.5. This trend is observed in most of the LP region with
of the significant trend zones over the entire LP region. The 97.7% and 97.4% ARs under RCP2.6 and RCP4.5, respectively, in
Mann–Kendall test reveals that the annual PET during 2011–2100 the west and north of the LP region with 36.3% AR under RCP6.0,
at each grid over the LP demonstrates a significantly increas- and in the west and east of the LP region with 30% AR under
ing trend under each RCP scenario, that is, the AR under each RCP8.5. The trend also increases significantly under the three sce-
RCP scenario is equal to 100% at the 95% confidence level. narios during 2041–2070. Specifically, the significantly increasing
The CV indicates that RCP2.6 shows the largest spatial varia- trend ranges from 12.7 mm/10yr to 49.3 mm/10yr with a mean of
tion (17.6%), while RCP8.5 shows the smallest spatial variation 29.9 mm/10yr for RCP4.5, from 10 mm/10yr to 42.5 mm/10yr with
(7.1%). The difference between Max and Min indicates that RCP8.5 a mean of 27.1 mm/10yr for RCP6.0, and from 12.9 mm/10yr to
shows the most extreme spatial variation (14.2 mm/10yr; rang- 46.2 mm/10yr with a mean of 28.5 mm/10yr for RCP8.5. This trend
ing from 22.9 mm/10yr to 37.1 mm/10yr), while RCP2.6 shows is observed in most of the LP region with 99.9%, 76.9%, and 95.7%
the most moderate spatial variation (12.9 mm/10yr; ranging from ARs under RCP4.5, RCP6.0, and RCP8.5, respectively. During this
3.1 mm/10yr to 16 mm/10yr). The Mean indicates that the averaged same period, the trend significantly decreases under the RCP2.6
magnitude of PET trend over the LP increases along with emis- scenario and it ranges from 8.6 mm/10yr to 19.8 mm/10yr with a
sion intensity; the averaged magnitude is equal to 10.4 mm/10yr mean of 14.8 mm/10yr. This trend is observed in north of the LP
for RCP2.6, 17.7 mm/10yr for RCP4.5, 21.0 mm/10yr for RCP6.0, and with a 0.3% AR. The trend significantly increases under the three
29.7 mm/10yr for RCP8.5. scenarios during 2071–2100; Specifically, the significantly increas-
Fig. 8 maps the spatial patterns of annual PET trends in the ing trend ranges from 7.5 mm/10yr to 33 mm/10yr with a mean
three future periods (i.e., 30-year span) under the four RCP sce- of 18.7 mm/10yr for RCP4.5, from 12.8 mm/10yr to 50.5 mm/10yr
narios, while Table 4 lists the corresponding indexes (Min, Max, with a mean of 26.7 mm/10yr for RCP6.0, and from 20.5 mm/10yr to
Mean, CV, and AR). The trend significantly increases under each 55.8 mm/10yr with a mean of 37.2 mm/10yr for RCP8.5. This trend
190 S. Peng et al. / Agricultural and Forest Meteorology 233 (2017) 183–194

Fig. 6. Geographic distributions of the changes in the mean annual PET (%) from the reference period (1961–1990) to beginning-of-century (a1, b1, c1, and d1: 2011–2040),
mid-century (a2, b2, c2, and d2: 2041–2070), and end-of-century (a3, b3, c3, and d3: 2071–2100) under four emission scenarios (rows).

Fig. 7. Geographic distributions of the trend for the annual PET (mm/10yr) during 2011–2100 under four emission scenarios.
 S. Peng et al. / Agricultural and Forest Meteorology 233 (2017) 183–194 191

Table 4
Spatial distribution characteristics of significant trend (mm/10yr) in annual PET in the future periods.

2011–2040 2041–2070 2071–2100 2011–2100

RCP2.6 RCP4.5 RCP6.0 RCP8.5 RCP2.6 RCP4.5 RCP6.0 RCP8.5 RCP2.6 RCP4.5 RCP6.0 RCP8.5 RCP2.6 RCP4.5 RCP6.0 RCP8.5

Min 10.9 9.8 7.7 7.4 8.6 12.7 10.0 12.9 10.2 7.5 12.8 20.5 3.1 10.9 14.0 22.9
Max 41.9 48.7 29.3 30.9 19.8 49.3 42.5 46.2 29.4 33.0 50.5 55.8 16.0 23.1 27.8 37.1
Mean 26.5 29.1 18.5 20.2 14.8 29.9 27.1 28.5 18.7 18.7 26.7 37.2 10.4 17.7 21.0 29.7
CV 18.2% 23.2% 15.8% 16.7% 10.4% 20.9% 14.9% 17.5% 11.6% 14% 16.2% 14.1% 17.6% 13.0% 10.3% 7.1%
AR 97.7% 97.4% 36.3% 30.0% 0.3% 99.9% 76.9% 95.7% 1.2% 10.3% 97.5% 100% 100% 100% 100% 100%

AR is the area ratio of the significant trend zones to the entire LP region. The boldface numbers indicate significant decreases at the 95% confidence level.

Fig. 8. Geographic distributions of the trend for the annual PET (mm/10yr) during the future periods (columns) under four emission scenarios (rows). The red zone indicates
where the trend significantly increases at the 95% confidence level, and the black zone indicates where the trend significantly decreases at the 95% confidence level. (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

is observed in the southeast and northwest of the LP with 10.3% magnitude of PET trend. This study assesses the PET changes and
and 97.5% ARs for RCP4.5 and RCP6.0, respectively. This trend is trends over the LP in future periods (i.e., 2011–2040, 2041–2070,
observed across the entire LP region under the RCP8.5 scenario. 2071–2100, and 2011–2100) based on the high spatiotemporal
During this same period, the trend significantly decreases under resolution future temperature data generated through the Delta
the RCP2.6 scenario; it ranges from 10.2 mm/10yr to 29.4 mm/10yr downscaling method and GCMs data of CMIP5. These spatiotem-
with a mean of 18.7 mm/10yr and is located in northeast of the LP poral results represent the PET changes and trends over the LP in
region with 1.2% AR. According to the CV analysis, RCP4.5 shows the detail and provide insights for developing flexible adaptation and
largest spatial variation in 2011–2040 (23.2%), while RCP2.6 shows mitigation strategies to combat the effects of global warming in this
the smallest spatial variation in 2041–2070 (10.4%). region.
The PET changes in each future period demonstrate a very strong
spatial variability (Figs. 5 and 6, Table 3). Therefore, high spatial
4. Summary and discussion
resolution temperature data are crucial in capturing PET varia-
tion at the fine scale. Large PET changes are also observed in some
Although the significant changes in the PET trend over the LP
high-elevation regions (Figs. 5 and 6). Based on the PET equation
have been studied in the literature (Li et al., 2012a), previous stud-
(i.e., Eq. (2)), these changes can be attributed to the fact that the
ies were performed at the station scale and did not provide the
192 S. Peng et al. / Agricultural and Forest Meteorology 233 (2017) 183–194

Fig. 9. Geographic distributions of the changes in the differences between MaxT and MinT (◦ C) from the reference period (1961–1990) to the future period (2011–2100)
under four emission scenarios.

high-elevation regions show a higher increment in the difference diverse spatial patterns and statistical results (Fig. 8 and Table 4).
between MaxT and MinT than the low-elevation regions, especially By comparing the three major uncertainties in climate predictions
for the Qilian Mountains located west of the LP region (Fig. 9). that result from future natural fluctuations, model responses, and
Moreover, the average annual PET over the LP region will increase emission scenarios (Hawkins and Sutton, 2009, 2011), the above
by 12.7%–23.9% from 1961 to 1990 to the end of this century diverseness can be attributed to inter-GCM (i.e., the spread of
(2071–2100) (Table 3), while Wang and Chen (2014) predicted that changes among GCMs under the same RCP scenario) and scenar-
the average annual precipitation over the LP region would increase ios uncertainties (Ning and Bradley, 2015, 2016). Given that this
by 10% in the same period. These results imply an increasing trend study adopted the data from two GCMs (i.e., NorESM1-M and GISS-
of water shortage over the LP region at the end of the century. Such E2-R models) to calculate the high-resolution future PET under the
accelerating water shortage may further decrease the crop yields four RCP scenarios, the inter-GCM and scenarios uncertainties have
and threaten the functioning of the native vegetation ecosystem. a conjunction effect on the PET trends. Moreover, the PET trends
Therefore, the necessary adaptation and mitigation strategies must under the RCP2.6 and RCP4.5 scenarios in 2011–2040 are larger
be adopted in the future to address these issues. than those under the RCP6.0 and RCP8.5 scenarios (Table 4), which
The magnitude of PET trend in each future period also demon- implies that the uncertainties resulting from the NorESM1-M and
strates a very strong spatial variability (Figs. 7 and 8, Table 4). The GISS-E2-R models are greater than those resulting from the differ-
significant trends in PET are mapped over the LP, and their distribu- ent scenarios in this period. By contrast, in 2071–2100, the total
tions present diverse characteristics, especially in terms of location uncertainties increase along with the emission scenarios, and the
and AR (Figs. 7 and 8). The Delta downscaling method reveals that largest trend is observed under the RCP8.5 scenario.
the above detailed spatial results differ from those presented in This study employed the Delta downscaling framework to gen-
other climate change studies, which have mapped the climate vari- erate the monthly mean, maximum, and minimum temperatures
ables (e.g., temperature, precipitation, and PET) based on multiple with a 1 km spatial resolution from January 1950 to December
weather stations data and by using geo-statistical interpolation 2100 based on the GCM dataset with a 0.5◦ spatial resolution and
methods, such as inverse distance weighted interpolation (Li et al., high-resolution climatology representing an orographic effect. The
2012a) and ordinary kriging interpolation (Atta-ur-Rahman and MAE and regression analysis between the downscaled data and
Dawood, 2016). Overall, unlike those from other studies, the spa- the independent surface station observations reflected the excel-
tial results in this study can draw the detailed orographic effects lent performance of the Delta downscaling method (Table 2 and
on MeanT, MaxT, MinT, and PET, as well as provide highly accu- Fig. 3). Wang and Chen (2014) used the Delta method to downscale
rate information, such as location, AR, and other statistical indexes. raw 35 GCM data to a 0.5◦ spatial resolution; the MAE of monthly
These differences can be attributed to two reasons. First, other mean temperature ranged from 1.6 ◦ C to 5.7 ◦ C with a mean of
studies directly interpolated their station data to reveal the spa- 3.2 ◦ C. Compared with those in Wang and Chen (2014), the MAE of
tial patterns of climate variables, while this study employed the monthly mean temperature in this study was smaller and ranged
Delta downscaling framework to detect such patterns. Second, this from 1.34 ◦ C to 1.6 ◦ C with a mean of 1.45 ◦ C (Table 2). Wang and
study adopted a high-resolution climatology that represented an Chen (2014) also identified EC-EARTH as the model with the small-
orographic effect with a 1 km spatial resolution in the Delta down- est MAE; this model generated a 0.5◦ spatial resolution temperature
scaling framework. in China. By contrast, this study identifies the NorESM1-M and
The significant PET trends in each future period (2011–2040, GISS-E2-R models as the best models for reproducing the monthly
2041–2070, and 2071–2100) under the four RCP scenarios show mean/maximum and minimum temperatures over the LP, respec-
 S. Peng et al. / Agricultural and Forest Meteorology 233 (2017) 183–194 193

tively. These differences may be attributed to the fact that this study atmospheric component AM3 of the GFDL global coupled model CM3. J. Clim.
adopted high-resolution climatology with a 1 km spatial resolution 24, 3484–3519.
Dosio, A., Panitz, H.-J., Schubert-Frisius, M., Lüthi, D., 2015. Dynamical downscaling
for downscaling and used surface station observations to evaluate of CMIP5 global circulation models over CORDEX-Africa with COSMO-CLM:
the downscaled results. These differences emphasized the impor- evaluation over the present climate and analysis of the added value. Clim. Dyn.
tance of high-resolution climatology and evaluation data in using 44, 2637–2661.
Dufresne, J.-L., Foujols, M.-A., Denvil, S., Caubel, A., Marti, O., Aumont, O., Balkanski,
the Delta downscaling framework. Until now, WorldClim (Mosier Y., Bekki, S., Bellenger, H., Benshila, R., Bony, S., Bopp, L., Braconnot, P.,
et al., 2014) and the employed climatology in this study are the best Brockmann, P., Cadule, P., Cheruy, F., Codron, F., Cozic, A., Cugnet, D., de Noblet,
available high-resolution climatologies in the world and in China, N., Duvel, J.-P., Ethé, C., Fairhead, L., Fichefet, T., Flavoni, S., Friedlingstein, P.,
Grandpeix, J.-Y., Guez, L., Guilyardi, E., Hauglustaine, D., Hourdin, F., Idelkadi, A.,
respectively. Both climatologies have a 1 km spatial resolution in
Ghattas, J., Joussaume, S., Kageyama, M., Krinner, G., Labetoulle, S., Lahellec, A.,
the Delta downscaling framework. Lefebvre, M.-P., Lefevre, F., Levy, C., Li, Z.X., Lloyd, J., Lott, F., Madec, G., Mancip,
M., Marchand, M., Masson, S., Meurdesoif, Y., Mignot, J., Musat, I., Parouty, S.,
Polcher, J., Rio, C., Schulz, M., Swingedouw, D., Szopa, S., Talandier, C., Terray, P.,
Acknowledgements Viovy, N., Vuichard, N., 2013. Climate change projections using the IPSL-CM5
earth system model: from CMIP3 to CMIP5. Clim. Dyn. 40, 2123–2165.
This work was supported by National Natural Science Foun- Dunne, J.P., John, J.G., Adcroft, A.J., Griffies, S.M., Hallberg, R.W., Shevliakova, E.,
Stouffer, R.J., Cooke, W., Dunne, K.A., Harrison, M.J., Krasting, J.P., Malyshev,
dation of China (41601058), CAS “Light of West China” Program S.L., Milly, P.C.D., Phillipps, P.J., Sentman, L.T., Samuels, B.L., Spelman, M.J.,
(XAB2015B07), China Special Fund for Meteorological Research in Winton, M., Wittenberg, A.T., Zadeh, N., 2012. GFDL’s ESM2 global coupled
the Public Interest (Major projects) (GYHY201506001-3), and Key Climate–Carbon earth system models. part I: physical formulation and
baseline simulation characteristics. J. Clim. 25, 6646–6665.
cultivation project of Chinese Academy of Sciences “The promotion
Giorgi, F., Jones, C., Asrar, G.R., 2009. Addressing climate information needs at the
and management of ecosystem functions of restored vegetation in regional level: the CORDEX framework. World Meteorol. Org. Bull. 58,
Loess Plateau, China”. The authors are grateful to the World Cli- 175–183.
Gocic, M., Trajkovic, S., 2013. Analysis of changes in meteorological variables using
mate Research Programme’s Working Group on Coupled Modelling,
Mann-Kendall and Sen’s slope estimator statistical tests in Serbia. Glob. Planet
which is responsible for CMIP, and the authors thank the climate Change 100, 172–182.
modeling groups (listed in Table 1 of this paper) for producing and Hargreaves, G.H., Samani, Z.A., 1982. Estimating potential evapotranspiration. J.
making available their model output. For CMIP the U.S. Department Irrig. Drain. Div. 108, 225–230.
Hargreaves, G.H., Samani, Z.A., 1985. Reference crop evapotranspiration from
of Energy’s Program for Climate Model Diagnosis and Intercom- temperature. Appl. Eng. Agric. 1, 96–99.
parison provides coordinating support and led development of Hawkins, E., Sutton, R., 2009. The potential to narrow uncertainty in regional
software infrastructure in partnership with the Global Organization climate predictions. Bull. Am. Meteorol. Soc. 90, 1095–1107.
Hawkins, E., Sutton, R., 2011. The potential to narrow uncertainty in projections of
for Earth System Science Portals. In addition, the authors wish to regional precipitation change. Clim. Dyn. 37, 407–418.
thank two anonymous reviewers for their constructive suggestions Hazeleger, W., Wang, X., Severijns, C., Ştefănescu, S., Bintanja, R., Sterl, A., Wyser,
to improve the quality of this article. K., Semmler, T., Yang, S., van den Hurk, B., van Noije, T., van der Linden, E., van
der Wiel, K., 2012. EC-Earth V2.2: description and validation of a new seamless
earth system prediction model. Clim. Dyn. 39, 2611–2629.
References IPCC, 2013. Climate Change 2013: The Physical Science Basis. Contribution of
Working Group I to the Fifth Assessment Report of the Intergovernmental
Ahmed, K.F., Wang, G., Silander, J., Wilson, A.M., Allen, J.M., Horton, R., Anyah, R., Panel on Climate Change. In: Stocker, T.F., Qin, D., Plattner, M., Tignor, S.K.,
2013. Statistical downscaling and bias correction of climate model outputs for Allen, J., Boschung, A., Nauels, Y., Bex, V., Midgley, P.M. (Eds.). Cambridge
climate change impact assessment in the U.S. northeast. Glob. Planet Change University Press, Cambridge, United Kingdom and New York, NY, USA, p. 1535.
100, 320–332. Ji, D., Wang, L., Feng, J., Wu, Q., Cheng, H., Zhang, Q., Yang, J., Dong, W., Dai, Y.,
Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop Gong, D., Zhang, R.H., Wang, X., Liu, J., Moore, J.C., Chen, D., Zhou, M., 2014.
Evapotranspiration-Guidelines for Computing Crop Water Requirements-FAO Description and basic evaluation of beijing normal university earth system
Irrigation and Drainage Paper 56, vol. 300. FAO, Rome, pp. D05109. model (BNU-ESM) version 1. Geosci. Model Dev. 7, 2039–2064.
Aouissi, J., Benabdallah, S., Chabaâne, Z.L., Cudennec, C., 2016. Evaluation of Li, Z., Zheng, F., Liu, W., 2012a. Spatiotemporal characteristics of reference
potential evapotranspiration assessment methods for hydrological modelling evapotranspiration during 1961–2009 and its projected changes during
with SWAT–Application in data-scarce rural Tunisia. Agric. Water Manage. 2011–2099 on the Loess Plateau of China. Agric. For. Meteorol. 154, 147–155.
174, 39–51. Li, Z., Zheng, F., Liu, W., Jiang, D., 2012b. Spatially downscaling GCMs outputs to
Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. Large area hydrologic project changes in extreme precipitation and temperature events on the Loess
modeling and assessment part I: model development1. J. Am. Water Resour. Plateau of China during the 21st Century. Glob. Planet Change 82, 65–73.
Assoc. 34, 73–89. Liu, W., Sang, T., 2013. Potential productivity of the Miscanthus energy crop in the
Atta-ur-Rahman, Dawood, M., 2016. Spatio-statistical analysis of temperature Loess Plateau of China under climate change. Environ. Res. Lett. 8, 044003.
fluctuation using Mann–Kendall and Sen’s slope approach. Clim. Dyn., http:// Liu, Z., Wang, Y., Shao, M., Jia, X., Li, X., 2016. Spatiotemporal analysis of multiscalar
dx.doi.org/10.1007/s00382-016-3110-y. drought characteristics across the Loess Plateau of China. J. Hydrol. 534,
Bentsen, M., Bethke, I., Debernard, J.B., Iversen, T., Kirkevåg, A., Seland, Ø., Drange, 281–299.
H., Roelandt, C., Seierstad, I.A., Hoose, C., Kristjánsson, J.E., 2013. The norwegian Long, M.C., Lindsay, K., Peacock, S., Moore, J.K., Doney, S.C., 2013.
earth system model, norESM1-M − part 1: description and basic evaluation of Twentieth-century oceanic carbon uptake and storage in CESM1(BGC). J. Clim.
the physical climate. Geosci. Model Dev. 6, 687–720. 26, 6775–6800.
Bi, H., Liu, B., Jie, W., Yun, L., Chen, Z., Cui, Z., 2009. Effects of precipitation and Marsland, S.J., Bi, D., Uotila, P., Hirst, A.C., 2013. Evaluation of ACCESS climate
landuse on runoff during the past 50 years in a typical watershed in Loess model ocean diagnostics in CMIP5 simulations. Aust. Meteorol. Oceanogr. 63,
Plateau, China. Int. J. Sediment Res. 24, 352–364. 101–119.
Brekke, L., Thrasher, B., Maurer, E., Pruitt, T., 2013. Downscaled CMIP3 and CMIP5 Maurer, E.P., Wood, A.W., Adam, J.C., Lettenmaier, D.P., Nijssen, B., 2002. A
Climate and Hydrology Projections: Release of Downscaled CMIP5 Climate long-term hydrologically based dataset of land surface fluxes and states for the
Projections, Comparison with Preceding Information, and Summary of User conterminous United States. J. Clim. 15, 3237–3251.
Needs. US Dept. of the Interior, Bureau of Reclamation, Technical Services Meehl, G., Covey, C., Delworth, T., Latif, M., McAvaney, B., Mitchell, J., Stouffer, R.,
Center, Denver, Colorado. Taylor, K., 2007. The WCRP CMIP3 multi-model dataset: a new era in climate
Chylek, P., Li, J., Dubey, M.K., Wang, M., Lesins, G., 2011. Observed and model change research. Bull. Am. Meteorol. Soc. 88, 1383–1394.
simulated 20th century arctic temperature variability: canadian earth system Miao, C., Ni, J., Borthwick, A.G., Yang, L., 2011. A preliminary estimate of human
model CanESM2. Atmos. Chem. Phys. Discuss. 2011, 22893–22907. and natural contributions to the changes in water discharge and sediment load
Collins, M., Tett, S.F.B., Cooper, C., 2001. The internal climate variability of HadCM3, in the Yellow River. Glob. Planet Change 76, 196–205.
a version of the Hadley Centre coupled model without flux adjustments. Clim. Mosier, T.M., Hill, D.F., Sharp, K.V., 2014. 30-Arcsecond monthly climate surfaces
Dyn. 17, 61–81. with global land coverage. Int. J. Climatol. 34, 2175–2188.
Donner, L.J., Wyman, B.L., Hemler, R.S., Horowitz, L.W., Ming, Y., Zhao, M., Golaz, Neale, R.B., Richter, J., Park, S., Lauritzen, P.H., Vavrus, S.J., Rasch, P.J., Zhang, M.,
J.-C., Ginoux, P., Lin, S.-J., Schwarzkopf, M.D., Austin, J., Alaka, G., Cooke, W.F., 2013. The mean climate of the community atmosphere model (CAM4) in
Delworth, T.L., Freidenreich, S.M., Gordon, C.T., Griffies, S.M., Held, I.M., Hurlin, forced SST and fully coupled experiments. J. Clim. 26, 5150–5168.
W.J., Klein, S.A., Knutson, T.R., Langenhorst, A.R., Lee, H.-C., Lin, Y., Magi, B.I., Ning, L., Bradley, R.S., 2015. Snow occurrence changes over the central and eastern
Malyshev, S.L., Milly, P.C.D., Naik, V., Nath, M.J., Pincus, R., Ploshay, J.J., United States under future warming scenarios. Sci. Rep. 5, 17073.
Ramaswamy, V., Seman, C.J., Shevliakova, E., Sirutis, J.J., Stern, W.F., Stouffer, Ning, L., Bradley, R.S., 2016. NAO and PNA influences on winter temperature and
R.J., Wilson, R.J., Winton, M., Wittenberg, A.T., Zeng, F., 2011. The dynamical precipitation over the eastern United States in CMIP5 GCMs. Clim. Dyn. 46,
core, physical parameterizations, and basic simulation characteristics of the 1257–1276.
194 S. Peng et al. / Agricultural and Forest Meteorology 233 (2017) 183–194

Ning, L., Riddle, E.E., Bradley, R.S., 2015. Projected changes in climate extremes Wang, Q., Fan, X., Qin, Z., Wang, M., 2012. Change trends of temperature and
over the northeastern United States. J. Clim. 28, 3289–3310. precipitation in the Loess Plateau Region of China, 1961–2010. Glob. Planet
Priestley, C., Taylor, R., 1972. On the assessment of surface heat flux and Change 92, 138–147.
evaporation using large scale parameters. Mon. Weather Rev. 100, 81–92. Wang, H.J., Chen, H.P., Liu, J.p., 2015. Arctic sea ice decline intensified haze
Qiao, F., Song, Z., Bao, Y., Song, Y., Shu, Q., Huang, C., Zhao, W., 2013. Development pollution in eastern China. Atmos. Ocean. Sci. Lett. 8, 1–9.
and evaluation of an Earth System Model with surface gravity waves. J. Watanabe, M., Suzuki, T., O’ishi, R., Komuro, Y., Watanabe, S., Emori, S., Takemura,
Geophys. Res. 118, 4514–4524. T., Chikira, M., Ogura, T., Sekiguchi, M., Takata, K., Yamazaki, D., Yokohata, T.,
Ren, H.L., Wu, J., Zhao, C.B., Cheng, Y.J., Liu, X.W., 2016. MJO ensemble prediction in Nozawa, T., Hasumi, H., Tatebe, H., Kimoto, M., 2010. Improved climate
BCC-CSM1.1(m) using different initialization schemes. Atmos. Ocean. Sci. Lett. simulation by MIROC5: mean states, variability, and climate sensitivity. J. Clim.
9, 60–65. 23, 6312–6335.
Rotstayn, L.D., Collier, M.A., Dix, M.R., Feng, Y., Gordon, H.B., O’Farrell, S.P., Smith, Watanabe, S., Hajima, T., Sudo, K., Nagashima, T., Takemura, T., Okajima, H.,
I.N., Syktus, J., 2010. Improved simulation of Australian climate and Nozawa, T., Kawase, H., Abe, M., Yokohata, T., Ise, T., Sato, H., Kato, E., Takata, K.,
ENSO-related rainfall variability in a global climate model with an interactive Emori, S., Kawamiya, M., 2011. MIROC-ESM 2010: model description and basic
aerosol treatment. Int. J. Climatol. 30, 1067–1088. results of CMIP5-20c3 m experiments. Geosci. Model Dev. 4, 845–872.
Schmidt, G.A., Ruedy, R., Hansen, J.E., Aleinov, I., Bell, N., Bauer, M., Bauer, S., Cairns, Willmott, C.J., Matsuura, K., 2005. Advantages of the mean absolute error (MAE)
B., Canuto, V., Cheng, Y., Genio, A.D., Faluvegi, G., Friend, A.D., Hall, T.M., Hu, Y., over the root mean square error (RMSE) in assessing average model
Kelley, M., Kiang, N.Y., Koch, D., Lacis, A.A., Lerner, J., Lo, K.K., Miller, R.L., performance. Clim. Res. 30, 79–82.
Nazarenko, L., Oinas, V., Perlwitz, J., Perlwitz, J., Rind, D., Romanou, A., Russell, Wood, A.W., Maurer, E.P., Kumar, A., Lettenmaier, D.P., 2002. Long-range
G.L., Sato, M., Shindell, D.T., Stone, P.H., Sun, S., Tausnev, N., Thresher, D., Yao, experimental hydrologic forecasting for the eastern United States. J. Geophys.
M.-S., 2006. Present-day atmospheric simulations using GISS ModelE: Res. 107, 4429.
comparison to In situ, satellite, and reanalysis data. J. Clim. 19, 153–192. Xin, X.G., Wu, T.W., Li, J.L., Wang, Z.Z., Li, W.P., Wu, F.H., 2013. How well does
Scoccimarro, E., Gualdi, S., Bellucci, A., Sanna, A., Fogli, P.G., Manzini, E., Vichi, M., BCC CSM1.1 reproduce the 20th century climate change over China? Atmos.
Oddo, P., Navarra, A., 2011. Effects of tropical cyclones on ocean heat transport Ocean. Sci. Lett. 6, 21–26.
in a high-resolution coupled general circulation model. J. Clim. 24, 4368–4384. Xu, J., Gao, Y., Chen, D., Xiao, L., Ou, T., 2016. Evaluation of global climate models for
Sun, Q., Miao, C., Duan, Q., Wang, Y., 2015. Temperature and precipitation changes downscaling applications centred over the Tibetan Plateau. Int. J. Climatol.,
over the Loess Plateau between 1961 and 2011, based on high-density gauge http://dx.doi.org/10.1002/joc.4731.
observations. Glob. Planet Change 132, 1–10. Yukimoto, S., Adachi, Y., Hosaka, M., Sakami, T., Yoshimura, H., Hirabara, M.,
Timm, O.E., Giambelluca, T.W., Diaz, H.F., 2015. Statistical downscaling of rainfall Tanaka, T.Y., Shindo, E., Tsujino, H., Deushi, M., Mizuta, R., Yabu, S., Obata, A.,
changes in Hawai‘i based on the CMIP5 global model projections. J. Geophys. Nakano, H., Koshiro, T., Ose, T., Kitoh, A., 2012. A new global climate model of
Res. 120, 92–112. the meteorological research institute: MRI-CGCM3 —Model description and
Voldoire, A., Sanchez-Gomez, E., Salas y Mélia, D., Decharme, B., Cassou, C., Sénési, basic performance—. J. Meteorol. Soc. Jpn. 90A, 23–64.
S., Valcke, S., Beau, I., Alias, A., Chevallier, M., Déqué, M., Deshayes, J., Douville, Zhang, X., Wang, Q., Mu, M., 2015. The impact of global warming on Kuroshio
H., Fernandez, E., Madec, G., Maisonnave, E., Moine, M.-P., Planton, S., Extension and its southern recirculation using CMIP5 experiments with a
Saint-Martin, D., Szopa, S., Tyteca, S., Alkama, R., Belamari, S., Braun, A., high-resolution climate model MIROC4 h. Theor. Appl. Climatol., http://dx.doi.
Coquart, L., Chauvin, F., 2013. The CNRM-CM5.1 global climate model: org/10.1007/s00704-015-1672-y.
description and basic evaluation. Clim. Dyn. 40, 2091–2121. Zhao, C., Nan, Z., Feng, Z., 2004. GIS-assisted spatially distributed modeling of the
Volodin, E.M., Dianskii, N.A., Gusev, A.V., 2010. Simulating present-day climate potential evapotranspiration in semi-arid climate of the Chinese Loess Plateau.
with the INMCM4.0 coupled model of the atmospheric and oceanic general J. Arid Environ. 58, 387–403.
circulations. Izv. Atmos. Ocean. Phys 46, 414–431. Zhou, T., Song, F., Chen, X., 2013. Historical evolution of global and regional surface
Wang, L., Chen, W., 2014. A CMIP5 multimodel projection of future temperature, air temperature simulated by FGOALS-s2 and FGOALS-g2: How reliable are the
precipitation, and climatological drought in China. Int. J. Climatol. 34, model results? Adv. Atmos. Sci. 30, 638–657.
2059–2078.