SCIENCE CHINA

Earth Sciences
•REVIEW•
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . https://doi.org/10.1007/s11430-018-9261-5
............................

Dynamical downscaling of regional climate: A review of methods

and limitations
1* 1 1,2
Zhongfeng XU , Ying HAN & Zongliang YANG
1
CAS Key Laboratory of Regional Climate-Environment for Temperate East Asia, Institute of Atmospheric Physics, Chinese Academy of
Sciences, Beijing 100029, China;
2
Department of Geological Sciences, Jackson School of Geosciences, The University of Texas at Austin, TX 78705, USA

Received March 12, 2018; revised May 14, 2018; accepted August 16, 2018; published online September 28, 2018

Abstract The traditional dynamical downscaling (TDD) method employs continuous integration of regional climate models
(RCM) with the general circulation model (GCM) providing the initial and lateral boundary conditions. Dynamical downscaling
simulations are constrained by physical principles and can generate a full set of climate information, providing one of the
important approaches to projecting fine spatial-scale future climate information. However, the systematic biases of climate
models often degrade the TDD simulations and hinder the application of dynamical downscaling in the climate-change related
studies. New methods developed over past decades improve the performance of dynamical downscaling simulations. These
methods can be divided into four groups: the TDD method, the pseudo global warming method, dynamical downscaling with
GCM bias corrections, and dynamical downscaling with both GCM and RCM bias corrections. These dynamical downscaling
methods are reviewed and compared in this paper. The merits and limitations of each dynamical downscaling method are also
discussed. In addition, the challenges and potential directions in progressing dynamical downscaling methods are stated.
Keywords Global circulation model, Bias correction, Regional climate model, Downscaling, Projection of regional climate

Citation: Xu Z, Han Y, Yang Z. 2018. Dynamical downscaling of regional climate: A review of methods and limitations. Science China Earth Sciences, 61,
https://doi.org/10.1007/s11430-018-9261-5

1. Introduction lation models (GCM) is 100–300 km. The coarse GCMs

cannot fully resolve the fine-scale orography, land cover, and
Climate change has far-reaching influence on natural eco- dynamics of the atmosphere, hindering their ability to si-
systems, the environment, and social economies. Projecting mulate extreme weather and climate events (Zhao et al.,
future climate change, especially regional climate change, is 2014). However, fine-scale weather and climate information
of great importance for policy-making. Compared with is urgently needed in many research areas, such as meteor-
previous Intergovernmental Panel on Climate Change ology, hydrology, agriculture, ecology, wind power, and the
(IPCC) assessment reports, the fifth report highlights the atmospheric environment. Therefore, the demand of down-
importance of regional scale climate change (IPCC, 2013). scaling is very extensive. The so-called “downscaling”
The climate and earth system models with complex bio- simply refers to the use of statistical or dynamical ap-
geochemical processes are the primary tools for projecting proaches to obtain regional or local scale climate information
future climate (Zeng et al., 2008; Prinn, 2013; IPCC, 2013). from the GCM output or reanalysis data (von Storch et al.,
Currently, the typical horizontal resolution of global circu- 1993, Liu et al., 2012).
The commonly used downscaling methods include statis-
tical, dynamical, and hybrid dynamical-statistical down-
* Corresponding author (email: [email protected])

© Science China Press and Springer-Verlag GmbH Germany, part of Springer Nature 2018 earth.scichina.com link.springer.com
2 Xu Z, et al. Sci China Earth Sci

scaling. Statistical downscaling relies on the statistical re- many methods have been proposed in recent years. In this
lationship between large-scale climate derived from GCMs paper, we review various dynamical downscaling methods
and local-scale climate obtained from observations. As- developed over recent years and summarize the merits and
suming such a relationship does not change with time, one limitations of each method. We also discuss the challenges
can downscale GCM output to individual stations. The ad- and potential directions of dynamical downscaling studies.
vantages of statistical downscaling are computational effi-
ciency and ease of use (Hanssen-Bauer et al., 2005; Fan et
al., 2005, 2013; Chen et al., 2016). However, statistical 2. The traditional dynamical downscaling
downscaling cannot capture relationships not present in method
history, and it is not suitable for the regions without a sig-
nificant relationship between the large-scale climate in- Given limited computation resources, Dickinson et al. (1989)
formation and local observations (Fan et al., 2005; Maraun et developed an RCM to investigate high-resolution climate.
al., 2010). In contrast, dynamical downscaling employs a The RCM only covers a limited area of the earth’s surface
regional climate model (RCM) nested within a GCM to and is nested in a GCM. The RCM is continuously integrated
generate fine-resolution climate information (Giorgi and with GCM output as initial and lateral boundary conditions,
Mearns, 1991). RCMs are formulated on the basis of phy- and this is the traditional dynamical downscaling (TDD)
sical principles. With the incorporation of biogeochemical method. Over the past two decades, TDD has been widely
processes, RCMs are growing into regional earth system used in a number of studies, such as meteorology (e.g.,
models that can simulate the interactions among the atmo- Giorgi et al., 1994a; Leung et al., 2003; Xu et al., 2006;
sphere, land, water, social economy, etc. (Adam et al., 2015; Bukovsky and Karoly, 2011; Gao X J et al., 2002, 2011; Bao
Yang, 2015). Therefore, compared with statistical down- et al., 2015; Niu et al., 2015; Tang et al., 2016a; El-Samra et
scaling, dynamical downscaling can resolve various pro- al., 2017; Sun et al., 2017; Yao et al., 2018), hydrology (e.g.,
cesses in climate or earth systems and their interactions. Giorgi et al., 1994b; Feng et al., 2011; Gao Y et al., 2011;
However, dynamical downscaling is a computation-intensive Shukla and Lettenmaier, 2013; Xiong and Yan, 2013; Bou-
and time-consuming approach limiting the application of this lard et al., 2016), and atmospheric chemistry and air quality
method. The dynamical-statistical downscaling method takes (e.g., Varghese et al., 2011; Tan et al., 2015; Khairunnisa et
advantage of both the dynamical and statistical downscaling al., 2017). However, as it is known, climate models contain
approaches (Chen et al., 2003; Han and Wei, 2010). Re- significant systematic biases that can significantly degrade
cently, Walton et al. (2015) applied dynamical downscaling the dynamical downscaling simulation and lead to large
to five GCM outputs followed by a simple statistical model uncertainties. This is one of the limitations of the TDD
to build the relationship between GCM and RCM output. methods. Ensemble downscaling simulations with various
The statistical relationship was in turn applied to 32 GCM combinations of RCMs and GCMs can provide a better es-
ensemble members to generate a number of downscaled si- timate of future climate change and an uncertainty range.
mulations. The full set of ensemble downscaling outputs Therefore, some international dynamical downscaling in-
allow for robust projection of future climate and the un- tercomparison projects were carried out, e.g., the Regional
certainty resulting from intermodal differences (Walton et Climate Model Intercomparison Project for Asia (RMIP; Fu
al., 2015; Sun et al., 2015). et al., 2005), Ensembles-Based Predictions of Climate
In the past two decades, downscaling performance has Changes and Their Impacts (ENSEMBLES; van der Linden
been investigated in a number of studies (e.g., Giorgi and and Mitchell, 2009), North American Regional Climate
Mearns, 1991; Wilby and Wigley, 1997; Wilby et al., 1998; Change Assessment Program (NARCCAP) (Mearns et al.,
Lo et al., 2008; Maraun et al., 2010; Wang and Kotamarthi, 2009), and Coordinated Regional Downscaling Experiment
2015; Tang et al., 2016b). Dynamical downscaling studies (CORDEX, Wilby et al., 1998). The ensemble downscaling
suggested that the RCM simulations are strongly affected by simulations require multi-institutional collaboration due to
many factors, such as the domain size and location, hor- the huge computation and manpower involved hindering the
izontal and vertical resolution, the quality of large-scale application of TDD in climate change and impact studies.
forcing data, and physical processes (Jacob and Podzun,
1997; Zhao and Luo, 1998; Liang et al., 2004; Alexandru et
al., 2007; Xue et al., 2007, 2014; Caron et al., 2011; Pielke 3. Pseudo global warming downscaling method
and Wilby, 2012; Chan et al., 2013). Recent studies indicated
that correcting GCM bias has the potential to significantly To project the climate response to increasing in greenhouse
improve dynamical downscaling simulations (e.g., Xu and gas concentrations, some studies employed the so-called
Yang, 2012; Bruyère et al., 2014). Bias corrections are be- pseudo global warming (PGW) downscaling method (Wu
coming a hotspot for dynamical downscaling studies and and Lynch, 2000; Sato et al., 2007; Cook and Vizy, 2008;
 Xu Z, et al. Sci China Earth Sci 3

Kawase et al., 2009; Lauer et al., 2013; Yu and Wang, 2014; Moreover, all GCMs contain significant systematic biases.
Kitoh et al., 2016). PGW methods construct RCM lateral The GCM biases can be passed to RCMs through LBCs and
boundary conditions (LBCs) by adding GCM projected degrade the RCM simulations (Wu and Lynch, 2000; Sato et
changes onto reanalysis data as follows: al., 2007). In addition, historical climate biases can propa-
gate into future climate projections during dynamical
BC F = RA H + GCM F GCM H, (1)
downscaling simulations (Liang et al., 2008). The GCM
where BCF represents the constructed RCM LBC at 6-hourly biases are one of the important sources of RCM biases and
intervals for the future. RAH is the 6-hourly reanalysis data uncertainty. To improve the performance of dynamical
over historical periods. “¯” represents the climatological downscaling, GCM bias correction methods were proposed
mean. The subscripts P and F indicate the historical (say in past years. The bias-corrected GCM data were fed to
1980–2010) and future (say 2050–2080) periods, respec- RCMs to generate high-resolution climate information. In
tively. GCM F GCM H is the mean climate change between the following sections, we introduce the GCM bias correc-
tion methods employed in dynamical downscaling simula-
the future and historical periods computed with monthly
tions in recent years.
data. The LBCs BC F = RA H + GCM F GCM H constructed
by the PGW method have a base climate provided by re-
analysis data and therefore exclude the influences of GCM 4.1 Mean bias correction
bias on dynamical downscaling simulations. One can also
replace single GCM data in eq. (1) with a multi-GCM en- Holland et al. (2010) noticed that the TDD significantly
semble to reduce the uncertainty induced by individual GCM underestimated the number of tropical cyclones over the
simulation, and this should provide a better projection of Atlantic Ocean. Further analysis suggested that the GCM
future climate. In addition to future climate projection, PGW significantly overestimated the vertical wind shear between
methods can also be applied to detection and attribution 850 and 200 hPa. Such a GCM bias was propagated into the
study (Xu and Yang, 2017). In terms of the detection of RCM and led to significant underestimation of the number of
external forcing, internal climate variabilities are regarded as tropical cyclones. To improve the dynamical downscaling
noise. The amplitudes of external forcing induced changes simulation of tropical cyclones, Holland et al. (2010) pro-
are usually weaker than internal climate variabilities, making posed a simple GCM bias correction method by removing
it harder to separate their influence. To effectively detect the climatological mean bias from the GCM data for the
external forcing signal, one can perform two dynamical future period:
downscaling simulations with LBCs constructed using the
PGW method. Two dynamical downscaling simulations BC F = GCM F ( GCM H RA H)
contain similar internal climate variability inherited from the = GCM F + GCM ( GCM H RA H). (2)
F
LBCs. Thus, the difference between two dynamical down-
scaling simulations can largely remove common internal The symbols in eq. (2) are the same as in eq. (1). The third
climate variability and retain external forcing signals. The term on the right-hand side represents the GCM mean bias
limitation of the PGW method is rooted in the construction of over the historical period. Assuming the GCM mean bias
LBCs. As shown in eq. (1), except the changes in climato- does not change with time, one can remove it from the GCM
logical mean, the temporal variations of variables in LBCs data over the future period. Clearly, eq. (2) retains the GCM
over the future period are all inherited from those in re- temporal variation while removing the mean bias. Therefore,
analysis data over the historical period. As the LBCs strongly this GCM bias correction method can be applied to project
modulate RCM internal variabilities, the RCM simulations the changes in temporal variation of future climate, such as
of the future period will contain historical temporal vari- changes in diurnal cycle, interannual and decadal vari-
abilities. Thus, the PGW is not well suited for the projection abilities, and this remedies the limitation of the PGW
of the changes in temporal variations, e.g., diurnal cycle, method. Compared with the TDD method, the dynamical
annual cycle, and interannual and decadal variations, over downscaling approach with GCM mean bias correction well
the future period. simulated the features of tropical cyclones over the historical
period (Figure 1; Holland et al., 2010; Done et al., 2015;
Bruyère et al., 2014). Similar bias correction methods were
4. Dynamical downscaling with GCM bias also applied to and significantly improved numerical weather
corrections and seasonal forecasting (Peng et al., 2013; Chang et al.,
2015; Ratnam et al., 2016). Recently, this method was also
It is known that the RCM simulations are strongly dominated applied to the projection of regional climate in China (Chen
by the LBCs provided by GCMs (Pielke and Wilby, 2012). et al., 2018).
4 Xu Z, et al. Sci China Earth Sci

bias corrections can well reproduce a distribution of daily

maximum temperature that is very close to the reference
experiment (Figure 2). This suggests that both the GCM
mean and variance bias corrections are necessary to improve
dynamical downscaling simulations. The IDD method was
applied to the downscaling study of East Asian climate and
air quality and significantly improved the RCM simulations
(Hong et al., 2017). The limitation of the IDD method is that
the variance bias correction may inappropriately exaggerate
or minify the GCM trend. To retain the GCM trend, Hoff-
mann et al. (2016) removed if before the variance bias cor-
rection and added it back afterwards. Thus, the GCM
variance bias was corrected and the long-term trend simu-
lated by GCM was retained.

4.3 Quantile-quantile correction

Colette et al. (2012) introduced quantile-quantile correction

of GCM data into dynamical downscaling simulations. As-
suming the cumulative distribution function (CDF) is sta-
tionary in time, one can correct the CDF of the GCM over the
future period with the quantile-matching technique using the
CDF derived from historical GCM and observation data
(Déqué, 2007; Michelangeli et al., 2009; Yang et al., 2017).
Figure 1 Tropical storm tracks derived from observation (a), the TDD The correction alters the shape of the probability distribution
methods over the period of 1995–2005 (b), and the dynamical downscaling and corrects all biases in the distribution of GCM data. Thus,
simulation with GCM mean bias correction (c). Modified based on Bruyère
et al. (2014).
the quantile-quantile correction considered not only the
GCM mean and variance biases but also the biases in the
frequencies of normal and extreme events. Compared with
4.2 Mean and variance bias corrections the TDD method, the dynamical downscaling simulation

Xu and Yang (2012) developed an improved dynamical

downscaling (IDD) method by taking both GCM mean and
variance bias corrections into account. Assuming the GCM
mean and variance biases remain stationary in time, one can
remove both biases from the GCM data by subtracting the
mean bias and scaling the variance based on reanalysis data:
BC F = GCM F ( GCM H RA H)
S RA H
+GCM FS , (3)
GCM H

where GCM F represents the GCM anomaly over the future

period. S RA and S GCM are the historical standard deviation
H H
of reanalysis and GCM data, respectively. The GCM bias
corrections involve shifting and scaling the GCM time series
to remove both the mean and variance biases, and they are
applied to 6-hourly GCM data. They also allow the temporal
variation to change similar to the method proposed by Hol- Figure 2 Frequency of daily maximum temperature in summer over the
land et al. (2010). Compared with the TDD method, the central United States and Canada (40°–50°N, 100°–85°W) generated by
GCM mean bias correction alone can significantly improve various dynamical downscaling methods. RCM_RA: reanalysis-driven
RCM simulation (reference experiment); RCM_GCM: TDD method;
the climatological mean state of the RCM. However, the RCM_GCMbc.ave: RCM simulation with GCM mean bias correction;
temperature frequency still shows significant bias. In con- RCM_GCMbc.std: RCM simulation with both GCM mean and variance
trast, the IDD method with both GCM mean and variance bias corrections. Modified based on Xu and Yang (2012).
 Xu Z, et al. Sci China Earth Sci 5

with quantile-quantile correction clearly improved the sur- dynamical downscaling (AOGCM-RCM). The 3-step dyna-
face air temperature and precipitation (Colette et al., 2012). mical downscaling corrected the SST and sea ice biases,
However, it failed to retain the intervariable dependencies which favors the improvement of atmospheric dynamics in
and introduced additional bias in the spatial gradient of both AGCM and RCM. However, correcting SST and sea ice
variables and spurious precipitation variability (White and biases may only have limited improvement to the down-
Toumi, 2013). Conversely, the GCM mean bias correction scaled climate because the AGCM biases in upper variables,
can retain the first-order spatial and intervariable de- such as air temperature, vector winds, and humidity, may still
pendencies. Therefore, White and Toumi (2013) re- degrade the RCM simulation. In addition, the 3-step dyna-
commended the GCM mean bias correction rather than the mical downscaling method may only apply to regions where
quantile-quantile correction for the purpose of dynamical the SST and sea ice affect the atmosphere, e.g., the eastern
downscaling. equatorial Pacific Ocean. The 3-step dynamical downscaling
method may not apply to the Asian monsoon regions where
4.4 Bias correction with physical consistency constraint the changes in SST are strongly affected by atmospheric
conditions, e.g., precipitation and clouds (Wang et al., 2005).
The bias correction methods introduced in Sections 4.1–4.3 Dai et al. (2017) argued that the inconsistency among in-
did not explicitly consider the physical constraint between dependently corrected forcing fields is likely to be small
variables such as the hydrostatic equilibrium and geostrophic because the adjustments are small compared with the var-
wind balance. Meyer and Jin (2015) developed a bias cor- iations in 6-hourly data. Additionally, the inconsistency
rection method based on the physical consistency constraint. could be smoothed out in the buffer zone by the RCM dy-
First, they corrected the air temperature, relative humidity, namics. Consequently, the influence of bias correction on
surface temperature and surface pressure by using a simple downscaled intervariable dependency could be negligible.
linear regression method. These variables can be used to Xu and Yang (2015) compared the ratio of ageostrophic wind
derive other variables based on their physical relationships. to total wind in various dynamical downscaling simulations. It
For example, geopotential height can be computed with air turns out that the GCM bias corrections only have a minor
temperature and humidity in terms of the hydrostatic equi- influence on the geographic balance. However, the ageos-
librium. Geostrophic wind can be calculated using geopo- trophic wind component significantly increases when spectral
tential height in terms of the geostrophic balance. The nudging was applied in the RCM inner domain, suggesting a
ageostrophic winds are corrected with the linear regression large disturbance in the intervariable dependencies. Further
method. Bias-corrected winds are obtained by adding the discussion will be presented in Section 5.
bias-corrected ageostrophic wind onto the geostrophic
winds. Compared with the TDD method, the RCM simula- 4.5 Bias correction of low frequency variability
tion driven by the GCM bias correction with physical con-
The bias correction methods presented in the previous sec-
sistency constraints significantly improved the downscaled
tions did not consider the frequency biases of GCM data. As
precipitation and convective effective potential energy over
known, many meteorological disasters, e.g., floods and
the southwestern United States and Mexico (Meyer and Jin,
drought, and their influences are closely related to low-fre-
2015). Note that the simple GCM mean bias correction can
quency variability. To improve the low-frequency variation,
also retain a zero or first-order relationship between vari-
Rocheta et al. (2017) introduced a nested bias correction
ables, such as the hydrostatic equilibrium and geostrophic
(NBC) method into dynamical downscaling study. The NBC
wind balance (White and Toumi, 2013).
method corrected GCM mean bias, variance bias, and low-
In addition, Hernández-Díaz et al. (2017) proposed a 3-
frequency bias. The GCM low-frequency bias was corrected
step dynamical downscaling method that can well retain the
by replacing the lag-1 autocorrelations with the observed
intervariable dependencies. First, they corrected the clima-
monthly lag-1 autocorrelations. The NBC method sig-
tological mean biases of sea surface temperature (SST) and
nificantly improves the climatological mean, variance, and
sea ice in the atmosphere-ocean coupled GCM (AOGCM)
low-frequency variability in GCM data. However, correcting
based on observational data. Second, an atmosphere GCM
the lag-1 autocorrelations through NBC did not significantly
(AGCM) was integrated with the bias-corrected SST and sea
improve the low-frequency precipitation variability in RCMs
ice data as boundary forcing data. The AGCM is formulated
over a simpler GCM bias correction (Rocheta et al., 2017). In
on the basis of physical principles and able to generate a set
addition, it is not clear to what extent the NBC method re-
of large-scale data and retain the intervariable dependencies.
tains the intervariable dependencies warranting further study.
Finally, the large scale data generated by the AGCM and the
bias corrected SST and sea ice data were used to drive 4.6 Bias correction of multi-model ensemble
RCMs. The so-called “3-step dynamical downscaling”
(AOGCM-AGCM-RCM) is relative to the common 2-step All bias correction methods presented in the previous sec-
6 Xu Z, et al. Sci China Earth Sci

tions were applied to one GCM. It is well-known that the suggested that both GCM and RCM biases can degrade the
projection of future climate based on one GCM contains dynamical downscaling simulations. The downscaling si-
large uncertainties. Multi-model ensembles (MME) can mulations show larger bias when the GCM and RCM contain
significantly reduce the uncertainties of climate projection. the same-sign biases. Conversely, the downscaling simula-
Therefore, Dai et al. (2017) proposed a bias correction tions may show smaller bias, when GCM and RCM biases
method by taking MME into account, which is similar to the are of opposite-sign and cancel out each other. Clearly, this is
GCM mean bias correction except that the single GCM data a “good” result for the wrong reasons.
in eq. (2) was replaced with the climatological mean of The TDD approach (RCM_GCM) did not employ any
MME. Note that MME will cancel out GCM internal technique to constrain GCM and RCM biases, thereby gen-
variability to a large extent, which in turn leads to under- erally showing the largest errors. Compared to the TDD
estimating the variability of RCM and limiting its ability to method, the dynamical downscaling simulation with spectral
simulate extreme events. To retain the amplitude of inter- nudging (RCM_GCM.Ng) showed no clear improvement in
annual variation in GCM data, the anomaly field is still de- the air temperature because spectral nudging forced the
rived from data from one GCM. Consequently, the bias RCM towards biased GCM data. The dynamical down-
corrected GCM data can be constructed as: scaling simulation with GCM bias corrections improved the
BC F = GCM EF + GCM F middle- and upper-tropospheric air temperature but no clear
improvement was found in the lower troposphere
( GCM EH RA H), (4)
(RCM_GCMbc). In contrast, the new dynamical down-
where the subscript E indicates MME. The other symbols in scaling method with GCM bias corrections and spectral
eq. (4) are the same as in eq. (2). Clearly, the climate change nudging (RCM_GCMbc.Ng or RCM_GCMbc.Nglow1 in
between future and historical periods is derived from MME, Figure 3) developed by Xu and Yang (2015) performs best
i.e., GCM EF GCM EH. Compared with the climate change among all GCM-driven downscaling simulations. Therefore,
derived from one GCM, i.e., GCM F GCM H, the MME are both the GCM and RCM biases need to be constrained
expected to provide more reliable projection of future cli- during dynamical downscaling simulations. In addition to
mate. MME-based bias correction eliminated the internal the climatological mean state, the new dynamical down-
climate variability and retained the response to external scaling method also significantly improved the temporal
forcing. Such a method is particularly suitable for the pro- variability statistics for temperature and precipitation as well
jection of externally forced future climate change (Dai et al., as extreme events (Xu and Yang, 2015).
2017). Note that the GCM F in eq. (4) represents the anomaly Xu and Yang (2015) computed the ratio of ageostrophic
of one GCM over the future period that may still contain wind to total wind using 3-hourly outputs to examine the
biases in the amplitude of interannual and low-frequency intervariable dependencies in various dynamical down-
variations.

5. Dynamical downscaling method with GCM

bias corrections and spectral nudging

In addition to the GCMs, RCMs also contain systematic

biases not considered in the dynamical downscaling methods
presented in Section 4. To constrain both the GCM and RCM
biases, Xu and Yang (2015) employed a spectral nudging
technique during RCM integration after GCM mean and
variance bias corrections. Spectral nudging, a four-dimen-
sional data assimilation technique (Stauffer and Seaman,
1990; von Storch et al., 2000), forces RCM simulation to-
ward the large-scale driving data while retaining the small-
scale features generated by the RCMs. Spectral nudging can
introduce the GCM bias corrections throughout the RCM
domain rather than just limiting them to the initial and lateral Figure 3 Annual mean root mean square errors (RMSEs) of downscaled
boundary conditions, thereby minimizing climate drift re- air temperature. The RMSEs were computed between the dynamical
downscaling simulations and North American regional reanalysis data over
sulting from both the GCM and RCM biases. The compar- the model domain from 1981 to 2010. Modified based on Xu and Yang
ison of eight 30-year dynamical downscaling simulations (2015).
 Xu Z, et al. Sci China Earth Sci 7

scaling simulations. The results suggested that atmospheric troduced in dynamical downscaling studies, including mean
imbalance induced by GCM bias corrections appears to be bias correction, mean and variance bias corrections, quantile-
less important to the downscaled climate. However, the quantile correction, nest bias correction, and bias correction
ageostrophic wind component remarkably increased in RCM of multi-GCM ensembles. The advantages and limitations of
when spectral nudging was applied. Therefore, the dynami- each dynamical downscaling method are summarized in
cal downscaling method with GCM bias corrections and Table 1. In general, the dynamical downscaling methods with
spectral nudging did significantly disturb the intervariable GCM bias corrections perform better than the TDD method.
dependencies, which was the limitation of the method. To Note that all bias correction methods assume the GCM
mitigate this limitation, Xu and Yang (2015) suggested re- systematic biases are stationary in time, which is not strictly
ducing the nudging strength from 0.0003 to 0.00003. In held. The GCM biases can be divided into mean state and
doing so, the RCM can significantly improve the simulation climate change bias. Two biases may cancel out each other to
of air temperature, geopotential height, vector winds, and a certain degree. Consequently, correcting mean state bias
specific humidity and constrain the precipitation bias as well. may not always improve the quality of GCM data. Bias
correction can usually improve the quality of GCM data and
generate more reasonable dynamical downscaling simulation
6. Discussion and conclusions when the GCM systematic biases are clearly larger than the
climate change biases.
We reviewed various dynamical downscaling methods de- It is difficult to quantitatively compare the performance of
veloped in recent decades. In general, the GCM bias cor- various dynamical downscaling methods due to the differ-
rections can prevent propagation of biases within the model ences in experimental design, climate models, domain sizes,
chain and improve the RCM simulations (Ott et al., 2013). parameterization schemes, and variables assessed. Further
Therefore, various GCM bias correction methods are in- studies are warranted with common experimental designs to

Table 1 Comparison of various dynamical downscaling methods

Methods and selected references Advantages and application Limitations
TDD can be applied to a wide-range of studies, including
TDD ensemble dynamical downscaling simulation with TDD did not constrain the GCM and RCM biases.
Multi-GCMs and multi-RCMs. One can measure the The TDD simulations often contain larger
(Giorgi and Mearns, 1999) uncertainty of climate projections with the ensemble biases and uncertainties
downscaling simulations
The PGW method only requires monthly GCM data rather
than 6-hourly data. The PGW method removed the The PGW method is not well suited for the
PGW projection of climate variabilities, e.g. the
influences of GCM systematic biases on the RCM
(Kawase et al., 2009) simulations and was suitable for investigating the regional relationship between the ENSO and monsoon
climate response to global warming or the changes in extreme events

This method removed the GCM mean bias and allowed the
GCM mean bias correction temporal variability to change. It is simple and can
significantly improve the downscaled statistics of tropical Only the GCM mean bias is corrected, and other
(Holland et al., 2010) cyclones and regional climate compared with the TDD GCM biases are untouched
method
GCM mean and variance bias This method is simple and able to significantly improve the
corrections The variance bias correction may inappropriately
RCM simulations in terms of both climatological mean and exaggerate or minify the GCM trend
(Xu and Yang, 2012) extreme events

This method corrects all biases in the probability This method failed to retain the intervariable
GCM Quantile-quantile correction dependencies and introduced additional bias
distribution of GCM data and is a more thorough correction
(Colette et al., 2012) in the spatial gradient of variable and spurious
than the GCM mean and/or variance bias corrections precipitation variability
GCM bias correction with physical This method explicitly considered the physical constraint This method is relatively complex and lacks
consistent constraint between various variables, i.e., the hydrostatic equilibrium comparison with other dynamical downscaling
(Meyer and Jin, 2015) and geostrophic wind balance methods, e.g., the one with GCM mean bias correction
NBC is relatively complex. This method does not
Nested bias correction In addition to the GCM mean and variance biases, this significantly improve the low-frequency variability of
(Rocheta et al., 2017) method also corrected GCM low-frequency biases precipitation in RCM over a simpler GCM bias
correction
Bias correction of multi-GCM This method is simple and takes advantage of MME. It is
ensemble particularly suited for the projection of externally forced Only the GCM mean bias is corrected,
future climate chang and other GCM biases are untouched
(Dai et al., 2017)
Both the GCM and RCM biases are constrained, which can The combination of GCM bias correction and
GCM+RCM bias corrections more effectively prevent propagation of biases within the spectral nudging disturbs the intervariable
(Xu and Yang, 2015) model chain. It significantly improved the downscaled dependencies, which could lead to significant
climate bias in the downscaled precipitation
8 Xu Z, et al. Sci China Earth Sci

compare different aspects of the downscaled climate, such as Rabinovitz et al., 2006). In recent years, the science com-
the climatological mean state, climate variability, extreme munity has made considerable advances in developing
event, and the intervariable dependencies. Current bias cor- variable-resolution models (Harris and Lin, 2013; Zarzycki
rection methods only take part of GCM biases into account, et al., 2015; Huang et al., 2016). The variable-resolution
e.g., climatological mean, variance, and probability dis- model is one of the important directions in climate model
tribution. The biases in temporal variations, e.g., frequency development and a powerful tool for dynamical downscaling
and phase, were not well considered. Although the NBC studies.
method did improve GCM low-frequency variation, the
subsequent RCM simulation showed less improvement in the Acknowledgements This work was supported by the National
Natural Science Foundation of China (Grant Nos. 91637103, 41675105,
low-frequency climate (Rocheta et al., 2017). No matter
41675080).
which bias correction method is used, the dynamical down-
scaling simulation still strongly relies on the performance of
GCMs and RCMs, especially the ability to simulate temporal References
climate variation. Therefore, it is extremely important to Adam J C, Stephens J C, Chung S H, Brady M P, Evans R D, Kruger C E,
continually develop and improve the climate models as a Lamb B K, Liu M, Stöckle C O, Vaughan J K, Rajagopalan K, Harrison
J A, Tague C L, Kalyanaraman A, Chen Y, Guenther A, Leung F Y,
fundamental way to improve the future projection of regional
Leung L R, Perleberg A B, Yoder J, Allen E, Anderson S, Chan-
climate. The improvement of climate model performance drasekharan B, Malek K, Mullis T, Miller C, Nergui T, Poinsatte J,
requires the long-term efforts of scientists world-wide, and Reyes J, Zhu J, Choate J S, Jiang X, Nelson R, Yoon J H, Yorgey G G,
the model biases cannot be fully removed. Thus, bias cor- Johnson K, Chinnayakanahalli K J, Hamlet A F, Nijssen B, Walden V.
2015. BioEarth: Envisioning and developing a new regional earth
rection is still an efficient way to improve downscaling si-
system model to inform natural and agricultural resource management.
mulations warranting further studies. Clim Change, 129: 555–571
Most previous dynamical downscaling simulations were Alexandru A, de Elia R, Laprise R. 2007. Internal variability in regional
integrated with given SST and sea ice data. The interaction climate downscaling at the seasonal scale. Mon Weather Rev, 135:
3221–3238
between atmosphere and ocean is neglected. As a matter of Bao J W, Feng J M, Wang Y L. 2015. Dynamical downscaling simulation
fact, the air-sea interaction is of great importance, especially and future projection of precipitation over China. J Geophys Res-At-
in the Asian monsoon region. The RCM simulations with air- mos, 120: 8227–8243
sea interaction demonstrate a significant improvement in the Boulard D, Castel T, Camberlin P, Sergent A S, Bréda N, Badeau V, Rossi
A, Pohl B. 2016. Capability of a regional climate model to simulate
monsoon circulation, precipitation, and intra-seasonal climate variables requested for water balance computation: A case study
variability (Yao and Zhang, 2008; Ratnam et al., 2009; Fang over northeastern France. Clim Dyn, 46: 2689–2716
et al., 2013; Liao and Zhang, 2013; Zou et al., 2016). Note Bruyère C L, Done J M, Holland G J, Fredrick S. 2014. Bias corrections of
that the SST simulated by an atmosphere-ocean RCM may global models for regional climate simulations of high-impact weather.
Clim Dyn, 43: 1847–1856
also contain significant biases. To constrain SST biases and Bukovsky M S, Karoly D J. 2011. A regional modeling study of climate
allow regional air-sea coupling, one may consider using change impacts on warm-season precipitation in the central United
spectral nudging, which also warrants further studies. States. J Clim, 24: 1985–2002
Caron L P, Jones C G, Winger K. 2011. Impact of resolution and down-
In addition, most previous dynamical downscaling simu-
scaling technique in simulating recent Atlantic tropical cylone activity.
lations employed one-way nesting, i.e., GCM output driving Clim Dyn, 37: 869–892
RCM. Under certain circumstances, the feedback from RCM Chan S C, Kendon E J, Fowler H J, Blenkinsop S, Ferro C A T, Stephenson
to GCM could be very important. For example, an RCM can D B. 2013. Does increasing the spatial resolution of a regional climate
model improve the simulated daily precipitation? Clim Dyn, 41: 1475–
better resolve the deep convection over the maritime con-
1495
tinents, significantly improving the larges-scale pattern of Chang J, Peng X D, Fan G Z, Che Y Z. 2015. Error correction of numerical
tropical temperature when the two-way interaction between weaterh prediction with historical data(in Chinese). Acta Meteorol Sin,
the GCM and RCM was considered (Lorenz and Jacob, 73: 341–354
Chen L J, Li W J, Zhang P Q, Wang J G. 2003. Application of a new
2005). There are two options when taking two-way interac- downscaling model to monthly precipitation forecast (in Chinese). J
tion into account. One is the two-way nesting between the Appl Meteorol Sci, 14: 648–655
GCM and RCM. Unfortunately, GCMs and RCMs are Chen L, Ma Z G, Li Z H, Wu L, Flemke J, Li Y P. 2018. Dynamical
usually developed and maintained by different institutions. It downscaling of temperature and precipitation extremes in China under
current and future climates. Atmos-Ocean, 56: 55–70
is hard to couple them together to fulfill two-way nesting. Chen J, Xu C Y, Guo S L, Chen H. 2016. Progress and chanllenge in
The variable-resolution model is another option. The vari- statistically downscaling climate model outputs (in Chinese). J Water
able-resolution model employs a uniform model structure Resour Res, 5: 299–313
that avoids the inconsistency of physical processes between Colette A, Vautard R, Vrac M. 2012. Regional climate downscaling with
prior statistical correction of the global climate forcing. Geophys Res
the GCM and RCM. Moreover, the variable-resolution Lett, 39: L13707
model can describe the interaction between the regions with Cook K H, Vizy E K. 2008. Effects of twenty-first-century climate change
different spatial resolution (McGregor and Dix, 2001; Fox- on the Amazon rain forest. J Clim, 21: 542–560
 Xu Z, et al. Sci China Earth Sci 9

Dai A G, Rasmussen R M, Ikeda K, Liu C H. 2017. A new approach to Conf., Houston. 20690
construct representative future forcing data for dynamic downscaling. Hoffmann P, Katzfey J J, McGregor J L, Thatcher M. 2016. Bias and
Clim Dyn, 43 variance correction of sea surface temperatures used for dynamical
Déqué M. 2007. Frequency of precipitation and temperature extremes over downscaling. J Geophys Res-Atmos, 121: 12877–12890
France in an anthropogenic scenario: Model results and statistical cor- Hong C P, Zhang Q, Zhang Y, Tang Y H, Tong D, He K B. 2017. Multi-
rection according to observed values. Glob Planet Change, 57: 16–26 year downscaling application of two-way coupled WRF v3.4 and
Dickinson R E, Errico R M, Giorgi F, Bates G T. 1989. A regional climate CMAQ v5.0.2 over East Asia for regional climate and air quality
model for the western United States. Clim Change, 15: 383–422 modeling: Model evaluation and aerosol direct effects. Geosci Model
Done J M, Holland G J, Bruyère C L, Leung L R, Suzuki-Parker A. 2015. Dev, 10: 2447–2470
Modeling high-impact weather and climate: Lessons from a tropical Huang X, Rhoades A M, Ullrich P A, Zarzycki C M. 2016. An evaluation
cyclone perspective. Clim Change, 129: 381–395 of the variable-resolution CESM for modeling California’s climate. J
El-Samra R, Bou-Zeid E, Bangalath H K, Stenchikov G, El-Fadel M. 2017. Adv Model Earth Syst, 8: 345–369
Future intensification of hydro-meteorological extremes: Downscaling IPCC. 2013. Climate Change 2013. The Physical Science Basis. Con-
using the weather research and forecasting model. Clim Dyn, 49: 3765– tribution of Working Group I to the Fifth Assessment Report of the
3785 Intergovernmental Panel on Climate Change. In: Stocker T F, Qin D,
Fan L, Fu C, Chen D. 2005. Review on creating future climate change Plattner G K, Tignor M, Allen S K, Boschung J, Nauels A, Xia Y, Bex
scenarios by statistical downscaling techniques (in Chinese). Adv Earth V, Midgley P M, eds. Cambridge: Cambridge University Press. 1535
Sci, 20: 320–329 Jacob D, Podzun R. 1997. Sensitivity studies with the regional climate
Fan L J, Chen D L, Fu C B, Yan Z W. 2013. Statistical downscaling of model REMO. Meteorl Atmos Phys, 63: 119–129
summer temperature extremes in northern China. Adv Atmos Sci, 30: Kawase H, Yoshikane T, Hara M, Kimura F, Yasunari T, Ailikun B, Ueda
1085–1095 H, Inoue T. 2009. Intermodel variability of future changes in the Baiu
Fang Y J, Zhang Y C, Huang A N, Li B. 2013. Seasonal and intraseasonal rainband estimated by the pseudo global warming downscaling method.
variations of East Asian summer monsoon precipitation simulated by a J Geophys Res, 114: D24110
regional air-sea coupled model. Adv Atmos Sci, 30: 315–329 Khairunnisa Y, Wang K, Campbell P, Chen Y, Glotfelty T, He J, Pirhalla
Feng L, Zhou T, Wu B, Li T, Luo J J. 2011. Projection of future pre- M, Zhang Y. 2017. Decadal application of WRF/Chem for regional air
cipitation change over China with a high-resolution global atmospheric quality and climate modeling over the U.S. under the representative
model. Adv Atmos Sci, 28: 464–476 concentration pathways scenarios. Part 1: Model evaluation and impact
Fox-Rabinovitz M, Côté J, Dugas B, Déqué M, McGregor J L. 2006. of downscaling. Atmos Environ, 152: 562–583
Variable resolution general circulation models: Stretched-grid model Kitoh A, Ose T, Takayabu I. 2016. Dynamical downscaling for climate
intercomparison project (SGMIP). J Geophys Res, 111: D16104 projection with high-resolution MRI AGCM-RCM. J Meteorol Soc Jpn,
Fu C, Wang S Y, Xiong Z, Gutowski W J, Lee D K, McGregor J L, Sato Y, 94A: 1–16
Kato H, Kim J W, Suh M S. 2005. Regional climate model inter- Lauer A, Zhang C, Elison-Timm O, Wang Y, Hamilton K. 2013. Down-
comparison project for Asia. Bull Amer Meteorol Soc, 86: 257–266 scaling of climate change in the Hawaii region using CMIP5 results: On
Gao X J, Shi Y, Giorgi F. 2011. A high resolution simulation of climate the choice of the forcing fields. J Clim, 26: 10006–10030
change over China. Sci China Earth Sci, 54: 462–472 Leung L R, Qian Y, Bian X. 2003. Hydroclimate of the western United
Gao X J, Zhao Z C, Giorgi F. 2002. Changes of extreme events in regional States based on observations and regional climate simulation of 1981–
climate simulations over East Asia. Adv Atmos Sci, 19: 927–942 2000. Part I: Seasonal statistics. J Clim, 16: 1892–1911
Gao Y, Vano J A, Zhu C, Lettenmaier D P. 2011. Evaluating climate Liang X Z, Li L, Kunkel K E, Ting M, Wang J X L. 2004. Regional climate
change over the Colorado River basin using regional climate models. J model simulation of U.S. precipitation during 1982–2002. Part I: An-
Geophys Res, 116: D13104 nual cycle. J Clim, 17: 3510–3529
Giorgi F, Mearns L O. 1991. Approaches to the simulation of regional Liang X Z, Kunkel K E, Meehl G A, Jones R G, Wang J X L. 2008.
climate change: A review. Rev Geophys, 29: 191–216 Regional climate models downscaling analysis of general circulation
Giorgi F, Mearns L O. 1999. Introduction to special section: Regional models present climate biases propagation into future change projec-
climate modeling revisited. J Geophys Res, 104: 6335–6352 tions. Geophys Res Lett, 35: L08709
Giorgi F, Jones G, Asrar G R. 2009. Addressing climate information needs Liao Z J, Zhang Y C. 2013. Simulation of a persistent snow storm over
at the regional level: The CORDEX framework. WMO Bull, 58: 175– Southern China with a regional atmosphere-ocean coupled model. Adv
183 Atmos Sci, 30: 425–447
Giorgi F, Shields Brodeur C, Bates G T. 1994a. Regional climate change Liu C M, Liu W B, Fu G B, Ouyang R L. 2012. A disscussion of some
scenarios over the United States produced with a nested regional cli- aspects of statistical downscaling in climate impacts assessment (in
mate model. J Clim, 7: 375–399 Chinese). Adv in Water Sci, 23: 427–437
Giorgi F, Hostetler S W, Brodeur C S. 1994b. Analysis of the surface Lo J C F, Yang Z L, Pielke Sr. R A. 2008. Assessment of three dynamical
hydrology in a regional climate model. Q J R Met Soc, 120: 161–183 climate downscaling methods using the Weather Research and Fore-
Han X, Wei F Y. 2010. The influence of vertical atmospheric circulation casting (WRF) model. J Geophys Res, 113: D09112
pattern over East Asia on summer precipitation in the east of China and Lorenz P, Jacob D. 2005. Influence of regional scale information on the
its forecasting test (in Chinese). Chin J Atmosp Sci, 34:533–547 global circulation: A two-way nesting climate simulation. Geophys Res
Hanssen-Bauer I, Achberger C, Benestad R, Chen D, Førland E. 2005. Lett, 32: L18706
Statistical downscaling of climate scenarios over Scandinavia. Clim Maraun D, Wetterhall F, Ireson A M, Chandler R E, Kendon E J, Widmann
Res, 29: 255–268 M, Brienen S, Rust H W, Sauter T, Themeßl M, Venema V K C, Chun
Harris L M, Lin S J. 2013. A two-way nested global-regional dynamical K P, Goodess C M, Jones R G, Onof C, Vrac M, Thiele-Eich I. 2010.
core on the cubed-sphere grid. Mon Weather Rev, 141: 283–306 Precipitation downscaling under climate change: Recent developments
Hernández-Díaz L, Laprise R, Nikiéma O, Winger K. 2017. 3-Step dyna- to bridge the gap between dynamical models and the end user. Rev
mical downscaling with empirical correction of sea-surface conditions: Geophys, 48: RG3003
Application to a CORDEX Africa simulation. Clim Dyn, 48: 2215– McGregor J L, Dix M R. 2001. The CSIRO conformal-cubic atmospheric
2233 GCM. In: Hodnett P F, ed. UTAM Symposium on Advances in Math-
Holland G J, Done J, Bruyere C, Cooper C, Suzuki A. 2010. Model in- ematical Modelling of Atmosphere and Ocean Dynamics. New York:
vestigations of the effects of climate variability and change on future- Springer. 197–202
Gulf of Mexico tropical cyclone activity. In: Proc. Offshore Technology Mearns L O, Gutowski W, Jones R, Leung R, McGinnis S, Nunes A, Qian
10 Xu Z, et al. Sci China Earth Sci

Y. 2009. A regional climate change assessment program for North regional scale model. Atmos Res, 101: 574–594
America. Eos Trans AGU, 90: 311 von Storch H, Zorita E, Cubasch U. 1993. Downscaling of global climate
Meyer J D D, Jin J. 2015. Bias correction of the CCSM4 for improved change estimates to regional scales: An application to Iberian rainfall in
regional climate modeling of the North American monsoon. Clim Dyn, wintertime. J Clim, 6: 1161–1171
46: 2961–2976 von Storch H, Langenberg H, Feser F. 2000. A spectral nudging technique
Michelangeli P A, Vrac M, Loukos H. 2009. Probabilistic downscaling for dynamical downscaling purposes. Mon Weather Rev, 128: 3664–
approaches: Application to wind cumulative distribution functions. 3673
Geophys Res Lett, 36: L11708 Walton D B, Sun F, Hall A, Capps S. 2015. A hybrid dynamical-statistical
Niu X R, Wang S Y, Tang J P, Lee D K, Gao X J, Wu J, Hong S Y, downscaling technique. Part I: Development and validation of the
Gutowski W J, McGregor J. 2015. Multimodel ensemble projection of technique. J Clim, 28: 4597–4617
precipitation in eastern China under A1B emission scenario. J Geophys Wang B, Ding Q, Fu X, Kang I S, Jin K, Shukla J, Doblas-Reyes F. 2005.
Res-Atmos, 120: 9965–9980 Fundamental challenge in simulation and prediction of summer mon-
Ott I, Duethmann D, Liebert J, Berg P, Feldmann H, Ihringer J, Kunstmann soon rainfall. Geophys Res Lett, 32: L15711
H, Merz B, Schaedler G, Wagner S. 2013. High-resolution climate Wang J, Kotamarthi V R. 2015. High-resolution dynamically downscaled
change impact analysis on medium-sized river catchments in Germany: projections of precipitation in the mid and late 21st century over North
An ensemble assessment. J Hydrometeorol, 14: 1175–1193 America. Earth’s Future, 3: 268–288
Pielke Sr R A, Wilby R L. 2012. Regional climate downscaling: What's the White R H, Toumi R. 2013. The limitations of bias correcting regional
point? Eos Trans AGU, 93: 52–53 climate model inputs. Geophys Res Lett, 40: 2907–2912
Peng X D, Che Y, Chang J. 2013. A novel approach to improve numerical Wilby R L, Wigley T M L. 1997. Downscaling general circulation model
weather prediction skills by using anomaly integration and historical output: A review of methods and limitations. Prog Phys Geography, 21:
data. J Geophys Res-Atmos, 118: 8814–8826 530–548
Prinn R G. 2013. Development and application of earth system models. Wilby R L, Wigley T M L, Conway D, Jones P D, Hewitson B C, Main J,
Proc Natl Acad Sci USA, 110: 3673–3680 Wilks D S. 1998. Statistical downscaling of general circulation model
Ratnam J V, Behera S K, Doi T, Ratna S B, Landman W A. 2016. Im- output: A comparison of methods. Water Resour Res, 34: 2995–3008
provements to the WRF seasonal hindcasts over south Africa by bias Wu W, Lynch A H. 2000. Response of the seasonal carbon cycle in high
correcting the driving SINTEX-F2v CGCM fields. J Clim, 29: 2815– latitudes to climate anomalies. J Geophys Res, 105: 22897–22908
2829 Xiong Z, Yan X D. 2013. Building a high-resolution regional climate model
Ratnam J V, Giorgi F, Kaginalkar A, Cozzini S. 2009. Simulation of the for the Heihe River Basin and simulating precipitation over this region.
Indian monsoon using the RegCM3-ROMS regional coupled model. Chin Sci Bull, 58: 4670–4678
Clim Dyn, 33: 119–139 Xu Y, Zhang Y, Lin E, Lin W, Dong W, Jones R, Hassell D, Wilson S.
Rocheta E, Evans J P, Sharma A. 2017. Can bias correction of regional 2006. Analyses on the climate change responses over China under
climate model lateral boundary conditions improve low-frequency SRES B2 scenario using PRECIS. Chin Sci Bull, 51: 2260–2267
rainfall variability? J Clim, 30: 9785–9806 Xu Z F, Yang Z L. 2012. An improved dynamical downscaling method
Sato T, Kimura F, Kitoh A. 2007. Projection of global warming onto with GCM bias corrections and its validation with 30 years of climate
regional precipitation over Mongolia using a regional climate model. J simulations. J Clim, 25: 6271–6286
Hydrol, 333: 144–154 Xu Z F, Yang Z L. 2015. A new dynamical downscaling approach with
Shukla S, Lettenmaier D P. 2013. Multi-RCM ensemble downscaling of GCM bias corrections and spectral nudging. J Geophys Res-Atmos,
NCEP CFS winter season forecasts: Implications for seasonal hydro- 120: 3063–3084
logic forecast skill. J Geophys Res-Atmos, 118: 10770–10790 Xu Z F, Yang Z L. 2017. Relative impacts of increased greenhouse gas
Stauffer D R, Seaman N L. 1990. Use of four-dimensional data assimilation concentrations and land cover change on the surface climate in arid and
in a limited-area mesoscale model. Part I: Experiments with synoptic- semi-arid regions of China. Clim Change, 144: 491–503
scale data. Mon Weather Rev, 118: 1250–1277 Xue Y, Vasic R, Janjic Z, Mesinger F, Mitchell K E. 2007. Assessment of
Sun F P, Walton D B, Hall A. 2015. A hybrid dynamical-statistical dynamic downscaling of the continental U.S. regional climate using the
downscaling technique. Part II: End-of-century warming projections Eta/SSiB Regional Climate Model. J Clim, 20: 4172–4193
predict a new climate state in the Los Angeles region. J Clim, 28: 4618– Xue Y K, Janjic Z, Dudhia J, Vasic R, De Sales F. 2014. A review on
4636 regional dynamical downscaling in intraseasonal to seasonal simulation/
Sun Y, Zhong Z, Li T, Yi L, Camargo S J, Hu Y, Liu K, Chen H, Liao Q, prediction and major factors that affect downscaling ability. Atmos Res,
Shi J. 2017. Impact of ocean warming on tropical cyclone track over the 147-148: 68–85
western north pacific: A numerical investigation based on two case Yang Z L. 2015. Foreword to the special issue: Regional earth system
studies. J Geophys Res-Atmos, 122: 8617–8630 modeling. Clim Change, 129: 365–368
Tan J, Zhang Y, Ma W, Yu Q, Wang J, Chen L. 2015. Impact of spatial Yang H, Jiang Z H, Li Z X, Wang X F, Cui C G. 2017. Applicability of a
resolution on air quality simulation: A case study in a highly in- quantile-quantile (Q-Q) bias-correction method for climate dynamical
dustrialized area in Shanghai, China. Atmos Pollut Res, 6: 322–333 downscaling at Beijing Station (in Chinese). Acta Meteorol Sin, 75:
Tang J P, Li Q, Wang S Y, Lee D K, Hui P H, Niu X R, Gutowski J W J, 460–470
Dairaku K, McGregor J, Katzfey J, Gao X J, Wu J, Hong S Y, Wang Y Yao S X, Zhang Y C. 2008. Simulation of China summer precipitation with
Q, Sasaki H. 2016a. Building Asian climate change scenario by multi- a regional air-sea coupled model (in Chinese). Acta Meteorol Sin, 66:
regional climate models ensemble. Part I: Surface air temperature. Int J 131–142
Climatol, 36: 4241–4252 Yao J C, Zhou T J, Zou L W. 2018. Dynamical downscaling of tropical
Tang J P, Niu X R, Wang S Y, Gao H, Wang X, Wu J. 2016b. Statistical cyclone and associated rainfall simulations of FGOALS-g2 (in Chi-
downscaling and dynamical downscaling of regional climate in China: nese). Chin J Atmosp Sci, 42: 150−163
Present climate evaluations and future climate projections. J Geophys Yu M, Wang G. 2014. Impacts of bias correction of lateral boundary
Res-Atmos, 121: 2110–2129 conditions on regional climate projections in West Africa. Clim Dyn,
van der Linden P, Mitchell J F B. 2009. ENSEMBLES: Climate change and 42: 2521–2538
its impacts: Summary of research and results from the ENSEMBLES Zarzycki C M, Jablonowski C, Thatcher D R, Taylor M A. 2015. Effects of
project. Met Office Hadley Centre Rep, 160 localized grid refinement on the general circulation and climatology in
Varghese S, Langmann B, Ceburnis D, O’Dowd C D. 2011. Effect of the Community Atmosphere Model. J Clim, 28: 2777–2803
horizontal resolution on meteorology and air-quality prediction with a Zeng Q C, Zhou G Q, Liao H, Lin C H, Liu H Z, Wang B Z, Xie Z H, Xu Y
 Xu Z, et al. Sci China Earth Sci 11

F, Xue F, Zeng X D, Zhang F. 2008. Research on the earth systemp affected by global warming (in Chinese)? Clim Change Res, 10: 388–
dynamic model and some related numerical simulations (in Chinese). 390
Chin J Atmosp Sci, 32: 653–690 Zou L W, Zhou T J, Peng D D. 2016. Dynamical downscaling of historical
Zhao Z C, Luo Y. 1998. Advances in regional climate simulations over climate over CORDEX East Asia domain: A comparison of regional
1990s (in Chinese). Acta Meteorol Sin, 56: 225–246 ocean-atmosphere coupled model to stand-alone RCM simulations. J
Zhao Z C, Luo Y, Huang J B. 2014. Are extreme weater and climate events Geophys Res-Atmos, 121: 1442–1458

(Responsible editor: Jianping HUANG)