Vol.:(0123456789)1 3

Clim Dyn

DOI 10.1007/s00382-016-3484-x

Evaluation of the regional climate response in Australia

to large-scale climate modes in the historical NARCliM

simulations

L. Fita1,2  · J. P. Evans1,3 · D. Argüeso1,3 · A. King1,3,4 · Y. Liu1 

Received: 13 October 2015 / Accepted: 30 November 2016

© Springer-Verlag Berlin Heidelberg 2016

response to large-scale modes better than the driving rea-

nalysis, though no regional model improves on all aspects

of the simulated climate.

Keywords Regional climate modelling · Australia ·

Climate modes

1 Introduction

Regional Climate Modelling (RCM) projects are designed

to: improve the regional representation of climate obtained

from Global Climate Models (GCM) and to provide added-

value (Giorgi and Mearns 1991) at the regional scale to the

climate change signal. Global models are run at coarse spa-

tial resolutions and they do not properly represent regional/

local features such as: topography, coastlines and land

cover changes. Using GCM outputs as driving conditions

for RCMs, which are run at higher resolution over a speci-

fied smaller area, allows one to obtain consistent climate

information, but with a better representation of some of the

details of the area under study (Hong and Kanamitsu 2014).

RCM outputs (‘high’ resolution climatological data-

sets) are commonly analyzed using surface observations

(mostly precipitation and minimum/maximum tempera-

tures) from station measurements (Fernández et al. 2007)

or gridded versions of them (Corney et al. 2013; Jiménez-

Guerrero et al. 2013; Ji et al. 2016). A common analysis

is to present a comparison of the simulated fields with

the observed fields, by means of statistical values such

as bias, root mean square error (RMSE), pattern correla-

tion (Evans and McCabe 2010; Argüeso et al. 2011) and

quantile plots (García-Díez et al. 2012). Beyond the basic

evaluation statistics a wide variety of analysis has been

performed focused on other aspects of the simulations

Abstract NARCliM (New South Wales (NSW)/Austral-

ian Capital Territory (ACT) Regional Climate Modelling

project) is a regional climate modeling project for the Aus-

tralian area. It is providing a comprehensive dynamically

downscaled climate dataset for the CORDEX-AustralAsia

region at 50-km resolution, and south-East Australia at

a resolution of 10 km. The first phase of NARCliM pro-

duced 60-year long reanalysis driven regional simulations

to allow evaluation of the regional model performance.

This long control period (1950–2009) was used so that the

model ability to capture the impact of large scale climate

modes on Australian climate could be examined. Simula-

tions are evaluated using a gridded observational dataset.

Results show that using model independence as a criteria

for choosing atmospheric model configuration from dif-

ferent possible sets of parameterizations may contribute to

the regional climate models having different overall biases.

The regional models generally capture the regional climate

Electronic supplementary material The online version of this

article (doi:10.1007/s00382-016-3484-x) contains supplementary

material, which is available to authorized users.

* L. Fita

lluis.fita@lmd.jussieu.fr

1 Climate Change Research Center, UNSW, Sydney, Australia

2 Present Address: Laboratoire de Météorologique Dynamique

(LMD), CNRS, UPMC-Jussieu, Universit de Paris 6,

Tour 45-55, 2nd floor, Postal Box 99 4, place Jussieu,

75252 Paris Cedex 05, France

3 ARC Centre of Excellence for Climate System Science,

UNSW, Sydney, Australia

4 ARC Centre of Excellence for Climate System Science,

School of Earth Sciences, University of Melbourne, Sydney,

Australia

L. Fita et al.

1 3

such as extremes (Herrera et al. 2010; Domínguez et al.

2013; Evans and McCabe 2013; Lee et al. 2014).

Before undertaking future climate projections, it is

important to know the systematic errors, accuracy and

overall performance of the RCMs. This is commonly

done by running the model for a ‘control’ period, where

the boundary conditions are obtained from a reanalysis

and the outcomes from the simulations can be compared

directly to observations. The information obtained from

such a comparison provides an evaluation of the models’

performance and a baseline of their reliability in repro-

ducing the climate of the region, which are required to

put future climate simulations in context.

In this study, we present an evaluation of the control

period runs of the NARCliM project. Along with some

standard climatological statistics we also examine the

simulated impact of various large-scale climate modes

on Australian rainfall and temperature. It is assumed that

if the RCM is able to represent the regional effect of a

climate mode then longterm characteristics of the cli-

matology would be, at least partially, well captured by

the model. It is assumed that a proper representation of

the regional effect of a climate mode during the current

period ensures the good representation of the atmos-

pheric situation which the climate mode represents. Thus,

future changes in the intensity/frequency of the climate

modes and their related atmospheric effects will be also

well represented by the RCM. However, this will be lim-

ited by at least two factors: accurate representation of the

climate mode by the given GCM, and the assumption that

the atmospheric dynamics related to the climate mode

do not substantially change in the future climate. Only

climate modes that are known to affect Australian cli-

mate are investigated (Risbey et al. 2009), these are rep-

resented by the indices: Niño 1+2, Niño 3.4, IPO, SOI,

DMI, Blocking and SAM (see more details further in the

text and Table 2).

A number of large regional climate modeling projects

have been performed, including the current World Cli-

mate Research Program (WCRP) initiative CORDEX

(Giorgi et  al. 2009). Similar previous exercises include

PRUDENCE (Christensen et  al. 2007) and ENSEMBLES

(Hewitt 2005) over Europe, NARCAPP (Mearns et  al.

2009) for north America, CLARIS (Boulanger et al. 2011)

for south America and RMIP for Asia (Fu et  al. 2005).

Results from these experiments have shown added value

in terms of representing regional characteristics that the

global climate projections are not able to capture. Improve-

ments have been found in the representation of the dynam-

ics of the region such as sea-breeze and topographic effects

as well as improvements in the simulation of extreme

events (Feser et  al. 2011; Luca et  al. 2012; Corney et  al.

2013; Solman 2013; Lee and Hong 2014).

Australia presents a richness of climates. The interior

is dominated by a dry and hot desert climate which covers

large areas of the country. To the north, a tropical area with

monsoon precipitation is found. On the east coast, in the

space between the Great Dividing Range and the sea, there

is an area of temperate climate which is also observed on

the southern island of Tasmania. The south-west displays

a Mediterranean type climate. This richness in climates

poses a challenge for RCMs due to the various dynamics

and features that shape the climate across the continent.

The representation of large-scale climate modes strongly

depends on the dynamics of the GCM. However, the impact

on a given region of the dynamics induced by the differ-

ent phases of the large scale mode might be influenced by

the RCM. Regional to local scale processes are likely better

represented by the RCMs due to their higher resolution and

higher complexity of parameterizations. These processes

include: eddy-induced circulations caused by SAM (Hen-

don et  al. 2014), surface thermal anomalies influencing

precipitation (Hendon et  al. 2007), realistic atmospheric

dynamics (Lim and Hendon 2015) and the generation and

interaction of Indian Rossby waves trains with Australian

baroclinicity and topography (Cai et al. 2011) among oth-

ers. The teleconnections between remote Indian and Pacific

ocean areas and their related indices, like IOD and ENSO

(Cai et al. 2012) will not be impacted by the RCM, since

these areas lie far outside its simulation area.

This article is structured as follows: the first section

describes the design of the experiment. The results of

a standard statistic analysis of model performance with

respect observations is shortly presented in Sect.  4. Fol-

lowed by Sect. 5 which examines the RCMs sensitivity to

the different climate modes, while the final section repre-

sents the conclusions.

2 The NARCliM project

The NARCliM (http://www.ccrc.unsw.edu.au/NARCliM/)

project is a collaboration with the NSW and ACT govern-

ments. It is intended to produce a comprehensive and con-

sistent set of regional climate change projections for use in

government planning processes. The experimental design

is detailed in Evans et al. (2014). Further information about

the project and access to the output data can be found on

the AdaptNSW website (http://climatechange.environment.

nsw.gov.au/Climate-projections-for-NSW/About-NAR-

CliM). Phase one of NARCliM involves using three RCMs

to simulate 60-years (1950-2009) with boundary conditions

obtained from the NCEP/NCAR reanalysis Project (Kal-

nay et al. 1996, NNRP 1). In the present article, Regional

Climate Model (RCM) refers to the use of a Limited Area

Model (LAM, here WRF) with a given specific physical

Evaluation of the regional climate response in Australia to large-scale climate modes in the…

1 3

configuration. In this case, all RCMs share the dynamical

core and differ in the combinations of physical parameteri-

zations. It has been shown that different combinations of

physical parameterizations of a given LAM can provide

valuable information on the climate variability over a given

area (Fernández et al. 2007) and previous findings suggest

that the spread of results from models with different physi-

cal parameterization are as large as the spread from mod-

els with different dynamical cores (Jerez et al. 2013). The

RCMs are run in a two domain, one-way nested configura-

tion with the outermost domain at a coarse resolution of 50

km corresponding to the CORDEX AustraliaAsia AUS44

domain. A second domain focused on the main area of

interest covering NSW and the ACT is run at higher 10-km

resolution. Weak spectral nudging (von Storch et al. 2000)

of winds and geopotential down to around 500 hPa is also

used in the outer nest only. A second phase performed cli-

mate projections using 4 different GCMs from the CMIP3

(Olson et al. 2016).

The three RCMs used were chosen from an ensemble of

36 models created by choosing various physical parameter-

izations available within the Weather Research and Fore-

casting (Skamarock et al. 2008, WRF version 3.3) model-

ling framework. It has been shown that changes in physics

schemes provide an effective way to produce an ensemble

of RCM simulations (Fernández et al. 2007). It was desired

that these models should be able to simulate the climate of

the region well, while spanning as much of the variability

in the full ensemble as possible. Thus models were chosen

through a two-step process: first, the ’models’ performance

was evaluated across a collection of variables and metrics

(Evans et al. 2012; Ji et al. 2014); second, the models are

ranked using a metric that accounts for model independ-

ence based on the covariance of model errors (Bishop and

Abramowitz 2013). See more details in Evans et al. (2013).

Table  1 shows the physical parametrizations of the three

models selected using this method. As a complementary

analysis, the mean of the ensemble of the 3 WRF mod-

els will also be evaluated (labeled hereafter, Rmean). The

main focus of the article is the analysis on the large climate

simulated modes, for that reason, only results from monthly

means are used.

3 Datasets defining climate regions

Monthly precipitation, minimum and maximum tempera-

tures from a gridded observational dataset at 0.05◦ spa-

tial resolution and developed within the Australian Water

Availability Project (Jones et  al. 2009, AWAP) were

selected as observational reference. This dataset has been

widely used in observational (King et al. 2013) and model-

ling studies in Australia (Evans and McCabe 2010; Pepler

et  al. 2014). The density of surface observations used in

this gridded product varies substantially across Australia

(Jones et al. 2009, see discussion there in). To ensure that a

given grid point has a well observed climatology, we mask

the gridded data based on the availability of station data in

both space and time (see the supplementary material for

more information).

Many indices have been derived to characterize the

large-scale climate modes, particularly in the tropical

Pacific ocean where many different El Niño indices are

available. The selection of the indices is based in an inde-

pendence and atmospheric mechanism criteria. The inde-

pendence of the indices is based on a pair-wise pearson-

correlation analysis (see Fig. 2 in supplementary material)

and only indices with low correlations (below 0.5 among

all indices) were chosen. The mehcanism criteria will

include those indices which provide some insights from a

different atmospheric mechanism. Seven different climate

indices from observational or model derived sources have

been chosen (see Table 2, for more details). Note that over-

all, there are only two pairs of indices with larger than 0.5

pearson correlations: Niño 3.4 and dmi (correlation of 0.8),

blocking and sam (corr. 0.6).

Niño indices measure the sea surface temperature anom-

aly in different equatorial Pacific regions. The Southern

Oscillation Index (SOI), is computed as a pressure differ-

ence between Darwin (AUS) and Tahiti, and is a comple-

mentary way to observe the flow anomalies over the Equa-

torial Pacific Ocean. These indices provide a measure of the

El Niño/Southern Oscillation (ENSO) phenomena which

can significantly impact Australian rainfall. The Inter-dec-

adal Pacific Oscillation (IPO) represents a multi-decadal

sea surface temperature pattern in the Pacific Ocean. Dipole

Mode Index (DMI), provides a measure of the sea surface

temperature dipole in the Indian ocean. The DMI has been

found to influence precipitation mostly in southern Aus-

tralia (Ummenhofer et al. 2009, 2011). The blocking index,

Table 1 WRF physical schemes selected to generate the ensemble of

RCMs

pbl/sfc   Planetary boundary layer and surface schemes, cu   cumulus,

mp  microphysics, sw/lw rad  short and long wave radiation schemes,

MYJ   Meyor–Yamada–Janjić (Janjić 1994), Eta    scheme used in the

NCEP-eta simulations (Janjić 1996), YSU    Yonsei university (Hong

et  al. 2006), KF    Kain-Fritsch (Kain 2004), BMJ    Betts–Miller–

Janjić (Janjić 1994, 2000), WDM5    WRF Double-Moment 5 class

(Hong et  al. 2004; Lim and Hong 2010), Dudhia    (Dudhia 1989),

RRTM  Radpid Radiative Transfer Method, (Mlawer et al. 1997) and

CAM  Community Atmospheric Model, CAM (Collins et al. 2004)

Name pbl/Sfc cu mp sw/lw rad

R1 MYJ/Eta KF WDM5 Dudhia/RRTM

R2 MYJ/Eta BMJ WDM5 Dudhia/RRTM

R3 YSU/MM5 KF WDM5 CAM/CAM

L. Fita et al.

1 3

measures the strength of an almost persistent high pressure

zone in the southeastern area of Australia, which dominates

the flow in this part of the continent. It has been computed

by P. Maher (Climate Change Research Center, Sydney)

using the NOAA 20CR-reanalysis (Compo et al. 2011) fol-

lowing Bureau of Meteorology definition (Pook and Gib-

son 1999, BI index). The Southern Annular Mode (SAM),

provides information about the atmospheric flow around

Antarctica which affects the southern hemisphere storm

tracks.

Risbey et  al. (2009) analyzed the influence of each of

these indices on Australian precipitation using observa-

tions. They explicitly analyze the relative contribution to

precipitation variability of these climate modes, depending

on season and location. Other studies have investigated the

impact of ENSO on Australian temperatures (Power et al.

1998; Jones and Trewin 2000). In this study the influence

of different large-scale modes on precipitation, and mini-

mum and maximum temperature was investigated.

A set of climate divisions were identified to study

regional characteristics of the model performance from a

temporal perspective. Such regions were determined using

a multi-step regionalization based on the method proposed

by Argüeso et  al. (2011) and applied to daily precipita-

tion, minimum and maximum temperature. The multi-step

methodology consists of seasonal Probability Distribution

Function (PDFs) calculated for each of the variables, an

agglomerative clustering and a non-hierarchical k-means

clustering. The methodology is designed to limit the num-

ber of subjective choices to the minimum possible, make

the most of the strengths of each step while reducing their

weaknesses. The method removes, as much as possible, the

information redundancy through PDFs, provides a suitable

number of regions and a first guess via the agglomerative

clustering, and allows for recombination of clusters through

k-means. As a result, it was possible to define 14 areas

for Australia (Fig.  1) with climate similarities offering an

appropriate way of analysing model outputs at regional

scales. The regions obtained are comparable to previously

defined climate areas (Stern et al. 2000), and those defined

in CCIA, (Climate Change in Australia, http://www.climat-

echangeinaustralia.gov.au/en/) with some differences due to

the addition of other information such as: biophysical fac-

tors and expected climate change signals.

4 Basic statistic analysis

A standard evaluation of model outputs over the long term

simulated period, using AWAP data-set as reference, is pre-

sented in this section. The most important results are here

described, whilst the full description can be found in the

supplementary material.

Spatial distribution of bias (see in supplementary mate-

rial figure 2) show mixture of results that depend on: geo-

graphical zone, variable and model. Coastal and tropical

area precipitation is over estimated by the RCMs, with bet-

ter results for the R2 physical parametrizations (the unique

RCM with BMJ as cumulus scheme) with a signature of

the role of orography (the Great Dividing Range along

the Eastern side of the continent). R3 (the unique run with

YSU/MM5 and CAM/CAM as pbl/surface layer and sw/lw

radiative schemes) has the smallest bias in minimum tem-

perature with a mixture of areas with positive and negative

bias. A north-south pattern in the bias is also obtained by

R1 and R2 with little bias in the north and an increasingly

negative bias in the south. All 3 RCMs produce an oppo-

site signed bias for maximum temperature with respect to

the driving re-analysis. No signature of the orography is

observed in the bias of temperatures. Overall the collection

of biases across the three variables differs for each RCM

demonstrating some independence in their model errors as

expected given the method for their selection (Evans et al.

2014).

Analysis of the seasonal bias (figures 8 and 9 in supple-

mentary material) shows a stronger seasonal sensitivity of

Table 2 Table with the period

covered and origin of the index

data used. The standardized

version of SOI has been used

KNMI   Koninklijk Nederlands Meteorologisch Institut (Holland), MO-HCCC   Met. Office Hadley Centre

for Climate Change, NOAA  National Oceanic and Atmospheric Administration (USA), JAMSTEC  Japan

Agency for Marine-Earth Science and Technology (Japan), CCRC  Climate Change Research Center (Aus-

tralia), BAS  British Antarctic Survey (UK)

Index Period Source

Niño 1+2 1950–2009 NOAA, http://www.esrl.noaa.gov/psd/data/correlation/nina1.data

Niño 3.4 1950–2009 KNMI, http://climexp.knmi.nl/data/inino5.dat

IPO 1950–2007 MO-HCCC www.iges.org/c20c/IPO_v2.doc

SOI 1951–2009 NOAA, http://www.cpc.ncep.noaa.gov/data/indices/soi

DMI 1958–2009 JAMSTEC, http://www.jamstec.go.jp/frcgc/research/d1/iod/

DATA/dmi_HadISST.txt

Blocking 1950–2009 CCRC, P. Maher using NOAA 20CR re-analysis

SAM 1957–2009 BAS, http://www.nerc-bas.ac.uk/icd/gjma/sam.html

Evaluation of the regional climate response in Australia to large-scale climate modes in the…

1 3

the re-analysis compared to the RCMs. For precipitation,

one of the RCMs (R2) shows a stronger seasonality in the

bias, and corresponds to the run with the overall lowest

bias. Except for the winter season (JJA), the RCMs show

an overall overestimation of the convection even in the

north during the dry season. For minimum temperature, R3

shows higher seasonal sensitivity in the bias, and is also the

RCM run with the overall lower bias (also true for all sea-

sons). There are no remarkable differences between RCM

runs for the bias in maximum temperature as they all show

a sensitivity with season dependency.

Examination of the precipitation distribution through

region based q-q plots (see in suppl. material figures 10–20)

reveals a mixture of results among seasons, regions, and

variables. While no RCM consistently performs better than

the other RCMs they do in general perform better than the

driving reanalysis. The best performance tends to occur in

summer (DJF) rather then winter (JJA). In particular they

show underestimations of precipitation in the south-west

and in western Tasmania and the highest mountains likely

due to mislocation of storm tracks and under-representation

of topographic effects respectively. They also tend to pro-

duce too much precipitation in the dry season (JJA) in the

northern areas affected by the Monsoon.

Time-series analysis for all period (figures  21–28 in

suppl. material) shows a good match between simulated

and observed time evolutions of the most significant events

(dry/wet spells, cold/warm periods). For temperature,

agreement between WRF simulated outcomes and observa-

tions is better than with the driving re-analysis. However,

precipitation tends to be higher in the models, particularly

in the tropical regions for R1 and R3 which use the Kain-

Fritsch convection parameterization.

Due to the long period simulated, a trend analysis has

been performed (see tables 2–4 in suppl. material). In gen-

eral RCM runs show a mix of results on the precipitation

trends, with no clear improvement over the driving reanaly-

sis. For temperature, WRF models consistently improve on

the trends found in the driving reanalysis, even though they

tend to overestimate trends in the minimum temperature

and underestimate them for the maximum temperature.

A complementary fourier-based spectrum analysis (see

in suppl. material figures 29 and 30) shows clear peaks on

5 and 10 year periods for all-variables and data-sets used in

the study, indicating the importance of considering decade

scale variability in precipitation over much of the Austral-

ian continent.

5 Response to climate modes

In this section we analyse the long term climatological

response of model outputs to various large-scale climate

modes. This is done through correlations of local precipita-

tion and temperature fields with indices related to the cli-

mate modes. The indices examined are described in Sect. 3

and Table 2. It should be noted that the centres of action for

the climate modes occur largely outside the RCM domain.

The effects of these climate modes are translated into the

RCM domain through the lateral boundary conditions.

Here we examine how successful the RCMs are at propa-

gating these boundary condition changes into precipitation

region climate

1 extreme dry desert

2 alpine temperate

3 tropical

4 desert with some tropical storm intrusions

5 semi-desert

6 desert with some sporadic storm intrusions

7 temperate wet

8 Mediterranean

9 alpine wet

10 coastal sub-tropical

11 coastal temperate

12 tropical

13 tropical

14 mountain sub-tropical

Fig. 1 Climate affinity spatial regions for temperature and precipitation following a two step method (with seasonality included) using the

AWAP data-set (see text for more details), and its comparison of derived regions and mean Australian climate characteristics

L. Fita et al.

1 3

and temperature changes over Australia. The results for

selected seasons and indices are shown here, the full range

of results are contained within the supplementary material.

Indices were obtained from various sources as shown in

Table 2.

5.1 Spatial distribution

Figure  2 shows the spring (SON) correlations (computed

for all the period of coincident data between index and sim-

ulations) between climate modes represented by the SOI

and the blocking indices, and the local precipitation, mini-

mum and maximum temperature. The SOI index shows the

highest correlation values over a large area during spring

(see top panels on Fig. 2). Niño 3.4 shows the second larg-

est correlations during spring (see supplementary material).

Blocking during spring has the third largest impact (based

on overall correlation mean) on the climate conditions (see

lower panels on Fig. 2).

For the standardized SOI index, precipitation is highly

correlated over most of Australia. It covers most of the

north-eastern half of Australia, except along the east coast

fringe known as the eastern seaboard. This clearly depicts

the different dynamics of precipitation in this area, includ-

ing the somewhat unique character of the eastern seaboard.

The RCM runs show a larger area with significant corre-

lations than observed (more overall colored values in the

maps), while NNRP shows weaker correlations particularly

in the east. It must be remembered that much of the centre-

west of the country has unreliable or missing observations.

For minimum temperature, all models are able to show the

dipole pattern of correlations between the south-west cor-

ner and the north-east. However, models overestimated the

area covered by positive correlations, along with a mixture

of magnitudes and extension of covered area at the anti-cor-

relation zone. For maximum temperature, correlations are

less significant and models generally fail to reproduce the

observed weak negative correlation over eastern Australia.

Rmean shows a better correlation map except for precipita-

tion with a clear stronger and wider positive correlation.

Positive correlation during SON between the blocking

index and precipitation is well represented by the models

with some differences between them (Fig. 2). Blocking is

mostly correlated with precipitation over the southern and

eastern portions of Australia. Results clearly show the

impact on the correlation with the increase of horizontal

resolution from NARCliM runs. WRF runs show a positive

correlation in the windward section of the mountain range

and a negative one on the lee side. For the minimum tem-

perature, NNRP re-analysis shows too much of the north

being positively correlated and not enough of the south

being anti-correlated. The RCMs also have difficulty rep-

licating the observed correlations with no areas of positive

correlation captured and only R3 reproducing the extent of

the negatively correlated region. The larger significant anti-

correlation pattern for the maximum temperature is well

captured by all the models. However, the RCM runs under-

estimate the strength of the correlation. For this index,

Rmean presents better correlation maps than the individual

models except for minimum temperatures.

While the RCM simulations generally do well at cap-

turing the spatial distribution of correlations with climate

mode indices there are examples when this is not the case.

These are: SON Niño 1+2 tasmax, JJA Niño 3.4 tasmax,

MAM SAM tasmin, SON SAM tasmax (see Fig.  3), JJA

SOI std tasmax, JJA DMI tasmax, MAM blocking tasmax

(see the rest of them in figure 50, ‘Poorly captured corre-

lations’ in supplementary material). It is noticeable that

NNRP re-analyses shows a good correlation structure in

these examples. These poor correlations might show some

of the limitations of the methodology and might be of con-

cern in climate projections. However these poorly captured

correlations are not in general high, nor the most important

in the area where the agreement is missed (see further text

and Fig.  6). However for MAM SAM tasmin (SW Aus-

tralia, suppl. material figure  55) and JJA SOI std tasmax

(North Australia, suppl. material figure 56) (both with cor-

relations lower than 40%) they are the most important ones

in some areas. This might introduce some doubts in the

projected climate changes in the affected areas.

Knowing the long term characteristics (decadal variabil-

ity) of the IPO index, yearly mean correlations maps are

also provided (see Fig. 4) which would better capture the

influence of the index. Observations show a stronger cor-

relation using yearly mean averaged fields than seasonal

fields (see supplementary material). Yearly precipitation is

correlated over much of the Eastern part of the continent.

While not as uniform as in the observations, the RCMs

do show significant correlations scattered over the eastern

part of the continent, an improvement over the NNRP. All

models have difficulty reproducing the observed yearly

IPO correlations with temperature. Rmean improves the

relationship with precipitation by producing larger coher-

ent anti-correlated regions in the east, though it does not

improve the temperature correlations.

5.2 Regional index correlation

Correlation analysis is reproduced for the time-series of

the spatial mean for each climate region. Only the results

for the spring (SON) season are shown (Fig. 5) since it is

the season with the largest influences from these modes.

Results for other seasons can be found in the supplemen-

tary material. IPO has a cycle with yearly characteristics,

thus it is excluded from this analysis (seasonal based).

Evaluation of the regional climate response in Australia to large-scale climate modes in the…

1 3

As noted previously, SOI, Niño 3.4 and blocking show

the highest correlations in spring (SON). To a certain

extent, all regions show significant correlations with some

of the analysed indices, although this is less consistent for

temperature variables. The RCMs are able to capture the

different regional correlations with each index. The RCMs

better reproduce the observed regional correlations with

precipitation when compared to reanalysis. However, they

Fig. 2 Temporal correlation

maps of the grid point seasonal

mean values with the time-

series of various indices. For the

SOI index (left column), and for

the blocking index (right col-

umn). Correlations for seasonal

accumulated precipitation (top

row), mean seasonal minimum

temperature (middle row)

and mean seasonal maximum

temperature (bottom row). Only

correlations with a significance

above 95% threshold are shown.

Number inside maps show

the mean significance corre-

lated value for the entire map

(used to order the influence of

the modes). Correlations are

computed for the full period

of coincidence between index

values and simulation runs

L. Fita et al.

1 3

Fig. 3 As in Fig. 2, but for the indices, seasons and variables where WRF runs do not well capture the significant driven effects. SON Niño 1+2

tasmax (top), JJA Niño 3.4 tasmax (bottom left), MAM sam tasmin (bottom right)

Evaluation of the regional climate response in Australia to large-scale climate modes in the…

1 3

Fig. 4 As for Fig. 2, but for correlation for yearly averages with IPO index

L. Fita et al.

1 3

overestimate the correlation with minimum temperature

and underestimate the correlation with the maximum tem-

perature. The SAM index shows no significant correlations

with observed minimum temperature.

A composite map with the index with the largest sig-

nificant correlation at each grid point is also provided.

Again only for the SON season (Fig.  6, see the supple-

mentary material for the other seasons). For precipita-

tion, the pattern of indices are generally well represented

by the models. However, all models have areas that differ

from each other. These differences may be, at least partly,

due to the RCMs being chosen based on the independ-

ence of their model errors. However, an over depend-

ence on the DMI index in certain parts of the continent

is given, whereas in others there is a mixture between

SOI and Niño 3.4 indices. However, the relationship with

DMI in these areas is well known (Ummenhofer et  al.

2009; Cai et  al. 2012). The gray area in the maps cor-

responds to grid-points where the index with the highest

correlation was not significant at 95% level.

Models tend to overestimate the correlation with the

minimum temperature of the standardized SOI, IPO and

DMI indices (see middle panel in Fig. 6). R3 shows the

best representation of the spatial influence of the indi-

ces on minimum temperature and clearly differs from

R1 and R2. Most of the country does show significant

correlations between maximum temperature and block-

ing, while northern Australia has a mix of many influ-

ences. All models properly represent this with significant

correlations over a larger portion of the country with a

stronger DMI influence in R3. A better representation

of the dominant modes is obtained with Rmean except

in precipitation, where the blocking influence is clearly

underestimated.

Fig. 5 SON Regional (see definition of regions in Fig.  1) spatial

mean (x axis) time-series correlation with different climatological

indices (y axis). For precipitation (top), minimum temperature (bot-

tom left) and maximum temperature (bottom right). Only correlation

values above the 95% of significance are plotted in colors, non-signif-

icant are plotted in grey

Evaluation of the regional climate response in Australia to large-scale climate modes in the…

1 3

6 Summary and conclusions

A series of regional climatological runs have been per-

formed within the NARCliM project in order to inform the

climate change impact and adaptation work of state govern-

ments. Results of the control period runs (1950–2009) of

three different WRF configurations (RCMs) are analysed

here.

6.1 Summary

This analysis has focused on seasonal and longer time

scales. A high resolution gridded observational data-

set has been taken as the reference. Analysis is done in

multiple ways for precipitation, minimum and maximum

temperature.

Fig. 6 Map of the indices with the maximum correlation (1950–

2009) during SON at each grid point. For precipitation (top), mini-

mum (bottom left) and maximum (bottom right) temperatures. Each

index is given by a different colour. Only correlated indices with

more than 95% significance are shown

L. Fita et al.

1 3

A general comparison of the model outcomes with the

observed gridded data-set in terms of biases, q–q plots

and trend analysis has been done (see more detail in sup-

plementary material). All WRF runs present a positive pre-

cipitation bias along the Great Dividing Range (mountain

range along the East coast). R1 and R3, which were run

with the same cumulus scheme (Kain-Fristch) also present

a positive bias in the north which is dominated by tropical

precipitation during the monsoon season. R1 and R2 (shar-

ing MYJ/ETa pbl and Dudhia/RRTM radiative schemes)

present a negative bias in minimum temperature whereas

R3 presents a positive one. All WRF runs show a nega-

tive bias in maximum temperature. In general, WRF biases

differ from those shown by the forcing re-analysis NNRP.

Seasonal precipitation biases are lower in winter (JJA), but

temperatures biases are larger. Model runs present a quan-

tile dependency of precipitation error only for the extreme

percentiles, being worst in tropical regions. Distribution

of regional temperatures show similar errors for all quan-

tiles. WRF is capable of reproducing the transient evolution

of precipitation and temperatures in all regions. Climate

trends in precipitation from WRF are different than those

given by NNRP suggesting strong local effects. WRF runs

overestimate minimum temperature trends and underesti-

mate those for maximum temperature.

Following a standard statistical analysis, correlation

with different indices representing climatological modes is

also performed. In general patterns of the most significant

indices (in order, standardized SOI, Niño 3.4 and blocking)

on the most impacted season (SON, spring), are well cap-

tured by the RCMs. In most cases, the correlation between

the climate mode indices and local climate variable is

improved in comparison with the driving re-analysis. How-

ever, there are some examples where the WRF simulations

do not represent the correlation with particular indices well.

Due to the long-term variability of the IPO index (dec-

ades), a complementary analysis based on yearly means is

also presented. In this case the RCMs are better able to cap-

ture the observed correlation with precipitation, compared

to the driving reanalysis, but not with temperature.

At the regional scale, seasonal simulated correlations

match the observed correlations well, with larger discrepan-

cies in the temperature fields. Maps of the most correlated

index at each grid point and season, are less well repre-

sented with the worst results for the minimum temperature.

Evaluating the mean of the ensemble of WRF models

presents a mix of results. The ensemble mean generally

improves the trends compared to any individual model. In

some cases it produces a better correlation map with vari-

ous large scale modes. However, any such improvement is

case specific and it remains unclear whether the ensemble

mean is able to provide a general improvement in terms of

the correlation with large scale climate modes.

It has been found that climate tendencies (see supple-

mentary material) in Tasmania and Victoria (regions 7 and

9, with few grid-points in NNRP) differ between NNRP

forcing and WRF runs. These differences in simulated

trends provide a note of caution when examining future

trends in these regions.

6.2 Conclusions

Overall, results show that dynamical downscaling of the

NNRP re-analysis using these RCMs improves the simu-

lated climate in a number of ways, though not all aspects

are improved. Due to the length of simulations performed

(60 years, 1950–2009), it is possible to examine the influ-

ence of large scale climate modes on local precipitation and

temperature. In general the RCMs improve the representa-

tion of these teleconnections compared to the driving rea-

nalysis due to their higher resolution and better represen-

tation of regional atmospheric dynamics. Though there are

seasons and climate modes where the reanalysis performs

better. Thus while many aspects of the simulations demon-

strate the added value provided by the RCMs, the improve-

ments are not found for all indices, variables and seasons.

It is worth noting that the RCMs presented in this study

differ in key parametrizations, specifically chosen as they

produced independent errors measured against a select

group of events (Evans et  al. 2012, 2014; Ji et  al. 2014).

This independence at short time scales has manifested

itself as independence of the errors at climatological time

scales, as seen in the long-term biases, q-q plots and tel-

econnections with climate modes. This behaviour vali-

dates the selection process in terms of producing a small

RCM ensemble where each model contributes independent

information rather than simply being near copies of each

other. This also conforms with previous findings that multi-

physics ensembles can produce ensemble spreads similar

to multi-model ensembles (Vautard et  al. 2013; Kotlarski

et al. 2014).

Overall these RCMs are able to capture the spatial struc-

ture of the correlations of Australian climate with indices

of climate modes including the spatial variability, inter-

annual variability and trends, and teleconnections with

large scale climate modes quite well. In many cases the

RCMs improve on the climate produced by the driving rea-

nalysis. The results also identify reasons for caution when

using these WRF models to produce climate projections

including the over-estimation of precipitation in the tropical

north of Australia by RCMs R1 and R3 and the reversal of

rainfall trends for two regions in the south-east. WRF mod-

els are able to show improvement in the climate representa-

tion through better resolving regional rainfall drivers such

as topography, coastal effects and blocking events which

are particularly important in the eastern seaboard and

Evaluation of the regional climate response in Australia to large-scale climate modes in the…

1 3

Tasmania as shown previously (Evans and McCabe 2010,

2013). The Eastern seaboard is also impacted by storm sys-

tems known as East Coast Lows which are better captured

at these resolutions (Luca et  al. 2015, 2016b). We also

note that these WRF models have been extensively tested

for added value compared to the driving global model and

found to provide improvements across most aspects of the

model tested however improvements in precipitation were

generally smaller than those found for temperature (Luca

et  al. 2016a). These simulations are being performed as

part of the NARCliM project in order to produce regional

scale future climate projections that can be used in climate

change impacts and adaptation research.

Acknowledgements This research was undertaken with the assis-

tance of resources provided at the ‘NCI National Facility’ through the

National Computational Merit Allocation Scheme supported by the

Australian Government. NARCliM is funded through a consortium

of project partners including NSW Office of Environment and Herit-

age (OEH), ACT Environment and Sustainable Development Direc-

torate, Sydney Water, Sydney Catchment Authority, Hunter Water,

NSW Department of Transport, NSW Department of Primary Indus-

try, NSW Office of Water. The work was supported by the Australian

Research Council as part of the Future Fellowship FT110100576 and

Linkage Project LP120200777. Authors are also gratefully to all the

different institutions which compute and provide the global climate

indices. Authors also thanks, P. Maher from the ARCCSS-CCRC for

the computations of the blocking index.

References

Argüeso D, Hidalgo-Muñoz JM, Gámiz-Fortis SR, Esteban-Parra MJ,

Castro-Díez Y, Dudhia J (2011) Evaluation of wrf parameteriza-

tions for climate studies over southern spain using a multi-step

regionalization. J Clim 24:5633–5651

Bishop CH, Abramowitz G (2013) Climate model dependence and the

replicate earth paradigm. Clim Dyn 41:885–900. doi:10.1007/

s00382-012-1610-y

Boulanger JP, Schlindwein S, Gentile E (2011) Claris lpb wp1: meta-

morphosis of the CLARIS LPB European project: from a mecha-

nistic to a systemic approach. CLIVAR Exch 16:7–10

Cai W, van Rensch P, Cowan T, Hendon HH (2011) Teleconnection

pathways of ENSO and the IOD and the mechanisms for impacts

on Australian rainfall. J Clim 24(15):3910–3923. doi:10.1175/20

11JCLI4129.1

Cai W, van Rensch P, Cowan T, Hendon HH (2012) An asymmetry

in the IOD and ENSO teleconnection pathway and its impact

on Australian climate. J Clim 25(18):6318–6329. doi:10.1175/

JCLI-D-11-00501.1

Christensen JH, Carter TR, Rummukainen M, Amanatidis G (2007)

Evaluating the performance and utility of regional climate mod-

els: the PRUDENCE project. Clim Change 81:1–6. doi:10.1007/

s10584-006-9211-6

Collins WD, Rasch PJ, Boville BA, Hack JJ, McCaa JR, William-

son DL, Kiehl JT, Briegleb B, Bitz C, Lin SJ, Zhang M, Dai Y

(2004) Description of the NCAR community atmosphere model

(CAM 3.0). NCAR Technical Note NCAR/TN-464+STR:226

Compo GP, Whitaker JS, Sardeshmukh PD, Matsui N, Allan RJ, Yin

X, Gleason BE, Vose RS, Rutledge G, Bessemoulin P, Brönni-

mann S, Brunet M, Crouthamel RI, Grant AN, Groisman PY,

Jones PD, Kruk M, Kruger AC, Marshall GJ, Maugeri M, Mok

HY, Nordli O, Ross TF, Trigo RM, Wang XL, Woodruff SD,

Worley SJ (2011) The twentieth century reanalysis project. Q J R

Meteorol Soc 137:1–28. doi:10.1002/qj.776

Corney S, Grose M, Bennett JC, White C, Katzfey J, McGregor J,

Holz G, Bindoff NL (2013) Performance of downscaled regional

climate simulations using a variable-resolution regional climate

model: Tasmania as a test case. J Geophys Res 118(21):11936–

11950. doi:10.1002/2013JD020087

Domínguez M, Romera R, Sánchez E, Fita L, Fernández J, Jiménez-

Guerrero P, Montávez JP, Cabos WD, Liguori G, Gaertner MA

(2013) Present climate precipitation and temperature extremes

over Spain from a set of high resolution RCMs. Climate research

58:149–164. doi:10.3354/cr01186

Dudhia J (1989) Numerical study of convection observed dur-

ing the winter monsoon experiment using a mesoscale

two-dimensional model. J Atmos Sci 46:3077–3107.

doi:10.1175/1520-0469(1989)046<3077:NSOCOD>2.0.CO;2

Evans JP, McCabe MF (2010) Regional climate simulation over Aus-

tralia’s Murray-Darling basin: a multitemporal assessment. J

Geophys Res 115:D14114. doi:10.1029/2010JD013816

Evans JP, McCabe MF (2013) Effect of model resolution impact on

regional climate and climate change using WRF over south-east

Australia. Clim Res 56:131–145. doi:10.3354/cr01151

Evans JP, Ekström M, Ji F (2012) Evaluating the performance of a

WRF physics ensemble over south-east Australia. Clim Dyn

39:1241–1258. doi:10.1007/s00382-011-1244-5

Evans JP, Ji F, Abramowitz G, Ekström M (2013) Optimally choosing

small ensemble members to produce robust climate simulations.

Environ Res Lett 8:044050. doi:10.1088/1748-9326/8/4/044050

Evans JP, Ji F, Lee C, Smith P, Argüeso D, Fita L (2014) A regional

climate modelling projection ensemble experiment–NARCliM.

Geosci Model Dev 7(2):621–629. doi:10.5194/gmd-7-621-2014

Fernández J, Montávez JP, Sáenz J, González-Rouco JF, Zorita E

(2007) Sensitivity of MM5 mesoscale model to physical param-

eterizations for regional climate studies: annual cycle. JGR

112(D04):101. doi:10.1029/2005JD00664

Feser F, Rockel B, von Storch H, Winterfeldt J, Zahn M (2011)

Regional climate models add value to global model data: A

review and selected examples. Bull Am Meteor Soc 92:1181–

1192. doi:10.1175/2011BAMS3061.1

Fu C, Wang S, Xiong Z, Gutowski WJ, Lee DK, McGregor JL,

Sato Y, Kato H, Kim JW, Suh MS (2005) Regional climate

model intercomparison project for Asia. Bull Am Meteor Soc

86(2):257–266

García-Díez M, Fernández J, Fita L, Yagüe C (2012) Seasonal

dependence of WRF model biases and sensitivity to pbl schemes

over Europe. Q J R Met Soc 139:501–514

Giorgi F, Mearns LO (1991) Approaches to the simulation of

regional climate change: a review. Rev Geophy 29(2):191–216.

doi:10.1029/90RG02636

Giorgi F, Jones C, Asrar G (2009) Addressing climate information

needs at the regional level: the CORDEX framework. WMO

Bull 58:175–183

Hendon HH, Thompson DWJ, Wheeler MC (2007) Australian rain-

fall and surface temperature variations associated with the

Southern hemisphere Annul Mode. J Clim 20(11):2452–2467.

doi:10.1175/JCLI4134.1

Hendon HH, Lim EP, Nguyen H (2014) Seasonal variations

of subtropical precipitation associated with the South-

ern Annular Mode. J Clim 27(9):3446–3460. doi:10.1175/

JCLI-D-13-00550.1

Herrera S, Fita L, Fernández J, Gutiérrez JM (2010) Evaluation of

the mean and extreme precipitation regimes from the ensembles

regional climate multimodel simulations over Spain. J Geophys

Res 115(D21):117

L. Fita et al.

1 3

Hewitt CD (2005) The ENSEMBLES project: providing ensemble-

based predictions of climate changes and their impacts. EGGS

Newsl 13:22–25

Hong SY, Kanamitsu M (2014) Dynamical downscaling: fundamental

issues from an NWP point of view and recommendations. Asia

Pac J Atmos Sci 50(1):83–104. doi:10.1007/s13143-014-0029-2

Hong SY, Dudhia J, Chen SH (2004) A revised approach to ice-

microphysical processes for the bulk parameterization of cloud

and precipitation. Mon Weather Rev 132:103–120

Hong SY, Noh Y, Dudhia J (2006) A new vertical diffusion package

with an explicit treatment of entrainment processes. Mon Wea

Rev 134:2318–2341

Janjić ZI (1994) The step-mountain eta coordinate model: further

developments of the convection. Mon Weather Rev 122:927–945

Janjić ZI (1996) The surface layer in the NCEP eta model. Elev-

enth Conference on Numerical Weather Prediction, Norfolk,

VA, 1923 August. American Meteor Society, Boston, MA, pp

354–355

Janjić ZI (2000) Comments on development and evaluation of a con-

vection scheme for use in climate models. J Atmos Sci 57:3686

Jerez S, Montávez JP, Gómez-Navarro JJ, Lorente-Plazas R, Garcia-

Valero JA, Jiménez-Guerrero P (2013) A multi-physics ensemble

of regional climate change projections over the Iberian peninsula.

Clim Dyn 41(7):1749–1768. doi:10.1007/s00382-012-1551-5

Ji F, Ekström M, Evans JP, Teng J (2014) Evaluating rainfall patterns

using physics scheme ensembles from a regional atmospheric

model. Theor Appl Climatol 115(1–2):297–304. doi:10.1007/

s00704-013-0904-2

Ji F, Evans JP, Teng J, Scorgie Y, Argüeso D, Luca AD, Olson R

(2016) Evaluation of long-term precipitation and temperature

WRF simulations for southeast Australia. Clim Res 67:99–115

Jiménez-Guerrero P, Montávez JP, Domínguez M, Romera R, Fita L,

Fernández J, Cabos WD, Liguori G, Gaertner MA (2013) Mean

fields and interannual variability in RCM simulations over Spain:

the ESCENA project. Clim Res 57:201–220

Jones DA, Trewin BC (2000) On the relationships between the

el Niño-southern oscillation and Australian land surface

temperature. Int J Clim 20(7):697–719. doi:10.1002/1097-

0088(20000615)20:7<697::AID-JOC499>3.0.CO;2-A

Jones DA, Wang W, Fawcett R (2009) High-quality spatial climate

data-sets for Australia. Aust Meteorol Mag 58:233248

Kain JS (2004) The Kain-Fritsch convective parameterization: an

update. J Appl Meteorol 43:170–181

Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin

L, Iredell M, Saha S, White G, Woollen J, Zhu Y, Leetmaa A,

Reynolds R, Chelliah M, Ebisuzaki W, Higgins W, Janowiak J,

Mo KC, Ropelewski C, Wang J, Jenne R, Joseph D (1996) The

NCEP/NCAR 40-year reanalysis project B. Am Meteorol Soc

77:437–471. doi:10.1175/1520-0477(1996)077<0437:TNYRP>

2.0.CO;2

King AD, Alexander LV, Donat MG (2013) Asymmetry in the

response of eastern Australia extreme rainfall to low-frequency

pacific variability. Geophys Res Lett 40(10):2271–2277.

doi:10.1002/grl.50427

Kotlarski S, Keuler K, Christensen OB, Colette A, Déqué M, Gob-

iet A, Goergen K, Jacob D, Lüthi D, van Meijgaard E, Niku-

lin G, Schär C, Teichmann C, Vautard R, Warrach-Sagi K,

Wulfmeyer V (2014) Regional climate modeling on European

scales: a joint standard evaluation of the Euro-CORDEX RCM

ensemble. Geosci Model Dev Discuss 7:217–293. doi:10.5194/

gmdd-7-217-2014

Lee JW, Hong SY (2014) Potential for added value to downscaled cli-

mate extremes over Korea by increased resolution of a regional

climate model. Theor Appl Climatol 117:667–677. doi:10.1007/

s00704-013-1034-6

Lee JW, Hong SY, Chang EC, Suh MS, Kang HS (2014) Assess-

ment of future climate change over east Asia due to the RCP

scenarios downscaled by GRIMS-RMP. Clim Dyn 42:733–747.

doi:10.1007/s00382-013-1841-6

Lim EP, Hendon HH (2015) Understanding the contrast of Aus-

tralian springtime rainfall of 1997 and 2002 in the frame of

two flavors of el Niño. J Clim 28(7):2804–2822. doi:10.1175/

JCLI-D-14-00582.1

Lim KSS, Hong SY (2010) Development of an effective double-

moment cloud microphysics scheme with prognostic cloud con-

densation nuclei (ccn) for weather and climate models. Mon

Weather Rev 138:1587–1612

Luca AD, de Elía R, Laprise R (2012) Potential for added value in

precipitation simulated by high-resolution nested regional cli-

mate models and observations. Clim Dyn 38(5–6):1229–1247.

doi:10.1007/s00382-011-1068-3

Luca AD, Evans JP, Pepler A, Alexander L, Argüeso D (2015) Res-

olution sensitivity of cyclone climatology over eastern Aus-

tralia using six reanalysis products. J Clim 28:9530–9549.

doi:10.1175/JCLI-D-14-00645.1

Luca AD, Argüeso D, Evans JP, de Elía R, Laprise R (2016a) Quan-

tifying the overall added value of dynamical downscaling and

the contribution from different spatial scales. J Geophys Res

121(4):1575–1590. doi:10.1002/2015JD024009

Luca AD, Evans JP, Pepler AS, Alexander L, Argüeso D (2016b)

Evaluating the representation of Australian east coast lows in

a regional climate model ensemble. J South Hemisphere Earth

Syst Sci 66:108–124

Mearns L, Gutowski WJ, Jones R, Leung R, McGinnis S, Nunes A,

Qian Y (2009) A regional climate change assessment program

for north America. EOS Trans AGU 90:311–312. doi:10.1029/

2009EO360002

Mlawer EJ, Taubman SJ, Brown PD, Iacono MJ, Clough SA (1997)

Radiative transfer for inhomogeneous atmosphere: RRTM, a

validated correlated-k model for the long wave. J Geophys Res

102(D14):16663–16682

Olson R, Evans JP, Luca AD, Argüeso D (2016) The NARCliM pro-

ject: Model agreement and significance of climate projections.

Clim Res 69:209–227

Pepler A, Timbal B, Rakich C, Coutts-Smith A (2014) Indian

Ocean Dipole overrides ensos influence on cool season rainfall

across the eastern seaboard of Australia. J Clim 27:3816–3826.

doi:10.1175/JCLI-D-13-00554.1

Pook MJ, Gibson T (1999) Atmospheric blocking and storm tracks

during sop-1 of the frost project. Aust Meteor Mag 1:51–60

Power S, Tseitkin F, Torok S, Lavery B, Dahni R, McAvaney B

(1998) Australian temperature, Australian rainfall and the south-

ern oscillation, 1910–1992: coherent variability and recent

changes. Aust Meteorol Mag 47:85–101

Risbey JS, Pook MJ, McIntosh PC, Wheeler MC, Hendon HH (2009)

On the remote drivers of rainfall variability in Australia. Mon

Wea Rev 137(10):3233–3253. doi:10.1175/2009MWR2861.1

Skamarock WC, Klemp JB, Dudhia J, Gill DO, Duda DMBMG,

Huang XY, Wang W, Powers JG (2008) A description of the

advanced research WRF version 3. NCAR Technical Note

475:NCAR/TN475+STR

Solman SA (2013) Regional climate modeling over south America: a

review. Adv Meteorol 504357:13. doi:10.1155/2013/504357

Stern H, de Hoedt G, Ernst J (2000) Objective classification of Aus-

tralian climates. Aust Met Mag 49:87–96

von Storch H, Langenberg H, Feser F (2000) A spectral nudging tech-

nique for dynamical downscaling purposes. Mon Weather Rev

128(10):3664–3673

Ummenhofer CC, Gupta AS, Taschetto AS, England MH (2009)

Modulation of Australian precipitation by meridional

Evaluation of the regional climate response in Australia to large-scale climate modes in the…

1 3

gradients in East Indian Ocean sea surface temperature. J Clim

22(21):5597–5610. doi:10.1175/2009JCLI3021.1

Ummenhofer CC, Gupta AS, Briggs PR, England MH, McIntosh PC,

Meyers GA, Pook MJ, Raupach MR, Risbey JS (2011) Indian

and Pacific Ocean Influences on Southeast Australian Drought

and Soil Moisture. J Clim 24(5):1313–1336. doi:10.1175/2010J

CLI3475.1

Vautard R, Gobiet A, Jacob D, Belda M, Colette A, Déqué M,

Fernández J, García-Díez M, Goergen K, Güttler I, Halenka T,

Karacostas T, Katragkou E, Keuler K, Kotlarski S, Mayer S, van

Meijgaard E, Nikulin G, Patarčić M, Scinocca J, Sobolowski S,

Suklitsch M, Teichmann C, Warrach-Sagi K, Wulfmeyer V, Yiou

P (2013) The simulation of European heat waves from an ensem-

ble of regional climate models within the Euro-CORDEX pro-

ject. Clim Dyn 41:2555–2575. doi:10.1007/s00382-013-1714-z