Cold tongue and warm pool ENSO Events in CMIP5: Mean state and future projections

Abstract

The representation of the El Nino-Southern Oscillation (ENSO) under historical forcing and future projections is analyzed in 34 models from the Coupled Model Intercomparison Project phase 5 (CMIP5). Most models realistically simulate the observed intensity and location of maximum sea surface temperature (SST) anomalies during ENSO events. However, there exist systematic biases in the westward extent of ENSO-related SST anomalies, driven by unrealistic westward displacement and enhancement of the equatorial wind stress in the western Pacific. Almost all CMIP5 models capture the observed asymmetry in magnitude between the warm and cold events (i.e., El Ninos are stronger than La Ninas) and between the two types of El Ninos: that is, cold tongue (CT) El Ninos are stronger than warm pool (WP) El Ninos. However, most models fail to reproduce the asymmetry between the two types of La Ninas, with CT stronger than WP events, which is opposite to observations. Most models capture the observed peak in ENSO amplitude around December; however, the seasonal evolution of ENSO has a large range of behavior across the models. The CMIP5 models generally reproduce the duration of CT El Ninos but have biases in the evolution of the other types of events. The evolution of WP El Ninos suggests that the decay of this event occurs through heat content discharge in the models rather than the advection of SST via anomalous zonal currents, as seems to occur in observations. No consistent changes are seen across the models in the location and magnitude of maximum SST anomalies, frequency, or temporal evolution of these events in a warmer world.

Full text

Cold Tongue and Warm Pool ENSO Events in CMIP5: Mean State and
Future Projections
ANDRE
´AS. TASCHETTO,ALEXANDER SEN GUPTA,NICOLAS C. JOURDAIN,AND AGUS SANTOSO
Climate Change Research Centre, and ARC Centre of Excellence for Climate System Science, University of New
South Wales, Sydney, New South Wales, Australia
CAROLINE C. UMMENHOFER
Department of Physical Oceanography, Woods Hole Oceanographic Institution, Woods Hole, Massachusetts
MATTHEW H. ENGLAND
Climate Change Research Centre, and ARC Centre of Excellence for Climate System Science, University of
New South Wales, Sydney, New South Wales, Australia
(Manuscript received 23 July 2013, in final form 12 November 2013)
ABSTRACT
The representation of the El Ni~
no–Southern Oscillation (ENSO) under historical forcing and future pro-
jections is analyzed in 34 models from the Coupled Model Intercomparison Project phase 5 (CMIP5). Most
models realistically simulate the observed intensity and location of maximum sea surface temperature (SST)
anomalies during ENSO events. However, there exist systematic biases in the westward extent of ENSO-
related SST anomalies, driven by unrealistic westward displacement and enhancement of the equatorial wind
stress in the western Pacific. Almost all CMIP5 models capture the observed asymmetry in magnitude be-
tween the warm and cold events (i.e., El Ni~
nos are stronger than La Ni~
nas) and between the two types of
El Ni~
nos: that is, cold tongue (CT) El Ni~
nos are stronger than warm pool (WP) El Ni~
nos. However, most
models fail to reproduce the asymmetry between the two types of La Ni~
nas, with CT stronger than WP events,
which is opposite to observations. Most models capture the observed peak in ENSO amplitude around
December; however, the seasonal evolution of ENSO has a large range of behavior across the models. The
CMIP5 models generally reproduce the duration of CT El Ni~
nos but have biases in the evolution of the other
types of events. The evolution of WP El Ni~
nos suggests that the decay of this event occurs through heat
content discharge in the models rather than the advection of SST via anomalous zonal currents, as seems to
occur in observations. No consistent changes are seen across the models in the location and magnitude of
maximum SST anomalies, frequency, or temporal evolution of these events in a warmer world.
1. Introduction
The environmental and societal impacts of the El Ni~
no–
Southern Oscillation (ENSO) set against a gradual
warming of the background climate has prompted con-
certed efforts to improve our understanding of ENSO
behavior. Our capacity to predict the onset and dura-
tion of ENSO events has benefitted from sustained
observing systems (e.g., McPhaden et al. 1998) coupled
with developments in ENSO theories (e.g., Jin 1997), as
well as ongoing improvements of climate models such
as those facilitated by the Climate Model Intercomparison
Project (CMIP). In the present study, we assess the fi-
delity of climate models submitted to CMIP phase 5
(CMIP5) in simulating the interannual SST variability
in the tropical Pacific that is largely associated with
ENSO and examine how this variability is projected to
change in the future.
Previous studies have shown that both atmospheric
and oceanic signatures of ENSO events are asymmetric
in intensity, frequency, duration, spatial distribution,
and in their large-scale atmospheric responses. For
Corresponding author address: Andr
ea S. Taschetto, Climate
Change Research Centre, and ARC Centre of Excellence for
Climate System Science, University of New South Wales, Sydney
NSW 2052, Australia.
E-mail: a.taschetto@unsw.edu.au
15 APRIL 2014 T A S C H E T T O E T A L . 2861
DOI: 10.1175/JCLI-D-13-00437.1
Ó2014 American Meteorological Society

example, Hoerling et al. (1997) noted that the nonlinear
response in Northern Hemisphere precipitation and
atmospheric circulation to the warm and cold phases
of the Southern Oscillation can be attributed to non-
linearities in deep convection to SST. Burgers and
Stephenson (1999) reported a skewness in equatorial
eastern Pacific SST anomalies that has shown to be
related to the different air–sea feedback interactions
during the warm and cold ENSO phases (Kang and
Kug 2002;Frauen and Dommenget 2010). Other stud-
ies, on the other hand, attribute ENSO asymmetry to
nonlinear oceanic processes (e.g., An and Jin 2004;Su
et al. 2010). For instance, An and Jin attributed the
warm–cold amplitude asymmetry to a strong nonlinear
dynamic heating that enhances the warm events, as oc-
curred in the 1982/83 and 1997/98 events, and weakens
subsequent cold events.
The asymmetric characteristics of ENSO also mani-
fest in the location of the associated maximum SST
anomalies. Canonical El Ni~
no events generally show
largest SST anomalies in the eastern Ni~
no-3 region. In
contrast, La Ni~
na anomalies tend to peak in the central
Pacific, within the Ni~
no-4 region (e.g., Schopf and
Burgman 2006;Sun and Yu 2009). This spatial asym-
metry may be partly related to the nonlinear wind stress
response to SST anomalies associated with opposite
phases of ENSO (e.g., Kang and Kug 2002;Frauen and
Dommenget 2010;Dommenget et al. 2013).
The transition between ENSO phases occurs via a
negative feedback involving ocean wave dynamics (e.g.,
Battisti and Hirst 1989;Suarez and Schopf 1988;Jin
1997). In a zonally integrated sense, the action of in-
ternal waves leads to a buildup of ocean heat content in
the equatorial Pacific as El Ni~
no develops and is sub-
sequently drained off the equator, leading to a delayed
negative feedback and phase reversal to La Ni~
na con-
ditions (Jin 1997). This so-called recharge–oscillator
paradigm has been confirmed by observations (Meinen
and McPhaden 2000) but can only explain the linear
component of ENSO transitions (McGregor et al. 2013).
In reality, the warm and cold event transition is not
regular and ENSO events are also asymmetric in dura-
tion (e.g., Larkin and Harrison 2002;McPhaden and
Zhang 2009;Okumura and Deser 2010;Ohba and Ueda
2009;Ohba et al. 2010; Okumura et al. 2011). Warm SST
anomalies associated with strong El Ni~
no events tend to
decay relatively quickly after their peak in December
and are followed by cold SST anomalies in the equato-
rial Pacific. On the other hand, strong La Ni~
na events
often persist through the following year.
In addition to the nonlinear duration between the
Pacific warm and cold events, asymmetric behavior is
also observed between strong and weak events of the
same ENSO phase. Previous studies have identified
inter–El Ni~
no variations when the maximum SST
anomalies concentrate in the central rather than the
eastern Pacific. This central warming pattern appears as
the second mode of tropical Pacific SST variability in an
empirical orthogonal function (EOF) or rotated-EOF
analysis (Lian and Chen 2012). Some studies have pos-
tulated that the first two modes of variability in tropical
Pacific SST anomalies represent dynamically indepen-
dent processes (e.g., Ashok et al. 2007). Others, however,
argue that the central Pacific events can be considered
as a nonlinear manifestation of the canonical ENSO
(e.g., Trenberth and Smith 2009;Takahashi et al. 2011;
Dommenget et al. 2013;Johnson 2013). Whether or not
a separate mode to canonical ENSOs, these central Pa-
cific events have drawn considerable attention as they
have occurred more frequently in the past few decades
(e.g., Ashok et al. 2007;Lee and McPhaden 2010;Na et al.
2011).
The mechanisms that give rise to enhanced central
Pacific anomalies are still not fully understood. Ashok
et al. (2007) proposed that the recent weakening of
equatorial easterlies in the central Pacific and enhanced
easterlies to the east have decreased the zonal SST
gradient and flattened the thermocline, resulting in
a climate state more favorable for the evolution of the
central Pacific events. Choi et al. (2011,2012) suggested
that decadal changes in climate can play an important
role in modulating the occurrence of El Ni~
no with dif-
ferent warming signatures. It has been proposed that the
asymmetries between the cold and warm phases of the
Southern Oscillation may produce a nonzero residual
effect on the time-mean state of the tropical Pacific that
in turn modulates ENSO amplitudes (Yeh and Kirtman
2004;Rodgers et al. 2004). Sun and Yu (2009) suggested
that the spatial asymmetries between El Ni~
no and
La Ni~
na lead to an ENSO cycle that shifts the tropical
Pacific mean climate from a state favorable for strong
ENSO activity to a state that sustains weak ENSO ac-
tivity, a mechanism that has been reproduced by 3 of 19
CMIP3 models according to Yu and Kim (2011).
Different names have been ascribed to central Pacific
ENSO events, despite referring to essentially the same
SST structure. We adopt the ‘‘cold tongue’’ (CT) and
‘‘warm pool’’ (WP) terminologies to refer to ENSO
events with maximum SST anomalies located in the
eastern and central equatorial Pacific, respectively: CT
El Ni~
no, CT La Ni~
na, WP El Ni~
no, and WP La Ni~
na.
Regardless of whether the CT and WP events are in-
dependent modes of variability or a manifestation of
ENSO nonlinearity, observations demonstrate that
such spatial asymmetries are part of the interannual
variability of the region and that distinct atmospheric
2862 JOURNAL OF CLIMATE VOLUME 27

teleconnections and associated climate impacts arise
when SST peaks in the central or eastern Pacific (e.g.,
Ashok et al. 2007;Weng et al. 2007;Taschetto and
England 2009;Taschetto et al. 2009,2010). As such, it is
important that climate models simulate the character-
istics of these different ENSO flavors.
Despite significant advances in climate models, sim-
ulating realistic ENSO characteristics is still a major
challenge (Guilyardi et al. 2009b), largely associated
with the difficulty in representing multiple competing
feedback processes (e.g., Collins et al. 2010;Kim and
Jin 2011;Bellenger et al. 2013). Leloup et al. (2008)
assessed 23 CMIP3 models and concluded that the
majority of the models are not able to simulate
the location of maximum amplitude of warm and cold
events; only half can properly simulate ENSO onset,
and none can represent the correct termination phases
of either El Ni~
no or La Ni~
na. Nevertheless, climate
models have shown some improvements in simulating
ENSO-related SST variability and trends since CMIP3
(e.g., Yu and Kim 2010;Kim and Yu 2012;Bellenger
et al. 2013;Guilyardi et al. 2012;Yeh et al. 2012;
Jourdain et al. 2013).
The relatively short observational record available to
date means that understanding the dynamics behind WP
ENSO events, which have become more frequent in
recent decades, relies more on the use of climate models
(e.g., Dewitte et al. 2012). Sparse observational records
and the lack of a dynamical theory also imply uncer-
tainty in determining whether the recent increase in the
frequency of WP relative to CT El Ni~
no can be attributed
to greenhouse warming (Wittenberg 2009;McPhaden
et al. 2011;Newman et al. 2011;Yeh et al. 2011;Kim et al.
2012). Assessing how well the CMIP models capture
the observed temporal and spatial characteristics of
ENSO may help to clarify the dynamics underlying the
two types of ENSO as well as their future projections.
Here we use a large pool of available CMIP5 models
to provide a comprehensive assessment on their perfor-
mance in representing the two types of ENSO regarding
1) their spatial characteristics; 2) their associated atmo-
spheric, ocean surface, and subsurface properties; 3) their
evolution and seasonality; and 4) their projections in
a warmer climate scenario.
2. Models and methods
a. Observations and reanalysis data
The SST dataset used here is the Hadley Centre Sea
Ice and Sea Surface Temperature dataset version 1
(HadISST1) (Rayner et al. 2003). Wind stress data
are from the National Centers for Environmental
Prediction–National Center for Atmospheric Research
(NCEP–NCAR) reanalysis (Kalnay et al. 1996). In this
study, we consider the period from December 1949 to
November 2012 for the SST and wind stress fields.
Subsurface ocean temperature data is from the Sim-
ple Ocean Data Assimilation (SODA version 2.1.6) re-
analysis (Carton and Giese 2008), covering the period
from December 1958 to November 2008. The upper-
ocean heat content accumulated in the top 300 m av-
eraged between 38S and 38N is used as a proxy for the
equatorial Pacific thermocline depth (Zebiak 1989).
b. CMIP5 models
We analyze outputs from 34 climate models taking part
in CMIP5 that were used to inform the Intergovernmental
Panel on Climate Change (IPCC) Fifth Assessment
Report (AR5). A summary of the climate models is
shown in Table 1.
We examine two scenarios in this study: 1) historical
simulations, which are integrations from around 1850 to
at least 2005 using realistic natural and anthropogenic
forcings, and 2) representative concentration pathway
8.5 (RCP8.5) simulations, which are subject to increas-
ing radiative forcing from the end of the historical sim-
ulation to 2100 when the radiative forcing reaches
;8.5 W m
22
. The last 50 years of the twenty-first century
are analyzed in the RCP8.5 simulations. A description of
the CMIP5 experiment design can be found in Taylor
et al. (2012).
c. Methodology
For all variables, anomalies are calculated by removing
the long-term monthly climatology over the entire period
analyzed here. Time series of observations and sim-
ulations are then linearly detrended. When required,
3-month averages are calculated for the examination of
particular seasons: namely, December–February (DJF),
March–May (MAM), June–August (JJA), and September–
November (SON).
When a mean across all CMIP5 models is consid-
ered, the spatial fields are interpolated onto a common
18318grid for comparison with observations. Ensemble
members for individual models are averaged prior to
computing multimodel mean.
Where necessary, the estimate of the confidence levels
or spread across CMIP5 models is calculated via the
standard deviation among the models. The estimate of
significance levels is computed via null hypothesis using
a Student’s ttest at the 0.05 significance level.
The selection of ENSO years is based on the DJF
season, when observed events typically peak. Classi-
fying ENSO events in models can be challenging, as
models contain spatial SST biases. However, defining
15 APRIL 2014 T A S C H E T T O E T A L . 2863

TABLE 1. List of the CMIP5 models, with respective institutes, variables, and number of ensemble members used in this study.
Model
acronym Model Institute, country
Variables
SST Wind
stress
Subsurface
temperature ThermoclineHIST RCP8.5
ACCESS1.0 Australian Community Climate and
Earth-System Simulator, version 1.0
Commonwealth Scientific and Industrial Research
Organisation (CSIRO)–Bureau of Meteorology
(BOM), Australia
111 1 1
ACCESS1.3 Australian Community Climate and
Earth-System Simulator, version 1.3
BOM, Australia 1 1 1 1 1
BCC-CSM1.1 Beijing Climate Center, Climate System
Model, version 1.1
Beijing Climate Center (BCC), Chinese Meteorological
Administration (CMA), China
313
CanESM2 Second Generation Canadian Earth
System Model
Canadian Centre for Climate Modelling
and Analysis (CCCma), Canada
553 1 1
CESM1 (CAM5) Community Earth System Model, version 1
(Community Atmosphere Model, version 5)
National Science Foundation (NSF)–U.S. Department
of Energy (DOE)–NCAR, United States
23
CESM1
(FASTCHEM)
Community Earth System Model, version 1
(with FASTCHEM)
NSF–DOE–NCAR, United States 3
CESM1
(WACCM)
Community Earth System Model, version 1
[with the Whole Atmosphere Community
Climate Model (WACCM)]
NSF–DOE–NCAR, United States 1 1
CCSM4 Community Climate System Model, version 4 NCAR, United States 4 1 1 1 1
CMCC-CM Centro Euro-Mediterraneo per I
Cambiamenti Climatici Climate Model
Centro Euro-Mediterraneo per I Cambiamenti
Climatici (CMCC), Italy
11
CNRM-CM5 Centre National de Recherches
M
et
eorologiques Coupled Global
Climate Model, version 5
Centre National de Recherches M
et
eorologiques
(CNRM)–Centre Europ
een de Recherche et de
Formation Avanc
ee en Calcul Scientifique
(CERFACS), France
10 5 9 9 9
CSIRO Mk3.6.0 Commonwealth Scientific and Industrial
Research Organisation Mark, version 3.6.0
CSIRO–Queensland Climate Change
Centre of Excellence (QCCCE), Australia
10 10 10 10 10
FGOALS-g2 Flexible Global Ocean–Atmosphere–Land
System Model gridpoint, version 2
State Key Laboratory of Numerical Modeling for
Atmospheric Sciences and Geophysical Fluid
Dynamics (LASG)–Center for Earth System
Science (CESS), China
22
FGOALS-s2 Flexible Global Ocean–Atmosphere–Land
System Model, second spectral version
LASG–Institute of Atmospheric Physics (IAP), China 2 2
FIO-ESM First Institute of Oceanography (FIO)
Earth System Model (ESM)
FIO, State Oceanic Administration (SOA), China 3 3
GFDL CM3 Geophysical Fluid Dynamics Laboratory
Climate Model, version 3
National Oceanic and Atmospheric
Administration (NOAA)/Geophysical Fluid
Dynamics Laboratory (GFDL), United States
51 1 1
GFDL-ESM2G Geophysical Fluid Dynamics Laboratory
Earth System Model with Generalized
Ocean Layer Dynamics (GOLD) component
NOAA/GFDL, United States 1 1 1 1
2864 JOURNAL OF CLIMATE VOLUME 27

TABLE 1. (Continued)
Model
acronym Model Institute, country
Variables
SST Wind
stress
Subsurface
temperature ThermoclineHIST RCP8.5
GFDL-ESM2M Geophysical Fluid Dynamics Laboratory Earth
System Model with Modular Ocean Model 4
(MOM4) component
NOAA/GFDL, United States 1 1 1 1
GISS-E2-H Goddard Institute for Space Studies Model E2,
coupled with the Hybrid Coordinate Ocean
Model (HYCOM)
National Aeronautics and Space Administration
(NASA) Goddard Institute for Space
Studies (GISS), United States
5211
GISS-E2-R Goddard Institute for Space Studies Model E2,
coupled with the Russell ocean model
NASA GISS, United States 5 1
HadCM3 Hadley Centre Coupled Model, version 3 Met Office (UKMO) Hadley Centre,
United Kingdom
9444
HadGEM2-AO Hadley Centre Global Environment Model,
version 2–Atmosphere and Ocean
National Institute of Meteorological Research
(NIMR), Korea Meteorological
Administration (KMA), South Korea
11
HadGEM2-CC Hadley Centre Global Environment Model,
version 2–Carbon Cycle
UKMO Hadley Centre, United Kingdom 2 3 2 1 1
HadGEM2-ES Hadley Centre Global Environment Model,
version 2–Earth System
UKMO Hadley Centre, United Kingdom 2 4 2 1 1
INM-CM4.0 Institute of Numerical Mathematics
Coupled Model, version 4.0
Institute of Numerical Mathematics (INM), Russia 1 1 1 1 1
IPSL-CM5A-LR L’Institut Pierre-Simon Laplace Coupled Model,
version 5A, coupled with Nucleus for European
Modelling of the Ocean (NEMO), low resolution
L’Institut Pierre-Simon Laplace (IPSL), France 4 4 4 2 2
IPSL-CM5B-LR L’Institut Pierre-Simon Laplace Coupled Model,
version 5B, coupled with NEMO, low resolution
IPSL, France 1 1 1 1
IPSL-CM5A-MR L’Institut Pierre-Simon Laplace Coupled Model,
version 5A, coupled with NEMO, mid resolution
IPSL, France 1 1 1 1 1
MIROC5 Model for Interdisciplinary Research on Climate,
version 5
Atmosphere and Ocean Research Institute
(AORI)–National Institute for Environmental
Studies (NIES)–Japan Agency for Marine-Earth
Science and Technology (JAMSTEC), Japan
333 1 1
MIROC-ESM Model for Interdisciplinary Research on
Climate, Earth System Model
AORI–NIES–JAMSTEC, Japan 3 3 3 3
MPI-ESM-LR Max Planck Institute Earth System Model, low
resolution
Max Planck Institute for Meteorology (MPI-M),
Germany
33 3 3
MPI-ESM-MR Max Planck Institute Earth System Model, medium
resolution
MPI-M, Germany 3 1 1 1
MRI-CGCM3 Meteorological Research Institute Coupled
Atmosphere–Ocean General Circulation
Model, version 3
Meteorological Research Institute (MRI), Japan 3 1 3 1 1
15 APRIL 2014 T A S C H E T T O E T A L . 2865

model-specific ENSO classifications to take into ac-
count model biases introduces subjective decisions. In
addition, allowing multiple ENSO classifications would
make model–observation intercomparison more diffi-
cult. As such, we adopt one common classification
based on the state of equatorial Pacific SST anomalies.
Even for observations, there is still no single method to
classify two types of the El Ni~
no pattern (e.g., Ashok
et al. 2007;Yeh et al. 2009).
Here we adopt the method of Ham and Kug (2012) to
classify ENSO events. The normalized DJF-averaged
Ni~
no indices are used as follows: An event is considered
aCTElNi
~
no if the Ni~
no-3 index is greater than one
standard deviation and the averaged SST anomaly in the
Ni~
no-3 region has a larger magnitude than the Ni~
no-4
SST anomaly. An event is classified as WP El Ni~
no if the
Ni~
no-4 index is above one standard deviation and the
magnitude of the SST anomaly in the Ni~
no-4 region is
larger than in Ni~
no-3. The opposite is used for La Ni~
na
events. Table 2 summarizes the ENSO classification for
the purpose of this paper and shows the years of dif-
ferent types of ENSO events from the observations.
ENSO events are selected for individual members of
each CMIP5 model when more than one ensemble
simulation is available.
Our analysis simplifies ENSO spatial pattern into CT
and WP categories. This does not imply that these dif-
ferent ENSO flavors are related to distinct dynamical
modes or that ENSO patterns are strictly bimodal. This
is just a convenient way to examine differences in sim-
ulated ENSO characteristics (compared to the obser-
vations) when the primary variability is shifted more to
the east or west.
To provide quantitative measures of the spatial struc-
ture of different ENSO flavors, we calculate the magni-
tude and location of the maximum SST anomaly (SSTA)
along the equator, as well as the westward extent of the
warm or cold anomalies. A model is considered to over-
estimate or underestimate the magnitude of El Ni~
no or
La Ni~
na events if jSSTAjmax
model exceeds (jSSTAjmax
observed 1
sSSTA
observed), where jSSTAjmax
model and jSSTAjmax
observed are the
maximum magnitude of the SST anomaly composites
along the equator (meridionally averaged between 58S
and 58N) for the models and observations, respectively,
and sSSTA
observed is the standard deviation of the maximum
SST anomaly over all events included in the observa-
tional composite. Similarly, a model is considered to
have a ‘‘realistic’’ representation of the ENSO position
if the longitude of the maximum SST anomaly falls
within one standard deviation of the mean longitude of
the observed composite events. The metric for the
westward extent of the SST anomalies is defined as the
most westward longitude where SSTA drops to half of
TABLE 1. (Continued)
Model
acronym Model Institute, country
Variables
SST Wind
stress
Subsurface
temperature ThermoclineHIST RCP8.5
NorESM1-M Norwegian Earth System Model, version 1
(intermediate resolution)
Norwegian Climate Centre (NCC), Norway 3 1 3 1 1
NorESM1-ME Norwegian Earth System Model, version 1
(intermediate resolution) with
interactive carbon cycle
NCC, Norway 1 1
2866 JOURNAL OF CLIMATE VOLUME 27

its maximum value. Note that, although this metric can
be subjective, it is not restrictive in the sense that it
accounts for spatial biases in each model in terms of the
magnitude of ENSO SST.
3. Results
a. Number of ENSO events
Figure 1 shows the number of events for each ENSO
type based on the historical simulations for each model,
the multimodel mean, and observations. To facilitate
comparison between models and with observations, the
number of events are shown per 100 years. For most of
the models, the number of El Ni~
no events is comparable
to the observations. For instance, there is a median of 11
CT El Ni~
no events (100 yr)
21
in historical simulations
versus approximately 10 events (100 yr)
21
in observa-
tions. There are 10 WP El Ni~
no (100 yr)
21
in both sim-
ulations and observations. The number of La Ni~
na
events is not as well captured as for El Ni~
no: there are 12
FIG. 1. Number of ENSO events (100yr)
21
in the historical simulations for each model, multimodel mean, and
observations. Number of events is averaged for models containing more than one member. (top) Warm events: CT
El Ni~
no (CTEN) represented by black circles and WP El Ni~
no (WPEN) by white squares. (bottom) Cold events: CT
La Ni~
na (CTLN) represented as black circles and WP La Ni~
na (WPLN) as white squares. Bars in the multimodel
mean indicate the interquartile range.
TABLE 2. Summary of criteria employed for the ENSO classification for observations and CMIP5 models, using the standardized DJF
Ni~
no-3 and Ni~
no-4 indices. Years refer to January–February. For further details see text.
Event Criteria Years selected in observations
Cold tongue El Ni~
no Ni~
no-3 .1.0 and Ni~
no-3 .Ni~
no-4 1966, 1973, 1983, 1987, 1992, and 1998
Warm pool El Ni~
no Ni~
no-4 .1.0 and Ni~
no-4 .Ni~
no-3 1958, 1969, 1988, 1995, 2003, and 2010
Cold tongue La Ni~
na Ni~
no-3 ,21.0 and Ni~
no-3 ,Ni~
no-4 1950 and 1985
Warm pool La Ni~
na Ni~
no-4 ,21.0 and Ni~
no-4 ,Ni~
no-3 1956, 1971, 1974, 1976, 1989, 1999, 2000, 2001, 2008, 2011, and 2012
15 APRIL 2014 T A S C H E T T O E T A L . 2867

WP La Ni~
na events (100 yr)
21
in simulations versus 17
events (100 yr)
21
in observations, and the median
number of simulated CT La Ni~
na events (100 yr)
21
is 8,
while the observed number is only 3 events (100 yr)
21
.
Despite overestimating the number of CT La Ni~
nas and
underestimating the number of WP La Ni~
nas, the ob-
served asymmetry in the number of cold events is rep-
resented in most of the models, with more WP than CT
La Ni~
na events.
The number of observed ENSO events shown in Fig. 1
suggests that the separation of cold events into CT and
WP La Ni~
nas is more difficult than for warm events,
agreeing with previous studies (e.g., Kug and Ham
2011). However, while there is a clear preference for WP
to CT La Ni~
na occurrence in observations, some CMIP5
models simulate similar numbers of the two types of cold
events (e.g., CSIRO Mk3.6.0, CanESM2, HadCM3, the
HadGEM2 models, the IPSL models, and MIROC-
ESM). Only approximately one-third of the CMIP5
models simulate the observed preference for WP La Ni~
na
events: that is, CCSM4, the CESM models, CMCC-CM,
CNRM-CM5, the FGOALS models, FIO-ESM, GFDL
CM3, GFDL-ESM2M, GISS-E2-R, and MIROC5.
b. Spatial pattern of ENSO
The multimodel composite of simulated SST anoma-
lies during austral summer (DJF) for both El Ni~
no and
La Ni~
na events (Fig. 2, center) shows a number of fea-
tures in common with the observations (Fig. 2, left). A
quantification of the magnitude and location of the
maximum SST anomaly as well as the westward extent
of the ENSO pattern are presented in Fig. 3 for each
model, the multimodel mean, and observations. The
numbers of models that underestimate, overestimate, or
‘‘realistically’’ (see methods) represent the ENSO types
are summarized in Table 3.
Most of the models simulate the magnitude of CT El
Ni~
no anomalies in the equatorial Pacific within the ob-
servational range. The magnitude of CT El Ni~
no events
is overestimated in only two models and underestimated
in 11 models. This does not necessarily imply biases in
the location or extension of SST anomalies during CT El
Ni~
no. In fact, the majority of the models (29 out of 34
models), regardless of the magnitude of maximum SST
anomalies, simulate a realistic maximum SST anomaly
position during CT El Ni~
no events between 1128and
FIG. 2. Composite of SST anomalies (8C) during the December–February season for ENSO events. (left) Observations from HadISST
based on the period December 1949–November 2012. (center) Multimodel mean based on 34 CMIP5 models. Areas within the thin gray
line are statistically significant across the observed events and the composited events of CMIP5 models at the 0.05 significance level based
on a Student’s ttest. (right) The December–February SST anomalies averaged over 58S–58N. The brown line represents the multimodel
mean, while the green line represents observations. The light brown shading indicates the standard deviation of simulated composites, an
estimate of the spread among CMIP5 models.
2868 JOURNAL OF CLIMATE VOLUME 27

FIG. 3. (a) Magnitude and (b) location of the maximum SST anomaly for each ENSO type. (c) Westward
extension of SST anomaly estimated as the location of half of the simulated maximum SST anomaly. Error bars
(also represented as dashed lines) in HadISST show the 90% confidence interval for the mean of observed
events. Error bars in the multimodel mean (MMM) represent the 90% confidence intervals for the mean of
composited events across the CMIP5 models. The mean and associated error of each type of ENSO event are
specified with the same color as in the legend for the MMM and observations. Models withinthe dashed lines are
considered to have a realistic simulation of the metric.
15 APRIL 2014 T A S C H E T T O E T A L . 2869

1468W(Table 3). However, the majority of the models
have an extension bias with the warm pattern extending
too far to the west. On the other hand, only four models
reveal a CT El Ni~
no pattern with an eastward bias:
namely, FIO-ESM and the HadGEM2 models.
The bias in the westward extension and magnitude of
SST anomalies is more severe for WP El Ni~
no com-
posites. In the observations the core of the warming is
relatively narrow in the zonal direction (Fig. 2b), while
the simulations reveal an elongated pattern with stron-
ger anomalies (Fig. 2j). Only nine CMIP5 models can
represent the relatively confined warming in the central
Pacific (Figs. 2f,b,3c) and only six of the models are able
to simulate the maximum SST anomalies within the
observed range (1598–1658W). Despite this, 18 out of 34
CMIP5 models simulate the magnitude of WP El Ni~
no
events within the observational range.
Despite biases in the magnitude and spatial extent
of SST anomalies, all of the models (except CSIRO
Mk3.6.0) reproduce the observed asymmetry between
CT and WP warm events: they simulate relatively strong
warm events in the east and relatively weak warm events
in the west (Fig. 3a). The exception is the CSIRO
Mk3.6.0 model that fails to capture the location and
magnitude of maximum SST anomalies during WP El
Ni~
no events and simulates weaker SST conditions dur-
ing CT instead of WP El Ni~
no (i.e., the asymmetry is in
the opposite sense to the observations).
While CMIP5 models simulate the observed asym-
metry in magnitude between the two types of warm
TABLE 3. Number of models out of 34 that underestimate (Y), overestimate ([), and reproduce ‘‘realistic’’ (0) types of ENSO. Largest
numbers are in boldface. See methodology for explanation.
CT El Ni~
no WP El Ni~
no CT La Ni~
na WP La Ni~
na
Y0[Y 0[Y 0[Y 0[
Magnitude of max SST anomaly 11 21 21018 6023 11 8 26 0
Location of max SST anomaly 0 29 510 618 034 0430 0
Extension of the western
0.5 times max magnitude
16 14 4 24 91034 028 24
FIG. 4. Composite of wind stress anomalies (zonal component shaded) during the December–February season for ENSO events.
Maximum vector length is 0.08 Pa. (left) NCEP–NCAR reanalysis. (center) Multimodel mean based on 20 CMIP5 models. Areas within
the thin gray line are statistically significant across the observed events and the composited events of CMIP5 models at the 0.05 significance
level based on a Student’s ttest. (right) The December–February zonal wind stress anomalies averaged over 38S–38N. The red (blue) line
represents the multimodel mean (reanalysis). A 158window running mean was applied to the curves. The light red shade indicates the
standard deviation of simulated composites, an estimate of the spread among CMIP5 models.
2870 JOURNAL OF CLIMATE VOLUME 27

events, this is not the case for the cold events. In contrast
to the observations, 22 out of 34 models simulate
stronger CT events than WP La Ni~
na (Figs. 2g,h,3a).
Nevertheless, the majority of the models (23 out of 34)
are able to simulate CT La Ni~
nas with a similar magni-
tude as in the observations (Fig. 2k). In addition, the
location and westward extension of CT La Ni~
na events
are within the observational range for most CMIP5
models. It is important to note that, due to the small
sample of observed CT La Ni~
na events, the comparison
with observations is limited.
There is a systematic bias in the extension of WP La
Ni~
na events, with 28 out of 34 models having cold
anomalies that extend farther west than in the obser-
vations. The bias in intensity and spatial structure of cold
events is reflected in the multimodel mean with mini-
mum SST anomaly of 21.48C at 1228W for CT La Ni~
na
(Figs. 2g,k) and 21.38C at 1538W for WP La Ni~
na (Figs.
2h,l) compared with 21.18C at 1328W(Figs. 2c,k) and
21.58C at 1528W(Figs. 2d,l) for observed CT La Ni~
na
and WP La Ni~
na, respectively.
The pattern of SST changes during ENSO events is
intimately tied to the changes in surface wind stress. As
a result we examine the composite DJF wind stress
anomalies for the different categories of ENSO (Fig. 4).
The maximum westerly (easterly) wind stress anomalies in
the central South Pacific during CT El Ni~
no (La Ni~
na)
events are reproduced in the multimodel mean (Figs. 4a,e,i
and 4c,g,k, respectively). Notable biases are, however,
apparent. For instance, the CT ENSO in the models is
associated with overall weaker than observed zonal
wind stress anomalies (Figs. 4a,e,i and 4c,g,k), consistent
with the weaker subsurface temperature and heat con-
tent anomalies in Figs. 5a,c and 6a,c. Large biases are
found during WP ENSO, where zonal wind stress
anomalies extend and peak in the far western Pacific
just to the east of Papua New Guinea with a larger
magnitude compared to reanalysis (Figs. 4b,f and 4d,h).
The maximum wind stress anomalies are located around
1458E (multimodel mean) in the WP ENSO compared to
1808in the reanalysis (Figs. 4j and 4l).
The overestimated zonal wind stress anomalies over
the western Pacific would generate excessive upwelling
(downwelling) during WP El Ni~
no (WP La Ni~
na). This
is consistent with the composites of subsurface temper-
ature anomalies along the equator shown in Fig. 5b
(Fig. 5d), which indicate colder (warmer) subsurface
temperature in the western Pacific at the depth of the
thermocline compared to reanalysis. Concurrently, the
overly strong westerly (easterly) wind anomalies tend
to excessively suppress (enhance) the climatological
equatorial upwelling during WP El Ni~
no (WP La Ni~
na),
FIG. 5. Multimodel-mean ocean temperature anomalies (8C, shaded) averaged across
38S–38N along the Pacific during the December–February season for ENSO events: (a) cold
tongue El Ni~
no, (b) warm pool El Ni~
no, (c) cold tongue La Ni~
na, and (d) warm pool La Ni~
na.
The difference between the multimodel mean and observations are contoured in 0.38C in-
tervals. Red (blue) contours are positive (negative) differences. The multimodel mean is based
on 23 CMIP5 models.
15 APRIL 2014 T A S C H E T T O E T A L . 2871

resulting in stronger warming (cooling) of the mixed
layer over the west-central Pacific (Figs. 5b,d). The dy-
namical effect of the winds associated with WP El Ni~
no
is to shoal the thermocline, which in turn enhances
stratification over the central Pacific, leading to cold
subsurface and warm mixed layer biases (Fig. 6b). The
reverse holds for WP La Ni~
na (Fig. 6d). The over-
estimated wind stress in the western Pacific can also
lead to a strengthened zonal current during WP events.
Although the thermocline in the western Pacific is rel-
atively deep, the zonal temperature gradients are gen-
erally strong, which could generate an efficient zonal
advective feedback to produce excessive warming (or
cooling) in that region during WP El Ni~
nos (La Ni~
nas).
These wind stress biases allow the warm water to
spread from the far west to the eastern equatorial Pacific
during WP El Ni~
no, as shown in the composite of sub-
surface temperature in Fig. 5b. In addition, the oppo-
site pattern is seen for WP La Ni~
na events (Fig. 5d)
when biases in zonal wind stress anomalies intensify
the easterlies too far west in the equatorial Pacific, al-
lowing an unrealistic extension of cold waters in the
equatorial Pacific. The response in the ocean surface is
such that the associated SST anomaly is too intense in
the western Pacific during WP La Ni~
na events (Fig. 2l).
c. Temporal evolution of ENSO
The temporal evolution of SST, wind stress, and heat
content anomalies associated with the four categories of
ENSO are shown in Figs. 7–9for the observations, re-
analysis, and CMIP5 multimodel mean. The evolution
of SST anomalies during ENSO events for individual
models is presented in Fig. 10, where SST anomalies are
averaged across the equatorial Pacific from 58Sto58N
and from 1508Eto808W. In Fig. 10, the evolution pat-
terns are ordered according to the models that exhibit
the highest to lowest correlations with the observed SST
evolution.
In general the CMIP5 models correctly reproduce the
timing of seasonal peaks in SST anomalies for all ENSO
types (Figs. 9 and 10). However, there is a range of be-
havior in terms of duration of ENSO events and tran-
sition from warm to cold or neutral SST conditions
(and vice versa). Here we describe the features associ-
ated with ENSO evolution separately for each type of
event.
FIG. 6. Composite of ocean heat content anomalies averaged across 38S–38N along the Pacific
and accumulated in the top 300 m during the December–February season for ENSO events:
(a) cold tongue El Ni~
no, (b) warm pool El Ni~
no, (c) cold tongue La Ni~
na, and (d) warm pool La
Ni~
na. The light gray curve is the multimodel mean heat content and the dark gray curve rep-
resents the SODA reanalysis. Light gray shade indicates the standard deviation of simulated
composites, an estimate of the spread among CMIP5 models. The curves were smoothed with
an 118longitude point running mean. The multimodel mean is based on 23 CMIP5 models.
2872 JOURNAL OF CLIMATE VOLUME 27

1) COLD TONGUE ELNIN
˜O
Overall, the multimodel mean evolution of CT El Ni~
no
events is well represented compared to observations.
The SST anomalies in the equatorial Pacific become
positive in February–March, peak in December, and
vanish in October of the following year (Fig. 9a). As
previously documented in the literature (e.g., Okumura
and Deser 2010), CT El Ni~
no events are generally fol-
lowed by a La Ni~
na event that starts in the following
FIG. 7. Hovmoeller diagram of the SST (8C, shaded) and wind stress (Pa, vectors) anomalies
averaged between 58S and 58N across the Pacific Ocean during ENSO events. Maximum vector
length is 0.05 Pa. (left) Observations from HadISST dataset and reanalysis from NCEP–
NCAR. (center) Multimodel mean of 20 CMIP5 models. (right) Evolution of zonal wind stress
anomalies (Pa) averaged between 58S and 58N, 1208E and 1108W. The red (blue) line is the
multimodel mean (NCEP–NCAR reanalysis), lines smoothed with an 11-month running mean.
The light red area represents the standard deviation of the multimodel mean as an estimate of
the spread across the models. (a),(e),(i) Cold tongue El Ni~
no; (b),(f),(j) warm pool El Ni~
no;
(c),(g),(k) cold tongue La Ni~
na; and (d),(h),(l) warm pool La Ni~
na.
15 APRIL 2014 T A S C H E T T O E T A L . 2873

year but peaks in December, 2 yr after the maximum
warming (Figs. 7a,9a). CT El Ni~
nos are also generally
preceded by cold anomalies 1 yr earlier. The evolution
of observed CT El Ni~
no events shown here is consistent
with previous results based on the extended recon-
structed SST, version 3 (ERSST.v3) data (Hu et al. 2012,
their Fig. 3). The models exhibit this observed evolution
with a wide range of fidelity. While some of the models
tend to simulate CT El Ni~
no events that are too long
(lasting longer than 2yr; e.g., GFDL-ESM2M, MIROC5,
and MPI-ESM-LR), some exhibit a rapid transition to a
strong La Ni~
na with a seasonal cycle that is more ex-
treme than in the observations [CCSM4, CESM1
(CAM5), CESM1 (FASTCHEM), CESM1 (WACCM),
FIO-ESM, GFDL CM3, GFDL-ESM2M, and MIROC5;
Fig. 10a]. Conversely, some models fail to simulate
the transition from warm to cold events altogether
(ACCESS1.0, IPSL-CM5A-LR, MIROC-ESM, MPI-
ESM-MR, HadGEM2-AO, and INM-CM4.0).
The simulated CT El Ni~
nos are preceded and fol-
lowed by much weaker than observed La Ni~
nas (Figs. 7e
and 7a). That the weak La Ni~
na following a CT El Ni~
no
occurs 1 yr earlier than observed is apparently associ-
ated with the more rapid thermocline adjustment in the
CMIP5 models (Figs. 8a,e), as indicated by the more
rapid transition of the basinwide wind anomalies from
westerly to easterly at the peak of El Ni~
no (Fig. 7i), as
well as the narrower meridional extent of the zonal
winds compared to NCEP reanalysis (Figs. 4a,e). A
narrower meridional extent of ENSO-related zonal wind
anomalies would tend to generate faster off-equatorial
Rossby waves, thus a more rapid phase transition
(Kirtman 1997), which is a bias also seen in CMIP3
models (Capotondi et al. 2006).
2) WARM POOL ELNIN
˜O
The multimodel mean for the WP El Ni~
no evolution
captures the correct initiation and peak of SST anoma-
lies in the central–western equatorial Pacific (Fig. 7f),
with SST anomalies becoming positive around October
and peaking in December of the following year (Fig. 9b).
However, in most of the models, the SST anomalies
associated with the simulated WP El Ni~
no last longer
(4 months longer in the multimodel mean, Fig. 9b) than
observed events (Figs. 7b,f,j). For example, CanESM2
and HadCM3 simulate overly strong and prolonged WP
El Ni~
no events, followed by slightly cold anomalies in
the following year (Fig. 10b). CCSM4, CESM1 (CAM5),
CESM1 (FASTCHEM), GFDL-ESM2M, IPSL-CM5B-LR,
and MIROC5 simulate prolonged warm SST anomalies
in the Pacific, followed by strong cold anomalies 2 yr
after the peak of WP El Ni~
no events (Fig. 10b). A few
models (MIROC-ESM and MPI-ESM-LR) represent
WP El Ni~
no with a much longer duration compared to
observations and do not show a transition to SST
anomalies of opposite sign either before or after the
peak of the event (Fig. 10b).
Overall, the CMIP5 models simulate similar durations
for WP and CT El Ni~
no events. In observations, how-
ever, CT El Ni~
nos tend to last longer than WP El Ni~
nos
(Figs. 7a,b)(Hu et al. 2012). This failure can be related
to biases in the wind stress anomaly field in the western
Pacific. Stronger than observed anomalous zonal wind
FIG. 8. Hovmoeller diagram of the ocean heat content (8C,
shaded) and zonal wind stress (Pa, contours) anomalies averaged
between 38S and 38N across the equatorial Pacific Ocean during
ENSO events: (a),(e) cold tongue El Ni~
no; (b),(f) warm pool El
Ni~
no; (c),(g) cold tongue La Ni~
na; and (d),(h) warm pool La Ni~
na.
Brown (green) contours are westerly (easterly) anomalies, plotted
at 0.003-Pa intervals. Multimodel mean of 16 CMIP5 models. An
11-month window running mean was applied to the data.
2874 JOURNAL OF CLIMATE VOLUME 27

stress is seen in the western Pacific approximately two
seasons before the peak of the WP El Ni~
no and lasts
a couple of months after its mature phase (Figs. 7b,f).
This results in a shallower than observed thermocline in
the west during the austral summer season (Fig. 8b). It is
currently thought that the spatial structure of WP El
Ni~
no does not favor a discharge process of the equatorial
heat content that is efficient enough to trigger a cold
event (Kug et al. 2009), which contrasts with CT El Ni~
no
events. Instead, the decay of WP El Ni~
no is thought to be
driven by zonal advection of mean SST gradients by
anomalous zonal currents. However, the cold condition
following the warm events and the eastward propagating
negative thermocline anomalies in the multimodel mean
(Fig. 8f) suggest that thermocline-related processes in-
fluence the evolution of WP El Ni~
no in CMIP5 models
to some degree. This is further supported by the south-
ward shift of westerly anomalies at the peak of the simu-
lated WP El Ni~
nos (marked by northerly anomalies across
the equator in Fig. 4f). This feature, which is clear in CT
but not WP El Ni~
nos in observations (Figs. 4a,b), is asso-
ciated with heat content discharge (McGregor et al. 2012).
3) LANIN
˜A
The temporal evolution of La Ni~
na events simulated
by CMIP5 is remarkably similar between CT and WP
events (Figs. 9c,d,10c,d). Overall, the CMIP5 models
simulate the peak of La Ni~
na events at the correct season;
however, the timing of the start of CT events is biased
compared to observations (Figs. 9c,10c). The multimodel
mean evolution of CT La Ni~
na shows a negative equa-
torial Pacific SST anomaly starting in April, that is,
7 months later than the observations (Figs. 7g,c,9c),
peaking correctly around January, and reaching neutral
conditions in April of the following year (Fig. 9c).
In observations, CT La Ni~
na events are preceded by
warm SST conditions in the equatorial Pacific 2 yr be-
fore their peak (Figs. 7c,9c, and 10c). CESM1 (CAM5)
is the only model that correctly simulates the timing and
magnitude of this warm event prior to CT La Ni~
na
FIG. 9. Evolution of averaged SST anomalies averaged across the equatorial Pacific (58S–58N, 1508E–808W):
observations (black curve); multimodel mean of historical simulations (blue curve); and the multimodel mean of
RCP8.5 scenario (red curve). Shading indicates the standard deviation of the multimodel mean for the historical
(blue) and RCP8.5 (red) simulations. Composite for (a) cold tongue El Ni~
no, (b) warm pool El Ni~
no, (c) cold
tongue La Ni~
na, and (d) warm pool La Ni~
na events.
15 APRIL 2014 T A S C H E T T O E T A L . 2875

FIG. 10. Evolution of SST anomalies (8C) averaged across the equatorial Pacific (58S–58N, 1508E–808W) for individual models: com-
posite for (a) cold tongue El Ni~
no, (b) warm pool El Ni~
no, (c) cold tongue La Ni~
na, and (d) warm pool La Ni~
na events. Models are ordered
according to correlations with observations.
2876 JOURNAL OF CLIMATE VOLUME 27

(Fig. 10c). In contrast, the CT La Ni~
na events simulated
by most of the CMIP5 models are preceded by a warm-
ing in the previous year (Figs. 7g,8g, and 9c), particu-
larly in CCSM4, CESM1 (FASTCHEM), CESM1
(WACCM), FIO-ESM, GFDL CM3, GFDL-ESM2M,
HadCM3, NorESM1-M, and NorESM1-ME (Fig. 10c).
It is important to note that a comparison of CT La Ni~
nas
between CMIP5 models and observations should be
treated with caution given the small sample of observed
events in the past 50 years.
Most of the models realistically simulate the timing
of the initiation of WP La Ni~
na events, with negative
SST anomalies in the equatorial Pacific starting around
March (Fig. 9d). However, most of the CMIP5 models
do not reproduce the cold SST anomalies that last
throughout the following year as in the observations
(Figs. 7d,h,8d,h), except CESM-CAM5 (Fig. 10d). In-
stead, the simulated multimodel mean WP La Ni~
na
events terminate 6 months earlier (Figs. 7h,8h). The
peak of observed and simulated WP La Ni~
na events
occurs in December and is preceded only by weak warm
anomalies (Figs. 7d,h). Exceptions are GFDL-ESM2M
and MIROC5 that simulate overly strong positive SST
anomalies 2 yr before the peak of the WP cold event in
the central Pacific (Fig. 10d).
The initiation timing of the zonal wind stress anomaly in
the central equatorial Pacific is well captured in the mul-
timodel mean (Fig. 7l). Unrealistically strong wind stress
anomalies appear in the western Pacific (Fig. 7h), favoring
a SST pattern that extends westward along the equator, as
previously discussed. Additionally, the simulated wind
stress anomaly persists throughout the year, resulting in
a rapid thermocline adjustment (Fig. 8h) and an early
termination of WP La Ni~
na events in the CMIP5 models.
d. Seasonality of ENSO
Some of the biases seen in the evolution of warm and
cold events may be related to an incorrect simulation of
the seasonality of ENSO. Figure 11 displays the stan-
dard deviation of Ni~
no indices for each model and ob-
servations. In general, most of the models show good
fidelity in the timing and amplitude of SST variability in
the central equatorial Pacific. Twenty-seven out of 34
CMIP5 models realistically represent the maximum
amplitude of ENSO during November–January in the
Ni~
no-3.4 region (Fig. 11b). Greater disagreement is ev-
ident in the seasonality of ENSO in the eastern and
western part of the tropical Pacific Ocean. For instance,
only 13 out of 34 models capture the correct timing
of maximum variability in the Ni~
no-3 region. The dis-
crepancies in the timing of the events among the models
are reflected in the notably weaker multimodel mean
ENSO seasonality compared to observations.
GFDL-ESM2M, GFDL CM3, and ACCESS1.3 ex-
hibit maximum variability that is 2 months late for the
Ni~
no-3.4 region compared to observations, and the
FIG. 11. Monthly standard deviation of (a) Ni~
no-4, (b) Ni~
no-3.4, and (c) Ni~
no-3 indices for CMIP5 models. For comparison purposes,
the monthly standard deviation is divided by the maximum value:this number is indicated in white in the month when it peaks. Models are
ordered according to correlations with observations.
15 APRIL 2014 T A S C H E T T O E T A L . 2877

BCC-CSM1.1 is 3 months too early. This phase bias is
even more extreme in some models, in particular IPSL-
CM5A-MR, IPSL-CM5A-LR, and CSIRO Mk3.6.0,
where the maximum variability occurs approximately
6 months after the observed peak of ENSO. In addition
to representing ENSO events in the wrong season, the
IPSL-CM5A-MR model has very weak seasonality, which
is also true for FIO-ESM, MPI-ESM-LR, MPI-ESM-MR,
IPSL-CM5B-LR, and CMCC-CM, although the ENSO
indices in these models tend to peak in the correct
season. These results are consistent with Bellenger
et al. (2013).
As shown in this analysis, the substantial spread in the
seasonal peak of warm and cold events compared to ob-
servations suggests that ENSO timing is one of the aspects
requiring improvement in future CMIP simulations.
4. Future projections
Here we analyze how the different types of ENSO
events may change in the future as projected by 27
CMIP5 models that had archived RCP8.5 simulations at
the time of writing. Figure 12 shows the multimodel
mean difference in equatorial Pacific SST anomalies
between the RCP8.5 and historical simulations for each
type of event. For the CT El Ni~
nos, the multimodel
mean shows significant cooling in the eastern South
Pacific and western Pacific but a warming in the eastern
North Pacific (Fig. 12a). The WP El Ni~
nos reveal
a slight warming in the central–west equatorial Pacific
and cooling on both sides of the equator, suggesting
a more confined warming in the future scenario than
over the historical period (Fig. 12b). The CT La Ni~
na
changes exhibit cooling in the west and warming in the
east equatorial Pacific (Fig. 12c). The WP La Ni~
na
pattern suggests strengthening of the cold events in a
warmer scenario (Fig. 12d). However, these future
changes in the amplitude of ENSO events are overall
small and not consistent across the models. Analysis of
the spatial metrics shown in Fig. 3 reveals no clear
change in the multimodel mean magnitude or location
of maximum SST anomaly. The westward extent of
FIG. 12. (a)–(d) Difference in the simulated ENSO SST anomaly (8C) composites during the
December–February season between the RCP8.5 and historical scenarios. Areas within the
thin gray line are statistically significant at the 0.05 significance level based on a Student’s ttest.
(e)–(h) SST anomaly averaged over the equatorial Pacific (58S–58N). The red (blue) line
represents the multimodel mean for RCP8.5 (historical simulation). Light red (blue) shade
indicates the standard deviation of simulated composites for the RCP8.5 (historical), an esti-
mate of the spread among CMIP5 models of each scenario. Based on 27 CMIP5 models.
2878 JOURNAL OF CLIMATE VOLUME 27

ENSO also does not show significant changes in the
future projections, except for CT La Ni~
na events that
extend 158westward on average, which is consistent
with the cooling around the date line shown in Fig. 12c.
The change in the amplitude of SST anomalies is also
quantified in Fig. 13 via the difference between the
standard deviation of Ni~
no indices from the historical
period to the RCP8.5 scenario. There is little agreement
in the projections of Ni~
no indices across the models,
indicating that the changes derived for the multimodel
mean (Fig. 12) are not statistically significant. Our
analysis based on 27 CMIP5 models does not reveal any
enhancement of WP to CT ENSO intensity from
historical to RCP8.5 scenario (Fig. 13d). This contradicts
the findings of Kim and Yu (2012), who reported
increased WP to CT intensity ratio from historical to
RCP4.5 scenario using a smaller set of CMIP5 models. In
particular, the Ni~
no-4 to Ni~
no-3 ratio averaged for 16 out
of 20 models in common with Kim and Yu (2012) exhibits
an 1.3% increase from the historical period to the RCP8.5
simulation; while the ratio averaged for their ‘‘7 models
that best represent CT and WP ENSO’’ (see asterisks in
Fig. 13) reveals a decrease of 0.9%; however, both
numbers are nonsignificant. It is also important to note
that Kim and Yu (2012) analyze RCP4.5 simulations
while we have assessed the RCP8.5 scenario.
FIG. 13. Difference in the standard deviation of (a) Ni~
no-3, (b) Ni~
no-3.4, and (c) Ni~
no-4
indices between RCP8.5 and historical simulations for 27 CMIP5 models. (d) Difference in the
ratio of the standard deviation between Ni~
no-4 and Ni~
no-3. The gray dashed line represents the
difference in the multimodel mean; zero appears as the black dashed line. Vertical bars rep-
resent the range of ensemble members when available and circles the respective ensemble
mean.
15 APRIL 2014 T A S C H E T T O E T A L . 2879

When individual models are considered, the WP to CT
ENSO asymmetry in regard to intensity can show signifi-
cant changes: for instance, there is a robust increase in the
Ni~
no-4/Ni~
no-3 ratio for all 10 members of the CSIRO
Mk3.6.0 model, 5 members of the CanESM2 model, and 3
members of the MIROC5 model. Our results support the
findings of Stevenson (2012), who found no robust change
in the multimodel mean SST difference between the Ni~
no-
4andNi
~
no-3.4 regions from twentieth century to RCP4.5,
except when individual models are examined: that is, 4 of
the 11 CMIP5 models containing more than three mem-
bers reveal statistically significant changes. This shows
the importance of considering large ensembles when ex-
amining the robustness of ENSO projections.
An evaluation of the frequency of ENSO events from
historical to RCP8.5 scenarios shows no significant result
(Fig. 14). The evolution of ENSO events also exhibits
little change in the future. On average, the timing of the
initiation, peak, and termination of the Pacific events
show similar behavior in the RCP8.5 scenario compared
to the historical period (Fig. 9b, red and blue curves).
5. Discussion and conclusions
Now as in the past there remain substantial problems
in the realistic simulation of ENSO in climate models,
despite good progress over the past decade. Of partic-
ular importance for ENSO teleconnections is the correct
FIG. 14. Difference in the number of events (100 yr)
21
between RCP8.5 and historical sim-
ulations for 27 CMIP5 models: (a) CT El Ni~
no, (b) WP El Ni~
no, (c) CT La Ni~
na, and (d) WP La
Ni~
na. The gray dashed line represents the difference in the multimodel mean; zero appears as
the black dashed line. Vertical bars represent the range of ensemble members when available
and circles the respective ensemble mean.
2880 JOURNAL OF CLIMATE VOLUME 27

simulation of the characteristics of different ENSO fla-
vors, classified here into cold tongue and warm pool
El Ni~
nos and La Ni~
nas. This study assesses ENSO in 34
CMIP5 models and finds that, while most models do
simulate events that can be classed as either CT or WP,
there is varying fidelity across the models.
Similar to observations, the CT and WP El Ni~
nos are
easily distinguishable in most CMIP5 models; however,
the two types of La Ni~
na are much less so. The scatter
diagram of Ni~
no indices in Fig. 15 illustrates this re-
lationship in the CMIP5 models: the larger the linearity
between Ni~
no-3 and Ni~
no-4 indices, the less indepen-
dent the events. This result corroborates the findings of
Kug and Ham (2011), who reported that CMIP3 models
simulate more distinct types of El Ni~
no than La Ni~
na.
However, when looking at individual models, over one-
third of CMIP5 models represent the two types of La
Ni~
na events (see models with similar numbers of CT and
WP La Ni~
nas in Fig. 1); for this reason we have assessed
La Ni~
nas separately in this study.
The CMIP5 models can simulate the intensity and
location of maximum SST anomalies during ENSO
events within the observational bounds. This result
is consistent with the findings by Kim and Yu (2012),
who reported a good representation of WP ENSO in-
tensity, with relatively more biases for CT ENSOs. Our
assessment of 34 CMIP5 models indicates that, while the
intensity of the four ENSO types is in general re-
alistically represented, the spatial pattern of warm and
cold events (particularly WP events) extend farther west
in the simulations compared to observations (Table 3).
The observed asymmetries in the intensity between
warm and cold events (i.e., El Ni~
nos stronger than La
Ni~
nas) and between warm events (i.e., CT stronger than
WP El Ni~
nos) are captured in most of the CMIP5
models analyzed here. However, most of the models fail
to reproduce the observed asymmetry between the cold
events: that is, simulated CT La Ni~
nas are stronger than
WP La Ni~
nas (note the limitation due to the small ob-
served number of CT La Ni~
nas).
Most CMIP5 models can simulate an evolution of CT
El Ni~
no events that is similar to that observed, with cor-
rect time of initiation, duration, and peak in December.
The simulated CT El Ni~
nos are often followed by cold
events one year after the peak of the warm event in
most of the models, while cold events more commonly
occur two years after in observations. The duration of
WP El Ni~
nos is overestimated for most of the models,
a bias related to the simulated wind stress anomalies in
the central to western equatorial Pacific being too strong
and persistent. In general, the evolution of cold events
also exhibits biases, with the simulated CT La Ni~
nas
starting about two seasons later than observed and WP
La Ni~
nas ending approximately six months earlier than
observed. It is important to note, however, that a fair
amount of variability exists in the life cycle of individual
events: thatis, not all El Ni~
nos and La Ni~
nas exhibit equal
duration. Thus, one should be cautious given the small
sample size of observed events (especially for CT La
Ni~
nas) and the different duration of individual events.
The seasonality of ENSO shows varying degrees of
fidelity depending on the Ni~
no region. Better agreement
in the timing of ENSO peak among CMIP5 models is
seen in the Ni~
no-3.4 region (27 of 34 models peak in the
correct season) while a large spread occurs in the Ni~
no-3
region (only ;1
/
3of the models peak in the correct sea-
son). Even in models where the peak of the Pacific SST
variability is simulated in December, good skill in
simulating other aspects of ENSO seasonality is not
guaranteed. Several models show an overly weak sea-
sonality, suggesting that many ENSO events are also
occurring at the wrong time of year. Particularly in the
Ni~
no-3 and Ni~
no-4 regions, ENSO events in many
CMIP5 models peak in the wrong seasons.
The seasonality of ENSO is an important feature that
determines the timing of the evolution of warm and cold
events. Lengaine and Vecchi (2010) showed that the
seasonality has been linked to the termination of strong
El Ni~
no events in CMIP3 models. This also seems to be
the case for CMIP5 models; that is, the larger the ENSO
seasonality, the better the timing of the termination,
particularly for CT and WP warm events (not shown). The
substantial spread in the seasonal peak and termination
timing of ENSO events in CMIP5 models compared to
FIG. 15. Scatterplot of averaged DJF Ni~
no-4 3Ni~
no-3 indices for
each model and each ENSO category: CT El Ni~
nos (red dots), WP
El Ni~
nos (yellow dots), CT La Ni~
nas (blue dots), and WP La Ni~
nas
(green dots); squares represent HadISST data and diamonds are
the multimodel mean. Dashed lines represent the spread across
CMIP5 models and observed events.
15 APRIL 2014 T A S C H E T T O E T A L . 2881

observations suggests that ENSO seasonality is still an
aspect that needs to be improved in models.
Most of the biases in the ENSO SST anomalies can be
linked to biases in the wind stress anomalies, which are
likely in turn related to mean state biases in the SST. For
all ENSO flavors the wind stress extends too far west-
ward, particularly during WP events. These biases in the
wind stress anomalies generate spurious thermocline
anomalies that propagate eastward as upwelling and
downwelling Kelvin waves, which can in turn influence
the evolution of ENSO events. It is noted here that
spurious thermocline anomalies over the central Pacific
may influence the evolution of warm pool events in the
models. Furthermore, the narrower meridional extent
of the CT ENSO wind stress anomalies seen in the
multimodel mean likely contributes to the more rapid
and somewhat more regular ENSO phase transition.
This systematic bias was also seen in CMIP3 models
(Capotondi et al. 2006), which is likely to be associated
with the classical cold tongue problem, requiring im-
provements in the representation of the physics of the
coupled climate system (e.g., Guilyardi et al. 2009a).
Lengaigne and Vecchi (2010) found that only four CMIP3
models are able to represent the observed termination of
moderate and extreme El Ni~
no events, with the westward
and eastward propagation of the SST anomalies, re-
spectively. In this study, all of the analyzed CMIP5 models
exhibit an eastward propagation of SST anomalies in the
termination phase of WP El Ni~
nos (not shown). Although
we have not explicitly separated each type of event into
strong or weak, our ENSO classification into WP and CT
El Ni~
nos also acts to differentiate events by strength in the
observations. In the CMIP5 models, however, the WP El
Ni~
nos are generally stronger than observed, and the dis-
sipation of the anomalies occurs through similar processes
in both events.
Ham and Kug (2012) and Kug et al. (2012) showed
that the representation of ENSO into CT and WP events
in climate models is sensitive to the atmospheric re-
sponse, in particular to the location of convection, to the
underlying SST anomaly patterns. Bellenger et al. (2013)
also demonstrated that convection parameterization in
climate models can strongly affect the seasonal phase
locking of ENSO. Watanabe et al. (2012) reported that
a wetter mean state with higher precipitation in the east-
ern Pacific leads to a larger ENSO amplitude as a result of
a stronger positive coupled feedback. In our analysis, the
biases in wind stress anomalies (also related to pre-
cipitation) may act to enhance the coupling between the
atmosphere and the SST in the CMIP5 models, resulting
in a larger than observed WP El Ni~
no amplitude in
western Pacific. As a consequence, the dissipation of the
WP events seems to occur via ocean heat discharge, in
a similar way to CT events, rather than via advection of
mean SST gradients by anomalous zonal currents as seen
in observations.
Recent studies have also attributed biases in ENSO
simulations to remote influences, particularly climate
over the Indian Ocean basin (Du et al. 2013;Santoso
et al. 2012, and references therein). Okumura and Deser
(2010) suggest that remote forcing from the Indian
Ocean can influence the asymmetry in the duration of
El Ni~
no and La Ni~
na. Thus, some of the modeled ENSO
biases reported here could relate to how CMIP5 models
simulate Indian Ocean climate and its variability.
The different types of ENSO do not show robust
changes in the spatial pattern, even when subject to
large changes in radiative forcing. In addition, no con-
sistent changes are seen in the WP–CT ratio or in the
frequency of ENSO events. This result is consistent with
previous studies based on CMIP3 that concluded there is
little agreement among the models for projected ENSO
changes (e.g., van Oldenborgh et al. 2005;Guilyardi
2006;Collins et al. 2010;Stevenson 2012) and is also
consistent with the idea that changes in ENSO ampli-
tude and frequency can be hard to detect given the level
of natural variability present in the climate system (e.g.,
Wittenberg 2009;Aiken et al. 2013).
In summary, our study suggests that CMIP5 models
can simulate the two types of ENSO with varying de-
grees of fidelity. The features that models represent well
include (i) stronger El Ni~
nos than La Ni~
nas; (ii) stronger
CT El Ni~
nos than WP El Ni~
nos; (iii) the location of
maximum SST anomaly for all events; (iv) the magni-
tude of CT and WP events; (v) the ENSO peak around
December; and (vi) the time evolution of CT El Ni~
no
events. On the other hand, the majority of models
(i) cannot simulate the asymmetry between cold events
(stronger CT than WP La Ni~
nas); (ii) overestimate wind
stress in the western Pacific; (iii) simulate SST anomalies
extending too far west in the equatorial Pacific; (iv) agree
poorly on the ENSO seasonal evolution; and (v) over-
estimate the termination duration of WP El Ni~
no and
underestimate for WP La Ni~
na. Finally, there are no ro-
bust changes in the future projections of the magnitude or
location of maximum SST anomalies, nor the frequency
of ENSO events. This study motivates further analyses to
understand the disagreement among models and pro-
jections, via assessments of climate simulations remote
from the tropical Pacific, as well as understanding the
feedback mechanisms operating in the Pacific region in
CMIP5 models.
Acknowledgments. We acknowledge the WCRP’s
Working Group on Coupled Modelling, which is re-
sponsible for CMIP, and we thank the climate modeling
2882 JOURNAL OF CLIMATE VOLUME 27

groups for producing and making available their model
output. The U.S. DOE’s PCMDI provided coordinating
support and led development of software infrastructure
in partnership with the Global Organization for Earth
System Science Portals. This work was also supported
by the NCI National FacilityattheANUviathepro-
vision of computing resources to the ARC Centre of
Excellence for Climate System Science. We acknowledge
all the Institutions responsible for the observations and
reanalysis products for having made their data available.
This project was supported by the Australian Research
Council.
REFERENCES
Aiken, C. M., A. Santoso, S. McGregor, and M. H. England, 2013:
The 1970’s shift in ENSO dynamics: A linear inverse model
perspective. Geophys. Res. Lett., 40, 1612–1617, doi:10.1002/
grl.50264.
An,S.-I.,andF.-F.Jin,2004:NonlinearityandasymmetryofENSO.
J. Climate, 17, 2399–2412, doi:10.1175/1520-0442(2004)017,2399:
NAAOE.2.0.CO;2.
Ashok, K., S. K. Behera, S. A. Rao, H. Weng, and T. Yamagata,
2007: El Ni~
no Modoki and its possible teleconnection. J. Geo-
phys. Res., 112, C11007, doi:10.1029/2006JC003798.
Battisti, D. S., and A. C. Hirst, 1989: Interannual variability in the
tropical atmosphere–ocean model: Influence of the basic state,
ocean geometry, and nonlinearity.J. Atmos. Sci., 46, 1687–1712,
doi:10.1175/1520-0469(1989)046,1687:IVIATA.2.0.CO;2.
Bellenger, H., E. Guilyardi, J. Leloup, M. Lengaigne, and J. Vialard,
2013: ENSO representation in climate models: From CMIP3
to CMIP5. Climate Dyn., doi:10.1007/s00382-013-1783-z.
Burgers, G., and D. B. Stephenson, 1999: The ‘‘normality’’ of El Ni~
no.
Geophys. Res. Lett., 26, 1027–1030, doi:10.1029/1999GL900161.
Capotondi, A., A. Wittenberg, and S. Masina, 2006: Spatial and
temporal structure of tropical Pacific interannual variability in
20th century coupled simulations. Ocean Modell., 15, 274–298,
doi:10.1016/j.ocemod.2006.02.004.
Carton, J. A., and B. S. Giese, 2008: A reanalysis of ocean climate
using Simple Ocean Data Assimilation (SODA). Mon. Wea.
Rev., 136, 2999–3017, doi:10.1175/2007MWR1978.1.
Choi, J., S.-I. An, J.-S. Kug, and S.-W. Yeh, 2011: The role of mean
state on changes in El Ni~
no’s flavor. Climate Dyn., 37, 1205–
1215, doi:10.1007/s00382-010-0912-1.
——, ——, and S.-W. Yeh, 2012: Decadal amplitude modulation of
two types of ENSO and its relationship with the mean state.
Climate Dyn., 38, 2631–2644, doi:10.1007/s00382-011-1186-y.
Collins, M., and Coauthors, 2010: The impact of global warming on
the tropical Pacific Ocean and El Ni~
no. Nat. Geosci., 3, 391–
397, doi:10.1038/ngeo868.
Dewitte, B., J. Choi, S.-I. An, and S. Thual, 2012: Vertical structure
variability and equatorial waves during central Pacific and
eastern Pacific El Ni~
nos in a coupled general circulation
model. Climate Dyn., 38, 2275–2289.
Dommenget, D., T. Bayr, and C. Frauen, 2013: Analysis of the
non-linearity in the pattern and time evolution of El Ni~
no
Southern Oscillation. Climate Dyn., 40, 2825–2847, doi:10.1007/
s00382-012-1475-0.
Du, Y., S.-P. Xie, Y.-L. Yang, X.-T. Zheng, L. Liu, and G. Huang,
2013: Indian Ocean variability in the CMIP5 multimodel
ensemble: The basin mode. J. Climate, 26, 7240–7266,
doi:10.1175/JCLI-D-12-00678.1.
Frauen, C., and D. Dommenget, 2010: El Ni~
no and La Ni~
na am-
plitude asymmetry caused by atmospheric feedbacks. Geo-
phys. Res. Lett., 37, L18801, doi:10.1029/2010GL044444.
Guilyardi, E., 2006: El Ni~
no–mean state–seasonal cycle interac-
tions in a multi-model ensemble. Climate Dyn., 26, 329–348,
doi:10.1007/s00382-005-0084-6.
——, P. Braconnot, F.-F. Jin, S. T. Kim, M. Kolasinski, T. Li, and
I. Musat, 2009a: Atmosphere feedbacks during ENSO in a
coupled GCM with a modified atmospheric convection scheme.
J. Climate, 22, 5698–5718, doi:10.1175/2009JCLI2815.1.
——, A. Wittenberg, A. Fedorov, M. Collins, C. Wang, A. Capotondi,
G. J. van Oldenborgh, and T. Stockdale, 2009b: Un-
derstanding El Ni~
no in ocean–atmosphere general circulation
models. Bull. Amer. Meteor. Soc., 90, 325–340, doi:10.1175/
2008BAMS2387.1.
——, W. Cai, M. Collins, A. Fedorov, F.-F. Jin, A. Kumar, D.-Z.
Sun, and A. Wittenberg, 2012: New strategies for evaluating
ENSO processes in climate models. Bull. Amer. Meteor. Soc.,
93, 235–238, doi:10.1175/BAMS-D-11-00106.1.
Ham, Y.-G., and J.-S. Kug, 2012: How well do current climate
models simulate two types of El Ni~
no? Climate Dyn., 39, 383–
398, doi:10.1007/s00382-011-1157-3.
Hoerling, M. P., A. Kumar, and M. Zhong, 1997: El Ni~
no, La
Ni~
na, and the nonlinearity of their teleconnections. J. Cli-
mate, 10, 1769–1786, doi:10.1175/1520-0442(1997)010,1769:
ENOLNA.2.0.CO;2.
Hu, Z.-Z., A. Kumar, B. Jha, W. Wang, B. Huang, and B. Huang,
2012: An analysis of warm pool and cold tongue El Ni~
nos: Air–
sea coupling processes, global influences, and recent trends.
Climate Dyn., 38, 2017–2035, doi:10.1007/s00382-011-1224-9.
Jin, F.-F., 1997: An equatorial ocean recharge paradigm for ENSO.
Part I: Conceptual model. J. Atmos. Sci., 54, 811–829, doi:10.1175/
1520-0469(1997)054,0811:AEORPF.2.0.CO;2.
Johnson, N. C., 2013: How many ENSO flavors can we distinguish?
J. Climate, 26, 4816–4827, doi:10.1175/JCLI-D-12-00649.1.
Jourdain, N. C., A. Sen Gupta, A. S. Taschetto, C. C. Ummenhofer,
A. F. Moise, and K. Ashok, 2013: The Indo-Australian
monsoon and its relationship to ENSO and IOD in re-
analysis data and the CMIP3/CMIP5 simulations. Climate
Dyn., 41, 3073–3102.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re-
analysis Project. Bull. Amer. Meteor. Soc., 77, 437–472,
doi:10.1175/1520-0477(1996)077,0437:TNYRP.2.0.CO;2.
Kang, I.-S., and J.-S. Kug, 2002: El Ni~
no and La Ni~
na sea surface
temperature anomalies: Asymmetry characteristics associated
with their wind stress anomalies. J. Geophys. Res., 107, 4372,
doi:10.1029/2001JD000393.
Kim, J., K. Kim, and S. Yeh, 2012: Statistical evidence for the
natural variation of the central Pacific El Ni~
no. J. Geophys.
Res., 117, C06014, doi:10.1029/2012JC008003.
Kim, S. T., and F. Jin, 2011: An ENSO stability analysis. Part II:
Results from the twentieth and twenty-first century simula-
tions of the CMIP3 models. Climate Dyn., 36, 1609–1627,
doi:10.1007/s00382-010-0872-5.
——, and J.-Y. Yu, 2012: The two types of ENSO inCMIP5 models.
Geophys. Res. Lett., 39, L11704, doi:10.1029/2012GL052006.
Kirtman, B. P., 1997: Oceanic Rossby wave dynamics and the
ENSO period in a coupled model. J. Climate, 10, 1690–1704,
doi:10.1175/1520-0442(1997)010,1690:ORWDAT.2.0.CO;2.
Kug, J.-S., and Y.-G. Ham, 2011: Are there two types of La Ni~
na?
Geophys. Res. Lett., 38, L16704, doi:10.1029/2011GL048237.
15 APRIL 2014 T A S C H E T T O E T A L . 2883

——, F.-F. Jin, and S.-I. An, 2009: Two types of El Ni~
no events:
Cold tongue El Ni~
no and warm pool El Ni~
no. J. Climate, 22,
1499–1515, doi:10.1175/2008JCLI2624.1.
——, Y.-G. Ham, J.-Y. Lee, and F.-F. Jin, 2012: Improved simu-
lation of two types of El Ni~
no in CMIP5 models. Environ. Res.
Lett., 7, 034002, doi:10.1088/1748-9326/7/3/034002.
Larkin, N. K., and D. E. Harrison, 2002: ENSO warm (El Ni~
no)
and cold (La Ni~
na) event life cycles: Ocean surface anomaly
patterns, their symmetries, asymmetries, and implications.
J. Climate, 15, 1118–1140, doi:10.1175/1520-0442(2002)015,1118:
EWENOA.2.0.CO;2.
Lee, T., and M. J. McPhaden, 2010: Increasing intensity of El Ni~
no
in the central-equatorial Pacific. Geophys. Res. Lett., 37,
L14603, doi:10.1029/2010GL044007.
Leloup, J., M. Lengaigne, and J.-P. Boulanger, 2008: Twentieth
century ENSO characteristics in the IPCC database. Climate
Dyn., 30, 277–291, doi:10.1007/s00382-007-0284-3.
Lengaigne, M., and G. Vecchi, 2010: Contrasting the termination of
moderate and extreme El Ni~
no events in coupled general
circulation models. Climate Dyn., 35, 299–313, doi:10.1007/
s00382-009-0562-3.
Lian, T., and D. Chen, 2012: An evaluation of rotated EOF analysis
and its application to tropical Pacific SST variability. J. Cli-
mate, 25, 5361–5373, doi:10.1175/JCLI-D-11-00663.1.
McGregor, S., A. Timmermann, S. Schneider, M. F. Stuecker, and
M. H. England, 2012: The effect of the South Pacific conver-
gence zone on the termination of El Ni~
no events and the
meridional asymmetry of ENSO. J. Climate, 25, 5566–5586,
doi:10.1175/JCLI-D-11-00332.1.
——, N. Ramesh, P. Spence, M. H. England, M. J. McPhaden, and
A. Santoso, 2013: Meridional movement of wind anomalies
during ENSO events and their role in event termination.
Geophys. Res. Lett., 40, 749–754, doi:10.1002/grl.50136.
McPhaden, M. J., and X. Zhang, 2009: Asymmetry in zonal phase
propagation of ENSO sea surface temperature anomalies.
Geophys. Res. Lett., 36, L13703, doi:10.1029/2009GL038774.
——, and Coauthors, 1998: The Tropical Ocean-Global Atmo-
sphere observing system: A decade of progress. J. Geophys.
Res., 103 (C7), 14 169–14 240, doi:10.1029/97JC02906.
——, T. Lee, and D. McClurg, 2011: El Ni~
no and its relationship to
changing background conditions in the tropical Pacific Ocean.
Geophys. Res. Lett., 38, L15709, doi:10.1029/2011GL048275.
Meinen, C. S., and M. J. McPhaden, 2000: Observations
of warm water volume changes in the equatorial Pacific
and their relationship to El Ni~
no and La Ni~
na. J. Climate,
13, 3551–3559, doi:10.1175/1520-0442(2000)013,3551:
OOWWVC.2.0.CO;2.
Na, H., B.-G. Jang, W.-M. Choi, and K.-Y. Kim, 2011: Statistical
simulations of the future 50-year statistics of cold-tongue El
Ni~
no and warm-pool El Ni~
no. Asia-Pac. J. Atmos. Sci., 47,
223–233, doi:10.1007/s13143-011-0011-1.
Newman, M., S.-I. Shin, and M. A. Alexander, 2011: Natural var-
iation in ENSO flavors. Geophys. Res. Lett., 38, L14705,
doi:10.1029/2011GL047658.
Ohba, M., and H. Ueda, 2009: Role of nonlinear atmospheric re-
sponse to SST on the asymmetric transition process of ENSO.
J. Climate, 22, 177–192, doi:10.1175/2008JCLI2334.1.
——, D. Nohara, and H. Ueda, 2010: Simulation of asymmetric
ENSO transition in WCRP CMIP3 multimodel experiments.
J. Climate, 23, 6051–6067, doi:10.1175/2010JCLI3608.1.
Okumura, Y. M., and C. Deser, 2010: Asymmetry in the duration of
El Ni~
no and La Ni~
na. J. Climate, 23, 5826–5843, doi:10.1175/
2010JCLI3592.1.
——, M. Ohba, C. Deser, and H. Ueda, 2011: A proposed mech-
anism for the asymmetric duration of El Ni~
no and La Ni~
na.
J. Climate, 24, 3822–3829, doi:10.1175/2011JCLI3999.1.
Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V.
Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan, 2003:
Global analyses of sea surface temperature, sea ice, and night
marine air temperature since the late nineteenth century.
J. Geophys. Res., 108, 4407, doi:10.1029/2002JD002670.
Rodgers, K. B., P. Friederichs, and M. Latif, 2004: Tropical Pacific
decadal variability and its relation to decadal modula-
tions of ENSO. J. Climate, 17, 3761–3774, doi:10.1175/
1520-0442(2004)017,3761:TPDVAI.2.0.CO;2.
Santoso, A., M. H. England, and W. Cai, 2012: Impact of Indo-
Pacific feedback interactions on ENSO dynamics diagnosed
using ensemble climate simulations. J. Climate, 25, 7743–7763.
Schopf, P. S., and R. J. Burgman, 2006: A simple mechanism for
ENSO residuals and asymmetry. J. Climate, 19, 3167–3179,
doi:10.1175/JCLI3765.1.
Stevenson, S. L., 2012: Significant changes to ENSO strength and
impacts in the twenty-first century: Results from CMIP5.
Geophys. Res. Lett., 39, L17703, doi:10.1029/2012GL052759.
Su, J., R. Zhang, T. Li, X. Rong, J.-S. Kug, and C.-C. Hong, 2010:
Causes of the El Ni~
no and La Ni~
na amplitude asymmetry in
the equatorial eastern Pacific. J. Climate, 23, 605–617,
doi:10.1175/2009JCLI2894.1.
Suarez, M. J., and P. S. Schopf, 1988: A delayed action oscil-
lator for ENSO. J. Atmos. Sci., 45, 3283–3287, doi:10.1175/
1520-0469(1988)045,3283:ADAOFE.2.0.CO;2.
Sun, F., and J.-Y. Yu, 2009: A 10–15-Yr modulation cycle of ENSO
intensity. J. Climate, 22, 1718–1735, doi:10.1175/2008JCLI2285.1.
Takahashi, K., A. Montecinos, K. Goubanova, and B. Dewitte,
2011: ENSO regimes: Reinterpreting the canonical and
Modoki El Ni~
no. Geophys. Res. Lett., 38, L10704, doi:10.1029/
2011GL047364.
Taschetto, A. S., and M. H. England, 2009: El Ni~
no Modoki im-
pacts on Australian rainfall. J. Climate, 22, 3167–3174,
doi:10.1175/2008JCLI2589.1.
——, C. C. Ummenhofer, A. Sen Gupta, and M. H. England, 2009:
Effect of anomalous warming in the central Pacific on the
Australian monsoon. Geophys. Res. Lett., 36, L12704,
doi:10.1029/2009GL038416.
——, R. J. Haarsma, A. Sen Gupta, C. C. Ummenhofer, K. J. Hill,
and M. H. England, 2010: Australian monsoon variability driven
by a Gill–Matsuno-type response to central west Pacific warm-
ing. J. Climate, 23, 4717–4736, doi:10.1175/2010JCLI3474.1.
Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An Overview
of CMIP5 and the experiment design. Bull. Amer. Meteor.
Soc., 93, 485–498, doi:10.1175/BAMS-D-11-00094.1.
Trenberth, K. E., and L. Smith, 2009: Variations in the three-
dimensional structure of the atmospheric circulation with
different flavors of El Ni~
no. J. Climate, 22, 2978–2991,
doi:10.1175/2008JCLI2691.1.
van Oldenborgh, G. J., S. Y. Philip, and M. Collins, 2005: El Ni~
no in
a changing climate: A multi-model study. Ocean Sci., 1, 81–95,
doi:10.5194/os-1-81-2005.
Watanabe, M., J.-S. Kug, F.-F. Jin, M. Collins, M. Ohba, and A. T.
Wittenberg, 2012: Uncertainty in the ENSO amplitude change
from the past to the future. Geophys. Res. Lett., 39, L20703,
doi:10.1029/2012GL053305.
Weng, H., K. Ashok, S. K. Behera, S. A. Rao, and T. Yamagata,
2007: Impacts of recent El Ni~
no Modoki on dry/wet conditions
in the Pacific Rim during boreal summer. Climate Dyn., 29,
113–129, doi:10.1007/s00382-007-0234-0.
2884 JOURNAL OF CLIMATE VOLUME 27

Wittenberg, A. T., 2009: Are historical records sufficient to con-
strain ENSO simulations? Geophys. Res. Lett., 36, L12702,
doi:10.1029/2009GL038710.
Yeh, S.-W., and B. P. Kirtman, 2004: Tropical Pacific decadal
variability and ENSO amplitude modulation in a CGCM.
J. Geophys. Res., 109, C11009, doi:10.1029/2004JC002442.
——, J.-S. Kug, B. Dewitte, M. H. Kwon, and B. P. Kirtman, 2009:
El Ni~
no in a changing climate. Nature, 461, 511–514, doi:10.1038/
nature08316.
——, B. P. Kirtman, J.-S. Kug, W. Park, and M. Latif, 2011: Natural
variability of the central Pacific El Ni~
no event on multi-
centennial timescales. Geophys. Res. Lett., 38, L02704,
doi:10.1029/2010GL045886.
——, Y.-G. Ham, and J.-Y. Lee, 2012: Changes in the tropical
Pacific SST trend from CMIP3 to CMIP5 and its implication of
ENSO. J. Climate, 25, 7764–7771, doi:10.1175/JCLI-D-12-00304.1.
Yu, J.-Y., and S. T. Kim, 2010: Identification of central-Pacific and
eastern-Pacific types of ENSO in CMIP3 models. Geophys.
Res. Lett., 37, L15705, doi:10.1029/2010GL044082.
——, and ——, 2011: Reversed spatial asymmetries between
El Ni~
no and La Ni~
na and their linkage to decadal ENSO
modulation in CMIP3 models. J. Climate, 24, 5423–5434,
doi:10.1175/JCLI-D-11-00024.1.
Zebiak, S. E., 1989: Oceanic heat content variability and El
Ni~
no cycles. J. Phys. Oceanogr., 19, 475–486, doi:10.1175/
1520-0485(1989)019,0475:OHCVAE.2.0.CO;2.
15 APRIL 2014 T A S C H E T T O E T A L . 2885