The response of ENSO flavors to mid-Holocene climate: Implications for proxy interpretation

The response of ENSO flavors to

mid-Holocene climate: Implications for proxy

interpretationChristina Karamperidou,

1Pedro N. Di Nezio and Axel Timmermann,

2

Fei-Fei Jin,1, and Kim M. Cobb,

3

Corresponding author: Christina Karamperidou, Department of Atmospheric Sciences, Uni-

versity of Hawai’i at Manoa, Honolulu, HI, USA. ([email protected])

1Department of Atmospheric Sciences,

University of Hawai’i at Manoa, Honolulu,

HI, USA.

2Department of Oceanography, University

of Hawai’i at Manoa, Honolulu, HI, USA.

3School of Earth and Atmospheric

Sciences, Georgia Institute of Technology,

Atlanta, GA, USA.

This article has been accepted for publication and undergone full peer review but has not been throughthe copyediting, typesetting, pagination and proofreading process, which may lead to differences be-tween this version and the Version of Record. Please cite this article as doi: 10.1002/2014PA002742

c©2015 American Geophysical Union. All Rights Reserved.

The response of El Nino/Southern Oscillation (ENSO) to mid-Holocene

boundary conditions remains an open question: paleoclimate proxies and cli-

mate model simulations do not agree in the magnitude of the reduction of

ENSO variability, while recent proxy evidence from fossil corals from the cen-

tral Pacific show that the reduction in mid-Holocene ENSO variability com-

pared to the end of the 20th century is not different from the reduction dur-

ing other Holocene periods. This is inconsistent with the interpretation of

lake and ocean sediment records from the eastern Pacific, which show a sig-

nificant reduction compared to all other Holocene periods. In order to rec-

oncile the seemingly conflicting proxy evidence from the eastern and central

Pacific, we hypothesize that ENSO remained active during the mid-Holocene;

however, there was a change in the spatial pattern of the sea surface tem-

perature anomalies, also known as ENSO flavors. Using NCAR’s Commu-

nity Climate System Model (CCSM4) model forced with mid-Holocene or-

bital conditions, we find that the frequency of occurrence of the strongest

Eastern Pacific (EP) events decreases in the mid-Holocene and their vari-

ance is reduced by 30%, while the frequency of Central Pacific (CP) events

slightly increases and their variances doesn’t change. We also find a shift in

the seasonality of EP events, but not in that of CP events. Lastly, mid-Holocene

EP events develop more slowly and decay faster. The differential response

of ENSO flavors to mid-Holocene forcing is remotely forced by the West Pa-

cific, where a weakening of the trade winds in early boreal spring in the mid-


Holocene initiates an anomalous downwelling annual Kelvin wave, which reaches

the eastern Pacific during the ENSO development season, weakens the upper-

ocean stratification and results in reduced ENSO upwelling feedback. The

simulated reduction in the EP flavor versus the CP flavor in the mid-Holocene

is consistent with proxy evidence: The teleconnection patterns of the two fla-

vors with temperature, precipitation and salinity are distinct, and proxies

from different regions of the Pacific might be recording variability associated

with only one of the flavors, or some combination of their relative effects.


1. Introduction

Neither models nor observations provide a conclusive answer as to whether El

Nino/Southern Oscillation (ENSO) is going to weaken or strengthen in response to green-

house warming [Meehl and Coauthors , 2007; Collins et al., 2010; DiNezio et al., 2012; Cai

et al., 2014], or whether ENSO sea-surface temperature (SST) anomalies will tend to be

located in the central rather than the eastern tropical Pacific [Yeh et al., 2009; Lee and

McPhaden, 2010; McPhaden et al., 2011]. Given that ENSO is the dominant mode of

tropical variability, the lack of model agreement is an important source of uncertainty for

projecting future regional climate change throughout the Pacific basin and beyond [Meehl

and Coauthors , 2007]. Paleoclimate reconstructions offer the possibility of testing the

theories of ENSO response to greenhouse warming, as well as the models used to project

this response. The climate of the mid-Holocene – about 6,000 years before present – has

received much attention owing to proxy interpretations suggesting a significant reduc-

tion in climate variability associated with ENSO [Rodbell et al., 1999; Moy et al., 2002;

Riedinger et al., 2002; Koutavas et al., 2006; Donders et al., 2008; Conroy et al., 2008].

Climate models of various complexity agree in key features of the response of the tropical

Pacific to mid-Holocene orbital forcing [Clement et al., 2000, 2001; Hewitt and Mitchell ,

1998; Bush, 2008; Liu et al., 2000; Kitoh and Murakami , 2002; Otto-Bliesner et al., 2003;

Brown et al., 2006, 2008]. Most coupled global climate models (GCMs) participating

in the Paleoclimate Modeling Intercomparison Projects (PMIP2 and PMIP3) simulate

a reduction of ENSO variability, along with a significant reduction in the strength of

the annual cycle of the eastern equatorial Pacific [Masson-Delmotte et al., 2013]. In


present-day climate ENSO is strongly influenced by the seasonal cycle of the Pacific cold

tongue. Therefore it is reasonable to expect that the orbitally-driven changes in seasonal

climate would alter ENSO, as proposed by previous intermediate-complexity or single

model studies [e.g. Clement et al., 2000; Salau et al., 2012].

Seasonal variations in the climate of the Pacific play a key role in the onset and ter-

mination of ENSO. First, the strength of the annual cycle of the Pacific cold tongue

modulates the strength of the Bjerknes feedback – the positive feedback loop responsible

for the growth of ENSO events [Jin et al., 1994, 1996; Tziperman et al., 1994, 1995; Wang

and Fang , 1996]. Second, the seasonal migration of the South Pacific Convergence Zone

(SPCZ) also plays a role, albeit in the termination of strong El Nino events [Harrison and

Vecchi , 1999; Stein et al., 2011; McGregor et al., 2012; Stuecker et al., 2013; Stein et al.,

2014]. Changes in processes away from the tropical Pacific could also influence ENSO.

For instance, mid-Holocene ENSO could be weakened due to a stronger Asian monsoon

through an alteration of the seasonal cycle of the tropical Pacific [Chang et al., 1994;

Liu, 2002; Pan et al., 2005; Timmermann et al., 2007], or through a strengthening of the

tropical Pacific trade winds [Liu et al., 2000; Brown et al., 2008; Marzin and Braconnot ,

2009]. An extratropical mechanism has also been proposed, in which reduced stochastic

forcing originating from the North Pacific weakens ENSO [Chiang et al., 2009].

There are, however, key inconsistencies between simulations and paleoclimate recon-

structions of mid-Holocene ENSO. First, the magnitude of the reduction in ENSO variabil-

ity in the models (approximately 10%) is not as large as suggested by some paleoclimate

data [Masson-Delmotte et al., 2013; Gagan et al., 2004; Donders et al., 2008]. Terrestrial


and marine proxies from the eastern tropical Pacific suggest abrupt changes within the

broadly-defined mid-Holocene period of 4-7 ka BP [Koutavas et al., 2002; Gagan et al.,

2004; Donders et al., 2005, 2007; Chazen et al., 2009; Koutavas and Joanides , 2012], while

the models simulate gradual changes in ENSO variability [Clement et al., 2000, 2001; Don-

ders et al., 2008; Chiang et al., 2009]. At the same time, the presence of noise, the short

length of many of the available proxy records (e.g. corals), and the multiple proxy resolu-

tions (from seasonally- to centennially resolved) create considerable uncertainty regarding

the magnitude of the suggested change, and further complicate model-proxy comparisons

[Wittenberg , 2009; Cobb et al., 2013]. At present, it is difficult to reject the hypothesis

that internal ENSO variability on decadal and centennial time scales dominates over the

forced orbital response during the Holocene [Wolff et al., 2011; Cobb et al., 2013]. In ad-

dition, model–proxy disagreement could be due to changes in the ENSO teleconnections

due to a northward shift of the climatological Intertropical Convergence Zone (ITCZ) in

the mid-Holocene [Woodroffe et al., 2003; Gagan et al., 2004; McGregor and Gagan, 2004;

Koutavas et al., 2006], and not necessarily due to changes in ENSO variability itself. Last,

while Eastern Pacific archives indicate that interannual variability in the mid-Holocene at

that location is weaker than even the case of completely vanished ENSO [Koutavas and

Joanides , 2012], recently-available proxy records from fossil corals from the central Pacific

show that ENSO’s strength during the mid-Holocene was comparable to that during the

last millennium [Cobb et al., 2013].

In this paper, we present a physical mechanism that could help reconcile the seemingly

conflicting proxy evidence of mid-Holocene ENSO from the eastern and the central Pacific


regions. Our hypothesis is motivated by the studies discussed above, which collectively

suggest a weaker ENSO in the eastern Pacific compared to all other Holocene periods

[e.g. Koutavas and Joanides , 2012], but question the presence of a reduction of similar

magnitude in the central Pacific, again as compared to other Holocene periods excepting

the post-1970 era [Cobb et al., 2013]. We hypothesize that the proxy data may reflect

a differential response of the spatial pattern of ENSO’s SST anomalies – also known as

ENSO flavors – to orbitally-driven changes in the seasonal cycle. That is, during the

mid-Holocene, ENSO events with SST anomalies concentrated in the eastern Pacific [the

Eastern Pacific (EP) flavor] were weaker and/or less frequent, while the ENSO events

with SST anomalies concentrated in the central Pacific [Central Pacific (CP) flavor] were

mostly insensitive to orbital forcing.

Various mechanisms have been proposed to explain the existence of ENSO flavors,

including the lack of thermocline-surface interactions in the central Pacific – leading to a

preponderance of the zonal advection feedback over the thermocline feedback [Yeh et al.,

2009]– and of a mechanism of transition to cold events [Kug et al., 2009; Kao and Yu,

2009; Yu and Kim, 2010; Newman et al., 2011a, b]; changes in intraseasonal variability

and its coupling with low-frequency atmospheric flow [Kug et al., 2009]; differences in the

location and intensity of westerly winds [Hu et al., 2012]; the timing of the onset of SST

anomalies [Xu and Chan, 2001]; Asian and Australian monsoon forcing [Yu et al., 2009];

and subtropical atmospheric forcing [Yu et al., 2010]. On the other hand, the occurrence

of EP events has been attributed to the nonlinear evolution of ENSO which does not

require CP and EP El Nino events to be different phenomena [Takahashi et al., 2011],


or conversely, to a statistical artifact due to the nonlinearity between El Nino and La

Nina [Monahan and Dai , 2004; L’Heureux et al., 2012], and natural variability [Yeh et al.,

2011; Newman et al., 2011a]. An investigation of the response of ENSO flavors to orbital

forcing could improve our understanding of their origins, especially in light of evidence

about their interaction with the annual cycle [McGregor et al., 2013b].

Should our hypothesis find support, it would not only reconcile proxy evidence from

around the Pacific, since CP ENSO events have distinct teleconnections from the EP

events [Larkin and Harrison, 2005; Hu et al., 2012; Ashok et al., 2007; Kim et al., 2009;

Yeh et al., 2009; di Lorenzo et al., 2010; Mo, 2010; Yu and Kim, 2010; Hoerling and

Kumar , 2002; Wang and Hendon, 2007; Weng et al., 2007], but would also be consistent

with a weaker annual cycle during the mid-Holocene, since EP events do interact with

the annual cycle, whereas CP events do not [McGregor et al., 2013b]. In fact, during

the preparation of the present paper, our hypothesis gained further support by the study

of Carre et al. [2014], who use fossil mollusk shells from Peru and fossil corals from the

Central Pacific to conclude that a predominance of CP events compared to EP events is

possible in the period 6.7 to 7.5 ka BP.

Here, we explore this hypothesis using pre-industrial control and mid-Holocene simu-

lations conducted with the Community Climate System Model V4 (CCSM4), developed

by the National Center for Atmospheric Research (NCAR). We provide a new detailed

representation of mid-Holocene tropical Pacific climate, which is largely consistent with

newly available records of SST variability from paleo-climate proxies. The remainder of

the paper is organized as follows: After a brief description of the model simulations (sec-


tion 2), we describe changes in ENSO flavors in the CCSM4 pre-industrial control and

mid-Holocene simulations (section 3). In section 4, we present orbitally-induced changes

in the seasonal cycle of the tropical Pacific, and in section 5, we study the consequent

changes in the main feedbacks that control ENSO event development and decay. Section

6 discusses the implications of our results for the interpretation of paleoclimate proxy

evidence. Summary and discussion close the paper (section 7).

2. Climate simulations

In this study, we use NCAR’s Climate System Model version 4 (CCSM4). CCSM4 is

a climate model consisting of coupled atmosphere and ocean general circulation models

(GCMs) and comprehensive land and cryosphere models. The reader is referred to Gent

and Coauthors [2011] for specific information about CCSM4. The pre-industrial Control

simulation (hereafter piControl) analyzed here spans 1300 years and includes interactions

between components of the climate system (ocean, atmosphere, cryosphere and land) con-

figured at nominal 1◦ latitude-longitude resolution and forced by constant pre-industrial

(1860) greenhouse gas concentrations. CCSM4 simulates ENSO realistically in the pre-

industrial experiment, including a 3 to 6 yr period, asymmetry between warm and cold

events, events with a range of amplitude and return times, and multi-decadal modulation

of ENSO [Deser and Coauthors , 2012]. As will be shown in section 3, CCSM4 is also

capable of simulating realistic SST patterns associated with ENSO’s EP and CP events

(Fig. 1).

The mid-Holocene simulation branches off year 800 of piControl, spans 500 years, and

was performed with the same version of CCSM4, following the experimental protocol of


the Paleoclimate Modeling Intercomparison Program 3 (PMIP3). The CO2 concentration

is set to 280ppm (pre-industrial values), the eccentricity is 0.018682 compared to 0.016724

in piControl, the obliquity is 24.105 degrees (piControl is 23.446), and the angle between

fall equinox and perihelion is 0.87 degrees compared to 102.04 degrees at piControl, in

order to reflect the effect of changes in the Earth’s orbit on insolation at about 6 ka BP.

The resulting changes in seasonal downwelling shortwave radiation are shown in Fig. A1

of the Appendix.

3. ENSO flavors in the mid-Holocene

3.1. Definition of ENSO flavors

It has long been recognized that at least two degrees of freedom are needed to describe

SST anomalies during the evolution of an ENSO cycle [Timmermann, 1999; Trenberth and

Stepaniak , 2001]. Some events, like the 1997-98 event, have SST anomalies localized in

the eastern Pacific, the so-called EP ENSO. Other events have maximum SST anomalies

located in the central Pacific, called dateline El Nino [Larkin and Harrison, 2005], warm-

pool El Nino [Kug et al., 2009], central Pacific (CP) El Nino [Kao and Yu, 2009], or El

Nino Modoki [Ashok et al., 2007].

Several indices have been used to characterize the two different flavors of ENSO. Ren

and Jin [2011] showed that the typical NINO3 and NINO4 indices are not an orthogonal

coordinate system to capture the two flavors of ENSO, and therefore developed new indices

based on combinations of the standard ones. Takahashi et al. [2011] also propose ENSO

indices that are effectively a rotation of the first (PC1) and second principal components


(PC2) of the tropical Pacific SST anomalies:

E − index =PC1− PC2√

2C − index =

PC1 + PC2√2

. (1)

Figure 1 shows the regression of SST anomalies on the E-index and the C-index in

the CCSM4 piControl simulation and in observations. The indices were computed as per

Takahashi et al. [2011]: The principal components of the SST anomalies in the region

[110◦E − 60◦W, 10◦S − 10◦N ] are based on the full 1300-yr climatology, are standardized

and passed through a 3-point (1-2-1) weighted running average filter. The first two EOF

patterns in CCSM4 explain 73.7 and 6.3% of the variance, respectively (the boreal spring

and boreal fall averages are shown in Fig. A1 of the Appendix). The regression patterns in

CCSM4 are in good agreement with the observed (explaining 66% and 10%, respectively);

note however that the simulated SST anomalies are latitudinally more constrained and

extend further to the west in the model, which is a known problem in GCM simulations

of ENSO anomalies [Bellenger et al., 2014]. Based on the regions of maximum SST

variance in Figure 1, the model’s standard ENSO regions are slightly shifted compared

to observations [also see Capotondi , 2013]. In this paper however, we avoid using these

standard ENSO indices, rather we base all our calculations on the E-index and the C-index

(eq. 1).

This decomposition of ENSO flavors presents two distinctive features: 1) the E-index

captures the strong EP events (see Fig. 1a and c ); and 2) the C-index captures CP

El Nino and all La Nina events (Fig. 1b and d). Figure 2a shows the scatterplot and

bivariate probability density function of the two leading principal components of monthly

SST anomalies (PC1 and PC2) for piControl, averaged over Oct-Apr. The model simulates


the nonlinearity in the relationship between PC1 and PC2 identified by Takahashi et al.

[2011] in observations. Superposed are Oct-Apr averages for events with the E-index

larger than two standard deviations from zero (solid black circles), and Oct-Apr averages

for events with C-index larger than one standard deviation away from zero (gray-filled

circles). Strong El Nino events lie along the E-index axis, while moderate warm El Nino

and La Nina events lie along the C-index axis.

3.2. Response to mid-Holocene orbital forcing

Strong EP events (solid circles) are significantly reduced in number in the mid-Holocene

simulation (Figure 2b), although they do not completely disappear. The Oct-Apr variance

of the E-index reduces from 0.72 in piControl to 0.52 in mid-Holocene (a 30% decrease).

In contrast, the variance of the C-index does not change between the two climates, and

is equal to 1.03. The reduction in EP events is more clearly seen in the difference of

the bivariate pdf’s between the two climates (Figure 2c). The reduction in the pdf mass

at the tails along the E-index axis (blue dashed contours) indicates that the frequency

of occurrence of the largest EP events decreases. Indeed, the average frequency of EP

events decreases from 3 per century in the piControl simulation to 1.8 per century in the

mid-Holocene simulation. On the contrary, the pdf along the C-index does not change

appreciably (Figure 2c), indicating that these types of events do not respond to mid-

Holocene forcing. The average frequency of CP events is 10 per century in piControl and

12 per century in mid-Holocene.

Figure 3 shows the time-longitude plots of SST anomalies for the composite EP and

CP events in piControl and mid-Holocene, as well as their difference (rightmost panel).


Anomalies are computed with respect to each simulation’s climatology. The composites

include 38 piControl EP events, 9 mid-Holocene EP events, 130 piControl CP events,

and 63 mid-Holocene CP events. The events used in the composites are shown in black

(EP) and gray (CP) circles in figure 2. Statistical significance for the difference in SST

(rightmost panel) is assessed as follows: Of the total of 47 EP events in both climates, we

randomly select 38 and 9 (without replacement) to be termed piControl and mid-Holocene

EP events, respectively. The difference time-longitude matrix is computed. We repeat

this process 1000 times, and compute the statistics of the resulting difference matrices. We

define the statistical significance level at the 5th and 95th percentile. The same process is

followed for the CP events. For the EP events, the significant differences lie within [-0.42

0.42] (stippled regions in Fig. 3), and for the CP events they lie within [-0.18 0.18].

In terms of the magnitude of events, EP events have almost the same magnitude – as

measured by the standard deviation of the NINO3 index – at approximately 3 ◦C in both

climates (also seen in Fig, 3a and b). The main difference, in addition to the reduction

in the frequency of EP events in the mid-Holocene, is in the length and timing of the

development and decay phase. The development of EP events in the mid-Holocene is slow

and starts more than 18 months before the peak of the event. The peak is delayed by

approximately two months into Feb-Mar. The decay of EP events in the mid-Holocene is

steeper and the system moves into a cold phase within 4 months after the peak. CP events

(lower panel of Fig. 3) also start developing earlier in the mid-Holocene and are slightly

more intense (by 0.6-0.8 ◦C). There is a westward propagation of SST anomalies in the

CP events, which is also enhanced in the mid-Holocene. We will show below that the shift


in the development and decay of EP events in the mid-Holocene can be attributed to the

a change in the seasonality of the peak of the events which is a consequence of changes in

the annual cycle of the tropical Pacific.

Figure 4a shows the percentage of winter (ONDFJMA) months with EP-event peaks by

month and climate. The identification of peak-months is done as follows: We select the

36-month period centered at DJF of each year in EP status as highlighted by black points

in figures 2a and b. The peak-month is defined as the month within this period in which

the E-index peaks. In piControl, the majority (40%) of EP events peak in December,

with April and February following with 29% and 24%, respectively. In mid-Holocene,

this percentage drops dramatically, with only 7% of events peaking in December, while

the majority peak in February (46%) and April (38%). These results indicate a shift

in the seasonality of EP events towards late winter and early spring (Feb-Apr) in the

mid-Holocene. Figure 4b shows the same calculation for CP events. The most notable

change is the drop in percentage of CP peaks in April (from 21% in piControl to 7% in

mid-Holocene). Increases in the percentage is found for October, November, December

and March, ranging from 2% to 8%. However, the distribution of peak months is not

significantly changing, as in the EP case.

In summary, CCSM4 simulates differential responses of the EP and CP flavors of ENSO

to orbital forcing. The three main features of this response are:

1. The frequency of occurrence of EP events decreases, whereas that of CP events

slightly increases.


2. The peak of EP events shifts by about two months, from December in piControl to

February in the mid-Holocene. The peak of CP events does not shift.

3. The EP events decay faster in the mid-Holocene climate compared with piControl

by approximately two months. CP events terminate faster in the mid-Holocene.

In the following sections we will show that these main features are a consequence of

changes in the seasonality of tropical Pacific SSTs and winds. These changes in the sea-

sonal cycle in turn influence the onset and termination of EP and CP events differentially,

resulting in weaker EP ENSO during the mid-Holocene.

4. Seasonal changes in tropical Pacific climate

In pre-industrial climate, CCSM4 simulates a fully-developed cold tongue during late

late summer/fall (JASO), along with stronger SE and NE trade winds, and a maximum

northward extent of the ITCZ (Fig. 5a colors, vectors, and contours respectively). Consis-

tent with observations, the simulated cold tongue vanishes during late boreal winter/spring

(FMAM) along with a slowing down of the SE trades and a weaker ITCZ (Fig. 5b).

The changes in the orbital parameters during the mid-Holocene result in increased in-

solation in the tropics in JASO (with a peak amplitude change in September compared

to piControl), and decreased insolation during FMAM, as shown in Fig. A1 of the Ap-

pendix. The tropical climate response to these changes in insolation is characterized by

a weakening of the seasonal cycle in the cold tongue region, with warmer SSTs there

during JASO and colder during FMAM (Fig. 5c and d, colors). In contrast, both the

NE and SE trades strengthen during JASO (Fig. 5c, vectors) – the season when they are

seasonally stronger – and they weaken in FMAM (Fig. 5d, vectors) – the season when


they are seasonally weaker. The response of the SE and NE trade winds does not follow

the changes in the cold tongue. Their response could be related to changes in the sub-

tropical highs, which strengthen forced by local and remote changes in diabatic heating

associated with the monsoons [Mantsis et al., 2013a]. The stronger trade winds during

JASO explain the widespread cooling of the tropical Pacific due to increased evaporative

cooling (excepting the cold tongue that warms), offsetting the warming due to increased

insolation (see Fig. A1c of the Appendix). The tropical Pacific also exhibits widespread

cooling during FMAM; however in this case due to decreased insolation (Fig. A1d of the

Appendix).

Ocean dynamics must be invoked to further explain the simulated response of the cold

tongue, which warms during JASO despite of stronger trade winds, and cools during

FMAM despite of the weaker trade winds. Figure 6a shows the time-longitude plot of

the difference in climatological trade wind stress in the tropical Pacific between the two

climates. Anomalous westerly winds in the west Pacific in FMAM result in an anoma-

lous ”annual Kelvin wave”, which is seen in the difference in thermocline depth between

mid-Holocene and piControl (Fig. 6b). This downwelling Kelvin wave (also seen in the

difference of surface ocean current velocity uos in Fig. 6c) is initiated during boreal spring

in the West Pacific, propagates eastward, reaches the Eastern Pacific during boreal fall,

and is responsible for a reduction of the stratification ∂T∂z

in the Eastern Pacific.

The seasonal evolution of the thermal stratification, measured by the difference be-

tween SST and ocean temperature at 50 m depth (Ts − Tsub), shows anomalous eastward

propagation in the mid Holocene associated with the “annual Kelvin wave” (Fig. 6d).


Over the eastern Pacific, CCSM4 simulates increased Ts− Tsub in boreal winter and early

spring during, with a maximum in Jan-Feb. This stratification increase in the eastern

Pacific results in enhanced vertical advection of colder subsurface waters, which acts to

cool the seasonally warmer SSTs (see FMAM in Fig. 5). Conversely, Ts − Tsub decreases

during boreal fall (6d), resulting in decreased vertical advection of colder waters, which

acts to warm the seasonally colder SSTs (see JASO in Fig. 5). These remotely forced

stratification changes manifest as a weakening of the annual cycle of cold tongue SSTs in

the mid-Holocene.

5. Seasonal controls of ENSO-flavor response to mid-Holocene forcing

5.1. The role of stratification in the East Pacific

The main differences in ENSO flavors between the two climates include a slower de-

velopment and faster decay of EP events in the mid-Holocene, as well as a shift in their

peak by approximately two months (section 3, Fig. 3). In this section, we show that

these differences result from changes in the main ENSO feedback mechanisms driven by

the orbitally-driven changes in the seasonal cycle discussed in Section 4.

We performed a heat budget analysis of CCSM4 output to diagnose the physical mech-

anisms involved in the ENSO changes described above. The heat budget consists of the

anomalous heat content tendency Q′t = ρ0cp∫ 0−H

∂T ′

∂tdz, which is related to the tendency of

(i.e. growth and decay) of interannual SST anomalies T ′ (ρ0 is a reference density of sea

water, cp is the specific heat of sea water, H is the thickness of the constant-depth layer

over which the terms are integrated). Q′t is approximately balanced by the anomalous ad-

vection of temperature by ocean currents (Q′adv) and by the net air-sea heat flux (Q′atm).


The dominant contributions to Q′adv during development and decay phases of El Nino

events are the zonal advection feedback Q′za = −ρ0cp∫ 0−H(u′ ∂T

∂x)dz, the thermocline feed-

back Q′tc = −ρ0cp∫ 0−H(w ∂T ′

∂z)dz, and the upwelling feedback Q′uw = −ρ0cp

∫ 0−H(w′ ∂T

∂z)dz,

where T is the climatological monthly-mean temperature. These feedbacks can be quan-

tified via an ENSO heat budget analysis, which is computed following the methodology

of DiNezio and Deser [2014]. Details may be found in the Appendix.

Figure 7 (upper panel) shows the time-longitude composite of total heat content ten-

dency Q′t for EP events in (a) piControl, and (b) mid-Holocene, as well as (c) their

difference. As expected, the development and decay of SST anomalies shown in Fig. 3

closely follows the total heat content tendency. In both climates the upwelling feedback

Q′uw explains most of the Q′t of EP events over the NINO1+2 and NINO3 regions – the

center of action of the EP events (lower panel of Fig. 7). Q′tc explains the remaining Q′t

(approximately 40-60 Wm−2) (not shown). The Q′t change during the mid-Holocene is

also entirely explained by the change in Q′uw (compare Fig. 7c and f). Q′uw and Q′t exhibit

a smaller magnitude in the mid-Holocene consistent with the slower development of the

events (figure 3). Moreover, Q′uw and Q′t become negative more sharply in mid-Holocene

than in piControl, consistent with a faster termination of EP events.

What causes the weakening of the upwelling feedback in the mid-Holocene? The change

in upwelling feedback (∆Q′uw) is predominantly due to the change in the background

stratification ∂∆T∂z

. In other words it is approximated by

∆Q′uw ≈ −ρ0cp

∫ 0

−Hw′∂∆T

∂zdz (2)


where ∆ indicates the difference between mid-Holocene and piControl. Note that we

ignore other contributions to the upwelling feedback that may arise from the convolution

of changes in ENSO and in background climatology. We refer the reader to the Appendix

for a detailed justification of this approximation.

In present-day climate ENSO events develop from April to October (i.e. their tendency

is largest during these months). This is the time of the year when CCSM4 simulates weaker

climatological stratification (∂∆T∂z

) during the mid-Holocene, represented by Ts − Tsub in

Fig. 6d. Therefore the upwelling feedback, which is dominant in the eastern Pacific,

becomes weaker during the growth phase of ENSO (Fig. 6d). Reduced upwelling feedback

is also evident in the same the development phase in the EP composite (Fig. 7f). We

therefore argue that this weakening of the upwelling feedback is the main mechanism

whereby EP events become less frequent or weaker during the mid-Holocene. Conversely,

CP events are insensitive to the seasonal shifts in stratification because they are primarily

governed by the zonal advection feedback (Fig. 8). The peak difference in zonal advection

feedback in the Central Pacific occurs in May-Jun of the year preceding the peak (Fig.

8f), which coincides with the ”annual downwelling Kelvin wave” reaching those longitudes

(Fig. 6c).

5.2. The role of climatological winds in the West Pacific

CCSM4 simulates EP events that decay faster in the mid-Holocene (Fig. 3c and 7c).

One of the mechanisms proposed to explain the termination of strong EP events involves

the shift of ENSO’s westerly wind anomalies off the equator into the Southern Hemisphere

following the seasonal migration of the SPCZ [Harrison and Vecchi , 1999, 2003, 2006;


Lengaigne et al., 2006; McGregor et al., 2012]. With this southward shift, the westerly

wind anomalies are no longer able to force the oceanic equatorial waveguide, which even-

tually leads to the termination of strong EP events.

CCSM4 simulates this mechanism realistically in piControl as seen in the composite of

wind stress (vectors) and wind stress curl (contours) anomalies (Fig. 9a). This shift also

occurs in the mid-Holocene simulation, albeit it happens earlier, as seen by comparing

figures 9a and b. The earlier shift is clearly seen in the difference composite as a dipole

feature (Fig. 9c, indicated by the thick black line connecting the stippled –i.e. statistically

significant– areas). This stronger and earlier southward shift of wind stress curl anomalies

in mid-Holocene would result in an earlier and stronger termination of the EP events, as

seen in Fig. 3c and 7c.

What is the cause of this earlier shift? We associate this feature again with the shift in

peak EP month. EP events primarily peak in JFMA in mid-Holocene. In these months,

the climatological wind stress in the Southern Hemisphere is stronger (Fig. 10b; eastward

difference indicated weaker climatological winds). This is associated with a stronger South

Pacific Convergence Zone (SPCZ), which can be seen in the difference in precipitation

climatology in Fig. 5. An enhancement and southward displacement of the SPCZ in the

mid-Holocene was also found by Mantsis et al. [2013b] in simulations of PMIP2 models. As

discussed in detail in McGregor et al. [2012], stronger climatological SPCZ and subsequent

weaker boundary layer wind speeds are key factors in the seasonal termination of EP events

and are consistent with the stronger and earlier southward shift of wind anomalies in the

mid-Holocene seen in Fig. 9c, and the stronger EP termination seen in Fig. 3c.


Conversely, this mechanism is much weaker for CP events (Fig. 9, compare upper

and lower panels), and is not important for their termination in either climate [also see

McGregor et al., 2013b].

6. Implications for interpreting paleo-ENSO proxies

Paleoclimate records in the eastern Pacific show a reduction of mid-Holocene ENSO

variance compared to the late 20th century [see Donders et al., 2008, for a comprehensive

review]. However, newly available coral records from the central Pacific exhibit variability

that is not significantly different from other periods of the Holocene that lack significant

orbital forcing [Cobb et al., 2013]. These findings shed some doubt on the hypothesis that

ENSO responds to orbital forcing, and leave open the possibility of a highly naturally

variable ENSO throughout the Holocene. In the previous sections, we provided modeling

support for an alternate hypothesis, namely that ENSO indeed responds to orbital forcing

yet this response is different for its two flavors. Thus, proxies from the eastern and

central Pacific might be reflecting these distinct responses of the ENSO flavors, as they

are simulated by CCSM4.

In modern climate EP El Nino events show SST anomalies in the eastern side of the

basin (Fig. 11a), and associated precipitation anomalies also shifted towards the east

(Fig. 11c). Note that only these events drive increased rainfall over the western coast of

South America. If these events weaken and become less frequent – as shown by CCSM4

for the mid-Holocene – then it is reasonable to expect that hydrological proxies from the

Eastern Pacific and equatorial South America will capture reduced ”ENSO” variability.

This is consistent with proxy evidence from lake sediment records from southern Ecuador


and the Galapagos Islands [Rodbell et al., 1999; Moy et al., 2002; Riedinger et al., 2002].

Lake sediment records capture fewer El Nino-related flood events in the mid-Holocene

in Ecuador [Rodbell et al., 1999; Moy et al., 2002] and Galapagos [Riedinger et al., 2002;

Conroy et al., 2008]. Furthermore, a marine sediment record off Galapagos Island (1.13◦S,

89.4◦W) also exhibits a drop in foraminifera population variance (reflective of a decrease

in annual and/or interannual SST extremes) that has been interpreted as weaker ENSO

[Koutavas et al., 2006]. Sandweiss et al. [1996, 2001] also interpret the presence of certain

mollusk species in geoarcheological evidence along the Peruvian coast as an absence of

ENSO variability prior to 5.8ka BP. The interpretation of the records from the eastern

Pacific is complex due to the nonlinear nature of the runoff proxies (EmileGeay and

Tingley, submitted); sensitivity of the archive to seasonal-interannual variability [e.g.

Koutavas and Joanides , 2012]; highland (garua) vs. lowland (convective) precipitation

signals recorded by Galapagos lake records [e.g. Trueman and d’Ozouville, 2010; Wolff ,

2010]; spatially heterogeneous nature of the climate signal [e.g. Liu et al., 2013]; however,

all proxy types point to a reduction in ENSO-related climate variability. Furthermore, all

these records are from locations where EP events have a hydrological (Fig 11c) and SST

(Fig 11a) signature uniquely different from CP events, therefore they could be explained

by a reduction in frequency/amplitude of EP events during the mid-Holocene as seen in

the CCSM4 simulation.

On the other hand, several δ18O records spanning the mid-Holocene, have been recently

obtained from corals from the Christmas (2◦N, 157◦W) and Fanning (4◦N, 160◦W) islands

[Cobb et al., 2013]. These records exhibit a reduction in ENSO variability in the mid-


Holocene (4–6ka BP), as measured by the standard deviation of the 2–7 yr band pass

filtered δ18O record, of 60% in Christmas Island and 40% in Fanning Island compared

to modern coral δ18O timeseries spanning the 1968-1998 period. While, the variability

is less than the observed during the aforementioned recent period, the level of variability

during the mid-Holocene is comparable to that of many multi-century periods, including

the last millennium. These proxies are thought to be able to record Central Pacific El

Nino events, as determined by calibration of modern coral δ18O to the NINO3.4 index

[Cobb et al., 2003; Nurhati et al., 2011]. In this region, El Nino events decrease coral δ18O

due to the combined temperature effect of warm SST and δ18O-depleted rainfall. Our

analysis of the observations shows that the EP and CP events have distinct signature on

SST, precipitation, and SSS in this region: CP events have a dominant impact on SST

(Fig. 11b) and precipitation (Fig. 11d)), while EP events have a dominant impact on SSS

(Fig. 11e). This contrast is more marked in Christmas, which is closer to the equator,

and could explain why coral δ18O records from these islands exhibit different sensitivities.

Thus a weaker EP ENSO during the mid-Holocene could lead to a reduction in δ18O

variability, as shown by these records, but primarily via the salinity effect that this type

of events have on this region. This reduction in δ18O variability could occur even if the

CP events did not change.

Other δ18O records from the central Pacific show greater reduction in ENSO-related

variability compared to the 20th century. McGregor et al. [2013a] present a 175-year-long,

monthly resolved oxygen isotope record, obtained from Christmas Island and dated at

around 4.3 ka BP, which shows that ENSO variance was persistently reduced by 79%,


compared to present day. In this study, the comparison with present climate is based on a

stacked modern coral record from the same region, spanning 1939-2007. This great reduc-

tion in interannual variability is consistent with the 60% reduction reported in Cobb et al.

[2013], and can be similarly explained, the difference of comparison periods notwithstand-

ing. As discussed above, due to its location, this proxy would have mixed SST and SSS

signals from both EP and CP events (Fig. 11): one would expect a mainly CP signature

and a less important EP signature in SST, as well as an EP signature in SSS.

McGregor et al. [2013a] also found that circa 4.3 ka BP El Nino events peaked two

months later, in agreement with our analysis of CCSM4, assuming that mid-Holocene con-

ditions relatively hold in this later period. Based on CCSM4’s simulation of mid-Holocene

climate, we showed that this shift could be caused by seasonal changes in stratification

in the Eastern Pacific, which are driven by changes in the trade winds over the western

basin and communicated to the eastern basin by a downwelling Kelvin wave. In contrast,

McGregor et al. [2013a] argue that the shift is driven by a strengthening of the cross-

equatorial winds during the boreal summer and early fall, enhancing eastern equatorial

Pacific upwelling, which in turn enhances the zonal equatorial SST gradient, enhances

the trade winds, and suppresses El Nino development; this mechanism is not compatible

with our modeling analyses. McGregor et al. [2013a] also report an enhancement of the

variability in their coral record within the annual band; however the model simulation

does not support a big change in the annual cycle in the central Pacific region (see Fig.

5c and d). Rather, the weakening of the annual cycle in the Eastern Pacific/cold tongue

region is a robust response in the PMIP models [Masson-Delmotte et al., 2013].


In the Warm Pool region, fossil coral records in Papua New Guinea (PNG) show reduced

variability on ENSO timescales. One of the records from Muschu Island (3.25◦S, 143◦E)

spanning the period 7.6-5.4ka BP exhibits an approximate 40% and 15% reduction in

ENSO frequency and amplitude compared to late 20th century, respectively [McGregor

and Gagan, 2004], while the record from Huon Peninsula (6.5◦S, 147.5◦E) also exhibits

similar reduction in ENSO variability in the period around 6.5 ka BP [Tudhope et al.,

2001]. A reduction of 20% in SST variability and 70% in precipitation variability at ENSO

periods is reported by Gagan et al. [2004], who use records from the Great Barrier Reef

and PNG. The western Pacific is a region where El Nino events have the opposite signature

on coral δ18O than in the central Pacific, owing to cold SST anomalies and reduced rainfall

during El Nino events. However, the observed SST patterns for both flavors are very small

there (Fig. 11a and 11b). In contrast, the rainfall/salinity anomalies could dominate the

δ18O signals in some sites. Moreover, the EP and CP events exhibit distinct rainfall/SSS

signatures there. EP events are associated with dry/saltier conditions off the coast of

PNG (Fig 11c and e). CP events, in contrast, have a rather muted hydrological response

because the nodal line of the rainfall anomalies straddles the coast of PNG (Fig 11d and

f). Note that the SSS anomalies are shifted to the west of the rainfall anomalies for both

types of events. This is because the South Equatorial Current advects the associated

freshwater flux westward, shifting the SSS anomalies. As a result, the PNG proxies are

likely to capture the saltier conditions associated with EP events. The CP events are

characterized by rainfall anomalies shifted further west than for EP events. The PNG

sites fall just in the nodal lines of the SSS pattern, therefore the PNG corals could also


be insensitive to the hydrological signature of CP events. In summary, the total signals

recorded in the Warm Pool sites are likely a mix of salinity and SST signals with influences

with varying strength from the two flavors; it is therefore possible that these sites cannot

provide further insight into the response of ENSO flavors to mid-Holocene forcing.

To conclude, the reduction in ENSO recorded by the proxies in the Eastern Pacific and

South America could be due to the significant reduction in ENSO’s EP flavor, shown

in section 3. Conversely, the CP flavor remained active, and only slightly enhanced in

terms of its strength and frequency according to the model, which is consistent with the

Fanning and Christmas Island records of Cobb et al. [2013]. The central Pacific ENSO

proxies could still be recording partially an EP signal, via the salinity effect, as discussed

above. Therefore, the simulated shift in the peak month of EP events a couple of months

later in the mid-Holocene, which can be explained by the climatological changes in SST,

precipitation, and wind stress resulting from orbital forcing, is in agreement with proxies

from Christmas Island by McGregor et al. [2013a].

It should be noted that focus of the present paper is to provide modeling support for the

idea of differential response of ENSO flavors to orbital forcing, and qualitatively compare

the modeling results to available proxy records. In the studies reported above, the defini-

tion of the mid-Holocene period as well as the reference period used to infer mid-Holocene

ENSO changes varies. However, the primary question that is addressed here is whether

ENSO responds to orbital forcing, i.e. whether there are significant changes between the

mid-Holocene period and other periods that lack significant anomalous orbital forcing,

such as the pre-industrial one. Proxy records from the eastern Pacific [e.g. Koutavas and


Joanides , 2012] answer this question positively, while records from the central Pacific [e.g.

Cobb et al., 2013] answer this question negatively or without ruling out the null hypothesis

of ENSO insensitivity to orbital forcing. It is therefore appropriate in the present study to

compare the mid-Holocene to the pre-industrial model simulation, i.e. a simulation with

anomalous orbital forcing to a simulation with present-day orbital forcing, and attempt to

discuss the proxies in this context, rather than compare to historical or late 20th century

simulations which are forced externally by greenhouse gases among other forcings.

7. Conclusions

Motivated by seemingly conflicting evidence from paleo-climate proxies from the East-

ern and Central Pacific, we studied the response of the two ENSO flavors, Eastern and

Central Pacific El Nino, to orbital forcing, using long simulations of pre-industrial and

mid-Holocene climate from NCAR’s CCSM4 model. We found a differential response of

the two flavors, which we attributed to changes in the seasonality of the cold tongue, and

the resulting changes in the heat budget during El Nino events of each flavor in the two

climates. Our main findings for the EP flavor can be summarized as follows:

• The frequency of occurrence of EP events significantly decreases in the mid-Holocene

(by 50%). The variance of the EP-event index decreases by 30%.

• The development of EP events in the mid-Holocene is slower and their decay is faster

compared to the pre-industrial climate.

• There is a shift in the seasonality of the EP events, as their peak is delayed by

approximately two months in the mid-Holocene.


• The determining factor for the development of strong EP events is the upwelling

feedback in the Eastern Pacific which is modulated by seasonal changes in stratification.

In the mid-Holocene, remote wind forcing in the western Pacific deepens the thermocline

over the NINO1+2 region during the boreal fall. The reduced stratification weakens the

upwelling feedback resulting in weaker and less frequent EP events.

• The faster decay of EP events in the mid-Holocene is associated with a stronger and

earlier southward shift of wind stress curl anomalies in mid-Holocene, which is in turn

associated with weaker climatological wind stress in boreal winter/early spring in the

Southern Hemisphere and a stronger SPCZ.

For the CP flavor, we found that:

• The frequency of occurrence of CP events slightly increases in the mid-Holocene

(from 10 to 12 per century on average), while the variance of the CP-event index remains

unchanged.

• CP events are stronger in the mid-Holocene.

• There is no shift in the seasonality of CP events.

• The dominant feedback term for the development of the CP events is the zonal

advection feedback, which is stronger in the mid-Holocene possibly in connection to the

downwelling Kelvin wave which forms in early spring and reaches the Eastern Pacific in

late summer-early fall.

• The mechanism for termination of EP events which involves the southward shift of

wind stress anomalies with the progression of seasons does not play a significant role for

CP events.


In summary, the differences in the development and decay of both ENSO flavors between

the pre-industrial and mid-Holocene climate is related to changes in the seasonality of the

trade winds. These changes are initiated over the western Pacific and communicated by

the ocean to the eastern Pacific. The proposed mechanism to explain the response of

ENSO to orbital forcing is presented schematically in Fig. 12.

The simulated reduction in the EP flavor and not the CP flavor in the mid-Holocene

are consistent with evidence from paleo-climate proxies from the Eastern and the Central

Pacific. The teleconnection patterns of the two flavors with temperature, precipitation

and salinity are distinct, and proxies from different regions in the Pacific might be record-

ing variability of only one of the two flavors, or various combinations of their relative

effects. Our model-based analysis suggests that the great reduction in ENSO variability

inferred by proxies in the Eastern Pacific may be due to a reduction in the EP flavor.

On the contrary, the absence of significant reduction in variability in the Central Pacific

compared to periods lacking orbital forcing is consistent with the model results that show

no significant changes in the CP flavor. Some reduction in the variability inferred by

Central Pacific ENSO proxies is still consistent with the model results since it could be

due to a mixed signal in temperature and salinity from both flavors, with the CP flavor

dominating in the temperature effect and the EP flavor dominating in the salinity effect.

The issue of mixed signals is particularly burdening in the case of the available Western

Pacific proxies, therefore one should be cautious in interpreting them in connection to the

two ENSO flavors. It should also be noted that the paleoclimate proxies, and especially

the short coral segments, could be sub-sampling periods of naturally-occurring variability


in ENSO flavors, which could be superposed on their response to orbital forcing (enhanc-

ing or attenuating the latter). For example, one cannot rule out the possibility that the

records of Cobb et al. [2013] are sampling a mid-Holocene period of enhanced EP and

CP activity superposed on an otherwise significantly muted ENSO-activity background,

which could result in a signal reduction that is not different from other Holocene periods.

Similarly, the records of McGregor et al. [2013a] could be sub-sampling from a period of

naturally decreased EP and CP activity on top of an orbitally-induced ENSO activity

reduction, which could result in the greater reduction they report in their study.

This study is possible thanks to the improved realism in the simulation of ENSO in

state-of-the-art climate models. The length of the CCSM4 simulations (1300 years of pi-

Control and 500 years of mid-Holocene) allows us to compute robust statistics and make

meaningful inter-climate comparisons of ENSO variables, especially at the tails of their

distributions. CCSM4 is considered one of the best climate models in terms of simu-

lation of ENSO [Deser and Coauthors , 2012], as well as its flavors (Fig. 1). However,

CCSM4 exhibits biases common to other models, such as the well-known ”double-ITCZ”

and ”cold-tongue” biases, which could influence our results. The cold-tongue bias is char-

acterized by a westward extension and lower temperatures of the cold tongue compared

to observations[Bellenger et al., 2014], while the double-ITCZ bias is characterized by

excessive precipitation over much of the Tropics (e.g., Northern Hemisphere ITCZ, South

Pacific convergence zone, Maritime Continent, and equatorial Indian Ocean), and insuf-

ficient precipitation over the equatorial Pacific, which may lead to overly strong trade

winds, excessive latent heat flux, insufficient shortwave radiation flux, and cold SST bi-


ases [Lin, 2007]. The cold tongue bias could favor the occurrence of CP events at the

expense of EP events, while the double-ITCZ leads to an unrealistic semi-annual cycle of

the cold tongue, which could influence the seasonal controls on ENSO flavors. However,

there are two other biases common the best climate models that could make the mecha-

nism proposed here more prominent in the real world. First, CCSM4 simulates an annual

cycle that is weaker than observed, therefore a much larger weakening of the annual cycle

would be expected for the mid-Holocene. Second, CCSM4 overestimates the amplitude of

ENSO SST variability by approximately 30%, thus if CCSM4’s ENSO was weaker, then

the simulated mid-Holocene reduction in EP ENSO (30%) could potentially lead to a

complete disappearance of EP ENSO in the mid-Holocene. After decades of research, the

mid-Holocene is still a challenging target for the simulation of ENSO in climate models.

Future progress understanding the mid-Holocene ENSO requires improvement of these

biases, along with new proxy records from undersampled regions, which include the west,

central and eastern equatorial Pacific.

Appendix A: Interpreting changes in the ENSO heat budget

The methodology used here to estimate the heat budget terms has been used to study

ENSO dynamics in coupled GCMs [DiNezio et al., 2009, 2012; Capotondi , 2013; DiNezio

and Deser , 2014]. The most important feature of this methodology is that a nearly

balanced heat budget can be obtained using monthly mean three dimentional velocity (u,

v, w) and temperature (T ) fields. Our analysis of the full heat budget shows that for


interannual anomalies, the heat budget can be approximated by:

Q′t = ρ0cp

∫ 0

−H

∂T ′

∂tdz = −ρ0cp

∫ 0

−H

(u∂T

∂x+ v

∂T

∂y+ w

∂T

∂z

)′dz +Q′atm. (A1)

The definitions of the variables in (A1) follow the convention, where primed variables

are anomalies with respect to the climatological monthly-mean seasonal cycle. Here,

anomalies are computed with respect to each simulation’s (mid-Holocene and piControl)

climatology. The right hand side of (A1) is the heat storage rate.The first term in the

left hand side is the advection of temperature by ocean currents. In our analysis in

section 5, we neglect several temperature advection terms, such as meridional advection

and most nonlinear terms, because their interannual variability is small. We focus on

the three main feedbacks in the NINO3.4 and NINO1+2 regions, i.e. the zonal advection

feedback Q′za = −ρ0cp∫ 0−H(u′ ∂T

∂x)dz, the thermocline feedback Q′tc = −ρ0cp

∫ 0−H(w ∂T ′

∂z)dz,

and the upwelling feedback Q′uw = −ρ0cp∫ 0−H(w′ ∂T

∂z)dz. Both the tendency and advection

terms are integrated over a constant-depth layer of thickness H=90m, which is taken

to be 20 m below the base of the ocean mixed layer (similar results are obtained using

10 to 30 m). T ′ is the ocean temperature anomaly averaged over depth H, and T is

the climatological monthly-mean temperature. Selecting H below the base of the mixed

layer allows us to neglect the effect of subgrid scale (SGS) processes, such as wind-driven

mixing and entrainment, and sun-light penetration on T ′ (see DiNezio and Deser [2014]

for further details). The heat budget in (A1) is completed with Q′atm, the net air-sea heat

flux (positive into the ocean). The remaining constants are ρ0, a reference density of sea

water, and cp, the specific heat of sea water.


The difference in total heat content tendency Q′t integrated over depth H between the

two climates is plotted in Fig. 7c and corresponds to

∆Q′t = −ρ0cp

[∫ 0

−H

∂T ′MH

∂tdz −

∫ 0

−H

∂T ′piControl

∂tdz

]= (A2)

= −ρ0cp

∫ 0

−H

∂(T ′MH − T ′piControl)

∂tdz = −ρ0cp

∫ 0

−H

∂∆T ′

∂tdz (A3)

As discussed in sections 3 and 4, there is a change in the seasonality of SSTs in the

tropical Pacific induced by the orbital forcing in the mid-Holocene, as well as a change

in the seasonality of the peak of EP events (Fig. 4). Since the composite plots presented

in section 5 are with respect to the peak of the events, and refer to anomalies with

respect to each simulation’s climatology, the difference plots are indeed descriptive of the

differences in the evolution of events within their respective SST (seasonal) background.

However, since the two systems (ENSO and background climatology) are not entirely

independent, the change in the heating feedbacks may include convoluted changes in

ENSO and background climatology that cannot be disentangled by the present analysis.

To illustrate this in detail, consider the difference in upwelling feedback

∆Q′uw = −ρ0cp∫ 0−H ∆

[w′ ∂T

∂z

]dz that is presented in Fig. 7f. The mid-Holocene w′MH can

be written as (w′ + ∆w′), and the mid-Holocene mean climatological temperature TMH

can be written as (T + ∆T ). Then, ∆Q′w can be written as:

∆Q′uw = −ρ0cp

∫ 0

−H∆

[w′∂T

∂z

]dz

= −ρ0cp

{ ∫ 0

−H

[(w′ + ∆w′)

∂(T + ∆T )

∂z

]dz −

∫ 0

−Hw′∂T

∂zdz}

(A4)

= −ρ0cp

∫ 0

−H

[w′∂T

∂z+ w′

∂∆T

∂z+ ∆w′

∂T

∂z+ ∆w′

∂∆T

∂z− w′∂T

∂z

]dz (A5)


The fist and last terms in the rhs cancel out, and the fourth term can be neglected because

it is the product of two O(∆) quantities. The third term has the same sign as a potential

“independent” ENSO change, therefore cannot be separated. Furthermore, we can ignore

this term because ENSO anomalies in the NINO1+2 region are not strictly coupled, rather

they are remotely driven from the central Pacific where ocean-atmosphere coupling is the

strongest. And since ENSO SSTs do not change considerably in the central Pacific, we

can ignore the ∆w′ · [...] terms of the heat budget.

We are then left with

∆Q′uw ≈ −ρ0cp

∫ 0

−Hw′∂∆T

∂zdz (A6)

We therefore argue that this term, i.e. the change in upwelling feedback that is due to the

change in the climatological stratification ∂∆T∂z

, is dominant in producing the reduction in

the upwelling feedback shown in Fig. 7f, and is hence the main cause of difference in the

development of ENSO events in the mid-Holocene.

Acknowledgments. This research is funded by US NSF Grant # OCN–1304910, and

US Department of Energy Grant # DESC005110. We are thankful to the three anony-

mous reviewers and the associate editor for their constructive feedback. Many thanks

are due to Bette Otto-Bliesner of NCAR for feedback and support. We wish to acknowl-

edge members of NCAR’s Climate Modeling Section, CESM Software Engineering Group

(CSEG), and Computation and Information Systems Laboratory (CISL) for their contri-

butions to the development of CESM and CCSM. All model results are publicly available

through NCAR’s portals, and all computed data are available upon request.


References

Adler, R., G. Huffman, A. Chang, R. Ferraro, P. Xie, J. Janowiak, B. Rudolf, U. Schnei-

der, S. Curtis, D. Bolvin, A. Gruber, J. Susskind, and P. Arkin (2003), The Version

2 Global Precipitation Climatology Project (GPCP) Monthly Precipitation Analysis

(1979-Present), J. Hydrometeor., 4, 1147–1167.

Ashok, K., S. Behera, S. Rao, H. Weng, and T. Yamagata (2007), El Nino Modoki and

its possible teleconnection, J. Geophys. Res.-Oceans, 112 (C11), 27.

Bellenger, H., E. Guilyardi, J. Leloup, M. Lengaigne, and J. Vialard (2014), ENSO

representation in climate models: from CMIP3 to CMIP5, Climate Dynamics, 42 (7–

8), 1999–2018.

Brown, J., M. Collins, and A. Tudhope (2006), Coupled model simulations of mid-

Holocene ENSO and comparisons with coral oxygen isotope records, Adv, in Geosci.,

6, 29–33.

Brown, J., M. Collins, A. Tudhop, and T. Toniazzo (2008), Modelling mid-Holocene

tropical climate and ENSO variability: Towards constraining predictions of future

change with paleo-data, Clim. Dyn., 30, 19–36.

Bush, A. (2008), Assessing the impact of mid-Holocene insolation on the atmosphere-

ocean system, Geophys. Res. Lett., 26 (1), 99–102, doi:10.1029/1998GL900261.

Cai, W., S. Borlace, M. Lengaigne, P. van Rensch, M. Collins, G. Vecchi, A. Timmer-

mann, A. Santoso, M. McPhaden, L. Wu, M. England, G. Wang, E. Guilyardi, and

F.-F. Jin (2014), Increasing frequency of extreme El Nino events due to greenhouse

warming, Nature Climate Change, 4 (1), 111–116.


Capotondi, A. (2013), ENSO diversity in the NCAR CCSM4 climate model, Journal of

Geophysical Research: Oceans, 118 (10), 4755–4770, doi:10.1002/jgrc.20335.

Carre, M., J. Sachs, S. Purca, A. Schauer, P.Braconnot, R. A. Falcon, M. Julien, and

D. Lavallee (2014), Holocene history of ENSO variance and asymmetry in the eastern

tropical Pacific, Science, 345 (6200), 1045–1048, doi:10.1126/science.1252220.

Chang, P., B. Wang, T. Li, and L. Ji (1994), Interactions between the seasonal cycle

and the Southern Oscillation-Frequency entrainment and chaos in a coupled ocean-

atmosphere model, Geophys. Res. Lett., 21, 2817–2820.

Chazen, C., M. Altabet, and T. Herbert (2009), Abrupt mid-Holocene onset of

centennial-scale climate variability on the Peru-Chile Margin, Geophys. Res. Lett, 36,

doi:10.1029/2009GL039749.

Chiang, J., Y. Fang, and P. Chang (2009), Pacific Climate Change and ENSO Activity

in the Mid-Holocene, Geophys. Res. Lett, 22, 923–939.

Clement, A., R. Seager, and M. Cane (2000), Suppression of El Nino during the mid-

Holocene by changes in the Earths orbit, Paleoceanography, 15 (6), 731–737.

Clement, A., M. Cane, and R. Seager (2001), An orbitally driven tropical source for

abrupt climate change, J. Climate, 14, 2369–2375.

Cobb, K., C. Charles, H. Cheng, and R. Edwards (2003), El Nino/Southern Oscillation

and tropical Pacific climate during the last millennium, Nature, 424, 271–276.

Cobb, K., N. Westphal, H. Sayani, E. D. Lorenzo, C. Charles, H. Cheng, and R. Ed-

wards (2013), Highly variable El Nino-Southern Oscillation throughout the Holocene,

Science, 339 (6115), 67–70.


Collins, M., S.-I. An, W. Cai, A. Ganachaud, E. Guilyardi, F.-F. Jin, M. Jochum,

M. Lengaigne, S. Power, A. Timmermann, G. Vecchi, and A. Wittenberg (2010), The

impact of global warming on the tropical Pacific Ocean and El Nino, Nat. Geo., 3,

391–397.

Conroy, J., J. Overpeck, J. Cole, T. Shanahan, and M. Steinitz-Kannan (2008), Holocene

changes in eastern tropical Pacific climate inferred from a Galapagos lake sediment

record, Quat Sci Rev, 27 (11–12), 1166–1180.

Delcroix, T., G. Alory, S. Cravatte, T. Correge, and M. McPhaden (2011), A gridded sea

surface salinity data set for the tropical Pacific with sample applications (1950-2008),

Deep Sea Res., 58, 38–48, doi:10.1016/j.dsr.2010.11.002.

Deser, C., and Coauthors (2012), ENSO and Pacific Decadal Variability in the Commu-

nity Climate System Model Version 4, J. Climate, 25, 2622–2651.

di Lorenzo, E., K. Cobb, J. Furtado, N. Schneider, B. Anderson, A. Bracco, M. Alexan-

der, and D. Vimont (2010), Central Pacific El Nino and decadal climate change in the

North Pacific Ocean, Nat. Geo., 3 (11), 762–765.

DiNezio, P., and C. Deser (2014), Nonlinear controls on the persistence of La Nina, J.

Climate, 27, 7335–7355.

DiNezio, P., A. Clement, G. Vecchi, B. Soden, B. Kirtman, and S.-K. Lee (2009), Climate

Response of the Equatorial Pacific to Global Warming, J. Climate, 22, 4873–4892.

DiNezio, P., B. J. Kirtman, A. Clement, S.-K. Lee, G. Vecchi, and A. Wittenberg (2012),

Mean Climate Controls on the Simulated Response of ENSO to Increasing Greenhouse

Gases, J. Climate, 25, 7399–7420.


Donders, T., F. Wagner, D. Dilcher, and H. Visscher (2005), Mid- to late-Holocene El

Nino-Southern Oscillation dynamics reected in the subtropical terrestrial realm, Proc.

Nat. Acad. Sci., 102 (31), 10,904–10,908.

Donders, T., S. Haberle, G. Hope, F. Wagner, and H. Visscher (2007), Pollen evidence

for the transition of the eastern Australian climate system from the post-glacial to

the present-day ENSO mode, Quater. Sci. Rev., 26 (11-12), 1621–1637.

Donders, T., F. Wagner-Cremer, and H. Visscher (2008), Integration of proxy data and

model scenarios for the mid-Holocene onset of modern ENSO variability, Quater. Sci.

Rev., 27, 571–579.

Gagan, M., S. H. E.J. Hendy, and W. Hantoro (2004), Postglacial evolution of the Indo-

Pacific Warm Pool and El Nino-Southern Oscillation, Quat. Int., 118–119, 127–143.

Gent, P., and Coauthors (2011), The Community Climate System Model Version 4, J.

Climate, 24, 4973–4991.

Harrison, D., and G. Vecchi (1999), On the termination of El Nino, Geophys. Res. Lett.,

26 (11), 1593–7.

Harrison, D., and G. Vecchi (2003), On the termination of the 2002-3 El Nino event,

Geophys Res. Lett., 30 (18), 1964–1967.

Harrison, D., and G. Vecchi (2006), The termination of the 1997-98 El Nino. Part II:

Mechanisms of Atmospheric Change., J. Climate, 19, 2647–2664.

Hewitt, C., and J. Mitchell (1998), A fully coupled GCM simulation of the climate of

the mid-Holocene, Geophys. Res. Lett., 25 (3), 361–364, doi:10.1029/97GL03721.


Hoerling, M., and A. Kumar (2002), Atmospheric response patterns associated with

tropical forcing, J. Climate, 15, 2184–2203.

Hu, Z.-Z., A. Kumar, B. Jha, W. Wang, B. Huang, and B. Huang (2012), An analysis

of warm pool and cold tongue El Ninos: airsea coupling processes, global influences,

and recent trends, Clim Dyn, 38, 2017–2035, doi:10.1007/s00382-011-1224-9.

Jin, F.-F., J. Neelin, and M. Ghil (1994), El Nino on the devil’s staircase: annual and

subharmonic steps to chaos, Science, 264, 70–72.

Jin, F.-F., J. Neelin, and M. Ghil (1996), El Nino/Southern Oscillation and the annual

cycle: subharmonic frequency-locking and aperiodicity, Physica D, 98, 442–465.

Kao, H.-Y., and J.-Y. Yu (2009), Contrasting Eastern-Pacific and Central-Pacific Types

of ENSO, J Climate, 22 (3), 615–632.

Kim, H.-M., P. Webster, and J. Curry (2009), Impact of Shifting Patterns of Pacific

Ocean Warming on North Atlantic Tropical Cyclones, Science, 325, 77–79.

Kitoh, A., and S. Murakami (2002), Tropical Pacific climate at the mid-Holocene and the

Last Glacial Maximum simulated by a coupled ocean-atmosphere general circulation

model, Paleoceanography, 17 (3), 1047, doi:10.1029/2001PA000724.

Koutavas, A., and S. Joanides (2012), El Nino-Southern Oscillation extrema in the

Holocene and Last Glacial Maximum, Paleoceanography, 27 (4), PA4208.

Koutavas, A., J. Lynch-Stieglitz, T. Marchitto, and J. Sachs (2002), El Nino-like pattern

in ice age tropical Pacic sea surface temperature, Science, 297 (5579), 226–230.

Koutavas, A., P. deMenocal, G. Olive, and J. Lynch-Stieglitz (2006), Mid-Holocene El

Nino-Southern Oscillation (ENSO) attenuation revealed by individual foraminifera in


eastern tropical Pacific sediments, Geology, 34, 993–996.

Kug, J.-S., F.-F. Jin, and S.-I. An (2009), Two Types of El Nino Events: Cold Tongue

El Nino and Warm Pool El Nino, J. Climate, 22 (6), 1499–1515.

Larkin, N., and D. Harrison (2005), On the definition of El Nino and associated

seasonal average U.S. weather anomalies, Geophys. Res. Lett, 32, L13,705, doi:

10.1029/2005GL022738.

Lee, T., and M. McPhaden (2010), Increasing intensity of El Nino in the central-

equatorial Pacific, Geophys. Res. Lett, 37, L14,603, doi:10.1029/2010GL044007.

Lengaigne, M., J.-P. Boulanger, C. Menkes, and H. Spencer (2006), Influence of the

Seasonal Cycle on the Termination of El Nino Events in a Coupled General Circulation

Model, J. Climate, 19, 1850–1868.

L’Heureux, M., D. Collins, and Z.-Z. Hu (2012), Linear trends in sea surface temperature

of the tropical Pacific Ocean and implications for the El Nino-Southern Oscillation,

Clim Dyn, doi:10.1007/s00382-012-1331-2.

Lin, J.-L. (2007), The Double-ITCZ Problem in IPCC AR4 Coupled GCMs:

OceanAtmosphere Feedback Analysis., J. Climate, 20, 4497–4525, doi:

http://dx.doi.org/10.1175/JCLI4272.1.

Liu, Y., L. Xie, J. Morrison, and D. Kamykowski (2013), Dynamic downscaling of the

impact of climate change on the ocean circulation in the Galapagos Archipelago,

Advances in Meteorology, 2013(ID 837432), 1–18.

Liu, Z. (2002), A simple model study of ENSO suppression by external periodic forcing,

J. Climate, 15, 1088–1098.


Liu, Z., J. Kutzbach, and L. Wu (2000), Modeling climate shift of El Nino variability

in the Holocene, Geophys. Res. Lett., 27, 2265–2268.

Mantsis, D., A. Clement, B. Kirtman, A. Broccoli, and M. Erb (2013a), Precessional

Cycles and their Influence on the North Pacific and North Atlantic Summer Anticy-

clones, J. Climate, 26, 4596–4611.

Mantsis, D., B. Lintner, A. Broccoli, and M. Khondri (2013b), Mechanisms of Mid-

Holocene Precipitation Changes in the South Pacific Convergence Zone, J. Climate,

26, 6937–6953.

Marzin, C., and P. Braconnot (2009), Variations of Indian and African monsoons in-

duced by insolation changes at 6 and 9.5 kyr BP, Clim. Dyn., 33 (2–3), 215–231.

Masson-Delmotte, V., M. Schulz, A. Abe-Ouchi, J. Beer, A. Ganopolski, J. G. Rouco,

E. Jansen, K. Lambeck, J. Luterbacher, T. Naish, T. Osborn, B. Otto-Bliesner,

T. Quinn, R. Ramesh, M. Rojas, X. Shao, and A. Timmermann (2013), Information

from paleoclimate archives, in Climate Change 2013: The Physical Science Basis. Con-

tribution of Working Group I to the Fifth Assessment Report of the Intergovernmental

Panel on Climate Changes, edited by T. Stocker, D. Qin, G.-K. Plattner, M. Tig-

nor, S. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex, and P. Midgley, Cambridge

University Press, Cambridge, United Kingdom and New York, NY, USA.

McGregor, H., and M. Gagan (2004), Western Pacific coral δ18O records of anoma-

lous Holocene variability in the El Nino-Southern Oscillation, Geophys. Res. Lett.,

31 (L11204).


McGregor, H., M. Fischer, M. Gagan, D. Fink, S. Phipps, H. Wong, and C. Woodroffe

(2013a), A weak El Nino/Southern Oscillation with delayed seasonal growth around

4300 years ago, Nature Geoscience, 6, 949–953.

McGregor, S., A. Timmermann, N. Schneider, M. Stuecker, and M. H. England (2012),

The Effect of the South Pacific Convergence Zone on the Termination of El Nino

Events and the Meridional Asymmetry of ENSO, J. Climate, 25, 5566–5586.

McGregor, S., N. Ramesh, P. Spence, M. H. England, M. J. McPhaden, and A. San-

toso (2013b), Meridional movement of wind anomalies during enso events and

their role in event termination, Geophysical Research Letters, 40 (4), 749–754, doi:

10.1002/grl.50136.

McPhaden, M., T. Lee, and D. McClurg (2011), El Nino and its relationship to changing

background conditions in the tropical Pacic Ocean., Geophys. Res. Lett, 38 (L15709).

Meehl, G., and Coauthors (2007), El Nino and its relationship to changing background

conditions in the tropical Pacic Ocean., Geophys. Res. Lett, 38 (L15709).

Mo, K. (2010), Interdecadal modulation of the impact of ENSO on precipitation and

temperature over the United States, J Climate, 23, 3639–56.

Monahan, A., and A. Dai (2004), The spatial and temporal structure of ENSO nonlin-

earity, J Climate, 17, 3026–3036.

Moy, C., G. Seltzer, D. Rodbell, and D. Anderson (2002), Variability of El NioSouthern

Oscillation activity at millennial timescales during the Holocene epoch, Nature, 420,

162–165, doi:10.1038/nature01194.


Newman, M., S.-I. Shin, and M. A. Alexander (2011a), Natural variation in ENSO

flavors, Geophys. Res. Lett., 420, doi:10.1029/2011GL047658.

Newman, M., M. Alexander, and J. D. Scott (2011b), An empirical model of tropical

ocean dynamics, Clim. Dyn., doi:10.1007/s00382-011-1034-0.

Nurhati, I., K. Cobb, and E. D. Lorenzo (2011), Decadal-Scale SST and Salinity Vari-

ations in the Central Tropical Pacific: Signatures of Natural and Anthropogenic Cli-

mate Change, J. Climate, 24, 3294–3308.

Otto-Bliesner, B., S. S. E. Brad and, Z. Liu, and C. Shield (2003), Modeling El Nino

and its tropical teleconnections during the last glacial-interglacial cycle, Geophys. Res.

Lett., 30 (23), 2198, doi:10.1029/2003GL018553.

Pan, A., Q. Liu, and Z. Y. Liu (2005), Periodic forcing and ENSO suppression in the

Cane-Zebiak model, J. Oceanogr., 61, 109–113.

Rayner, N., D. Parker, E. Horton, C. Folland, L. Alexander, D. Rowell, E. 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.

Ren, H.-L., and F.-F. Jin (2011), Nino indices for two types of ENSO, Geophys. Res.

Lett., 38, doi:10.1029/2010GL046031.

Riedinger, M., M. Steinitz-Kannan, W. Last, and M. Brenner (2002), A 6100 14C yr

record of El Nino activity from the Galapagos Islands, J. Paleolimnol., 27, 1–7.

Rodbell, D., G. Seltzer, D. Anderson, M. Abbott, D. Enfield, and J. Newman (1999),

An 15,000-year record of El Nino-driven alluviation in southwestern Ecuador, Science,


283, 516–520, doi:10.1126/science.283.5401.516.

Salau, O., B. Schneider, W. Park, V. Khon, and M. Latif (2012), Modeling the ENSO im-

pact of orbitally induced mean state climate changes, J. Geophys. Res., 117, C05,043,

doi:10.1029/2011JC007742.

Sandweiss, D., J. R. III, E. Reitz, H. Rollins, and K. Maasch (1996), Geoarchaeological

evidence from Peru for a 5000 years BP onset of El Nino, Science, 273, 1531–1533.

Sandweiss, D., K. Maasch, R. Burger, J. Richardson, H. Rollins, and A. Clement (2001),

Variation in Holocene El Nino frequencies: climate records and cultural consequences

in ancient Peru, Geology, 29 (7), 603–606.

Stein, K., A. Timmermann, and N. Schneider (2011), Phase Synchronization of the El

Nino-Southern Oscillation with the Annual Cycle, Phys. Rev. Lett., 107 (12), 128,501.

Stein, K., A. Timmermann, N. Schneider, F.-F. Jin, and M. Stuecker (2014), ENSO

seasonal synchronization theory, J. Climate, 27 (14), 5285–5310.

Stuecker, M., A. Timmermann, F.-F. Jin, S. McGregor, and H.-L. Ren (2013), A Com-

bination Mode of Annual Cycle and the El Nino-Southern Oscillation, Nature Geo-

science, 6, 540–544.

Takahashi, K., A. Montecinos, K. Goubanova, and B. Dewitte (2011), ENSO regimes:

Reinterpreting the canonical and Modoki El Nino, Geophys. Res. Lett., 38, doi:

10.1029/2011GL047364.

Timmermann, A. (1999), Detecting the Nonstationary Response of ENSO to Greenhouse

Warming, J. Atmos. Sci., 56, 2313–2325.


Timmermann, A., S. Lorenz, S. An, A. Clement, and S.-P. Xie (2007), The effect of

orbital forcing on the mean climate and variability of the tropical Pacific, J. Climate,

20, 4147–4159.

Trenberth, K., and D. Stepaniak (2001), Indices of El Nino Evolution, J. Climate, 14,

1697–1701.

Trueman, M., and N. d’Ozouville (2010), Characterizing the Galapagos terrestrial cli-

mate in the face of global climate change, Galapagos Research, 67, 26–36.

Tudhope, A., C. Chilcott, M. McCulloch, E. Cook, J. Chappell, R. Ellam, D. Lea,

J. Lough, and G. B. Shimmield (2001), Variability in the El Nino-Southern

Oscillation through a glacial-interglacial cycle, Science, 291, 1511–1517, doi:

10.1126/science.1057969.

Tziperman, E., L. Stone, M. Cane, and H. Jarosh (1994), El-Nino chaos: overlapping

of resonances between the seasonal cycle and the Pacific ocean-atmosphere oscillator,

Science, 264 (5155), 72–74.

Tziperman, E., M. Cane, and S. Zebiak (1995), Irregularity and locking to the seasonal

cycle in an ENSO prediction model as explained by the quasi-periodicity route to

chaos, J. Atmos. Sci., 52 (3), 293–306.

Wang, B., and Z. Fang (1996), Chaotic Oscillation of the tropical climate: A dynamic

system theory for ENSO, J. Atmos. Sci., 53, 2786–2802.

Wang, G., and H. Hendon (2007), Sensitivity of Australian rainfall to inter-El Nino

variations, J. Climate, 20, 4211–4226.


Weng, H., K. Ashok, S. Behera, S. Rao, and T. Yamagata (2007), Impacts of recent

El Nino Modoki dry/wet conditions in the Pacific rim during Boreal Summer, Clim.

Dyn., doi:10.1007/s00382-007-0234-0.

Wittenberg, A. (2009), Are historical records sufficient to constrain ENSO simulations?,

Geophys. Res. Lett., 36, doi:10.1029/2009GL038710.

Wolff, C., G. Haug, A. Timmermann, J. S. Damste, A. Brauer, D. Sigman, M. Cane, and

D. Verschuren (2011), Reduced interannual rainfall variability in east Africa during

the Last Ice Age, Science, 333, 743–747.

Wolff, M. (2010), Galapagos does not show recent warming but increased seasonality,

Galapagos Research, 67, 38–44.

Woodroffe, C., M. Beech, and M. Gagan (2003), Mid-late Holocene El Nino variability

in the equatorial Pacific from coral microatolls, Geophys. Res. Lett., 30 (7), 1358, doi:

10.1029/2002GL015868.

Xu, J., and J. Chan (2001), The Role of the AsianAustralian Monsoon System in the

Onset Time of El Nino Events, J. Climate, 14, 418–433.

Yeh, S.-W., J.-S. Kug, B. Dewitte, M.-H. Kwon, B. Kirtman, and F.-F. Jin (2009), El

Nino in a changing climate, Nature, 461, 511–515.

Yeh, S.-W., B. P. Kirtman, J. Kug, W. Park, and M. Latif (2011), Natural variability of

the central Pacific El Nino event on multicentennial timescales, Geophys. Res. Lett.,

38, doi:10.1029/2010GL045886.

Yu, J.-Y., and S. T. Kim (2010), Three Evolution Patterns of Central-Pacific El Nino,

Geophys. Res. Lett., 37, doi:10.1029/2010GL042810.


Yu, J.-Y., F. Sun, and H.-Y. Kao (2009), Contributions of Indian Ocean and monsoon

biases to the excessive biennial ENSO in CCSM3, J. Climate, 22, 1850–1858.

Yu, J.-Y., H.-Y. Kao, and T.Lee (2010), Subtropics-Related Interannual Sea Surface

Temperature Variability in the Central Equatorial Pacific. Journal of Climate, J.

Climate, 23 (11), 2869–2884.


a)

b)

c)

d)

HadISST

NINO1+2

NINO3

Figure 1. Linear regression coefficients (◦C, shading) between SST anomalies and the E-index

and C-index from CCSM4’s control simulation (a,b) and observations (c,d). The indeces result

from a 45-degree rotation of the fist two principal components of tropical Pacific SST anomalies

(eq. 1), and capture the EP and CP events, respectively.


probability density

warmer eastco

lder c

entral P

aci�c

warmer c

entral P

aci�c

warmer eastco

lder c

entral P

aci�c

warmer c

entral P

aci�c

0.1 0.2 0.3

Figure 2. Scatterplots and bivariate probability density function of the two leading principal

components (monthly values, averaged Oct-Apr) in a) piControl, b) mid-Holocene, and c) their

difference. The diagonals indicate the E-index and C-index axes. Red solid (blue dashed) con-

tours indicate positive (negative) differences in pdf mass in c. The pdf is computed via kernel

density estimation. Solid black circles denote Oct-Apr averages of the principal components for

events with the E-index larger than two standard deviations from zero, and gray-filled circles

denote Oct-Apr averages for events with C-index larger than one standard deviation away from

zero, which are the events defined as EP and CP, respectively, in our analysis.


K

K

K

K

a) b) c)

e) f )

Figure 3. Composite SST anomaly plots for EP (upper panel) and CP (lower panel) events in

piControl (a,d), mid-Holocene (b,e), and their difference (c,f). Stippled areas in figures c and f

denote statistical significance. All fields are averaged between 5◦S-5◦N . Anomalies are computed

with respect to each simulation’s climatology.


Figure 4. The percentage of winter (ONDJFMA) months with a) EP-event peaks and b)

CP-event peaks by month and climate.


mid-Holocene minus piControl

mid-Holocene minus piControla)

b)

c)

d)

Figure 5. Climatological mean monthly SST (shading), precipitation (contours) and winds

(vectors) during boreal summer/fall (JASO) and boreal winter/spring (FMAM) and in piControl

(left panel), as well as the change in the mid-Holocene (right panel).


Figure 6. Time-longitude plot of the difference between piControl and mid-Holocene in

climatological a) eastward wind stress, b) thermocline depth, and c) stratification, i.e. Ts− Tsub,

where Ts is the surface temperature and Tsub is the temperature at depth of 50m. All fields are

averaged between 5◦S-5◦N .


Q’uw

(Wm-2)

Qt

(Wm-2)Qt

(Wm-2)a) b) c)

d) e) f )Q’uw

(Wm-2)

Figure 7. Composites of EP-event total heat tendency (upper panel) and tendency due to

the upwelling feedback (lower panel) in piControl (a,d), mid-Holocene (b,e), and their difference

(c,f). All fields are averaged between 5◦S-5◦N . Stippled areas in figures c and f denote statistical

significance.


Q’za

(Wm-2) Q’za

(Wm-2)

Qt

(Wm-2)Qt

(Wm-2)a) b) c)

d) e) f )

Figure 8. Composites of CP-event total heat tendency (upper panel) and tendency due to

the zonal advection feedback (lower panel) in piControl (a,d), mid-Holocene (b,e), and their

difference (c,f). All fields are averaged between 5◦S-5◦N . Stippled areas in figures c and f denote

statistical significance.


a) b) c)

d) e) f )

x10-7 Pa.m-1 x10-7 Pa.m-1

Figure 9. Composite wind stress anomaly (vectors, Pa) and wind stress curl anomaly (con-

tours, 10−7Pa · m−1) for EP (upper panel) and CP (lower panel) events in piControl (a,d),

mid-Holocene (b,e), and their difference (c,f). All fields are averaged between 150◦E-160◦W .

Stippled areas in figures c and f denote statistical significance. The thick black line connects

statistically significant areas to indicate the dipole feature of the change in c.


x10-7 Pa.m-1

Figure 10. a) Climatological time-latitude plot of wind stress (vectors, Pa) and wind stress curl(contours, 10−7Pa · m−1) in piControl. b) Difference with the mid-Holocene.All fields are averagedbetween 150◦E-120◦W .


20˚S

0˚

20˚N (a) EP ENSO

90˚E 120˚E 150˚E 180˚ 150˚W 120˚W 90˚W

20˚S

0˚

20˚N (b) CP ENSO

−1.0−0.8−0.6−0.4−0.2

0.00.20.40.60.81.0

SS

T (

K)

20˚S

0˚

20˚N (c) EP ENSO

90˚E 120˚E 150˚E 180˚ 150˚W 120˚W 90˚W

20˚S

0˚

20˚N (d) CP ENSO

−2.0−1.0−0.5−0.2−0.1

0.00.10.20.51.02.0

prec

ipita

tion

(mm

/day

)20˚S

0˚

20˚N (e) EP ENSO

90˚E 120˚E 150˚E 180˚ 150˚W 120˚W 90˚W

20˚S

0˚

20˚N (f) CP ENSO

−0.25−0.20−0.15−0.10−0.05

0.000.050.100.150.200.25

SS

S(p

su)

Figure 11. Observed sea surface temperature (SST) (a,b), rainfall (c,d), and sea surface salinity(SSS) (e,f) anomalies associated with the Central (CP) and Eastern Pacific (EP) flavors of ENSO.These patterns are the regression of the SST, rainfall, and SSS anomalies on the E- and C-indecesrespectively. Stippling indicates regressions that are not statistically significant (p < 0.33). Note thatthe precipitation color scale is not linear. Observations covering the 1979-2009 period were used. SSTdata are from HadISST [Rayner et al., 2003], rainfall data are from GCPCv2 [Adler et al., 2003], andSSS is Delcroix’s gridded observational dataset [Delcroix et al., 2011]. Circles indicate the location ofproxies of mid-Holocene ENSO variability.


Figure 12. Schematic of the proposed mechanism for the ENSO response to mid-Holocene

forcing. The term ”anomalous” for the winds and the Kelvin wave are with respect to the mean

seasonal pre-industrial climate.


Figure A1. Climatological mean monthly Surface Downwelling Clear-Sky Shortwave Radiation

during boreal summer/fall (JASO) and boreal winter/spring (FMAM) and in piControl (left

panel), as well as the change in the mid-Holocene (right panel). Zonal deviations are due to the

presence of water vapor.


The response of ENSO flavors to mid-Holocene climate: Implications for proxy interpretation

Documents