Processes Controlling the Mean Tropical Pacific Precipitation Pattern. Part I:
The Andes and the Eastern Pacific ITCZ
KEN TAKAHASHI AND DAVID S. BATTISTI
Department of Atmospheric Sciences, University of Washington, Seattle, Washington
(Manuscript received 12 May 2006, in final form 6 October 2006)
ABSTRACT
The question of why the intertropical convergence zone (ITCZ) is generally north of the equator in the
tropical Pacific is addressed. Experiments with an atmospheric general circulation model coupled to ide-
alized representations of the ocean show that the presence of the Andes is enough to lower sea surface
temperature (SST) off the west coast of South America through evaporation, thus promoting a north–south
asymmetry, with the ITCZ north of the equator, which is amplified by interactions between the ocean and
the atmosphere. The evaporative cooling results mainly from the subsidence of low specific humidity air,
which is due in turn to the mechanical effect of the Andes on the zonal mean flow. The positive feedback
from low-level clouds on SST is an important factor for the efficiency of the mechanism described.
West of 120°W, the presence of the Rockies and Himalayas produces a comparable forcing to that of the
Andes, but this is not enough to reverse or neutralize the north–south asymmetry set by the Andes. It is
shown that the longitudinal offset between the forcings in both hemispheres allows the Andes to prefer-
entially set the north–south asymmetry, which propagates westward into the rest of the Pacific.
Asymmetry in the observed ocean heat transports (more heat transport convergence in the Northern
Hemisphere) associated with the Kuroshio was found to reinforce the effect of the Andes, although it is not
a strong forcing by itself. Sensitivity experiments indicate that the north–south asymmetry of the ITCZ
caused (evaporatively) by the Andes is robust to the presence of a strong equatorial cold tongue and to
seasonality in insolation.
1. Introduction
A striking characteristic of the mean climatic state of
the tropical Pacific is that the intertropical convergence
zone (ITCZ) is generally located to the north of the
equator. The ITCZ is associated with an underlying sea
surface temperature (SST) that is higher than that in
the Southern Hemisphere (SH) at the same latitudes
(Fig. 1). An explanation for this north–south asymme-
try cannot be found in the external forcing of the earth
system, that is, insolation, since it is essentially symmet-
ric about the equator, with the exception of a slight bias
toward the SH, which should produce the opposite
asymmetry from what is observed. Therefore, asymme-
tries within the system must be responsible for the
asymmetry in the longitudinal position of the Pacific
ITCZ. The fluid dynamics of the ocean and atmosphere
do not predict a preference for either hemisphere, so
the northern location of the ITCZ can ultimately be
explained only by the geography of the continents and
oceans.
An important piece of the puzzle was provided by
Xie and Philander (1994, hereafter XP94), who de-
scribed a mechanism by which air–sea interactions in a
simplified model allowed stable equilibrium states with
the ITCZ in one hemisphere and colder water in the
other. This so-called wind–evaporation–SST (WES)
mechanism assumes that convection occurs preferen-
tially over warm water and that, if it is initially stronger
in one hemisphere, it would then drive stronger surface
wind speeds in the other hemisphere, which would cool
the water there by evaporation, reinforcing the initial
location of the convection. That this asymmetry is a
stable state of their unforced system (and that the sym-
metric state is unstable) indicates that the forcing re-
quired to select either of the two possible asymmetric
configurations need not be very strong. Wang and
Wang (1999), using a somewhat more complex model,
found that the symmetric state in their model was stable
Corresponding author address: Ken Takahashi, NOAA/
Geophysical Fluid Dynamics Laboratory, Princeton, NJ 08540.
E-mail: ken.takahashi@noaa.gov
3434 J O U R N A L O F C L I M A T E VOLUME 20
DOI: 10.1175/JCLI4198.1
© 2007 American Meteorological Society
JCLI4198
under equinoxial insolation. When the seasonal cycle of
insolation was included in their model, the two asym-
metric states, similar to those of XP94, became the
stable states.
Philander et al. (1996, hereafter P96) addressed the
question of what causes the Pacific climate to be domi-
nated by a single ITCZ that is located north of the
equator. Using general circulation models (GCMs),
they concluded that the most important geographical
asymmetry was that the orientation of the coastline of
South America is more favorable for coastal upwelling
than the one of North and Central America. This, they
argued, would cool the SH preferentially and drive the
climate system into an asymmetric state with the ITCZ
north of the equator. Interestingly, their coupled
ocean–atmosphere model produced only a weak asym-
metry and they argued that the feedback from low-level
clouds, ubiquitous in the southeastern Pacific (Mitchell
and Wallace 1992; Klein and Hartmann 1993, hereafter
KH93) but absent from their model, might be a crucial
ingredient that could enhance the asymmetry. A diffi-
culty with the P96 hypothesis is the mismatch of scales:
coastal upwelling has a spatial scale of about 50 km,
whereas the observed low SST region has a much larger
scale, as discussed below.
A different result was obtained by Giese and Carton
(1994), who also ran a coupled GCM under very similar
conditions but obtained a stronger north–south asym-
metry in the eastern Pacific than P96, even though the
coastal SST off South America in their model was
among the highest in the Pacific, indicating a serious
misrepresentation of coastal cooling processes, such as
upwelling. Their model also had deficiencies in low-
level clouds, but a fundamental difference from the P96
study was that they considered realistic topography,
whereas P96 had flat land only. Altogether, the above
results suggest that a process involving mountains, dif-
ferent from coastal upwelling, might be acting to create
the observed north–south asymmetry.
An examination of the climatological SST distribu-
tion off the coast of South America (Fig. 1) reveals two
distinct cool regions: a narrow coastal one off Peru
(coolest around 15°S) directly associated with coastal
upwelling (e.g., Takahashi 2005), and a broad cold re-
gion extending from the central coast of Chile (around
30°S) toward the northwest, and to about 3000 km off-
shore. We hypothesize that the latter results from the
presence of the South Pacific subtropical anticyclone,
which cools the ocean surface through evaporation, ef-
fectively suppressing deep convection in the Southern
FIG. 1. (top) Climatological mean rainfall rate (shaded; mm day
1; data from the Global
Precipitation Climatology Project; Huffman et al. 1997) and SST (contours; °C; data from NCEP
OI SST v2; Reynolds et al. 2002). (bottom) Climatological mean 925-mb streamlines and mean
sea level pressure (shaded; mb) from NCEP–NCAR reanalysis (Kalnay et al. 1996; Kistler et al.
2001). Land elevation greater than 1000 m is lightly shaded and bounded by a thin contour.
15 JULY 2007 T A K A H A S H I A N D B A T T I S T I 3435
Hemisphere and leading to the northerly location of the
ITCZ. As argued by Rodwell and Hoskins (2001; here-
after RH01), the Andes, to a large extent, force the
subtropical southeastern Pacific anticyclone into place.
However, farther to the west a similar, although
weaker, cool region can be seen extending to the south-
west from the coast of California (Fig. 1). A crucial
asymmetry is that there is no counterpart to this system
in the Northern Hemisphere (NH) at the same longi-
tudes and this allows the north–south asymmetry to de-
velop in the eastern basin and propagate westward (Xie
1996), where the forcing by the NH anticyclone, which
is “downstream” of the SH anticyclone, is unable to
reverse it.
In this paper, we will test our hypotheses within the
context of a numerical model, and the processes re-
sponsible for the north–south asymmetry will be inves-
tigated. The following section will describe the numeri-
cal models used. Section 3 will assess the effect of the
presence of the Andes on local SST and in driving the
north–south asymmetry. In section 4, the effects of the
Rockies and Himalaya, as well as of the asymmetry in
ocean heat transport, will be assessed. Section 5 pre-
sents additional experiments to test the importance of
low-level cloud feedback and the robustness of the ef-
fect of the Andes to the presence of a strong equatorial
cold tongue and seasonality. Finally, further discussion
of the results and conclusions will be presented.
2. Models
a. Atmosphere
The model used is based on a primitive-equation dy-
namical core with T63 spectral truncation and seven
vertical levels and a suite of simplified physical param-
eterization schemes optimized for this vertical resolu-
tion (Molteni 2003). The climatology of a T30 version
of this model, forced with observed SST, is reasonably
realistic in the tropical Pacific, comparable to many
state-of-the-art general circulation models (Molteni
2003).1
A limitation of the model that is potentially impor-
tant is the misrepresentation of low stratiform clouds in
subtropical subsidence regions, which play a crucial
role in mean tropical climate (KH93; Mechoso et al.
1995). Therefore, a scheme for the diagnosis of the
cover associated with these clouds was implemented
based on the linear relation between cloud cover (CC)
and low-level stability found by KH93 (see the appendix).
The only orography used in the experiments was that
associated with the Andes, the Rockies, and the Hima-
laya (Fig. 2). The Andes required special considerations
due to their narrowness, which, at the T63 truncation,
produced a strong Gibbs effect. This had the conse-
quence of producing large undulations in the surface
height over the ocean, with amplitudes of up to 500 m,
which significantly affected the surface heat fluxes. To
ameliorate this effect, the Andes were widened by ap-
proximately 50% in the zonal direction, while maintain-
ing the height unchanged. This reduced the amplitude
of the undulations to around 100 m, with extreme val-
ues of around 
200 m (Fig. 2). This procedure should
not affect the representation of interaction between the
Andes and the flow over the southeast Pacific, since the
mechanism is nonlinear and consists mainly on blocking
of the westerly flow (RH01), which is a function of the
height and not the width of the mountains. This is sup-
ported by idealized experiments, reported in more de-
tail in Takahashi and Battisti (2007, hereafter Part II),
which indicate that the subsidence rate is insensitive to
the width of the Andes.
When orography was included, only the surfaces of
the mountain ranges were considered to be land (Fig.
2). On these, the temperature was fixed and given the
value of the reference SST distribution adjusted by a
“surface” lapse rate of 4.8°C km
1, which was esti-
mated from observational data.2 This lapse rate was
considered instead of one representative of the free
troposphere because the way isentropes intersect the
surface appears to be an important factor for the inter-
action of the large-scale flow and the mountains in sub-
sidence regions (RH01). The soil moisture was set to
zero, so the mountains do not act as moisture sources.
Unless otherwise stated, the runs were made with
annual mean insolation symmetrical about the equator.
All coupled runs are 10 yr long and the averages from
the last 6 yr are shown. The runs with fixed SST were 48
months long, and we show the averages from the last 40
months.
An aquaplanet run (AQUA) with no orography and
fixed (reference) SST was performed to calculate both
the Q-fluxes required for the mixed layer model (see
next section) and a reference atmospheric state for the
1 We used T63 instead of the original T30 spectral truncation to
better resolve the Andes and to improve the representation of the
midlatitude westerly flow, which is sensitive to resolution (e.g.,
Held and Phillips 1993).
2 The land temperature climatology is from the Climatic Re-
search Unit (CRU) of the University of East Anglia; the orogra-
phy is from the 5-min gridded elevation (ETOPO5) dataset ob-
tained from the National Geophysical Data Center of the Na-
tional Oceanic and Atmospheric Administration (NOAA).
3436 J O U R N A L O F C L I M A T E VOLUME 20
analysis in section 3c. The resulting zonal mean precipi-
tation from this run features a double ITCZ, with peaks
of around 7 mm day
1 at approximately 7° latitude
(Fig. 3). The surface heat fluxes peak at the equator
with values similar to observational estimates (cf.
dashed and dotted curves in Fig. 3), although the model
produces much stronger fluxes into the tropical ocean
off the equator. This model bias is associated with a
shortwave (SW) bias of around 30 W m
2 at the top of
the atmosphere, which implies a planetary albedo of
around 0.2, two-thirds of the observed. This, in turn,
appears to be associated with biases in cloud albedo,
which are particularly strong in the off-equatorial Tropics.
b. Ocean
The ocean was represented in a highly simplified
manner. Between 40°S and 40°N, either SST was pre-
scribed to match the reference distribution, which con-
sisted of the zonal- and annual-mean SST symmetrized
about the equator (Fig. 3), or an interactive mixed layer
(ML) was used, forced by the net surface heat fluxes
produced by the atmospheric model and prescribed Q-
fluxes that are symmetric about the equator and con-
stant in longitude. Poleward of 40°, SST was prescribed.
The temperature T in the interactive mixed layer is
determined by the heat budget equation:
C
dT
dt
 SW  LW  LHF  SHF  Q, 1
FIG. 2. Topography for the Andes, Rockies, and Himalayas used in the experiments (solid contours
every 1000 m with zero omitted for clarity, and 
100-m contour dashed). The shaded areas are treated
as land.
FIG. 3. Reference SST distribution (solid line; °C), and zonal
mean precipitation (shaded; mm day
1) and net surface heat flux
(dashed line; W m
2) from the AQUA run. Also, an estimate of
the zonal average surface heat flux for the Pacific Ocean based on
data from NCEP–NCAR reanalysis (Trenberth and Caron 2001)
assuming a basin width of 140° longitude (dotted).
15 JULY 2007 T A K A H A S H I A N D B A T T I S T I 3437
where SW and LW are the net surface shortwave and
longwave (LW) radiative fluxes, respectively, LHF and
SHF are the net surface latent and sensible heat fluxes
(LHFs and SHFs), respectively, and C is the heat ca-
pacity of the mixed layer, taken to be 50 m deep.
The Q-fluxes (Q), which act as a proxy for the effect
of ocean heat transports, were determined from the
zonal and temporal averages of the net surface heat flux
from the AQUA run (Fig. 3), symmetrized about the
equator. Since, in equilibrium, the Q-fluxes must bal-
ance this surface heat flux, the former are estimated by
multiplying the latter by 
1.
3. Effect of the Andes
a. Experiment 1: Fixed SST
To assess the direct effect of the Andes on the other-
wise equatorially and zonally symmetric atmospheric
flow seen in the AQUA run, we run the model with the
same aquaplanet configuration but including the Andes.
In this experiment (ANDES_FIXED) the only orography
is that shown in Fig. 2 over South America and only the
surface of these mountains was considered to be land.
SST was prescribed to match the reference distribution,
so ocean–atmosphere interactions are not present.
The resulting distribution of low-level horizontal and
vertical motion around the Andes features the five cen-
ters of vertical motion described by RH01 as the signa-
ture of nonlinear adiabatic interaction between the
zonal mean flow and mountains (cf. our Fig. 4, left, and
Fig. 12 in RH01), except for an additional region of
strong ascent on the eastern flank of the Andes north of
20°S in our model (Fig. 4, left), which is associated with
deep convection. The distribution of 850-mb vertical
motion is remarkably similar to the vertical motion
field in the National Centers for Environmental Predic-
tion–National Center for Atmospheric Research
(NCEP–NCAR) reanalysis and 40-yr European Centre
for Medium-Range Weather Forecasts (ECMWF) Re-
Analysis (ERA-40; Fig. 4, center and right). The hori-
zontal flow is also very similar, although the equator-
ward flow is weaker in our model (Fig. 4). According to
the theory of RH01, the subtropical subsidence is es-
sentially adiabatic and is a result of the blocking of the
westerly flow by the Andes, which is then deflected
equatorward along the downward-sloping isentropes
(RH01). However, in the Tropics, radiative cooling
plays an important role in compensating for the warm-
ing due to subsidence. Following RH01, the vertical and
horizontal advective tendencies of potential tempera-
ture at 600 mb from our results are shown in Fig. 5;
evident in these figures is the presence of nearly adia-
batic descent poleward of 20°S (where vertical and
horizontal advection cancel one another) and diabatic
descent equatorward of that latitude (where horizontal
advection is unimportant).
Within the oceanic subsidence region, surface cool-
ing is enhanced by around 30 W m
2 in the region 20°–
10°S, 90°–80°W (Fig. 6, left), of which about 80% is due
to the enhancement of the latent heat fluxes (Fig. 6,
right).
FIG. 4. 850-mb pressure velocity (Pa s
1, shaded) and horizontal wind (m s
1) from the (left) ANDES_FIXED
run, (middle) NCEP–NCAR reanalysis, and (right) ERA-40. The 850-mb surface isobar is dashed. Spatial smooth-
ing was applied to the vertical motion field in the left panel.
3438 J O U R N A L O F C L I M A T E VOLUME 20
The increase in evaporative cooling appears to be the
result of the combination of the drying of the boundary
layer (Fig. 6, left) and the enhancement of surface
wind speed3 (Fig. 6, right). It is of particular interest to
determine whether the drying is due to advection of
drier air from aloft or from higher latitudes. In Fig. 7 we
show the advective tendencies of specific humidity
calculated at 887.5 mb. Between 20° and 5°S, where the
descent is diabatic, the dominant term is vertical ad-
vection of dry air. On the other hand, poleward of
20°S, both terms are of similar magnitude, albeit small.
These results suggest that most of the drying in the
boundary layer is due to the enhancement in sub-
sidence rate.
Interestingly, despite the drying effect of the Andes
on the low-level conditions in the southeast Pacific, the
southern branch of the ITCZ was enhanced in this ex-
periment (Fig. 8), presumably due to the enhancement
in moisture influx associated with the strengthening of
the southerly flow due to the Andes. Also, the influence
of the Andes on the rainfall distribution extends west-
ward only to 120°W. However, as shown in the next
section, coupling to the ocean changes this dramati-
cally.
b. Experiment 2: Interactive mixed layer
The perturbation in surface heat fluxes associated
with the presence of the Andes will produce changes in
the surface temperature of the ocean, which in turn will
affect the atmospheric circulation. In particular, it is
expected that the ocean west of the Andes will cool,
leading to a suppression of convection in this region
and to a north–south asymmetry with the ITCZ in the
NH. To test this idea, we performed an experiment
(ANDES), similar to the previous one, except that the
atmosphere interacts with the ML ocean model instead
of fixed SST.
The results exhibit a strong north–south asymmetry
in precipitation throughout the Pacific. In the east Pa-
cific, east of the date line, a single ITCZ is located north
of the equator (Fig. 9, top), which is significantly more
intense than those in the AQUA case. Farther to the
west, a weaker ITCZ is also found in the SH. Interest-
ingly, the results feature a realistically located South
Pacific convergence zone (SPCZ), although it is nar-
rower and less horizontally tilted than the observed (cf.
Figs. 1 and 9, top). The processes controlling the SPCZ
and the geometry of the dry zone to the east of it are
addressed in Part II.
Associated with this equatorial asymmetry, SST is
reduced in the southeast Pacific with respect to the ref-
erence state by as much as 3.5 K, with the cooling ex-
tending into the equatorial region, producing a weak
equatorial cold tongue (Fig. 9, top).
3 The trade winds are very steady (e.g., Riehl 1979), so the
magnitude of the mean vector wind provides an approximate
measure of wind speed.
FIG. 5. Difference in (left) vertical and (right) horizontal advective tendencies of po-
tential temperature (10
5 K s
1) at 600 mb between the ANDES_FIXED and the AQUA
runs. Note the reversal in shading scheme. Spatial smoothing was applied to both fields.
15 JULY 2007 T A K A H A S H I A N D B A T T I S T I 3439
The eastern flank of the South Pacific anticyclone
is realistic (cf. Figs. 1 and 9, bottom), and the equa-
torward flow is strengthened by the presence of
the mountains (by 4 m s
1 within approximately
1000 km of the coast), with respect to the AQUA run.
The low-level flow features a northward component
on the equator throughout the Tropics, which is pre-
sumably a consequence of the westward propagation of
the north–south asymmetry (Xie 1996; Xie and Saito
2001).
FIG. 6. Difference in (left) 925-mb specific humidity (g kg
1; shaded) and net surface
heat flux (W m
2; contours) and (right) magnitude of the surface wind vector (m s
1;
shaded) and surface latent heat flux (W m
2; contours) between the ANDES_FIXED and
the AQUA runs. Spatial smoothing was applied to all fields.
FIG. 7. (left) Vertical and (right) horizontal advective tendencies of specific humidity
(10
5 g kg
1 s
1) at 887.5 mb in the ANDES_FIXED run. Spatial smoothing was applied
to both fields.
3440 J O U R N A L O F C L I M A T E VOLUME 20
c. Surface heat flux analysis
To diagnose the mechanism through which the pres-
ence of the Andes forces the north–south asymmetry in
the eastern Pacific, we will start by comparing the in-
dividual components of the surface heat flux. Table 1
presents the averaged SST and surface fluxes from the
ANDES run averaged over four regions. The SHoff box
(20°–10°S, 90°–80°W; see Fig. 9 for location of the
boxes) corresponds to the maximum surface cooling
and the NHoff box covers the same latitudinal and lon-
gitudinal range but in the opposite hemisphere (10°–
20°N, 90°–80°W). The NHeq box corresponds to the
region occupied by the ITCZ in the NH (5°–10°N,
140°–100°W), while the SHeq box covers the same lati-
tudes and longitudes in the opposite hemisphere (10°–
5°S, 140°–100°W).
The area averages of the SST and surface heat flux
components, shown in Table 1 as departures from the
corresponding zonal mean and hemispherically symme-
trized fluxes from the AQUA run, are discussed next.4
It can be seen that there is a general cooling in the
eastern Pacific, but the changes in the SH SST are
larger than in the NH. In general, the changes in the
components of the surface heat flux are small. In the
off-equatorial box in the NH (NHoff) the changes are
negligible in general, while in the NH ITCZ (NHeq)
4 The small negative residual in the budgets (on the order of 1
W m
2) indicates that equilibrium was nearly, but not fully,
achieved.
FIG. 8. Precipitation (shading; mm day
1) and 925-mb
streamlines in the ANDES_FIXED run.
FIG. 9. (top) Mean rainfall rate (shading; mm day
1) and SST (contours; °C) for the ANDES
run. (bottom) Mean sea level pressure (shading; mb) and 925-mb streamlines for the ANDES
run. In both panels, the intersection of the surface with the 925-mb isobar is indicated by the
dashed contour. The coastlines are included for reference only. The boxes are described in the text.
15 JULY 2007 T A K A H A S H I A N D B A T T I S T I 3441
there seems to an increase in shading from insolation by
convective clouds, which is compensated by reduced
evaporative and longwave cooling.
In the SHeq box, the reduction in convective clouds
leads to an increase in insolation, which is mostly com-
pensated by increased longwave cooling and, to a lesser
extent, by evaporation. However, changes in net long-
wave cooling at the surface in the Tropics act as a posi-
tive feedback through changes in water vapor (Hart-
mann and Michelsen 1993), so it is not obvious what
drives the cooling in the first place.
The question of what drives the cooling is particu-
larly relevant to the off-equatorial box (SHoff), where
the strongest cooling is found. The changes in heat
fluxes are relatively small and the balance is mainly
between decreased insolation and reduced evaporative
cooling. Naively interpreted, this might suggest that the
presence of stratus clouds is responsible for the cooling.
However, these clouds act mainly as a feedback on SST
rather than as a forcing (e.g., Philander et al. 1996;
Takahashi 2005). As we will show next, despite its small
change in equilibrium, it is evaporation that produces
the reduction in SST.
d. Diagnosis of the forcing of SST
To assess the relative importance of different pro-
cesses in the determination of the surface cooling in the
SH due to the introduction of the Andes, we will con-
sider the changes in the equilibrium energy budget of
the ocean mixed layer:

SW  
LW  
LHF  
SHF  0, 2
where s indicate differences between the run with
Andes (ANDES) and without (AQUA). Note that, by
construction, Q  0.
As an approximation, we write the changes in radia-
tive fluxes as a forcing plus a feedback on SST changes:

SW  
LW  
F  
T, 3
where F is an externally imposed radiative forcing
(e.g., the impact of the Andes on the downward long-
wave forcing at the surface by reducing free tropo-
sphere moisture) and  is a radiative feedback param-
eter, which includes the effects of low-level clouds and the
water vapor feedback. As we will see, the actual value
of  is of no consequence to the following analysis.
We use the following bulk formulas5 for representing
latent and sensible heat fluxes:
LHF  
LCeUqs 
 q 4
SHF  
cpCsUT 
 Ta, 5
where T stands for SST, Ta for surface air temperature,
Ce and Cs are the exchange coefficients, U is the surface
wind speed, q is the near-surface specific humidity, and
qs is the near-surface saturation specific humidity. For
our analysis we will assume L	Ce, cpCs, and 
 to be
constants.
Substituting into (2), we get

F 
T
 
LCeUqs 
 q 
 
cpCsUT
 Ta  0.
6
As we saw in the run with fixed SST (section 3a), the
presence of the Andes changes the low-level circulation
near the surface, which implies changes in near-surface
wind speed, as well as in temperature and specific hu-
midity of the air masses advected into the Tropics. Al-
though there is a strong influence from the ocean on air
temperature and specific humidity through sensible
heat flux and evaporation, respectively, the changes in
these variables are ultimately determined by the
changes in the atmospheric circulation. Therefore, we
will consider the changes in U, q, and Ta, together with
F, as manifestations of the external “forcings” (land
or mountains).
Assuming small changes, we make a linear approxi-
mation of (6):

F  
T 
 LCe
UrqsTr
T 
 
q qsr 
 qr
U 
 cpCsUr
T 
 
Ta  Tr 
 Tar
U  0,
7which can be rewritten using (4) and (5) as

T  r
 
UUr  11  Br 
qqsr 
 qr  Br1  Br 
TaTr 
 Tar 
 
F1  BrLHFr, 8
5 The linear dependence of our bulk formulas on wind speed,
instead of quadratic, is consistent with the formulation of the
atmospheric model.
TABLE 1. Area-averaged deviations in SST (°C) and net surface
SWs, LWs, LHFs, and SHFs (W m
2) for the ANDES minus
AQUA run. Fluxes are positive into the ocean. See the text for
the location of the boxes.
Box SST SW LW LHF SHF
NHoff 
0.2 1.3 
1.4 0.7 
1.4
NHeq 
1.3 
14.7 7.2 6.9 0.3
SHeq 
2.6 11.4 
8.2 
3.7 0.0
SHoff 
3.3 
7.2 
2.5 6.0 2.9
3442 J O U R N A L O F C L I M A T E VOLUME 20
where
r  1  Br LHFr  1qsr 
 qr qsTr  BrTr 
 Tar

1
9
and B  SHF/LHF is the Bowen ratio. The subscript r
indicates the reference state (corresponding to the
AQUA run). Equation (8) enables us to estimate the
change in SST (T ) associated with changes in the at-
mospheric variables ultimately due to the presence of
the mountains. Although estimating the actual value of
T would require determining r, at this point we are
only interested in the relative importance of the pro-
cesses represented by U, q, and Ta in changing SST
through the latent and sensible heat fluxes and the di-
rect effect of F, which can be assessed by comparing
the four terms in the parentheses on the rhs of (8) (U
term, q term, Ta term, and F term, respectively).
The values of these terms calculated over the same
SH boxes as for Table 1, are shown in Table 2. We
roughly estimated the F term, corresponding to long-
wave radiative forcing associated with the changes in
upper-tropospheric humidity produced by the presence
of the Andes, by taking F 
 LW from Table 1, while
SW is assumed to participate only as a feedback.6
In the SHoff box, which experiences the most direct
forcing from the mountains, the dominant process is, by
far, drying (q term), which contributes 92% of the cool-
ing (Table 2). This indicates that, in our experiment, the
means by which the presence of the Andes, and the
resulting subtropical anticyclone, cools the ocean is by
drying the tropical near-surface boundary layer. It
should be noted that these results do not imply that
low-level clouds are not important for the cooling near
the continent: since they act as a feedback, their con-
tribution is through the term involving  in r [defined
in (9)].
In the SHeq box, where the SH ITCZ was obliterated
by the presence of the Andes, the cooling is produced
mainly by the increase in wind speed (56% of the cool-
ing), but also to a significant extent by the drying effect
(37%). XP94 argued that the changes in wind speed are
the dominant factor in setting up the north–south asym-
metry, but our results suggest that the effects of changes
in specific humidity are also important.
The simplest model for how the SST is changed in the
SHoff box, based on the previous results, is one in which
evaporative cooling associated with dry air advection
cannot be balanced, in equilibrium, except by the low-
ering of SST, which decreases the evaporation rate:

LHF  
LCeU
qs 
 
q 
 0, 10
which, in the linear approximation, leads to

T  qsT
1
q. 11
Using the Clausius–Clapeyron equation with a tem-
perature of 300 K and a pressure of 1000 mb, we obtain
T 
 0.8q, where T is in Kelvins and q is in g kg
1. In
fact, we can see an excellent spatial correlation between
the T and q fields (Fig. 10), and linear regression
analysis for the region 35°S–0°, 120°–80°W for the
model yields T 
 0.7q, which is close to what is
obtained from (11).
It is worth noting that, if we consider separately the
two  terms in (10) and take qs  q  3 g kg

1 (cf.
Fig. 10), the magnitude of each term is on the order of
100 W m
2. Therefore, the net change in latent heat
fluxes shown in Table 1 results from the cancellation
6 In the uncoupled run (ANDES_FIXED), LW is somewhat
stronger in the SHoff box (
6 W m

2), which indicates that it
participates in a negative feedback to the ocean cooling. However,
the results are qualitatively unchanged and the effect on T re-
mains small. In the SHeq box we do not expect a direct forcing by
the Andes, so our estimate for the F term provides an upper
bound.
FIG. 10. Difference in SST (shaded; K) and 925-mb specific
humidity (contours, g kg
1) between the ANDES and the AQUA
runs.
TABLE 2. Area-averaged relative contributions to SST changes,
through latent and sensible heat fluxes, by changes in surface wind
speed (U term), in specific humidity (q term), in surface air tem-
perature (Ta term), and by changes in radiative fluxes (F term).
Box U term q term Ta term F term
SHeq 
1.3 
0.9 
0.1 
0.1
SHoff 0.1 
1.4 
0.2 
0.0
15 JULY 2007 T A K A H A S H I A N D B A T T I S T I 3443
between two large terms and does not reflect the im-
portance of evaporation in producing the changes in
SST.
The above considerations can be summarized in a
more intuitive way by considering the adjustment pro-
cess to the instantaneous introduction of the Andes.
Dry air from above is forced to subside and is entrained
into the boundary layer over the ocean to the west of
the Andes, thereby strongly enhancing surface evapo-
rative cooling. As the upper ocean cools, the efficiency
with which it does so decreases with the saturation va-
por pressure according to the Clausius–Clapeyron
equation (the strongest negative feedback), although,
as indicated next, this reduction is significantly offset by
the increase in low-level clouds, which blocks insolation
and, thus, allows the cooling to proceed further. When
equilibrium is reached, however, the net change in
evaporation is small because the initial increase in
evaporation is tempered by the reduction in the surface
saturation vapor pressure (due to lower SST).
Note that if we include the positive cloud feedback in
the equilibrium surface energy budget equation, then
(10) becomes

LHF  
T 
 0. 12
Hence, though the Andes drives the subsidence that
dries and cools the ocean surface through evaporation,
(12) predicts that the net effect is a decrease in the
evaporation rate (LHF  0), consistent with what we
see in Table 1.
The diagnostic relation (11) does not provide infor-
mation on the importance of the low-cloud feedback,
since both q and T are affected by it. An approxi-
mate estimate of the magnitude of this feedback can be
made from the regression coefficients in Figs. 3 and 13
in KH93 by assuming that low-troposphere stability is
controlled by SST alone and the cloud radiative forcing
is approximately the same at the surface as at the top of
the atmosphere. This yields the estimate  
 6.6 W m
2
K
1, which—for typical values of the other negative
feedbacks in (9)—would double r (e.g., Takahashi
2006). Thus, this feedback can lead to twice the mag-
nitudes of both q and T. The importance of low-
cloud feedback is also highlighted in the experiment
reported in section 3a, in which the feedback is re-
moved.
Assuming (10) to be true, we can make an estimate of
the cooling (T) we would expect from the changes in
surface latent heat fluxes that resulted from the un-
coupled (ANDES_FIXED) run (Fig. 6, right). This
cooling would correspond to the case in which nothing
but T is allowed to change (i.e., no reorganization of the
atmospheric flow). Since by construction T does not
change in the uncoupled run, using (4) we can write

LHF  
LCeU
q, 13
because changes in U are of secondary importance.
Placing (13) into (11), we obtain

T  LCeU qsT
1
LHF. 14
From Fig. 6 (right), we can estimate LHF to be 
30
W m
2. Using typical values for the other parameters
on the rhs of 14, we obtain  
 
1 . This value is a
third of what the coupled run (ANDES) produces,
which again points to the importance of ocean–
atmosphere feedbacks for amplifying the response.
4. Northern Hemisphere forcing
a. Rockies and Himalaya
Although the influence of the Rocky and Himalaya
mountain ranges on upper-tropospheric flow, in par-
ticular their role on forcing stationary eddies, has been
extensively studied (see Held et al. 2002, for a review),
their influence on the near-surface conditions, which
control air–sea interactions in the tropical Pacific, is not
obvious.
To assess the influence of the presence of the Rocky
and Himalaya mountain ranges on the asymmetry in
the tropical Pacific, an experiment [Andes–Rockies–
Himalayas (ARH)] was carried out in which these
ranges were included in addition to the Andes, as de-
scribed in section 2a (see Fig. 2).
Adding the Rockies and Himalaya results in the in-
tensification of the North Pacific subtropical anticy-
clone. This simulation has two anticyclonic center re-
sults (Fig. 11, bottom), whereas observations indicate a
single, weaker, anticyclonic center on the eastern side
of the basin (Fig. 1, bottom).
The stronger anticyclonic circulation has a cooling
effect on local SST in the northeast Pacific that is simi-
lar to that produced by the Andes in the southeast Pa-
cific (ANDES run). However, despite the apparent
overestimation of this effect, it is not strong enough to
reverse the north–south asymmetry forced by the
Andes in the Pacific, and the ITCZ remains predomi-
nantly in the NH east of the date line, although it shifts
to the SH farther to the west (cf. Figs. 9 and 11).
We propose that the forcing in the NH is ineffective
in altering the north–south asymmetry because of the
longitudinal offset between the forced surface cooling
in both hemispheres. This hypothesis will be elaborated
on and tested in the following section.
3444 J O U R N A L O F C L I M A T E VOLUME 20
b. Longitudinal offset
The spatial distribution of orography around the Pa-
cific is such that, east of 120°W, the anticyclonic circu-
lation in the North Pacific does not directly affect the
near-equatorial Pacific, so the asymmetry in this region
is set only by the Andes, whereas, farther to the west, as
the north–south asymmetry propagates westward (Xie
1996), the orography in both hemispheres has a direct
influence on the asymmetry. We argue that, assuming
the forcings in both hemispheres have the same mag-
nitude, the asymmetry will be dominated by the forcing
farthest to the east (“upstream”).
To test this hypothesis, an experiment was carried
out (2ANDES) in which the Andes (as represented in
the ANDES experiment but cropped at 5°S) were mir-
rored about the equator into the NH, to produce equa-
torially symmetric topography with Andes in both
hemispheres, and then the NH Andes were longitudi-
nally shifted by approximately 40° to the west, roughly
to the longitude of the Rockies. This experiment was
designed to produce a forcing in the NH, a proxy for the
Rockies and Himalayas, but eliminating the interhemi-
spheric differences in the shapes of the mountains (i.e.,
the relative “strengths” of the forcings).
The results show that the NH Andes have a similar
local effect over the ocean to the west as their SH coun-
terpart, enhancing the equatorward flow and lowering
the underlying SST by approximately the same
amounts near 20° latitude (Fig. 12, top and bottom).
However, in the Pacific the ITCZ remains north of the
equator and the near-equatorial (within 20° latitude)
SST remains lower, south of the equator (Fig. 12, top),
in support of our hypothesis. Therefore, the fact that
the Andes are situated to the east of the Rockies prob-
ably plays an important role in the maintenance of the
ITCZ north of the equator west of 120°W.
c. Ocean heat transport
A detailed inspection of the observed zonally inte-
grated convergence of the meridional ocean heat trans-
port depicted in Fig. 3 shows some asymmetry with
respect to the equator. Between 15° and 45° latitude,
there is an average of 16 W m
2 less energy going into
the ocean in the NH than in the SH, which is a conse-
quence of evaporative and sensible heat fluxes due to
airflow over the Kuroshio off the east coast of Asia.
This asymmetry can be considered a result of continen-
tal geometry.
To assess whether the heat transport asymmetry as-
sociated with the Kuroshio can force a north–south
asymmetry comparably to the Andes, an experiment
was carried out in which the asymmetry in the observed
oceanic meridional heat transport convergence (Fig. 3)
was added to the Q-fluxes between 15° and 45° latitude
in both hemispheres (warming in NH, cooling in SH) at
all longitudes. The orographic setup was similar to that
FIG. 11. Same as in Fig. 9, but for the ARH run.
15 JULY 2007 T A K A H A S H I A N D B A T T I S T I 3445
of the ARH experiment, but the Andes were removed.
In this way, we can test whether the Kuroshio forcing is
comparable to that due to the Andes.
The results (not shown) indicate that the ocean forc-
ing can significantly reduce the strength of the forcing
due to the Rockies and the Himalaya, although it is not
enough to prevent the ITCZ from remaining in the SH.
Therefore, this asymmetry in ocean heat transport ap-
pears to be a contributing factor to the tropical north–
south asymmetry in the tropical Pacific, but of second-
ary importance relative to the Andes.
5. Other experiments
a. Importance of low-cloud feedback
To assess the importance of the low-cloud feedback
in the north–south asymmetry driven by the ANDES,
we performed an experiment (ARHNOSTRATUS)
with the same setup as ARH but without the low-cloud
parameterization (i.e., with the original cloud scheme
of SPEEDY only). The Q-fluxes were recalculated, as
in section 2b, from an aquaplanet run similar to AQUA
but without the low-cloud parameterization scheme.
The results are quite different. The forcing by the
Andes is not able to drive the asymmetry in the east
Pacific. In fact, the forcing from the NH orography,
particularly the Rockies, creates NH subtropical highs
and northeasterly trades that are even stronger than in
the ARH experiment. As a result, the ITCZ in the
central Pacific is now in the SH (Fig. 13). Yet, the forc-
ing by the Andes is able to maintain a double ITCZ
structure in the east Pacific (out to 120°W), whereas a
single ITCZ is found south of the equator elsewhere
(Fig. 13). Although, admittedly, the NH forcing is un-
realistically strong in this model, this result highlights
the need to adequately depict the low-cloud feedback,
in agreement with current views on the subject (e.g.,
Mechoso et al. 1995).
b. Equatorial cold tongue and the east–west
asymmetry
The experiments with realistic mountain forcings fea-
ture a strong north–south asymmetry and an east–west
contrast, albeit with a zonal SST gradient that is too
weak in the equatorial Pacific (2 K) compared to ob-
servations (5 K). This is most likely the result of the
absence of equatorial ocean dynamics in our model,
which is essential for coupled ocean–atmosphere pro-
cesses that strongly contribute to the observed zonal
asymmetry (Dijkstra and Neelin 1995).
The presence of a dynamically induced equatorial
cold tongue over the eastern Pacific would presumably
favor the occurrence of deep atmospheric convection in
the equatorial west Pacific, which is something the
FIG. 12. Same as in Fig. 9, but for the 2ANDES run.
3446 J O U R N A L O F C L I M A T E VOLUME 20
aforementioned experiments lacked. On the other
hand, the presence of a stronger cold tongue can have
a symmetrizing effect in the eastern Pacific by driving
low-level convergence south of the equator (Lietke et
al. 2001; Gu et al. 2005).
To test the effect of a more realistic east–west asym-
metry with a stronger equatorial cold tongue, a run
(ARH  CT) was carried out with the same configu-
ration as the ARH run, but with a perturbation added
to the Q-fluxes as a proxy for the effect of the ther-
mocline tilt on equatorial upwelling. This perturbation
was calculated by fitting a sine function in longitude
between 120°E and 80°W to the annual mean surface
net heat fluxes from the Southampton Oceanography
Centre (SOC) dataset (Josey et al. 1999) within 10° of
the equator, and then symmetrizing the result with re-
spect to the equator. The resulting perturbation has a
peak-to-peak amplitude of 64 W m
2 at the equator
(Fig. 14).
The results feature a well-defined east Pacific cold
tongue (Fig. 15), although considerably farther to the
west than observed (cf. Fig. 1). The total east–west dif-
ference in SST is around 4 K, which is closer than in the
ARH run to the observed. Furthermore, the results fea-
ture deep convection on the equator between 120° and
160°E, which was absent in the ARH run, although the
western Pacific warm pool is smaller than observed.
The modeled equatorial cold tongue is strong enough
FIG. 14. Observed annual mean Q-flux (
1  SOC net surface heat flux; W m
2;
shaded) and the Q-flux perturbation used in the ARHCT run (contours; interval 5 W
m
2; zero omitted, negative dashed).
FIG. 13. Same as in Fig. 9, but for the ARHNOSTRATUS run.
15 JULY 2007 T A K A H A S H I A N D B A T T I S T I 3447
to produce southward SST gradients between the equa-
tor and 5°S east of 120°W (Fig. 15). Such gradients
cause low-level convergence south of the equator
through thermally induced pressure gradients (Lindzen
and Nigam 1987), which could lead to an ITCZ in the
SH, as mentioned previously. Indeed, the 925-mb di-
vergence is reduced by around 5  10
7 s
1 (20%) near
5°S between 130° and 90°W, but not enough to promote
deep convection. However, it is possible that a stronger
cold tongue or a convective parameterization scheme
that is more sensitive to low-level moisture conver-
gence could lead to a double ITCZ configuration.
c. Seasonal cycle
Giese and Carton (1994) showed that the seasonal
cycle of insolation could have a profound effect on the
mean state if the seasonality in SST in the SH were
strong enough to reverse the north–south asymmetry
during the warm season, which would lead to an annual
mean state featuring a double ITCZ.
To test the robustness to seasonality of the north–
south asymmetry forced by the Andes, a run (ARH 
CT  SEAS) was performed with the same configura-
tion as the ARH  CT experiment, except that the
insolation was allowed to vary seasonally, with zero
orbital eccentricity. By construction, the annual mean
incident insolation at the top of the atmosphere was the
same as in the previous experiments.
The resulting mean state from this run (not shown) is
similar to the one from the ARH  CT run. The mod-
eled seasonal cycle was similar to the observed, albeit
weaker (not shown). In the southeast Pacific, this weak-
ness might be due to an underestimation of the low-
cloud/SST feedback in the model (Takahashi 2005).
The north–south asymmetry is essentially unaffected
by the introduction of the seasonal cycle. Both the
mean ITCZ and SPCZ are displaced slightly toward the
poles, and the eastern Pacific is cooler (by around 1 K).
The rainfall rate in the ITCZ is less than in the ARH 
CT run, due to seasonal meridional displacements of
the eastern Pacific ITCZ of around 5° latitude, and a
weakening during March and April, when it is closest to
the equator. Some rainfall does occur south of the
equator (around 4°S) between March and May, but it is
smaller than that in the NH at that time and the annual
mean rainfall distribution does not feature a double
ITCZ.
d. Equatorial ITCZ
The establishment of the north–south asymmetry in
our experiments can be described as the eradication, in
the eastern Pacific, of the southern branch of the
double ITCZ found in the reference state. To test
whether our results depend on having this double ITCZ
configuration in the reference state, we performed two
experiments. The first experiment (AQUA_EQ) is a
FIG. 15. Same as in Fig. 9, but for the ARHCT run.
3448 J O U R N A L O F C L I M A T E VOLUME 20
prescribed SST aquaplanet run with the same setting as
the AQUA run, except that a perturbation was added
to the reference SST to remove the equatorial mini-
mum (i.e., the cold tongue) and make it a maximum.
This resulted in a zonal mean climate with a strong
single ITCZ located on the equator (not shown). The
results from the AQUA_EQ run were then used to
determine the Q-fluxes for a coupled experiment
(ARH_EQ), which is otherwise the same as the ARH
run (i.e., includes the Andes, the Rockies, and the
Himalayas). The results feature a significant asymmetry
in the eastern Pacific out to 160°W (not shown), albeit
unsurprisingly weaker than in the ARH run. In particu-
lar, the eastern Pacific ITCZ is located around 3°N,
while the southeastern Pacific is cooled by around 2°C
(the respective numbers for the ARH run were about
6°N and 3.5°C). This experiment shows that the Andes
are sufficient to establish the north–south asymmetry in
the eastern Pacific from a reference state without an
equatorial cold tongue and with a single equatorial
ITCZ, even in the presence of the Rockies and Hima-
layas.
6. Discussion
The evaporative cooling associated with large-scale
subsidence appears to play a major role in producing
the relatively low temperatures seen off the western
coast of South America, causing the ITCZ to reside in
the NH. In the tropical Atlantic and Indian Oceans,
mountain forcing is probably secondary in importance
to land–sea thermal contrasts in determining the north–
south asymmetries. However, the ocean cooling by the
dry subsidence associated with subtropical anticyclones
is likely to be at work.
Given that our model has a crude representation of
boundary layer processes, one might question whether
the drying mechanism, which depends on the entrain-
ment of subsiding air into the boundary layer, can effi-
ciently occur in nature. The observationally based esti-
mates of entrainment rates at the boundary layer top in
the stratus region in the southeast Pacific are around
3–4 mm s
1, or about 0.04 Pa s
1 (Wood and Brether-
ton 2004; Caldwell et al. 2005); in equilibrium, the en-
trainment rate must be balanced by the subsidence rate,
which in our model is 0.05 Pa s
1. Hence, we believe
that our model is adequately representing the impor-
tance of this process.
The WES mechanism proposed by Xie and Philander
(1994), by which air–sea interactions reinforce an asym-
metric state with the ITCZ in one hemisphere, is a
useful concept for explaining our results. However, ad-
ditional aquaplanet experiments indicate that the asym-
metric (symmetric) state in our model is not as stable
(unstable) as in theirs. This is partly the consequence of
cloud radiative effects, namely shortwave shading by
convective clouds, which in our model (and presumably
in nature) vary in opposition to changes in evaporative
cooling. Thus, our results suggest that the WES mecha-
nism (Xie and Philander 1994) might be more fruitfully
considered as a feedback that enhances the response to
asymmetric forcing, such as that by the Andes, and not
as a strong symmetry-breaking mechanism in its own
right.
Our results are not inconsistent with other recent
studies dealing with the effect of orography, particu-
larly the Andes, on the eastern Pacific climate. Kitoh
(2002) examined the contributions of the (global) orog-
raphy to the climate simulated using the MRI flux-
corrected coupled ocean–atmosphere GCM, and his re-
sults show a response in the tropical Pacific to the pres-
ence of mountains that is consistent with ours, although
the flux correction precluded the appreciation of the
full impact of the removal of the Andes in their experi-
ment. A similar issue arises with the study by Xu et al.
(2004), who removed the Andes in an experiment with
a regional atmospheric model with prescribed SST and
argued that the presence of the Andes favored low-
level clouds in the southeast Pacific by blocking warm
advection from the east. Although their results are con-
sistent with our interpretation, their experimental de-
sign (fixed SST that already included the subsidence-
induced cooling) did not enable them to identify the
most important effect of the Andes.
It is interesting to note that the presence of the South
American landmass, even in the absence of the Andes,
can lead to a north–south asymmetry, albeit much
weaker than that produced by the Andes, due to the
dryness of the easterly flow entering the southeast Pa-
cific. This was observed in an experiment at T30 reso-
lution with idealized landmasses and the interactive
ocean mixed layer. We believe that this process, rather
than coastal upwelling, could have been responsible for
the (weak) asymmetry found by Philander et al. (1996)
in their experiments.
The annual cycle in upper-ocean heat content in the
southeast tropical Pacific, which is closely related to
SST (Takahashi 2005), is dominated by the seasonality
of insolation, which is strongly enhanced by a feedback
involving low-level clouds (Mitchell and Wallace 1992;
Takahashi 2005). The similarity between the spatial
pattern of the annual mean reduction of SST due to the
Andes (Fig. 10) and the pattern of the climatological
May to August transition in SST (Fig. 7 in Mitchell and
Wallace 1992) is probably not a coincidence, as the
low-cloud feedback that enhances the annual cycle of
SST will be more effective where the annual mean SST
15 JULY 2007 T A K A H A S H I A N D B A T T I S T I 3449
is low because of the Andes. This is because low SST
implies high low-level stability, which favors the pres-
ence of the low-level clouds (KH93).
7. Summary and conclusions
The hypothesis that the distribution of orography de-
termines the north–south asymmetry in the tropical Pa-
cific through atmosphere-only processes was tested by
performing experiments with an atmospheric general
circulation model coupled to an idealized model of the
ocean with no built-in asymmetries, either in the east–
west or the north–south directions.
We showed that subsidence of low specific humidity
air to the west of the Andes exerts a strong local control
on SST through evaporation. The presence of the
Andes is found to be sufficient for generating such a
circulation, in agreement with Rodwell and Hoskins
(2001), and the associated surface cooling is sufficient
to create a north–south asymmetry, with the ITCZ in
the Northern Hemisphere in the east Pacific, through
feedbacks involving air–sea interactions (WES feed-
back; Xie and Philander 1994).
Furthermore, we showed that despite the existence
of similar processes in the North Pacific due to anticy-
clonic flow forced by the Himalayas and Rockies, the
fact that the Andes are located farther to the east than
any of these ranges allows the Andes to set the north–
south asymmetry, which then propagates westward into
the Pacific (Xie 1996).
The north–south asymmetry forced by the Andes ap-
pears to be robust to the presence of an equatorial cold
tongue and to the seasonality in insolation. It was also
found that north–south asymmetries in ocean heat
transport (associated with the Kuroshio) are of second-
ary importance, but that can contribute to reinforcing
the asymmetry.
Some simple calculations and a numerical experi-
ment indicate that the feedback involving low-cloud
cover and SST (Mitchell and Wallace 1992; Philander et
al. 1996) plays a significant role in enhancing the cool-
ing off South America driven by the Andes and allow-
ing the associated north–south asymmetry to dominate
in the Pacific. It should be noted that processes other
than those addressed in this paper, such as upwelling
along the coast of South America, feedbacks between
low-level clouds and meridional winds (Nigam 1997;
Wang et al. 2005), and summertime convection in the
Amazon (Rodwell and Hoskins 2001), may play a sec-
ondary role in reinforcing the asymmetry.
Current state-of-the-art coupled ocean–atmosphere
general circulation models still have problems repre-
senting the observed north–south asymmetry (e.g.,
Mechoso et al. 1995). Despite different physical param-
eterization schemes (e.g., convective and boundary
layer) and different representations of the ocean pro-
cesses, the physical processes by which the presence of
the Andes cools the southeast tropical Pacific are likely
to be robust and reproducible in other models as long
as the interaction between the zonal mean flow and the
Andes and the humidity of the subsiding air are ad-
equately represented. It should be noted, however, that
the preparation of the orography used in this study em-
phasized the preservation of the height of the Andes to
ensure that the nonlinear interaction with the zonal
mean flow (Rodwell and Hoskins 2001) was well rep-
resented, which is not a common practice in climate
modeling. We will pursue collaborations with other
modeling groups to assess the reproducibility and ro-
bustness of these results.
Acknowledgments. This study is part of the doctoral
dissertation of KT. The authors thank C. S. Bretherton,
T. P. Mitchell, J. M. Wallace, R. Wood, and three
anonymous reviewers for useful comments and sugges-
tions.
APPENDIX
Low-Level Cloud Scheme
Low-level stability is diagnosed similarly to KH93:

    0.685 
 surface  5 K,
where  is potential temperature and the addition of 5
K is a correction for the low vertical resolution of the
model, while   p/ps, where ps is surface pressure, is
the vertical coordinate of the model.
Stratiform CC is placed over the ocean where  
14 K, where there is subsidence at   0.685, and where
the diagnosed stratus cloud cover would be higher than
that predicted by the original cloud scheme. In that
case, stratus clouds would be placed between   0.77
and   0.60, with CC (in %) calculated following the
linear relation found by KH93:
CC  5.7
 
 55.73.
This diagnosed cloud cover is used only for the short-
wave calculations and does not directly affect the mois-
ture distribution.
REFERENCES
Caldwell, P., C. S. Bretherton, and R. Wood, 2005: Mixed-layer
budget analysis of the diurnal cycle of entrainment in south-
east Pacific stratocumulus. J. Atmos. Sci., 62, 3775–3791.
Dijkstra, H. A., and J. D. Neelin, 1995: Ocean–atmosphere inter-
3450 J O U R N A L O F C L I M A T E VOLUME 20
action and the tropical climatology. Part II: Why the Pacific
cold tongue is in the east. J. Climate, 8, 1343–1359.
Giese, B. S., and J. A. Carton, 1994: The seasonal cycle in a
coupled ocean–atmosphere model. J. Climate, 7, 1208–1217.
Gu, G., R. F. Adler, and A. H. Sobel, 2005: The eastern Pacific
ITCZ during the boreal spring. J. Atmos. Sci., 62, 1157–1174.
Hartmann, D. L., and M. L. Michelsen, 1993: Large-scale effects
on the regulation of tropical sea surface temperature. J. Cli-
mate, 6, 2049–2062.
Held, I. M., and P. J. Phillips, 1993: Sensitivity of the eddy mo-
mentum flux to meridional resolution in atmospheric GCMs.
J. Climate, 6, 499–507.
——, M. Ting, and H. Wang, 2002: Northern winter stationary
waves: Theory and modeling. J. Climate, 15, 2125–2144.
Huffman, G. J., and Coauthors, 1997: The Global Precipitation
Climatology Project (GPCP) combined precipitation dataset.
Bull. Amer. Meteor. Soc., 78, 5–20.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re-
analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Kistler, R., and Coauthors, 2001: The NCEP–NCAR 50-Year Re-
analysis: Monthly means CD-ROM and documentation. Bull.
Amer. Meteor. Soc., 82, 247–267.
Kitoh, A., 2002: Effects of large-scale mountains on surface cli-
mate—A coupled ocean-atmosphere general circulation
model study. J. Meteor. Soc. Japan, 80, 1165–1181.
Klein, S. A., and D. L. Hartmann, 1993: The seasonal cycle of low
stratiform clouds. J. Climate, 6, 1587–1606.
Josey, S. A., E. C. Kent, and P. K. Taylor, 1999: The Southampton
Oceanography Centre (SOC) Ocean–Atmosphere Heat, Mo-
mentum and Freshwater Flux Atlas. Southampton Oceanog-
raphy Centre, Southampton, United Kingdom, 30 pp. plus
figures.
Lietzke, C. E., C. Deser, and T. H. Vonder Haar, 2001: Evolu-
tionary structure of the eastern Pacific double ITCZ based on
satellite moisture profile retrievals. J. Climate, 14, 743–751.
Lindzen, R. S., and S. Nigam, 1987: On the role of sea surface
temperature gradients in forcing low-level winds and conver-
gence in the tropics. J. Atmos. Sci., 44, 2418–2436.
Mechoso, C. R., and Coauthors, 1995: The seasonal cycle over the
tropical Pacific in coupled ocean–atmosphere general circu-
lation models. Mon. Wea. Rev., 123, 2825–2838.
Mitchell, T. P., and J. M. Wallace, 1992: The annual cycle in equa-
torial convection and sea surface temperature. J. Climate, 5,
1140–1156.
Molteni, F., 2003: Atmospheric simulations using a GCM with
simplified physical parameterizations. I: Model climatology
and variability in multi-decadal experiments. Climate Dyn.,
20, 175–191.
Nigam, S., 1997: The annual warm to cold phase transition in the
eastern equatorial Pacific: Diagnosis of the role of stratus
cloud-top cooling. J. Climate, 10, 2447–2467.
Philander, S. G. H., D. Gu, D. Halpern, G. Lambert, N.-C. Lau, T.
Li, and R. C. Pacanowski, 1996: Why the ITCZ is mostly
north of the equator. J. Climate, 9, 2958–2972.
Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W.
Wang, 2002: An improved in situ and satellite SST analysis
for climate. J. Climate, 15, 1609–1625.
Riehl, H., 1979: Climate and Weather in the Tropics. Academic
Press, 611 pp.
Rodwell, M. J., and B. J. Hoskins, 2001: Subtropical anticyclones
and summer monsoons. J. Climate, 14, 3192–3211.
Takahashi, K., 2005: The annual cycle in the Peru Current region.
J. Climate, 18, 4937–4954.
——, 2006: Processes controlling the tropical Pacific mean pre-
cipitation pattern. Ph.D. dissertation, University of Washing-
ton, 135 pp.
——, and D. S. Battisti, 2007: Processes controlling the mean
tropical Pacific precipitation pattern. Part II: The SPCZ and
the Southeast Pacific Dry Zone. J. Climate, in press.
Trenberth, K. E., and J. M. Caron, 2001: Estimates of meridional
atmosphere and ocean heat transports. J. Climate, 14, 3433–
3443.
Wang, B., and Y. Wang, 1999: Dynamics of the ITCZ–equatorial
cold tongue complex and causes of the climate asymmetry. J.
Climate, 12, 1830–1847.
Wang, Y., S.-P. Xie, B. Wang, and H. Xu, 2005: Large-scale at-
mospheric forcing by southeast Pacific boundary layer
clouds: A regional model study. J. Climate, 18, 934–951.
Wood, R., and C. S. Bretherton, 2004: Boundary layer depth, en-
trainment and decoupling in the cloud-capped subtropical
and tropical marine boundary layer. J. Climate, 17, 3576–
3588.
Xie, S.-P., 1996: Westward propagation of latitudinal asymmetry
in a coupled ocean–atmosphere model. J. Atmos. Sci., 53,
3236–3250.
——, and S. G. H. Philander, 1994: A coupled ocean-atmosphere
model of relevance to the ITCZ in the eastern Pacific. Tellus,
46A, 340–350.
——, and K. Saito, 2001: Formation and variability of a northerly
ITCZ in a hybrid coupled AGCM: Continental forcing and
oceanic–atmospheric feedback. J. Climate, 14, 1262–1276.
Xu, H., Y. Wang, and S.-P. Xie, 2004: Effect of the Andes on
eastern Pacific climate: A regional atmospheric model study.
J. Climate, 17, 589–602.
15 JULY 2007 T A K A H A S H I A N D B A T T I S T I 3451