Zonal structure and seasonal variability of the Atlantic Equatorial Undercurrent

Zonal structure and seasonal variability of the Atlantic EquatorialUndercurrent

W. E. Johns • P. Brandt • B. Bourles •

A. Tantet • A. Papapostolou • A. Houk

Received: 2 August 2013 / Accepted: 2 April 2014

� Springer-Verlag Berlin Heidelberg 2014

Abstract Simultaneous mooring arrays were maintained

along the path of the Equatorial Undercurrent (EUC) at

three longitudes (23�W, 10�W, and 0�E), from October

2007 to June 2011, as part of the CLIVAR Tropical

Atlantic Climate Experiment. The measurements allow for

the first time a description of the seasonal cycle and

interannual variability of the EUC across the Atlantic

basin. The mean transport of the EUC at 23�W is

14.3 ± 0.6 Sv, decreasing to 12.1 ± 0.9 and 9.4 ± 0.6 Sv

at 10�W and 0�E, respectively. The EUC shows a changing

seasonal cycle across the basin: at 23�W, the strongest

EUC transport occurs in boreal fall in association with

maximum easterly wind stress, at 10�W the EUC transport

shows a semiannual cycle with a maximum in boreal spring

and fall, while at 0�E the EUC has a single spring maxi-

mum. At all locations the EUC core exhibits a similar

seasonal vertical migration, with shallowest core depths

occurring in boreal spring and deepest core depths in boreal

fall. The maximum core intensity occurs in boreal spring

all across the basin, when the EUC is shallow, during the

annual wind relaxation. The weakest EUC core intensity

occurs during the boreal summer cold tongue phase,

especially in the eastern part of the basin. At both 23�W

and 10�W, a deep extension of the EUC occurs in boreal

summer, which increases the transport in the lower ther-

mocline and partially offsets the weaker upper EUC

transport during boreal summer. No clear linkage could be

established between the interannual variability of the EUC

in the eastern part of the basin and the intensity of the

summer cold tongue, despite evidence for such a linkage in

the western part of the basin.

1 Introduction

The Equatorial Undercurrent (EUC) is a quasi-permanent

feature of the zonal equatorial circulation in both the

Atlantic and Pacific oceans. Its main role in both oceans is

to supply thermocline waters from the shallow subduction

zones in the subtropics to the main upwelling zones in the

central and eastern part of the equatorial basins (Schott

et al. 1998). In these so-called ‘‘subtropical cells’’ (STCs;

McCreary and Lu 1994), surface waters that are subducted

near 20�N and S move westward and equatorward, even-

tually reaching the equator near the western boundary

where they supply the EUC with salty waters that are

carried eastward by the EUC in a narrow ribbon across the

This paper is a contribution to the special issue on tropical Atlantic

variability and coupled model climate biases that have been the focus

of the recently completed Tropical Atlantic Climate Experiment

(TACE), an international CLIVAR program (http://www.clivar.org/

organization/atlantic/tace). This special issue is coordinated by Wil-

liam Johns, Peter Brandt, and Ping Chang, representatives of the

TACE Observations and TACE Modeling and Synthesis working

groups.

W. E. Johns (&) � A. Papapostolou � A. Houk

Division of Meteorology and Physical Oceanography, Rosenstiel

School of Marine and Atmospheric Science, University of

Miami, 4600 Rickenbacker Causeway, Miami, FL 33149, USA

e-mail: [email protected]

P. Brandt

GEOMAR Helmholtz-Zentrum fur Ozeanforschung Kiel,

Kiel, Germany

B. Bourles

IRD/LEGOS, Brest, France

A. Tantet

Institute for Marine and Atmospheric Research,

Utrecht University, Utrecht, The Netherlands

123

Clim Dyn

DOI 10.1007/s00382-014-2136-2

http://www.clivar.org/organization/atlantic/tace




entire width of the basin (Zhang et al. 2003; Hazeleger

et al. 2003). The EUC therefore provides both cool and

salty waters to the central and eastern equatorial regions

that affect both the heat and salt budget of the mixed layer

through upwelling and vertical mixing processes (Jouanno

et al. 2011; Kolodziejczyk et al. 2013). In the Atlantic

ocean, the symmetry of the STCs is broken by the mean

northward upper ocean flow of the Atlantic Meridional

Overturning Circulation, which results in the southern

hemisphere thermocline water being the dominant source

for the EUC (Fratantoni et al. 2000; Malanotte-Rizzoli

et al. 2000; Zhang et al. 2003).

Coupled climate models currently show severe biases in

sea surface temperature (SST) in the eastern tropical

Atlantic, which have been linked mainly to westerly wind

biases and corresponding ocean–atmosphere feedbacks

(Richter and Xie 2008). Equatorial easterlies are consis-

tently too weak in the models, resulting in too little

upwelling, an excessively deep thermocline in the central

and eastern parts of the basin, and a corresponding warm

SST bias (Davey et al. 2002; deWitt 2005). The EUC also

tends to be poorly reproduced in these models (Chang et al.

2007), being either much too weak or penetrating only a

short distance eastward into the basin from the western

boundary. As a result, the strong vertical shear that exists in

the real ocean between the westward South Equatorial

Current (SEC) at the surface and the underlying EUC,

which drives strong vertical mixing and surface cooling in

the central and eastern equatorial regions, is not present in

the models, which could further contribute to the warm

SST bias.

While in the Pacific Ocean the EUC is rather well-

described from over a decade of intensive shipboard and

time-series observations in the TOGA and TAO/TRITON

programs (Johnson et al. 2002), the EUC in the Atlantic has

remained more poorly sampled and neither its mean

structure across the basin or its seasonal-to-interannual

variability is understood. Particularly in the eastern part of

the basin, where the EUC decays and appears to exhibit

strong variability in its eastward penetration, observations

are sparse and very few time series measurements have

been collected. In the Pacific, the variability of the EUC is

closely linked to SST variations in the eastern cold tongue

region on both seasonal and interannual (El-Nino) time

scales, and it is anticipated that similar behavior may occur

in the Atlantic in association with the seasonal cycle and

the analogous ‘‘Atlantic Nino’’ phenomenon.

Despite the more limited measurements of the EUC in

the Atlantic, a substantial increase in the understanding of

the EUC in the western and central Atlantic has developed

over the past decade. At two longitudes, near 35�W and

23�W, a sufficient number of shipboard transects across the

equator have now been acquired to afford reasonable

estimates of the mean EUC structure and transport at these

locations (Schott et al. 2003; Brandt et al. 2006). At 35�W

the EUC transports approximately 20 Sv above the density

surface rh = 26.8, whereas by 23�W this transport is

reduced to approximately 14 Sv. The core of the EUC is at

100 m depth at 35�W and shoals to about 85 m depth at

23�W. At both locations instantaneous core velocities are

typically in the range of 80–100 cm/s.

Farther east, at 10�W, available estimates of the EUC

transport from cross-equatorial sections suggest a mean

value near 12 Sv (Kolodziejczyk et al. 2009), which is

surprisingly similar to the 14 Sv value at 23�W in view of

the expected eastward decay of the EUC in the Atlantic.

Here the EUC has a core depth near 60 m, having shoaled

some 40 m from the western part of the basin.

Farther yet to the east the available measurements are

sparse and estimates of the EUC transport vary widely.

Reported estimates from individual cruises include those of

Mercier et al. (2003) at 7�W (24.6 Sv) and 2�E (12.6 Sv),

Gouriou and Reverdin (1992) at 4�W (10.2 Sv), and

Bourles et al. (2002) at 0�E (6 Sv), At 0�E, Bourles et al.

(2002) found the EUC core at 50 m depth, with maximum

core speeds of only 40 cm/s. More recently, Kolodziejczyk

et al. (2013) found transports at 1�E ranging from 5 to

15 Sv, and at 6�E from 0 to 7 Sv.

Observational and modeling studies of the EUC in both

the Atlantic and Pacific oceans indicate that EUC behavior

results from a complicated mix of local and remote forcing,

and linear and nonlinear dynamics (Philander and Paca-

nowski 1986; Wacongne 1989; Philander and Chao 1991;

Qiao and Weisberg 1997; Yu and McPhaden 1999a; Ke-

enlyside and Kleeman 2002). The driving force for the

EUC is the eastward baroclinic zonal pressure gradient set

up by the forced response of the thermocline to the trade

winds. The main retarding forces on the EUC include

vertical stress due to downward mixing from the overlying

westward SEC, lateral dissipation, and both vertical and

horizontal nonlinear momentum advection. In an equilib-

rium state the EUC should be strongest where the easterly

winds along the equator are strongest and the associated

zonal pressure gradient is a maximum. In the Atlantic the

strongest mean winds occur in the far western part of the

basin whereas in the Pacific they occur in midbasin. Thus

the EUC is expected to be strongest in midbasin in the

Pacific and strongest near the western boundary in the

Atlantic, which is in accord with observations (Johnson

et al. 2002; Brandt et al. 2006).

In the Pacific Ocean, the seasonal cycle of the EUC is

characterized by a springtime maximum (the so-called

‘‘springtime surge’’; Keenlyside and Kleeman 2002) that

occurs in May in the eastern Pacific and progressively later

in the west. The amplitude of the variation is approxi-

mately 15 Sv across the width of the basin. Associated with

W. E. Johns et al.

123

this transport variation is a shoaling of the core of the EUC

that also propagates westward. These variations are rea-

sonably well explained by linear models of the local and

remotely forced Rossby/Kelvin wave response in the

Pacific (Yu and McPhaden 1999b), with some modifica-

tions due to nonlinear effects (Keenlyside and Kleeman

2002).

Considerably less is known about the seasonal cycle of

the EUC transport in the Atlantic. It is known that a sea-

sonal cycle of shoaling and deepening of the EUC occurs

that is similar to the behavior in the Pacific (Giarolla et al.

2005; Brandt et al. 2006). The EUC at 23�W shoals to its

minimum depth in the central Atlantic (23�W) in April

(*60 m), and reaches its maximum depth in about October

(*90 m). A similar cycle occurs at 10�W (Kolodziejczyk

et al. 2009) and also at 35�W, where the EUC routinely

‘‘surfaces’’ in boreal spring when the winds are weak.

However, models suggest a rather different transport cycle

in the Atlantic than the Pacific, characterized by two

maxima—one (the primary maximum) in fall, and another

(weaker) maximum in spring (Philander and Pacanowski

1986; Hormann and Brandt 2007). The transport maximum

in the fall is related to the maximum in easterly wind stress

that occurs in September–October in the far western part of

the basin and the associated response of the zonal pressure

gradient. By contrast, at 10�W, the recent analysis by

Kolodziejczyk et al. (2009) indicates a boreal summer

maximum of the EUC transport, with a minimum occurring

in boreal fall. The exact nature of the seasonal cycle has

been difficult to determine from the available ship-based

sections due to large intraseasonal variability related to

processes including transient wind forcing or Tropical

Instability Waves.

Here, we present new time series measurements of the

EUC collected from moorings deployed at 23�W, 10�W,

and 0� for almost a 4 year period, from 2007 to 2011, and

use them to describe the mean transport and seasonal cycle

of the EUC across the basin. We show that a relatively

simple technique can be used to reconstruct the EUC

transport and vertical structure from a limited set of

moorings at each longitude and produce robust estimates of

its seasonal cycle and variability. The changes in EUC

transport across the basin and their linkage to forcing

mechanisms are also discussed, and preliminary findings

on the interannual EUC variability during the TACE period

are presented.

2 Data and methods

Several different data sets are used in this study to inves-

tigate the zonal structure and variability of the EUC,

including moored Acoustic Doppler Current Profiler

(ADCP) time series, shipboard ADCP and CTD sections,

and temperature and salinity profile observations from

Argo profiling floats and PIRATA (Prediction and

Research Moored Array in the Atlantic; Bourles et al.

2008) moorings. The main analysis is focused on the

moored ADCP observations, while the other observations

are used primarily to validate our methods for estimating

the EUC transport from the moored observations, and to

determine the distribution of the EUC transport in different

density classes.

2.1 Moored ADCP observations

From October 2007 to May 2011, an array of moorings

equipped with upward-looking ADCPs was maintained

along 23�W, 10�W, and 0�E to monitor the temporal var-

iation of the EUC at each longitude (Fig. 1). These

moorings were deployed by different groups as part of a

coordinated program during the 2007–2011 International

CLIVAR Tropical Atlantic Climate Experiment (TACE;

Brandt et al. 2013a, b). The moorings at 23�W were

maintained by GEOMAR (Germany), those at 10�W were

maintained by the University of Miami (US) and by IRD

(Institut de recherche pour le developpement, France, as

part of the PIRATA program), and those at 0�E by the

University of Miami (US) (Fig. 1). The moorings consisted

of either single ADCPs or dual ADCP systems, where the

dual systems were designed to profile a larger extent of the

water column and yield higher vertical resolution near the

surface. At 23�W, dual ADCP systems were used at the

equator, with an upward-looking 150 kHz at about 150 m

and either upward or downward-looking 75 kHz ADCPs

below, while single upward-looking ADCPs of either 150

or 75 kHz were used at the off-equatorial sites. At 0�E,

dual ADCP systems were used consisting of an upward-

looking 300 kHz ADCP at 100 m and an upward-looking

150 kHz ADCP at 380 m depth. At 10�W, the PIRATA

ADCP mooring on the equator was deployed for most of

the period as a single upward-looking 300 kHz ADCP

mounted near 100 m depth (except for the last 8 months of

the experiment, October 2011–May 2012, where a dual

ADCP system similar to that at 0�E was used), and the off-

equatorial moorings were single upward-looking 150 kHz

ADCPs moored at 300 m depth. At each longitude, one

mooring was maintained at the equator, and the other

moorings were located at ±0.75� on either side of the

equator (at 23�W and 10�W), and at 0.75� south of the

equator at 0�E. Vertical profiling resolution was nominally

8 m for the 150 and 75 kHz ADCPs and 4 m for the

300 kHz ADCPs, with sampling configured so that hourly-

averaged velocities had uncertainties B0.02 ms-1. The

vector velocity profiles were subsequently interpolated to a

uniform 5 m vertical spacing at all sites. The profiling

Atlantic Equatorial Undercurrent

123

range typically extended to within 30 m of the surface, and

somewhat closer (*20 m) for the 0�E moorings and the

10�W equatorial mooring. The time series of the zonal

velocity profiles obtained at 10�W and 0�E are shown in

Fig. 2, after low-pass filtering to twice-daily values with a

40 h Butterworth filter; those from 23�W are shown in

Brandt et al. (2013b) where further details on the mooring

configurations and results at 23�W can be found. The

locations and vertical extent of the moored ADCP mea-

surements at each longitude are indicated in Fig. 3.

Full data sets were obtained at both 23�W and 0�E for

the duration of the experiment, but at 10�W there were two

significant gaps. The first occurred at the 10�W, 0.75�N

mooring when the mooring broke loose shortly after its

initial deployment in October 2007, and could not be

reinstalled until fully 1 year later (Fig. 2a). The second gap

occurred at the 10�W, 0�N PIRATA mooring, from

December 2009 to October 2010, due to a failure of the

ADCP. The manner in which these gaps are dealt with in

the subsequent analysis are described in Sect. 2.4.

2.2 Shipboard ADCP and CTD sections

During the last 20 years a large number of cross-equatorial

shipboard ADCP sections have been acquired through vari-

ous national and international programs that have provided

repeated sampling at (or near) four main longitudes: 35�W,

23�W, 10�W, and 0�E. Most of these results have been pre-

viously published: e.g., at 35�W by Schott et al. (2003) and

Brandt et al. (2006); at 23–28�W by Brandt et al. (2006) and

Brandt et al. (2013b); and at 10�W by Kolodziejczyk et al.

(2009). In addition to these sections, we use a similar com-

pilation of sections near 0�E from recent cruises to construct

mean sections of the near-equatorial zonal currents at these

longitudes (Fig. 3). At the three westernmost longitudes there

are at least 15 individual sections that go into these averages

(15 sections at 35�W, 20 sections at 23�W, and 17 sections at

10�W), while at 0�E only eight sections are available. For

each section, near-surface velocities were extrapolated

upward from the shallowest ADCP measurement level (typ-

ically 30 m) using the monthly surface (15 m) drift clima-

tology of Lumpkin and Garraffo (2005) (http://www.aoml.

noaa.gov/phod/dac/drifter_climatology.html), prior to aver-

aging. The seasonal sampling of the sections is relatively

uniform at 35�W and 23�W, but at 10�W and 0�E the sam-

pling is biased toward the boreal summer months, owing to

the annual servicing schedule of the PIRATA moorings in the

Gulf of Guinea, and the focus on the summer cold tongue of

the 2005–2007 EGEE program (Etude de la circulation oc-

eanique et des echanges ocean-atmosphere dans le Golfe de

Guinee; Bourles et al. 2007), on which many of these sections

were acquired. Thus, there could be a significant seasonal bias

in the mean section at 10�W (Kolodziejczyk et al. 2009), and

especially at 0�E where all of the available sections were

acquired between June and September. Nevertheless, these

mean sections provide a baseline description of the zonal

changes of the EUC structure across the basin.

On many of these cruises, CTD stations were also

occupied at a spatial resolution of at least 0.5� between 2�S

and 2�N, which have been used in the above references to

determine the EUC transport in different density classes.

Here, we use a number of these available sections at 23�W

(nine sections) and 10�W (eight sections) to validate

methods for estimating the EUC transport in density clas-

ses from a combination of the moored ADCP measure-

ments and equatorial density profiles derived from Argo

and PIRATA observations.

2.3 PIRATA and Argo data

Surface meteorological buoys with temperature and salin-

ity sensors through the upper water column were

Fig. 1 Locations of the ADCP moorings deployed along 23�W

(German), 10�W (US/France), and 0�E (US), superimposed on the

climatological SST for July (from 2007 to 2011) over the tropical

Atlantic, based on TMI satellite retrievals. The mean velocity vectors

near the core of the EUC (80 m at 23�W, 70 m at 10�W, and 60 m at

0�E) between 2�S and 2�N, derived from the mean of available

shipboard ADCP sections at each longitude (see text and Fig. 3), are

also shown

W. E. Johns et al.

123

http://www.aoml.noaa.gov/phod/dac/drifter_climatology.html

http://www.aoml.noaa.gov/phod/dac/drifter_climatology.html

maintained by the PIRATA program at each of the three

longitudes, 23�W, 10�W, and 0�E, during the period of the

experiment. Data recovery from these sites was generally

good, except for some gaps in subsurface temperate and

salinity data during 2008 and 2010 at the 0�E site. The

surface (1 m) and subsurface temperature and salinity

measurements at various depths on these moorings are used

together with Argo profiling float data to reconstruct

equatorial density profiles at each of the longitudes. The

approach for merging the PIRATA and Argo data is

described in Sect. 3.1.

The Argo data used in this study is taken from the

global monthly analysis produced by the Scripps Institu-

tion of Oceanography (SIO; http://sio-argo.ucsd.edu/RG_

a

b

Fig. 2 a Zonal velocity profiles

from the moored ADCP records

at 10�W for the period of the

observations (bottom 0.75�S;

middle: 0�N, top 0.75�N).

Positive velocities are eastward;

color scale is in m/s. b Zonal

velocity profiles from the

moored ADCP records at 0�E

(bottom 0.75�S; top 0�N)


123

http://sio-argo.ucsd.edu/RG_Climatology.html

Climatology.html), which provides objectively analyzed,

monthly temperature and salinity profiles over the top

2,000 m on a 1� by 1� grid. Details of the analysis method-

ology can be found in Roemmich and Gilson (2009). Similar

global analysis products are available from other groups, and

two of these, from JAMSTEC (http://www.jamstec.go.jp/

ARGO/argo_web/MapQ/Mapdataset_e.html), and IPRC

(http://apdrc.soest.hawaii.edu/projects/Argo/), were also

evaluated in this study. It was found that the Scripps product

had the smallest bias and RMS deviation from the actual

temperature and salinity data measured by the PIRATA

moorings, which led us to adopt this as the preferred Argo

product for this study. The reasons for this are unknown, but it

may be related to the spatially asymmetric (zonally

elongated) decorrelation scales used in the tropics in the

Scripps analysis, that may help to effectively extend the data

coverage. The Argo coverage for the equatorial Atlantic

during the TACE period included, on average, about 14 floats

within ±1� of the equator between 25�W and the African

coast, which corresponds to an average float spacing of

approximately 2� longitude.

2.4 Reconstruction of EUC transport from moorings

The strategy for reconstructing the EUC transport from a

limited set of discrete moorings is based on a relatively

simple approach, in which the zonal transport profile

integrated across the width of the EUC:

Fig. 3 Averaged shipboard ADCP sections across the EUC at 35�W,

23�W, 10�W, and 0�E (see text for details on the number of sections

used at each longitude). The locations of the ADCP moorings at each

longitude and the vertical extent of the moored velocity measure-

ments is indicated by the dashed black lines

W. E. Johns et al.

123

http://sio-argo.ucsd.edu/RG_Climatology.html

http://apdrc.soest.hawaii.edu/projects/Argo/

UðzÞ ¼Zy0

�y0

uðy; zÞdy ð1Þ

is assumed to be represented by

UðzÞ ¼X

Wn � unðzÞ ð2Þ

where un(z) are the zonal velocity profiles measured by the

moorings and Wn are ‘‘optimal widths’’ associated with

each mooring. This corresponds to simply assigning a fixed

width to each mooring that accounts for its respective

contribution to the total EUC transport, similar to an

approach used earlier in the Pacific by Knox and Halpern

(1982). Our decision to place moorings 0.75� north and

south of the equator at 10�W and 23�W was based on an

initial assessment of this approach prior to the deploy-

ments, using the available ADCP sections as a test bed.

This analysis showed that a single equatorial mooring (with

a mean meridional decay scale determined by a fit to the

section transports) could track the variation in the EUC

transport reasonably well, but was subject to errors as large

as 5 Sv (or order 30 % errors) due to meandering of the

EUC core off the equator. Using three moorings spaced

between 0.6� and 1.0� off the equator reduced this error con-

siderably (to about 10 % error). There was little sensitivity to

the choice of the spacing within this range, and therefore a

spacing of 0.758 was adopted for the moored arrays.

To determine the final best values of the optimal widths

(Wn) for the arrays, and the associated uncertainty of the

method, tests were performed using all of the available

shipboard ADCP sections at each longitude. The zonal

velocity profiles at the exact mooring locations were

extracted from each section, and the transport computed

from (2) was compared to the transport derived from the

actual shipboard ADCP sections. In these calculations, only

positive (eastward) zonal velocities are included in the

integrations and sums in (1) and (2), to exclude any con-

tributions from westward flows adjacent to the EUC. A

least squares minimization across all cruises provided the

best fit values for Wn. These Wn’s are then applied to the

actual moored ADCP profiles to estimate the time-varying

EUC transport profile at each longitude. The moored

ADCP profiles were extrapolated upward to the surface

from their shallowest measurement depth (typically

20–30 m) using the same method as applied to the section

data, namely, using the interpolated monthly surface drift

climatology of Lumpkin and Garraffo (2005). We refer to

this method hereafter as the Optimal Width (OW) method.

The Wn’s were estimated using both fixed limits for the

width of the EUC (y0 = ± 1.2�) and variable limits based

on the actual northern and southern limits of the EUC in

each section determined from visual inspection. The results

were very similar for both cases, with the optimal

Wn’s varying by less than 10 %. Therefore the Wn’s based

on the fixed EUC limits of ±1.2� were used. Two other

methods for reconstructing the EUC transport were also

evaluated, one based on a fitting the moored velocity

profiles to complex EOF’s determined from the available

ADCP sections at each longitude, and the other based on a

nonlinear fit of the data to a two-dimensional Gaussian

velocity structure. The advantage of these approaches is

that they can provide additional information on EUC

properties (e.g., EUC core velocity, width, and lateral

position relative to the equator), but the OW method was

found overall to yield the most accurate reconstruction of

the EUC transport (see Brandt et al. (2013b) for a com-

parison of the OW and complex EOF methods applied to

the 23�W observations). One problem with these methods

is that they are more sensitive to incomplete velocity

profile data, such as at 10�W where only a shallow equa-

torial ADCP profile was available, and are difficult to apply

at 0�E where only two moorings were deployed and there

are too few shipboard ADCP sections available to permit a

representative EOF reconstruction. Therefore, we use the

OW method throughout this paper, and restrict our focus to

the EUC bulk transport and vertical structure rather than its

meridional structure or variability.

At 23�W, where velocity profiles spanning the full depth

of the EUC were continuously available at all three lati-

tudes (0.75�N, 0�N, and 0.75�S), the OW approach can be

applied in a straightforward manner and results in optimal

Wn’s of [0.76�, 0.74�, 0.79�] latitude, respectively (see

further discussion in Brandt et al. 2013b). These widths

correspond fairly closely to the physical separation of the

moorings, and are slightly less than the 0.8� widths that

would correspond to even partitioning of the domain

between 1.2�S and 1.2�N, where the bulk of the EUC is

typically found.

At 0�E, there are only two moorings available for the

reconstruction, at the equator and 0.75�S. The OW recon-

struction yields optimal Wn’s of [0.91, 0.99] for the [0�N,

0.75�S] moorings, respectively. Examples of the recon-

struction are shown in Fig. 4 for two representative sec-

tions. In these two cases the upper part of the EUC

transport profile (\200 m) is reproduced well, but the

transport below 200 m is underestimated. The overall

accuracy of the reconstruction EUC transport based on the

available ADCP sections at 0�E is ±1.0 Sv (Fig. 6).

At 10�W, a modified version of the OW method was

required due to the fact that only a shallow equatorial

ADCP profile is available for most of the period, as well as

the data gaps. For the period between October 2007 and

September 2008—when the 0.75�N mooring was miss-

ing—we do not attempt to produce EUC transports,

because the 0.75�S profile and shallow equatorial profile, in

themselves, are not sufficient for a robust reconstruction.


123

For the remainder of the record we make two different

reconstructions: one derived from just the two full ADCP

profiles at 0.75�S and 0.75�N, and a second that uses the

data from all three moorings when the equatorial ADCP

profile is available. This 3-mooring reconstruction is lim-

ited to depths B100 m (from October 2008 to November

2009) and to depths B230 m (from October 2010 to May

2011; Fig. 2a). For these periods the final EUC transport

profile is obtained by combining the upper 3-mooring

transport profile with the 2-mooring reconstructed profile

over the deeper part of the water column (which we refer to

as the ‘‘merged’’ transport profile). For the period from

November 2009 to October 2010, the results rely only on

the 2-mooring reconstruction. The optimal widths for the

3-mooring reconstruction were [0.71 0.81 0.79] for 0.75�N,

0�N, and 0.75�S, respectively, and for the 2-mooring

reconstruction using only the off-equatorial moorings at

0.75�N and 0.75�S they were [1.31 1.39]. The larger

optimal widths for the 2-mooring reconstruction are con-

sistent with expectations, since these moorings will typi-

cally miss the maximum EUC core, and this is also

reflected in the larger net meridional scale (sum of the

Wn’s) for the 2-mooring reconstruction (*2.7� latitude)

versus that for the 3-mooring reconstruction (*2.3�).

Examples of the reconstruction at 10�W are shown in

Fig. 5. In the first case the merged profile captures the

observed transport profile very well, and the 2-mooring

reconstruction only slightly underestimates the transport. In

Fig. 4 Shipboard ADCP sections across the EUC at 0�E in June 2006

(EGEE-3; top) and September 2007 (EGEE-6; bottom), where the

locations of the moored ADCP’s are indicated by dashed black lines.

At right are shown the corresponding EUC transport profiles derived

from the shipboard ADCP sections (blue), the respective transport

contributions from the extracted velocity profile at the equator (green)

and 0.75�S (red) based on the OW method, and their sum (dashed

black line), which represents the approximated total transport profile.

The cumulative transport from the surface downwards is shown at the

far right for the full shipboard ADCP section and the OW

approximation

W. E. Johns et al.

123

the second case, the 2-mooring reconstruction does poorly

through the core of the EUC, while the merged profile

again captures the observed profile well. This is an extreme

case in which the EUC is displaced unusually far south of

the equator, and it is remarkable that the 2-mooring

reconstruction performs as well as it does, having a total

transport error of only *1 Sv. The overall accuracy of the

reconstruction at 10�W is ±0.8 Sv for the merged profile

estimates, and ±2.0 Sv for the 2-mooring estimates

(Fig. 6), where most of the latter uncertainty comes from a

few outlying estimates. A similar analysis performed for

23�W shows an overall accuracy of ±0.65 Sv for the OW

method (Brandt et al. 2013b). Thus, we conclude that the

OW method provides relatively accurate (to *1 Sv)

estimates of the EUC transport and vertical structure,

except for the 1-year period from November 2009 to

October 2010 when the estimates for 10�W may have

slightly larger error.

3 Results

3.1 EUC transport and vertical structure

3.1.1 Shipboard sections

A first view of the changing structure of the EUC across the

basin is provided by the averaged shipboard ADCP

Fig. 5 As in Fig. 4, but for shipboard ADCP sections across the EUC

at 10�W in November 2006 (EGEE-4; top) and September 2005

(EGEE-2; bottom). At right are shown the corresponding EUC

transport profiles from the full shipboard ADCP sections (blue), and

from the 2-mooring OW reconstruction (using velocity profiles only

from 0.75�S to 0.75�N; green), and from the ‘‘merged’’ reconstruction

including additionally the equatorial velocity profile (red) over the

upper 100 m of the water column. The cumulative transport profiles

from the surface downward are compared for each case at the far right


123

sections (Fig. 3). These mean sections have been described

previously in several studies (Schott et al. 2003; Brandt

et al. 2006; Kolodziejczyk et al. 2009), but it is the first

time that they have been shown together in one place. The

mean section at 0�E compiled in this study is a new

addition and is based mostly on sections acquired during

the 2005–2007 French EGEE program in the Gulf of

Guinea, as part of the AMMA program (e.g., Bourles et al.

2007).

Consistent with previous observations, the EUC core

shoals across the basin, from a core depth of about 90 m at

35�W to 65 m at 0�E. The width of the EUC also decreases

across the basin and the mean core velocity decreases

slightly from[0.6 ms-1 to about 0.5 ms-1. These sections

represent geographic (Eulerian) averages and therefore the

peak velocities at the EUC core are considerably weaker

than seen in individual sections, where the velocity max-

ima are typically between 0.8 and 1.0 ms-1. The meridi-

onally-elongated velocity core and much larger overall

width at 35�W is due in part to broadening of the mean

flow distribution at 35�W by significant lateral meandering

of the EUC just after it retroflects eastward from the North

Brazil Undercurrent. In these sections, the EUC core is

found slightly south of the equator at the three easternmost

longitudes, suggesting only a small and relatively uniform

displacement of about 0.2� from the equator, even in the

eastern part of the basin. The strong and coherent zonal

flow pattern associated with the EUC is contained mostly

above 200–250 m, and its lower limit also shoals along

with the velocity core, to depths of about 150 m at 0�E.

3.1.2 Mooring data

The time series of the EUC transport profile constructed

from the moorings using the OW method described in Sect.

2.4 are shown in Fig. 7, and show several notable features.

First, at each longitude, the core of the EUC exhibits a

seasonal vertical migration, being shallowest in boreal

spring months (March–May) and deepest in boreal fall

(September–October). This behavior is most pronounced in

the west (23�W) and generally decreases toward the east.

Associated with this deepening EUC core in fall is a much

deeper extension of the eastward flow below the EUC core,

which is clearly evident at 23�W and 10�W but not clearly

at 0�E. This deeper EUC structure can also be seen in the

individual zonal velocity profile at 10�W, 0.75�S during

2008 (Fig. 2a), even though we do not produce a EUC

transport reconstruction at 10�W for this period. The timing

and duration of this deep extension varies somewhat from

year to year, but it generally emerges in boreal summer

(July–August) and lasts through about the end of October.

During these periods, significant eastward transport

extends to depths of C300 m, while in boreal spring the

eastward EUC flow is confined mostly above 150 m.

A second feature that can be noticed in Fig. 7 is that the

EUC core intensity is generally weakest in boreal summer

(June–August) at all longitudes, which coincides with the

onset and seasonal development of the Atlantic could

tongue. This behavior is more pronounced in the east,

especially at 0�E, and also at 10�W, where it follows a

sustained period of maximum EUC core intensity in boreal

spring (March–May).

In addition to these seasonal changes there is consider-

able short-term variability throughout the records. The

dominant time scales of this variability are generally

between 12 and 60 days, associated with Tropical Insta-

bility Waves and other modes of equatorial variability that

have been previously described (e.g., von Schuckmann

et al. 2008; Athie and Marin 2008; Athie et al. 2009; Perez

et al. 2012). The meridional component of velocity at each

of the longitudes shows relatively high coherence

throughout this band, and is mostly symmetric about the

equator (i.e., in-phase between the equatorial and off-

equatorial sites). Zonal velocity anomalies are mainly out

of phase across the equator on these time scales, which, to

first order, reflects the meandering of the EUC in response

to these meridional velocity perturbations. This out of

phase relationship is difficult to see in Fig. 2 due to the

long time span of the records, but is clearly evident when

the time scale is expanded to more closely examine

Fig. 6 Comparison of the total

EUC transport estimates from

the OW-method reconstruction

with the full EUC section-

derived transports, at 10�W

(left) and 0�E (right), for all of

the available shipboard ADCP

sections at each longitude

W. E. Johns et al.

123

individual events. On longer time scales from 80 days up to

annual, the zonal velocity anomalies become more in-

phase with each other and reflect seasonally-coherent

changes in the intensity and structure of the EUC. It is

noteworthy also that the ADCP records at 0�E—which

represent the first long-term records of the EUC in the

eastern Gulf of Guinea—show that the EUC remains

essentially tied to the equator at this location and that it has

not shifted significantly south of the equator at this location

as is depicted in some models.

The time-mean profiles of the EUC transport at each

longitude are shown in Fig. 8, where it can be seen that the

EUC core shoals progressively to the east, from 75 m at

23�W to 55 m at 0�E. The transport at the EUC core is

largest at 23�W and decreases by about 10 and 30 %,

respectively, at 10�W and 0�E, relative to that at 23�W.

Below the EUC core the transport remains higher at 23�W

than 10�W until about 200 m, while at 0�E the transport

profile shows almost a uniform reduction of *0.01 Sv/m

relative to 10�W. In the region above the EUC core the

transport profiles are nearly identical across all longitudes.

The mean transports for the EUC derived from these

measurements, integrated to 300 m, are 14.3 ± 0.6,

12.1 ± 0.9, and 9.4 ± 0.6 Sv, at 23�W, 10�W, and 0�E,

respectively (Table 1), where the given uncertainties rep-

resent standard errors. These uncertainties are based on the

number of available degrees of freedom in each transport

time series, determined by the length of the time series

divided by twice the integral time scale of the transport

variability (which is approximately 27 days at 23�W,

24 days at 10�W, and 21 days at 0�E). The associated

standard deviations of the transport are 3.2 Sv at 23�W,

4.1 Sv at 10�W, and 3.0 Sv at 0�E. The larger uncertainty

at 10�W is a result of both its larger transport variability

and the shorter length of time series available at that lon-

gitude. If one includes in these error estimates a random

measurement uncertainty of ±1 Sv, then the overall

uncertainties for the above mean transports increase by

only about 5 %, which indicates that the errors in the

transport reconstruction associated with the OW method

add little to the total uncertainty. The uncertainties in the

mean transport at each longitude are therefore essentially

governed by the natural variability of the EUC transport.

To compare these results with previous estimates of the

EUC transport, it is desirable to break the total transport

down into different density classes, rather than a single

depth integrated value. In order to accomplish this, we

utilize the SIO monthly Argo analysis described in Sect.

2.3, together with the available PIRATA moored temper-

ature and salinity data during the measurement period, to

construct density profiles at the equator at each longitude.

The Argo data provide high-resolution vertical profiles of

temperature and salinity, but may occasionally be inaccu-

rate due to sparse regional sampling, whereas the PIRATA

data are from discrete depths and sometimes have temporal

gaps. To obtain the most accurate possible density profiles,

Dep

th (

m)

EUC Transport Profiles (Sv/m)

23oW

N J M M J S N J M M J S N J M M J S N J M M

0

100

200

300 0

0.05

0.1

0.15

0.2

0.25

Dep

th (

m)

10oW


0

100

200

300 0

0.05

0.1

0.15

0.2

0.25

Dep

th (

m)

2008 2009 2010 2011| | | |

0oE


0

100

200

300 0

0.05

0.1

0.15

0.2

0.25

Fig. 7 EUC transport profiles

(Sv/m) derived from the

moorings at 23�W, 10�W, and

0�E. The EUC transport

reconstruction is not attempted

at 10�W for the first year of the

observations, when the 0.75�N,

10�W mooring was missing


123

a simple correction scheme was applied to the monthly

Argo profiles, as follows. At each PIRATA measurement

depth, the differences between the Argo and monthly-mean

PIRATA temperature and salinity data are used to con-

struct an ‘‘error’’ profile for the Argo data, that is linearly

interpolated between the PIRATA measurement depths.

This error profile is then added back into the Argo profiles

to obtain a corrected Argo profile. These corrected tem-

perature and salinity profiles pass through all the measured

PIRATA points, but are otherwise consistent with the

vertical structure of the continuous Argo profiles. The

required corrections to the Argo profiles were small, less

than 1 �C RMS for temperature and 0.2 RMS for salinity,

which indicates that the Argo profiles themselves are

remarkably accurate (even though the SIO analysis does

not utilize any of the PIRATA data). During periods when

no simultaneous PIRATA data were available for this

correction, the mean error (i.e., the mean error profile from

all of the available Argo-PIRATA comparisons at that

longitude) was used instead to correct those Argo profiles.

Monthly density profiles at the equator were then con-

structed at each longitude from these corrected Argo

temperature and salinity profiles. These density profiles are

then used to transform the EUC transport profiles, mea-

sured as a function of depth, into density coordinates.

To verify that this approximation works adequately, the

available shipboard sections were again used as a test bed.

For each section that included both CTD and ADCP

sampling across the equator, the directly integrated trans-

ports in density coordinates using the full sections were

compared against those obtained by multiplying the me-

ridionally-integrated EUC transport profile for the section

with just the equatorial density profile from the CTD sec-

tion (Fig. 9). This comparison shows that there is no sig-

nificant bias from this approximation. Therefore, for

monthly averages, this technique seems appropriate to

derive representative transports in density classes.

The corresponding mean EUC transport profiles in

density coordinates at each longitude are shown in Fig. 8b.

The peak transport occurs near rh = 25.1 at 23�W and

near rh = 25.5 at both 10�W and 0�E. Thus while the EUC

core is physically deeper at 23�W (Fig. 8a), it occurs at a

lighter mean density. The eastward shoaling of the EUC

core in depth space—but trending toward higher densi-

ties—reflects a more rapid shoaling of the main pycnocline

toward the east than the EUC velocity core, which is

analogous to the observed EUC structure in the Pacific

(Johnson et al. 2002). An alternate presentation of the same

results is shown in Fig. 10, where the transports are

accumulated into even density classes of 0.1 kg/m-3, as in

Fig. 9. The maximum EUC transport in this representation

occurs in the density class rh = 26.3 at each longitude,

which reflects the greater thickness of isopycnal layers

0 0.05 0.1 0.15

0

50

100

150

200

250

300

Transport per unit depth (Sv/m)

Dep

th (

m)

23oW

10oW

0oE

0 0.05 0.1 0.15

21

22

23

24

25

26

27

Transport per unit depth (Sv/m)

Pot

entia

l Den

sity

230W

10oW

0oE

Fig. 8 Mean EUC transport

profiles (Sv/m) at 23�W, 10�W,

and 0�E, plotted versus depth

(left) and versus potential

density (rh) (right)

Table 1 Mean EUC transport estimates for 23�W, 10�W, and 0�E,

and associated standard deviations and standard errors (in Sv)

Longitude Mean transport Standard deviation Standard error

23�W 14.3 3.2 0.6

10�W 12.1 4.1 0.9

0�E 9.4 3.0 0.6

W. E. Johns et al.

123

below the core of the EUC. The main differences between

the profiles are above rh = 25.5 where the transports

decrease toward the east. In particular there is significant

transport above rh = 24.5 at 23�W which decreases

markedly by both 10�W and 0�E. Below the transport

maximum at rh = 26.3, the transport at 0�E also decreases

relative to that at 23�W and 10�W.

The transports in four main density classes that have

been used to describe the regional characteristics of the

EUC in previous studies (e.g., Brandt et al. 2006; Kol-

odziejczyk et al. 2009) are listed in Table 2. These classes

correspond to a ‘‘surface’’ layer (rh \ 24.5), an ‘‘upper

thermocline’’ layer (24.5 \rh \ 25.5), a ‘‘lower thermo-

cline’’ layer (25.5 \ rh \ 26.5), and a ‘‘deep thermocline’’

layer (26.5 \ rh \ 26.8), after Kolodziejczyk et al. (2009).

The upper and lower thermocline layers contain the main

contributions to the EUC transport in the regions above and

below the EUC velocity core, respectively, and these are

often grouped together into a single ‘‘thermocline layer’’

EUC transport (e.g., Schott et al. 2003; Brandt et al. 2006).

The transport in the thermocline layer is identical at 23�W

and 10�W (10.2 Sv), but there is a relative decrease

(increase) in the upper (lower) thermocline component at

10�W. At 0�E the thermocline transport is reduced to

8.0 Sv, with both the upper and lower thermocline contri-

butions decreasing from 10�W. The results obtained from

our mooring-based analysis compare very well with the

earlier estimates at 23�W and 10�W derived from ship

sections (Table 2), essentially confirming the results of

Brandt et al. (2006) and Kolodziejczyk et al. (2009) at

these longitudes, and indicating that those estimates did not

Fig. 9 Mean transport in

density classes, at 0.1 kg/m3

intervals, from all of the

available shipboard CTD/ADCP

sections at 23�W (top) and

10�W (bottom), compared with

the approximation derived by

multiplying the meridionally-

integrated transport profile from

the ADCP sections with the

observed density profile at the

equator. At right are shown the

mean bias from this

approximation (heavy black

lines) and the standard deviation

envelope of the transport profile

error for all of the sections

0 0.5 1 1.5 227

26.5

26

25.5

25

24.5

24

23.5

23

22.5

Transport (Sv)

Pot

entia

l Den

sity

23oW

10oW

0oE

Fig. 10 Mean EUC transport profile in density classes, at 0.1 kg/m3

intervals, at 23�W, 10�W, and 0�E, derived from the mooring-based

EUC transport profiles and PIRATA-corrected ARGO density profiles

at each longitude


123

suffer greatly from aliasing of the ship sections by intra-

seasonal variability. The measurements at 0�E, however,

provide fundamentally new information on the EUC

transport and vertical structure in the eastern Gulf of

Guinea, where the available ship sections have been too

limited to construct representative annual estimates.

3.2 Seasonal cycle of the EUC

A climatological seasonal cycle is derived for the EUC

transport profile at each longitude by averaging together all

of the data from the available records by the month of

observation (Fig. 11). These seasonal cycles represent the

climatological average over 3.7 years of data at 23�W and

0�E but only over 2.7 years at 10�W. The features descri-

bed earlier in the time series are clearly evident: (1) a semi-

annual cycle in the intensity of the EUC core transport,

with maxima in boreal spring and fall; (2) an annual ver-

tical migration of the EUC core with shallowest (deepest)

depths in boreal spring (fall); and (3) a deep extension of

the eastward flow beneath the core of the EUC in boreal

summer to early fall at both 23�W and 10�W.

The semi-annual cycle of the core intensity is most

pronounced in the eastern part of the basin, where the two

Dep

th (

m)

2323.5 24

24 2425 25

2525.5 25.525.526 26

2626.2 26.226.2

26.4 26.4 26.4

26.626.6 26.6

26.826.826.8

23oW

J F M A M J J A S O N D

0

100

200

300 0

0.05

0.1

0.15

0.2D

epth

(m

)

2323.5 2424 2425

25

2525.525.5

25.526

26

2626.226.2

26.2

26.426.4

26.4

26.6 26.626.6

26.826.8

10oW


0

100

200

300 0

0.05

0.1

0.15

0.2

Dep

th (

m)

22 2323 2323.523.5 23.524

242425

252525.5

25.5

25.526

26

2626.2

26.2

26.2

26.4

26.4

26.4

26.6 26.6

26.626.826.8

26.8

0oE


0

100

200

300 0

0.05

0.1

0.15

0.2

Fig. 11 Monthly-mean EUC

transport profiles (Sv/m) at

23�W, 10�W, and 0�E. Density

contours derived from the

PIRATA-corrected ARGO

temperature and salinity data,

averaged over the TACE time

period (2007–2011), are shown

in thin white lines

Table 2 Transports (in Sv) in

density classes, from this study

compared with previous

estimates from averaged ship

sections at the various

longitudes (from Brandt et al.

2006 and Kolodziejczyk et al.

2009)

Density class r-range This study Brandt et al. (2006) Kolodziejczyk et al. (2009)

23�W 10�W 0�E 35�W 23�W 10�W

Surface \24.5 3.0 1.1 0.8 5.4 3.1 0.4

Upper thermocline 24.5–25.5 3.1 2.4 1.6 14.6 10.7 2.5

Lower thermocline 25.5–26.5 7.1 7.8 6.4 8.3

Deep thermocline 26.5–26.8 1.1 0.8 0.6 0.9

Total 14.3 12.1 9.4 20.0 13.8 12.1

W. E. Johns et al.

123

maxima are separated by a progressively weakening core

intensity during boreal summer. The spring maximum

tends to be better sustained across the basin while the fall

maximum decays more sharply, especially at 0�E. There-

fore in terms of the absolute EUC core intensity, as well as

its seasonal behavior across the basin, the spring maximum

is considered the primary maximum and the fall maximum

a secondary maximum.

The density contours overlain on Fig. 11 show that the

EUC core is generally found throughout the year near

rh = 25.0–25.5, and its seasonal vertical migration follows

the seasonal displacement of the main thermocline. The

exception to this is in the boreal summer cold tongue phase

when the density surfaces are uplifted to their shallowest

depths of the year and the EUC core is found at greater

density. This again is most pronounced in the east where

during July–August the EUC core occurs at about

rh = 26.0 at 10�W and rh = 26.2 at 0�E.

The deep extension of the EUC during boreal summer

had been previously observed by Kolodziejczyk et al.

(2009) at 10�W, who noted its importance to the overall

seasonal cycle of the EUC. Until now this had been gen-

erally viewed as a feature unique to the region near 10�W,

but it is clear from Fig. 11 that this deep extension also

occurs in at 23�W, although with somewhat less magni-

tude. Brandt et al. (2006) earlier noted that the annual

harmonic of the eastward velocity on the equator at 23�W

showed a maximum in August at 250 m, which is consis-

tent with the behavior seen here. The maximum of this

deep flow extension appears to take place at about the same

time at both 10�W and 23�W, in August–September, but

there are indications that the phase of this feature leads

slightly at 10�W relative to 23�W, with both its maximum

and initial onset occurring about one month earlier at

10�W. At 10�W there is also some suggestion of upward

vertical propagation of this deeper flow, where its onset can

be traced to as early as April or May at depths near 300 m.

Interestingly, there is also a much weaker, but otherwise

similar, deep flow anomaly at 0�E that occurs in boreal

spring, centered in about May. This is not a robust feature

of the seasonal cycle there since it does not occur as sys-

tematically from year to year as at 10�W or 23�W (Fig. 7).

Nevertheless, at all three longitudes there are indications

that a seasonal increase in eastward flow develops below

the main core of the EUC, and that this signal may prop-

agate westward across the basin from boreal spring to

boreal summer.

The seasonal cycle of the total (0–300 m) EUC transport

at each longitude is illustrated in Fig. 12 where it is

superimposed on the transport time series from each

measurement year. At 23�W the transport exhibits a single

broad maximum in September, with otherwise nearly

constant total transport from boreal winter through spring

(December to June). The reverse is true at 0�E, where the

seasonal maximum occurs in boreal spring (April–May),

and the transport is nearly constant from August through

January. In between these longitudes, at 10�W, a semi-

annual cycle is evident, with maxima in September (pri-

mary) and April (secondary), and a relatively sharp mini-

mum in November–December. The amplitude of the

seasonal variability is about 5 Sv (peak-to-peak) at 23�W

and 0�E and about 10 Sv at 10�W. The instantaneous

transports range from about 5 to 25 Sv at 23�W and 10�W,

consistent with the range of observed transports from

individual ship sections at these longitudes (Brandt et al.

2006; Kolodziejczyk et al. 2009). At 0�E the transports

vary from about 5 to 20 Sv.

The seasonal cycles at all three longitudes are overlain

in Fig. 13, where it can be seen that the transports at all

locations are roughly the same during boreal spring,

approximately 12–13 Sv, when the EUC core is relatively

intense and shallowest across the whole basin (Fig. 11).

During boreal fall, the transports at 23�W and 10�W are

again similar, at about 18 Sv, but the transport at 0�E is

nearly 10 Sv weaker at this time. There is some indication

of a secondary maximum in the September monthly mean

transport at 0�E, associated with the secondary fall

J F M A M J J A S O N D0

5

10

15

20

25

Tra

nspo

rt (

Sv)

23oW

2007 2008 2009 2010 2011


5

10

15

20

25

10oW

Tra

nspo

rt (

Sv)


5

10

15

20

25

0oE

Tra

nspo

rt (

Sv)

Fig. 12 EUC transport (0–300 m) at each of the observed longitudes,

for individual years (colors; see legend), with monthly means and

standard deviations shown as black symbols with error bars. The

respective seasonal cycles derived from fits of the data to an annual

plus semi-annual harmonic are shown in the bold solid lines


123

maximum in the upper EUC core at that time, but it is not

clearly present in all years and is not a feature of the sea-

sonal harmonic cycle.

The depth-integrated transport of the EUC does not

reveal the changing vertical structure of the EUC transport,

which is illustrated more clearly in Fig. 14 where the

transports are broken down into density classes. The

transports in the surface and upper thermocline layer all

reach a minimum in boreal summer, and nearly vanish at

0�E during July–August. This reduction is compensated at

both 23�W and 10�W by an increase in the lower ther-

mocline layer transport that begins in June and peaks in

August. Most of this transport occurs in the deeper part of

the thermocline (rh = 26.0–26.5), and there is also a fur-

ther contribution from rh [ 26.5 at this time. At 0�E the

behavior is somewhat different, with the peak in lower

thermocline transport actually occurring in May, but

showing a less pronounced seasonal variation than at 23�W

or 10�W. Thus, during the summer upwelling season, the

transport in the upper portion of the EUC is substantially

reduced at all locations, and this reduction becomes pro-

gressively larger toward the east so that by 0�E almost all

of the transport occurs in the lower thermocline.

3.3 Forcing of the EUC seasonal cycle

Dynamically, these changes in the EUC transport and

vertical structure must be linked to changes in the

momentum balance of the EUC. The zonal momentum

balance along the equator is given by:

ou

otþ u~ � r~u ¼ �1

qop

oxþ 1

qosx

ozþ AHr2u

where the left hand side includes the local rate of change

and nonlinear momentum advection, and terms on the right

hand side represent the zonal pressure gradient, the tur-

bulent vertical stress, and lateral friction, respectively. A

detailed analysis of the EUC momentum balance is not

undertaken in this paper and is left to a future study.

However, some relevant aspects are illustrated in Fig. 15,

where the Cross Calibrated Multi-platform Product

(CCMP) zonal wind stress (Atlas et al. 2011) and AVISO

sea surface height anomaly along the equator are shown for

the period of the TACE measurements.

Climatologically, the seasonal wind pattern in the

Atlantic closely resembles the eastern third or so of the

Pacific (Fig. 15c). During boreal spring the Intertropical

Convergence Zone (ITCZ) migrates northward, bringing

intensifying easterly wind stress to the equator following a

late wintertime lull. This transition occurs first in the east

and progressively later in the west. The seasonal response

of the EUC to these changes should reflect two primary

processes: (1) the variations in locally-forced downward

diffusion of momentum from the surface layer, and (2) the

buildup of the zonal pressure gradient (ZPG) associated

with the zonal redistribution of mass caused by wind-dri-

ven advection and equatorial wave processes. The evolu-

tion in time of the latter depends on the width of the basin


10

15

20

Tra

nspo

rt (

Sv)

23oW

10oW

0oE

Fig. 13 Monthly means (symbols) and seasonal harmonic cycle of

the 0–300 m EUC transport, overlain from Fig. 12 at each of the

longitudes

Fig. 14 Seasonal cycle of the EUC transport in density classes,

shown as cumulative transport from the lowest (shallow) to highest

(deeper) density classes. rh \ 24.5 corresponds to the ‘‘surface’’

layer, rh = 24.5–25.5 to the ‘‘upper thermocline’’ layer,

rh = 25.5–26.5 to the ‘‘lower thermocline’’ layer (split here into

two sub-layers), and rh = 26.5–26.8 to the ‘‘deep thermocline’’,

following Kolodziejczyk et al. (2009). The dots with error bars at the

top of each plot show the total EUC transport and its standard error

for each month

W. E. Johns et al.

123

and detailed pattern of the wind stress forcing across the

basin.

From a linear perspective, the basic understanding of the

EUC dynamics in both the Atlantic and Pacific is that the

seasonal cycle of the ZPG in the surface and upper ther-

mocline is established rapidly by low vertical mode (first

and second mode) Kelvin and first meridional mode

Rossby waves that are forced directly by the winds (Cane

and Sarachik 1981; Yu and McPhaden 1999b; Brandt and

Eden 2005). The response time for the equatorial Atlantic

is on the order of 30 days for the zonal currents in the

upper *100 m, where the response of the zonal currents

on the equator tends to be dominated by the equatorial

Rossby waves (Yu and McPhaden 1999b).

Near 23�W the surface zonal pressure gradient (hereafter

ZPG) begins to increase in May and reaches its maximum

strength in about August (Fig. 15d). The EUC transport

maximum at 23�W occurs in September (Fig. 13), consis-

tent with a slightly lagged response to the acceleration

driven by the ZPG. The ZPG minimum at 23�W occurs in

March–April which is nearly the same time as the zonal

wind stress minimum, and the ZPG is generally weak across

the entire central/eastern part of the basin at this time. It is

remarkable that the EUC velocity core remains relatively

strong across the whole basin during boreal spring (Fig. 11),

and even has its maximum transport at 0�E then. This

suggests that the reduction in downward westerly momen-

tum transport associated with the weak zonal winds in

spring is sufficient to offset the weaker acceleration by the

ZPG and maintain a relatively strong EUC during this

period. The shoaling of the EUC during spring is also

consistent with the reduced penetration of the westward

surface stress, as has been noted in previous studies (Phi-

lander and Pacanowski 1986; Arhan et al. 2006).

An eastward ZPG begins to develop at both 10�W and

0�E in late boreal spring, but this quickly weakens at both

locations in boreal summer as the cold tongue (evidenced

by the low SSH anomaly) spreads westward and the sea

surface becomes almost flat across the whole eastern part

of the basin. The summer minimum in EUC intensity is

probably explained by this summer relaxation of the ZPG

as well as the enhanced westward wind stress, that peaks in

the eastern equatorial region in June. Enhanced vertical

mixing due to increased shear between the EUC and

overlying westward SEC may also lead to more effective

downward transport of westward momentum into the EUC

core in boreal summer (Hummels et al. 2013; Jouanno et al.

2011).

Fig. 15 Left Zonal wind stress along the equator, from CCMP winds

(in N/m2), for the 2007–2011 TACE period (top) and the seasonal

climatology for the same period (bottom). Right AVISO sea surface

height anomaly (cm) along the equator, (top), and seasonal climatol-

ogy (bottom)


123

While the upper layers, in and above the thermocline,

appear to be largely consistent with a forced near-equilib-

rium response, seasonal fluctuations of the deeper zonal

velocity along the equator may be less connected to this

near-equilibrium response. Model results suggest that the

zonal velocity fluctuations below the thermocline are

related to third and higher vertical mode equatorial Rossby

waves that are generated by reflection of forced annual

Kelvin waves from the eastern boundary (Brandt and Eden

2005), which has some support in observations (Brandt and

Eden 2005; Lukas and Firing 1985). According to linear

theory (McCreary 1984), these higher mode Rossby waves

propagate as downward beams from the eastern boundary,

with corresponding upward phase propagation, and they

can take many months to cross a basin the size of the

Atlantic. Brandt and Eden (2005) diagnosed the presence

of these waves in a model forced with a realistic seasonal

wind cycle and found that they dominated the zonal flow

variability along the equator below about 200 m, with

zonal velocity amplitudes of 0.1–0.2 m/s and a mean zonal

propagation speed of about 0.3 m/s. The seasonal cycle of

the zonal currents reconstructed from their model analysis

shows, in the lower thermocline layer of the EUC, from

about 150–300 m, a maximum in eastward zonal velocity

that occurs earliest at 0�E in late boreal spring to early

summer (May–June), and progressively later to the west [in

about July–August at 10�W and September at 23�W;

Fig. 12 of Brandt and Eden (2005)]. We hypothesize that

the ‘‘deep extension’’ of the EUC seen in our observations

at both 23�W and 10�W in boreal summer and early fall,

and the similar but weaker deep signal seen at 0�W in

May–June, are expressions of these higher mode equatorial

Rossby waves as they progress annually across the basin.

According to the Brandt and Eden (2005) study, the max-

imum amplitude of the zonal velocity associated with these

waves should be largest near 10�W, which appears to be

consistent with the largest zonal amplitude of the lower

thermocline transport occurring in our observations at

10�W (Fig. 14).

3.4 Interannual variability

In this section we describe some aspects of the interannual

EUC variability observed during TACE. The interannual

variability of SST in the Atlantic cold tongue region as

represented by the ATL3 SST index is shown in Fig. 16 for

the period from 1984 to 2013, including the 2007–2011

TACE period. Relatively warm temperatures have occurred

in the equatorial Atlantic since the late 1990’s, associated

with a broad increase in North Atlantic SST’s at that time,

reflected in the AMO (Atlantic Multidecadal Oscillation)

index (e.g., Enfield and Cid-Serrano 2010), that has per-

sisted until now. The TACE program occurred during a

period of particularly warm SST’s in the equatorial

Atlantic, which was broken only by a single cold anomaly

in boreal summer of 2009. During the planning of TACE it

had been hoped that the deployed arrays would sample an

‘‘Atlantic Nino’’ event during one of the measurement

years, so that the response of the EUC to such an event

could be studied in detail. The Atlantic Nino, often referred

to as the ‘‘zonal mode’’ of tropical Atlantic interannual

variability (Ruiz-Barradas et al. 2000)—and so-named

because of its similarity to El Nino events in the Pacific—is

a coupled ocean–atmosphere mode that manifests itself in

warm SST anomalies in the central and eastern equatorial

during the boreal summer months that frequently exceed

1–2 �C. During these events, the equatorial trade winds

relax west of 20�W, while farther eastward the northward

cross-equatorial winds associated with the North African

summer monsoon also weaken (Horel et al. 1986; Zebiak

1993; Ruiz-Barradas et al. 2000). Corresponding increases

in diabatic heating in the mid-troposphere occur along with

a southward shift of tropical convection (Carton et al.

1996; Giannini et al. 2003), often causing extensive rainfall

and flooding in the coastal cities of the Gulf of Guinea.

A pronounced variability of the EUC in the Pacific is

known to occur in association with El Nino events,

including a complete shutoff of the EUC during the

largest events (Johnson et al. 2002; Izumo 2005). Model

results suggest that the thermocline EUC transport in the

central and eastern Atlantic is also linked to the cold

tongue variability, with weaker EUC transports correlated

with a warmer cold tongue (Hormann and Brandt 2007).

Relative to the climatological state of the last decade or

so, the summers of 2008 and 2010 can be characterized as

mildly warm years for the cold tongue (Fig. 16), while

summer 2009 was a significant cold event, clearly the

most significant SST event during TACE. Brandt et al.

(2013b) have recently shown, from EUC measurements at

23�W obtained for a longer time period than considered

here, that the EUC at 23�W shows the expected rela-

tionship to the cold tongue: a weaker EUC in early boreal

summer (June) is correlated with an anomalously warm

cold tongue. They further showed that weaker (stronger)

EUC transports are associated with weaker (stronger)

westward wind stress across most of the equatorial

Atlantic in boreal spring. These winds act to precondition

the ZPG to a weaker or stronger state, correspondingly

driving a weaker (stronger) EUC at the onset of the

summer upwelling season. Brandt et al. (2013b) also

showed that the westward surface currents on the equator

in early boreal summer strengthen in concert with the

EUC, in response to these same wind anomalies, and they

hypothesize that the resulting larger vertical shear pro-

duces stronger vertical mixing that ultimately leads to a

colder cold tongue during these events.

W. E. Johns et al.

123

The cold event during 2009 was atypical with regard to

the above scenario in several respects—as noted also by

Brandt et al. (2013b). The normal spring preconditioning

did not occur; in fact the easterly winds along the equator

were anomalously weak in spring 2009 as noted in

Fig. 15. Foltz and McPhaden (2010) have suggested

instead that the anomalous cold event in 2009 was related

to off-equatorial wind stress anomalies north of the

equator in spring 2009, which forced an upwelling Rossby

wave which then propagated to the western boundary, and

then equatorward to generate an upwelling Kelvin wave

that crossed the Atlantic and led to the anomalous cool-

ing. Comparing the years 2008 2009, and 2010 in our

data, the EUC at 23�W is not obviously stronger during

summer of 2009; in fact it is relatively weak compared to

2010 and comparable to 2008 (Figs. 7, 12). At 10�W, we

are only able to compare 2009 and 2010, and the EUC is

also relatively weaker during 2009. Judging from the

longer 0.75�S ADCP record at 10�W (Fig. 2a), the EUC

was also relatively strong during summer of 2008 com-

pared to 2009. This would suggest—from the TACE

period alone—that the EUC at 10�W is weaker during

strong cold tongue events than during warm or ‘‘normal’’

years, although this may again be biased by the anoma-

lous nature of the 2009 cold event. At 0�E the boreal

summer EUC is actually weakest, in terms of its upper

velocity core, in 2008, and comparable in 2009–2010,

which is also generally reflected in the total transports

(Fig. 12). Thus, from the TACE measurement period we

cannot draw any firm conclusions about how the EUC

strength varies across the basin in association with

anomalously cold or warm cold tongue years.

It is interesting to note that at both 23�W and 10�W, the

interannual variability of the EUC during summer, based on

the standard deviations of the monthly means (Fig. 12), is

actually weaker than in either boreal spring or fall. However,

this appears to be due mainly to differences in the phasing or

timing of intraseasonal fluctuations rather than significant

seasonal differences between the years.

The spring wind relaxation across the tropical Atlantic

was more pronounced in 2008 and 2009 (Fig. 15), and in

particular during 2009 the relaxation was more sustained in

the central Atlantic and began earlier. At 23�W, the boreal

spring (March–May) EUC appears relatively strongest in

2010, weakest in 2009, and intermediate strength in 2008

(Fig. 7). This suggests that the early and more sustained

wind relaxation in 2009 led to a reduced ZPG driving the

EUC. At 10�W this same behavior is not apparent, with

similar strengths seen for the upper EUC in both 2009 and

2010. At 0�E there is some suggestion of a stronger boreal

spring EUC in 2008 and 2010 than in 2009, but the dif-

ferences are subtle. Thus, we find overall that there were

not dramatic interannual fluctuations of the EUC during the

TACE period. This is perhaps advantageous from the

viewpoint of obtaining a representative seasonal cycle of

the EUC from our observations. However, resolving its

interannual variability in the eastern equatorial region and

its relationship to interannual wind forcing and the cold

tongue development, as done by Brandt et al. (2013b) for

23�W, will require a longer measurement period than

obtained here.

4 Discussion and conclusions

In this paper we present new observations of the Equatorial

Undercurrent in the central and eastern part of the Atlantic

based on moored current measurements collected during

2007–2011 along 23�W, 10�W, and 0�E. These observa-

tions provide, for the first time, a clear description of the

seasonal cycle of the EUC across the basin, which had

before relied mainly on ship sections that can be affected

by seasonal sampling biases and strong intraseasonal

fluctuations.

The mean transport of the EUC at 23�W is

14.3 ± 0.6 Sv, decreasing to 12.1 ± 0.9 Sv at 10�W and

9.4 ± 0.6 Sv at 0�E. The character of the seasonal cycle

changes moving eastward: at 23�W the strongest EUC

transport occurs in boreal fall, at 10�W the EUC transport

shows a semiannual cycle with a maximum in boreal spring

and fall, while at 0�E the EUC has a single spring maxi-

mum. The maximum core intensity within the upper part of

Fig. 16 Weekly ATL-3 SST

index for the equatorial Atlantic

region (0�E–20�W, 3�S–3�N),

from 1984 to 2013 (NOAA/

AOML; http://www.aoml.noaa.

gov/phod/regsatprod/atl3/sst_

anm.php). The TACE period is

indicated


123

http://www.aoml.noaa.gov/phod/regsatprod/atl3/sst_anm.php



the EUC occurs in boreal spring at all longitudes, and at

this time (April) the EUC transport is almost uniform, at

about 12–13 Sv, all across the basin. The weakest EUC

core intensity occurs during the boreal summer cold tongue

phase at all locations.

A number of modeling studies are now available that

have made predictions on the seasonal cycle of the EUC in

the Atlantic, including early studies by Philander and

Pacanowski (1986) and Schott and Boning (1991), and

more recent studies by Hazeleger et al. (2003), Arhan et al.

(2006), and Hormann and Brandt (2007). All of these

models tend to show a common behavior in the central part

of the basin, where they exhibit an EUC transport maxi-

mum (or relative maxima) in boreal fall. However, in both

the western and eastern parts of the basin, there are dif-

ferences among the models, and particularly in the east

there are significant differences with respect to the obser-

vations. Arhan et al. (2006) find a maximum EUC transport

in boreal spring (April–May) in the western part of the

basin (west of about 20�W), and a secondary maximum in

fall (August–September). By about 10�W their modeled

EUC becomes mainly annual in character with a primary

fall maximum and a much weaker spring maxima. Hor-

mann and Brandt (2007) show a similar seasonal cycle as

Arhan et al. (2006) at 10�W, but at 23�W and 35�W they

show a primary fall transport maximum and a secondary

spring maximum, in better agreement with our observa-

tions. Hazeleger et al.’s (2003) model results are interme-

diate between these, showing nearly equal spring and fall

maxima in the western part of the basin (35�W), and a

singe fall maximum at 20�W. In the eastern part of the

basin, Arhan et al. (2006) show mainly an annual cycle east

of 10�W, with a fall maximum and a second weak maxima

in January, while Hormann and Brandt (2007) show, at

3�E, a single seasonal maximum in February. Both the

mean EUC transport and its seasonal maximum in the

central part of the basin are reproduced reasonably well by

these models, but both of these models—and generally

others—underestimate the mean EUC transport in the east

due to a too-rapid eastward decay of the EUC. Signifi-

cantly, none of the available models appears to properly

capture the observed spring maximum in EUC transport in

the eastern part of the basin, as revealed by the measure-

ments at 0�E.

As noted by Arhan et al. (2006), there is potential

confusion in the literature when discussing the seasonal

cycle of the EUC, depending on whether it is the total

transport or the maximum intensity (core speed) that is

being considered. Our observations show that the maxi-

mum in core intensity occurs across the basin in boreal

spring (April), while there is a secondary maximum in the

fall (October) at both 23�W and 10�W, and a transport

maximum at 23�W and 10�W that occurs slightly earlier, in

September. This transport maximum is associated in part

with the intensified upper core of the EUC in boreal fall,

but derives in large part from a deeper extension of the

EUC into the lower thermocline in late summer and early

fall at those longitudes. The study by Philander and

Pacanowski (1986), in fact, does show two seasonal max-

ima in the velocity core of the EUC, which take place in

April–May and November at 30�W and have about equal

strength (*0.8 m/s), and in October and February at 0�E,

where the October maximum is significantly stronger

(*0.6 vs. 0.3 m/s in February). These were the only two

longitudes studied in that paper. On the other hand, Arhan

et al. (2006) find only a fall maximum in core intensity, and

the core intensity is actually a minimum across the whole

basin in boreal spring (their Fig. 6). Furthermore, the

models that do predict a secondary maximum in the EUC

transport (Hormann and Brandt 2007; Arhan et al. 2006) or

EUC core intensity (Philander and Pacanowski 1986) in the

eastern part of the basin, seem to get this at the wrong time,

in January or February instead of April. Therefore, we

conclude that the observed spring maximum in EUC core

intensity across the whole basin, and the fact that this leads

to an actual transport maximum in spring at 0�E, is not a

feature that is correctly reproduced by the available

models.

The existence of a spring maximum in the EUC in the

eastern part of the basin also has support in hydrographic

observations. Since the SEQUAL–FOCAL experiment (e.g.,

Hisard and Henin 1987), it has been known that the thermo-

cline salinity maximum associated with the EUC is strongest

in the eastern equatorial Atlantic in late winter to spring, and

we find in our analysis of the Argo data at 0�E (Fig. 17) that

this consistently occurs between March–May, exactly when

the EUC core intensity and transport are a maximum there.

This is consistent with the notion that the high salinity core of

the EUC—originating from the subtropics and mostly

entering the EUC at the western boundary—is more effec-

tively transported across the basin during boreal spring when

the EUC core intensity is at a maximum. A second salinity

maximum occurs in October–November at 0�E (and at 10�W)

when the EUC re-accelerates after its summer minimum

(Figs. 11, 17). The Argo data, and previous observations, also

show clearly that the EUC salinity maximum is strongly

eroded, or can even disappear, in boreal summer, due to the

strong mixing that occurs at the top of the EUC during the

development of the cold tongue. Surface mixed layer salini-

ties also reach their seasonal maximum over most of the

central and eastern Atlantic between June–November

(Fig. 17). This is consistent with the weaker upper EUC

observed in boreal summer, which is presumably also retar-

ded by downward mixing of westward surface momentum

and also by the relaxed ZPG in the eastern part of the basin

during summer.

W. E. Johns et al.

123

The fate of the high salinity waters carried in the EUC,

especially in boreal spring when they are carried farthest

eastward into the Gulf of Guinea, remains uncertain. His-

torical studies of the EUC in the eastern Gulf of Guinea have

generally concluded that the EUC penetrates in the mean to

the eastern boundary where it feeds coastal undercurrents

both to the north and south of the equator (Hisard and Henin

1987; Wacongne and Piton 1992). However, the more recent

cruises conducted in the Gulf of Guinea, including a US

cruise in June 2009 in support of our mooring operations that

extensively sampled the EUC across the basin, have shown

the presence of westward flowing, high-salinity cores flank-

ing the EUC in the Gulf of Guinea during boreal summer and

fall (Kolodziejczyk et al. 2013). The salinity in these west-

ward flows is comparable to the salinity of the EUC core

itself, and can only come from the EUC, since there is no other

source of such high salinity waters in the Gulf of Guinea. We

hypothesize that these high salinity waters are remnants of the

high salinity EUC core waters that are advected most strongly

into the eastern Gulf of Guinea during boreal spring, which

are then recirculated back in these westward flows toward the

central part of the basin. The portion of the EUC fed into these

westward recirculations may actually be larger than the

amount that reaches the African coast and can escape the Gulf

of Guinea through the Gabon-Congo Undercurrent flowing

southward along the eastern boundary (Wacongne and Piton

1992; Mercier et al. 2003), as suggested by Kolodziejczyk

et al. (2013).

The deep extension of the EUC in boreal summer and

fall that occurs at 23�W and 108W is a very curious feature,

which to our knowledge has no counterpart in the Pacific.

What drives it remains unclear. During the above-men-

tioned June 2009 cruise, our shipboard CTD/ADCP section

at 108W revealed a core of eastward flow at depths of

200–400 m that seemed to be clearly separated from the

overlying EUC core, which can also be seen in the ADCP

records at 10�W in 2009 (Fig. 2a). The evidence suggests

that this flow may initially develop as a separate subsurface

core in boreal spring which then merges into the base of the

EUC to form the ‘‘deep reaching’’ EUC observed in sum-

mer/fall. Evidence for similar behavior was seen at 10�W

during both 2004 and 2005 by Kolodziejczyk et al. (2009),

but it apparently does not occur in all years [for example,

there is no clear evidence of a separate subsurface core

occurring prior to the deep extension of the EUC in 2008 or

2010 (Fig. 2a)]. A possible dynamical explanation for this

seasonal feature in the deeper part of the EUC is the

propagation of higher vertical mode equatorial Rossby

waves from the eastern boundary, that arise from reflection

of equatorial Kelvin waves forced by the annual wind

stress variation in the western part of the basin (Brandt and

Eden 2005). These waves have downward energy propa-

gation and upward phase propagation, and could poten-

tially explain the development of the deeper current core

and its upward migration into the lower part of the EUC.

The interannual variability of the EUC during the time

period of the TACE observations was found to be relatively

small, and no consistent variability pattern could be

established in relation to the interannual variability of the

cold tongue. In particular, the cold SST event that occurred

in the central and eastern equatorial Atlantic in 2009

appears to have been analogous to a ‘‘non-canonical’’ cold

event (Richter et al. 2013), which lacked the usual pre-

conditioning by stronger westward wind stress in the

western parts of the basin during boreal spring. The EUC

was actually weakest across most of the basin during this

event, which differs from the statistical results obtained by

Brandt et al. (2013b), based on longer records, where years

with a stronger cold tongue are found to be correlated with

a stronger EUC at 23�W, which is consistent with expec-

tations from models (Hormann and Brandt 2007). The

nature of the EUC’s response in the eastern part of the

basin to anomalous cold tongue variability associated with

‘‘Atlantic Nino’’ events therefore remains to be established.

Longer ADCP records from a single equatorial mooring

exist at 10�W prior to the TACE time period (Bunge et al.

2006; Kolodziejczyk et al. 2009) and are currently being

maintained by the PIRATA program, which may enable

these relationships to be established in the future. There are

also developing plans to maintain the 0�E equatorial ADCP

beyond the time period collected here, so that linkages

Dep

th (

m)

23

23.5

24

24 24

2525

2525.525.5

25.526 26

2626.2 26.2

26.226.4

26.426.4

23oW


0

50

100

150

200 34.5

35

35.5

36

36.5D

epth

(m

)

2323.5

24

2424

2525

2525.525.5

25.52626

2626.226.2

26.2

26.4 26.4 26.4

10oW


0

50

100

150

200 34.5

35

35.5

36

36.5

Dep

th (

m)

22 2323 23

23.5

23.523.5

2424

242525

2525.5

25.5

25.526

26

2626.2

26.2

26.2

26.426.4

26.40oE


0

50

100

150

200 34.5

35

35.5

36

36.5

Fig. 17 Monthly-mean salinity profile variability at the equator for

the TACE period (2007–2011), from 0 to 200 m, derived from the

PIRATA-corrected Argo data at 23�W, 10�W, and 0�E. Density

contours are overlain in black


123

between the eastern equatorial Atlantic and the coastal

upwelling regions off Southwest Africa—another impor-

tant region of interannual SST variability—can be better

understood.

Acknowledgments This research was supported by the U.

S. National Science Foundation under awards OCE0623552 and

OCE1129874, and by the Deutsche Bundesministerium fur Bildung

und Forschung (BMBF) as part of the projects NORDATLANTIK

(03F0443B), RACE (03F0651B), MIKLIP (01LP1114A) and by the

Deutsche Forschungsgemeinschaft through several research cruises

with RV Meteor and RV Maria S. Merian, and as part of the Son-

derforschungsbereich 754 ‘‘Climate–Biogeochemistry Interactions in

the Tropical Ocean’’. Moored velocity observations were acquired in

cooperation with the PIRATA project. The authors thank the PIRATA

program for their timely and free provision of data to the scientific

community. Special thanks go to Mark Graham and Robert Jones

(RSMAS), and Jacques Grelet and Fabrice Roubaud (IRD) who

contributed to the RSMAS ADCP mooring maintenance at 0�E and

10�W during PIRATA-FR and US/RSMAS cruises.

References

Arhan M, Treguier AM, Bourles B, Michel S (2006) Diagnosing the

annual cycle of the Equatorial Undercurrent in the Atlantic

Ocean from a general circulation model. J Phys Oceanogr

36:1502–1522

Athie G, Marin F (2008) Cross-equatorial structure and temporal

modulation of intraseasonal variability at the surface of the

Tropical Atlantic Ocean. J Geophys Res 113:C08020. doi:10.

1029/2007JC004332

Athie G, Marin F, Treguier A-M, Bourles B, Guiavarc’h C (2009)

Sensitivity of near surface tropical instability waves to sub-

monthly wind forcing in the tropical Atlantic. Ocean Model

30:241–255

Atlas R, Hoffman RN, Ardizzone J, Leidner SM, Jusem JC, Smith

DK, Gombos D (2011) A cross-calibrated multiplatform ocean

surface wind velocity product for meteorological and oceano-

graphic applications. Bull Am Meteorol Soc 92:157–174

Bourles B, D’Orgeville M, Eldin G, Gouriou Y, Chuchla R,

DuPenhoat Y, Arnault S (2002) On the evolution of the

thermocline and subthermocline eastward currents in the equa-

torial Atlantic. Geophys Res Lett 29(1785):2002G. doi:10.1029/

LO15098

Bourles B, Brandt P, Caniaux G, Dengler M, Gouriou Y, Key E,

Lumpkin R, Marin F, Molinari RL, Schmid C (2007) African

Monsoon Multidisciplinary Analysis (AMMA): special mea-

surements in the tropical Atlantic. CLIVAR Exch Lett 41(Vol.

12, n�2): 7–9

Bourles B et al (2008) The PIRATA program: history, accomplish-

ments, and future directions. Bull Am Meteor Soc 89:1111–1125

Brandt P, Eden C (2005) Annual cycle and interannual variability of

the mid-depth tropical Atlantic Ocean. Deep Sea Res Part I

52:199–219

Brandt P, Schott F, Provost C, Kartavtseff A, Hormann V, Bourles B,

Fischer J (2006) Circulation in the central equatorial Atlantic—

mean and intraseasonal to seasonal variability. Geophys Res Lett

33:L07609. doi:10.1029/2005GL025498

Brandt P, Araujo M, Bourles B, Chang P, Dengler M, Johns WE,

Lazar A, Lumpkin CF, McPhaden MJ, Nobre P, Terray L (2013a)

Tropical Atlantic Climate Experiment (TACE). CLIVAR Exch

18(61; 1):26–31. ISSN 1026-0471

Brandt P, Funk A, Tantet A, Johns WE, Fischer J (2013b) The

Equatorial Undercurrent in the central Atlantic and its relation to

tropical Atlantic variability. Clim Dyn. doi:10.1007/s00382-014-

2061-4

Bunge L, Provost C, Lilly J, D’Orgeville M, Kartavtseff A, Melice JL

(2006) Variability of the horizontal velocity structure in the

upper 1600 m of the water column on the equator at 10 W.

J Phys Oceanogr 36:1287–1304

Cane MA, Sarachik ES (1981) The response of a linear baroclinic

equatorial ocean to periodic forcing. J Mar Res 39:651–693

Carton JA, Cao XH, Giese BS, daSilva AM (1996) Decadal and

interannual SST variability in the tropical Atlantic Ocean. J Phys

Oceanogr 26(7):1165–1175

Chang Ching-Yee, Carton JA, Grodsky SA, Nigam S (2007) Seasonal

climate of the tropical Atlantic sector in the NCAR community

climate system model 3: error structure and probable causes of

errors. J Clim 20:1053–1070

Davey M et al (2002) STOIC: a study of coupled model climatology

and variability in tropical ocean regions. Clim Dyn 18:403–420

DeWitt DG (2005) Diagnosis of the tropical Atlantic near-equatorial

SST bias in a directly coupled atmosphere-ocean general

circulation model. Geophys Res Lett 32:L01703. doi:10.1029/

2004GL021707

Enfield DB, Cid-Serrano L (2010) Secular and multidecadal war-

mings in the North Atlantic and their relationships with major

hurricane activity. Int J Climatol 30:174–184

Foltz GR, McPhaden MJ (2010) Abrupt equatorial wave-induced

cooling of the Atlantic cold tongue in 2009. Geophys Res Lett

37:L24605. doi:10.1029/2010GL045522

Fratantoni DM, Johns WE, Townsend TL, Hurlburt HE (2000) Low-

latitude circulation and mass transport pathways in a model of

the tropical Atlantic Ocean. J Phys Oceanogr 30:1944–1966

Giannini A, Saravanan R, Chang P (2003) Oceanic forcing of Sahel

rainfall on interannual to interdecadal time scales. Science

302(5647):1027–1030

Giarolla E, Nobre P, Malagutti M, Pezzi L (2005) The Atlantic

Equatorial Undercurrent: PIRATA observations and simulations

with GFDL Modular Ocean model at CPTEC. Geophys Res Lett

32(10):L10617. doi:10.1029/2004GL022206

Gouriou Y, Reverdin G (1992) Isopycnic and diapynal circulation of

the upper equatorial Atlantic Ocean in 1983–1984. J Geophys

Res Ocean 97:3543–3572

Hazeleger W, de Vries P, Friocourt Y (2003) Sources of the

Equatorial Undercurrent in the Atlantic in a high-resolution

ocean model. J Phys Oceanogr 33:677–693

Hisard P, Henin C (1987) Response of the equatorial Atlantic Ocean

to the 1983–1984 wind from the Programme Francais Ocean et

Climat dans l’Atantique Equatorial cruise data set. J Geophys

Res 92(C4):3759–3768

Horel JD, Kousky VE, Kagano MT (1986) Atmospheric conditions in

the Atlantic sector during 1983 and 1984. Nature

322(6076):248–251

Hormann V, Brandt P (2007) Atlantic Equatorial Undercurrent and

associated cold tongue variability. J Geophys Res 112:C06017.

doi:10.1029/2006JC003931

Hummels R, Dengler M, Bourles B (2013) Seasonal and regional

variability of upper ocean diapycnal heat flux in the Atlantic cold

tongue. Prog Oceanogr 111:52–74

Izumo T (2005) The Equatorial Undercurrent, meridional overturning

circulation, and their roles in mass and heat exchanges during ElNino events in the tropical Pacific ocean. Ocean Dyn

55(2):110–123

Johnson GC, Sloyan BM, Kessler WS, McTaggart KE (2002) Direct

measurements of upper ocean currents and water properties

across the tropical Pacific during the 1990s. Prog Oceanogr

52(1):31–61

W. E. Johns et al.

123

http://dx.doi.org/10.1029/2007JC004332

http://dx.doi.org/10.1029/2007JC004332

http://dx.doi.org/10.1029/LO15098

http://dx.doi.org/10.1029/LO15098

http://dx.doi.org/10.1029/2005GL025498

http://dx.doi.org/10.1007/s00382-014-2061-4

http://dx.doi.org/10.1007/s00382-014-2061-4

http://dx.doi.org/10.1029/2004GL021707

http://dx.doi.org/10.1029/2004GL021707

http://dx.doi.org/10.1029/2010GL045522

http://dx.doi.org/10.1029/2004GL022206

http://dx.doi.org/10.1029/2006JC003931

Jouanno J, Marin F, du Penhoat Y, Sheinbaum J, Molines J-M (2011)

Seasonal heat balance in the upper 100 m of the equatorial

Atlantic Ocean. J Geophys Res-Ocean 116:C09003. doi:10.1029/

2010JC006912

Keenlyside N, Kleeman R (2002) Annual cycle of equatorial zonal

currents in the Pacific. J Geophys Res 107(C8):Art. No. 3093

Knox RA, Halpern D (1982) Long range Kelvin wave propagation of

transport variations in Pacific Ocean equatorial currents. J Mar

Res 40(329):39

Kolodziejczyk N, Bourles B, Marin F, Grelet J, Chuchla R (2009)

Seasonal variability of the Equatorial Undercurrent at 10 degrees

W as inferred from recent in situ observations. J Geophys Res

Ocean 114:C06014. doi:10.1029/2008JC004976

Kolodziejczyk N, Marin F, Bourles B, Gouriou Y, Berger H (2013)

Seasonal variability of the equatorial undercurrent termination

and associated salinity maximum in the Gulf of Guinea.

Submitted in Clim Dyn

Lukas R, Firing E (1985) The annual Rossby wave in the central

equatorial Pacific Ocean. J Phys Oceanogr 15:55–67

Lumpkin R, Garraffo Z (2005) Evaluating the decomposition of

tropical Atlantic drifter observations. J Atmos Ocean Technol

22:1403–1415

Malanotte-Rizzoli P, Hedstrom K, Arango H, Haidvogel DB (2000)

Water mass pathways between the subtropical and tropical ocean

in a climatological simulation of the North Atlantic Ocean

circulation. Dyn Atmos Oceans 32:331–371

McCreary JP (1984) Equatorial beams. J Mar Res 42:395–430

McCreary JP, Lu P (1994) On the interaction between the subtropical

and equatorial ocean circulation: the subtropical cell. J Phys

Oceanogr 24:466–497

Mercier H, Arhan M, Lutjeharms JRE (2003) Upper-layer circulation

in the eastern Equatorial and South Atlantic Ocean in January–

March 1995. Deep-Sea Res 50(7):863–887

Perez RC, Lumpkin R, Johns WE, Foltz GR, Hormann V (2012)

Interannual variations of Atlantic tropical instability waves.

J Geophys Res-Ocean 117:C03011. doi:10.1029/2011JC007584

Philander SGH, Chao Y (1991) On the contrast between the seasonal

cycles of the equatorial Atlantic and Pacific Oceans. J Phys

Oceanogr 21(9):1399–1406

Philander SGH, Pacanowski RC (1986) A model of the seasonal cycle

in the tropical Atlantic Ocean. J Geophys Res

91(C12):14192–14206

Qiao L, Weisberg R (1997) The zonal momentum balance of the

Equatorial Undercurrent in the central Pacific. J Phys Oceanogr

27(6):1094–1119

Richter I, Xie S-P (2008) On the origin of equatorial Atlantic biases in

coupled general circulation models. Clim Dyn 31:587–598

Richter I, Behera SK, Masumoto Y, Taguchi B, Sasaki H, Yamagata

T (2013) Multiple causes of interannual sea surface temperature

variability in the equatorial Atlantic Ocean. Nat Geosci

6(1):43–47. doi:10.1038/Ngeo1660

Roemmich D, Gilson J (2009) The 2004–2008 mean and annual cycle

of temperature, salinity, and steric height in the global ocean

from the Argo Program. Prog Oceanogr 82:81–100

Ruiz-Barradas A, Carton J, Nigam S (2000) Structure of interannual-

to-decadal climate variability in the tropical Atlantic sector.

J Clim 13(18):3285–3297

Schott FA, Boning CW (1991) The WOCE model in the western

equatorial Atlantic—upper layer circulation. J Geophys Res

Oceans 96:6993–7004

Schott FA, Fischer J, Stramma L (1998) Transports and pathways of

the upper-layer circulation in the western tropical Atlantic.

J Phys Oceanogr 28:1904–1928

Schott F, Dengler M, Brandt P, Affler K, Fischer J, Bourles B,

Gouriou Y, Molinari R, Rhein M (2003) The zonal currents and

transports at 35 W in the tropical Atlantic. Geophys Res Lett

30(7):1349. doi:10.1029/2002GL016849

von Schuckmann KV, Brandt P, Eden C (2008) Generation of tropical

instability waves in the Atlantic Ocean. J Geophys Res

113:C08034. doi:10.1029/2007JC004712

Wacongne S (1989) Dynamical regimes of a fully nonlinear stratified

model of the Atlantic Equatorial Undercurrent. J Geophys Res

94(C4):4801–4815

Wacongne S, Piton B (1992) The near-surface circulation in the

northeastern corner of the South Atlantic Ocean. Deep-Sea Res

39:1273–1298

Yu XR, McPhaden MJ (1999a) Dynamical analysis of seasonal and

interannual variability in the equatorial Pacific. J Phys Oceanogr

29(9):2350–2369

Yu X, McPhaden MJ (1999b) Seasonal variability in the equatorial

Pacific. J Phys Oceanogr 29:925–947

Zebiak SE (1993) Air-sea interaction in the Equatorial Atlantic

region. J Clim 6(8):1567–1568

Zhang D, McPhaden MJ, Johns WE (2003) Observational evidence

for flow between the subtropical and tropical Atlantic: the

Atlantic subtropical cells. J Phys Oceanogr 33(8):1783–1797


123

http://dx.doi.org/10.1029/2010JC006912

http://dx.doi.org/10.1029/2010JC006912

http://dx.doi.org/10.1029/2008JC004976

http://dx.doi.org/10.1029/2011JC007584

http://dx.doi.org/10.1038/Ngeo1660

http://dx.doi.org/10.1029/2002GL016849

http://dx.doi.org/10.1029/2007JC004712

Documents

Zonal structure and seasonal variability of the Atlantic Equatorial Undercurrent