Article

Mixed layer heat and salinity budgets during the onset of the 2011 Atlantic cold tongue

Abstract and Figures

The mixed layer (ML) temperature and salinity changes in the central tropical Atlantic have been studied by a dedicated experiment (Cold Tongue Experiment (CTE)) carried out from May-July 2011. The CTE was based on two successive research cruises, a glider swarm, and moored observations. The acquired in-situ datasets together with satellite, reanalysis, and assimilation model data were used to evaluate box-averaged ML heat and salinity budgets for two sub-regions: 1) the western equatorial Atlantic Cold Tongue (ACT) (23°-10°W) and 2) the region north of the ACT. The strong ML heat loss in the ACT region during the CTE was found to be the result of the balance of warming due to net surface heat flux and cooling due to zonal advection and diapycnal mixing. The northern region was characterized by weak cooling and the dominant balance of net surface heat flux and zonal advection. A strong salinity increase occurred at the equator, 10°W, just before the CTE. During the CTE, ML salinity in the ACT region slightly increased. Largest contributions to the ML salinity budget were zonal advection and the net surface freshwater flux. While essential for the ML heat budget in the ACT region, diapycnal mixing played only a minor role for the ML salinity budget. In the region north of the ACT, the ML freshened at the beginning of the CTE due to precipitation, followed by a weak salinity increase. Zonal advection changed sign contributing to ML freshening at the beginning of the CTE and salinity increase afterward.
Content may be subject to copyright.
RESEARCH ARTICLE
10.1002/2014JC010021
Mixed layer heat and salinity budgets during the onset of the
2011 Atlantic cold tongue
Michael Schlundt
1
, Peter Brandt
1
, Marcus Dengler
1
, Rebecca Hummels
1
, Tim Fischer
1
, Karl Bumke
1
,
Gerd Krahmann
1
, and Johannes Karstensen
1
1
GEOMAR Helmholtz-Zentrum f
ur Ozeanforschung Kiel, Kiel, Germany
Abstract The mixed layer (ML) temperature and salinity changes in the central tropical Atlantic have
been studied by a dedicated experiment (Cold Tongue Experiment (CTE)) carried out from May to July 2011.
The CTE was based on two successive research cruises, a glider swarm, and moored observations. The
acquired in situ data sets together with satellite, reanalysis, and assimilation model data were used to evalu-
ate box-averaged ML heat and salinity budgets for two subregions: (1) the western equatorial Atlantic cold
tongue (ACT) (23–10W) and (2) the region north of the ACT. The strong ML heat loss in the ACT region
during the CTE was found to be the result of the balance of warming due to net surface heat flux and cool-
ing due to zonal advection and diapycnal mixing. The northern region was characterized by weak cooling
and the dominant balance of net surface heat flux and zonal advection. A strong salinity increase occurred
at the equator, 10W, just before the CTE. During the CTE, ML salinity in the ACT region slightly increased.
Largest contributions to the ML salinity budget were zonal advection and the net surface freshwater flux.
While essential for the ML heat budget in the ACT region, diapycnal mixing played only a minor role for the
ML salinity budget. In the region north of the ACT, the ML freshened at the beginning of the CTE due to pre-
cipitation, followed by a weak salinity increase. Zonal advection changed sign contributing to ML freshening
at the beginning of the CTE and salinity increase afterward.
1. Introduction
Large-scale ocean-atmosphere interaction in the tropical Atlantic is an important driver of climate variability.
It undergoes strong changes under current global warming conditions. Increasing sea surface temperatures
(SSTs) [Deser et al., 2010; Xie et al., 2010] are associated with distinct changes in sea surface salinity (SSS) pat-
tern [e.g., Durack and Wijffels, 2010], which were predicted by climate models as the result of an increased
hydrological cycle [e.g., Allen and Ingram, 2002]. However, the physical processes dominating the oceans’
heat and salinity balances in the mixed layer (ML), which interact with the overlying atmosphere, are often
poorly understood. A particular phenomenon in the Eastern Equatorial Atlantic (EEA) is the annual develop-
ment of a region of cold SSTs. This so-called Atlantic cold tongue (ACT) forms in boreal spring/summer,
when the southeasterly trades intensify [Philander and Pacanowski, 1981], and retracts toward the end of
the year resulting in uniformly warm SSTs within the tropical Atlantic. Minimum temperatures of about 22C
are reached within the ‘‘center’’ of the ACT at approximately 10W[Jouanno et al., 2011a]. This is a reduction
of about 6C compared to maximum SSTs occurring during late boreal winter and spring, before the onset
of the ACT. The seasonal cycle of SST is most pronounced at this location. Toward the western and southern
edges of the ACT, the seasonal cycle of SST is still evident, but temperatures do not reach the minimum val-
ues found at 10W at the equator. The interannual variability of the SSTs within the ACT is small in ampli-
tude compared to the seasonal cycle. Nevertheless, it is of climatic relevance: significant correlation was
found between interannual variability of the ACT and the West African Monsoon (WAM) [Brandt et al., 2011;
Caniaux et al., 2011].
The seasonal migration of the Intertropical Convergence Zone (ITCZ), in association with the corresponding
wind field, is the most prominent phenomenon in tropical ocean-atmosphere interaction. In the EEA, the
southeasterlies cross the equator throughout the year and the ITCZ is located to the north of the equator
[Philander et al., 1996]. The upper-ocean current system in the equatorial region is dominated by zonal
flows: the eastward flowing Equatorial Undercurrent (EUC) and the westward South Equatorial Current (SEC)
Key Points:
Atlantic cold tongue development
from May to July 2011 was examined
Diapycnal mixing is key process for
cooling in western cold tongue
region
Zonal advection is main contributor
to mixed layer salinity changes
Correspondence to:
M. Schlundt,
mschlundt@geomar.de
Citation:
Schlundt, M., P. Brandt, M. Dengler,
R. Hummels, T. Fischer, K. Bumke,
G. Krahmann, and J. Karstensen (2014),
Mixed layer heat and salinity budgets
during the onset of the 2011 Atlantic
cold tongue, J. Geophys. Res. Oceans,
119, 7882–7910, doi:10.1002/
2014JC010021.
Received 4 APR 2014
Accepted 29 SEP 2014
Accepted article online 6 OCT 2014
Published online 24 NOV 2014
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7882
Journal of Geophysical Research: Oceans
PUBLICATIONS
at the sea surface. One consequence of the equatorial current structure is elevated vertical shear of horizon-
tal velocity due to these opposing currents leading to elevated mixing in the upper thermocline [e.g.,
Hummels et al., 2013]. Additionally, variability in the currents and in hydrography is caused by the passage
of Tropical Instability Waves (TIWs), propagating westward at and near the equator [Duing et al., 1975;
Legeckis, 1977]. These waves are generated via barotropic and baroclinic instabilities of the zonal current
system [Philander, 1978; von Schuckmann et al., 2008].
A number of different processes lead to spatial and temporal variability of upper-ocean temperatures and
salinities on diurnal, intraseasonal, seasonal, interannual, and longer time scales. Besides atmospheric forc-
ing and horizontal advection through the currents, the ML is also influenced by vertical entrainment and
diffusion through the ML base. However, the role of the different processes in the ML heat and ML salinity
(MLS) budgets is still under debate. For the western ACT region, Foltz et al. [2003] showed the importance
of horizontal advection by evaluating the different terms of the ML heat budget at PIRATA (Prediction and
Research Moored Array in the Tropical Atlantic) [Bourles et al., 2008] buoy locations. However, the authors
were not able to close the ML heat budget for the equatorial sites at 23W and 10W, similar to the majority
of other studies conducted within the ACT region [e.g., Wade et al., 2011]. They speculated that unac-
counted diapycnal mixing at the base of the ML may contribute to their unexplained residual. This contribu-
tion was recently estimated by Hummels et al. [2013]. Using an extensive set of microstructure observations,
they showed that indeed the diapycnal heat flux through the ML base is a dominant cooling term during
ACT development on the equator at 10W. This is also in general agreement with results obtained from
ocean general circulation models as was recently reported by Jouanno et al. [2011a].
A distinct feature of the upper equatorial thermocline in the central and eastern tropical Atlantic is a pro-
nounced seasonal cycle of salinity [Kolodziejczyk et al., 2014]. During early boreal summer, saline water
masses of the upper thermocline are transported eastward within the EUC to the African coast and recircu-
late westward within the SEC on both sides of the EUC. During mid-boreal summer, the velocity maximum
of the EUC weakens and the equatorial subsurface salinity maximum disappears [Gouriou and Reverdin,
1992; Hisard and Morlie
`re, 1973; Johns et al., 2014; Kolodziejczyk et al., 2014].
In analogy to the ML heat content variability, the MLS variability is driven by horizontal salinity (or fresh-
water) advection, vertical entrainment at the ML base, surface freshwater flux (here defined as difference of
evaporation and precipitation: E-P), and salinity diffusion. In fact, Kolodziejczyk et al. [2014] as well as Johns
et al. [2014] recently suggested that intense mixing of the high-saline upper thermocline waters with sur-
face waters in the eastern Gulf of Guinea (GG) is responsible for the erosion of the upper thermocline salin-
ity maximum during late spring and summer. In general, the major difference in the evolution of ML heat
and freshwater anomalies is the lack of a direct feedback between ocean and atmosphere for freshwater
anomalies. Comparing the impact on density changes, the freshwater anomalies might be smaller in magni-
tude than heat anomalies, but their persistence can be longer [Hall and Manabe, 1997].
Analyses of the MLS budget in the western tropical North Atlantic indicated that horizontal advection domi-
nantly contributes to the seasonal cycle in MLS [Foltz and McPhaden, 2008; Foltz et al., 2004]. In the central
and eastern tropical North Atlantic, seasonal variability is dominated by the seasonal cycle in precipitation
[Dessier and Donguy, 1994; Foltz and McPhaden, 2008]. Similar results were found by Da-Allada et al. [2013]
using a combination of in situ, satellite, and reanalysis data to constrain a simplified ML model. They
showed that horizontal advection, entrainment, and precipitation dominantly contribute to the MLS vari-
ability in the eastern tropical Atlantic (ETA), in the GG, and in the Congo region (CO). In their study, the con-
tribution of diapycnal mixing to the MLS budget was not considered due to a lack of observational data.
Recently, ocean remote sensing has yielded a large increase in available SSS data, particularly important to
study the SSS variability of open ocean regions which were so far insufficiently sampled in the past by
research vessels, thermosalinographs on voluntary observing ships and Argo floats. Currently, two satellite
missions monitoring SSS are operating in parallel, namely the Soil Moisture Ocean Salinity (SMOS) mission
of the European Space Agency [Berger et al., 2002; Font et al., 2012], which started in November 2009, and
the joint U.S./Argentinean Aquarius/Sat
elite de Aplicaciones Cientificas (SAC)-D mission [Lagerloef et al.,
2008], which started in June 2011. The high-resolution SSS data have provided new insights into oceanic
freshwater cycles as well as surface ocean dynamics [e.g., Lee et al., 2012; Alory et al., 2012; Tzortzi et al.,
2013] and can be used to improve estimates of the MLS budget.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7883
In this study, an extensive in situ data set is used to investigate the ML heat and salinity budgets concur-
rently. Hydrographic, oceanic microstructure, and atmospheric data collected during two expeditions on R/
V Maria S. Merian (MSM) in spring/summer 2011 have been combined with simultaneous high-resolution
temperature and salinity data from a glider swarm experiment. During this glider swarm experiment, six
gliders were deployed to measure hydrographic properties between 2S and 2N (one glider track was
extended to 4S) and between 23W and 10W. In the following, we refer to the observational experiment
as the ‘‘Cold Tongue Experiment’’ (CTE). The CTE data set is further augmented by temperature and salinity
profiles from Argo floats and time series from PIRATA buoys and subsurface moorings.
In contrast to former studies, which concentrated on single mooring locations or empirically defined boxes and
examined seasonal ML heat budgets [Foltz et al., 2003; Hummels et al., 2013; Jouanno et al., 2011a; Peter et al.,
2006; Wade et al., 2011], seasonal MLS budgets [Da-Allada et al., 2013], seasonal SST variability [Carton and
Zhou, 1997], or seasonal SSS variability [Bingham et al., 2012; Dessier and Donguy, 1994], this study aims to esti-
mate all terms contributing to the ML heat and MLS budgets as an average over the entire region associated
with the western ACT (Figure 1). Due to the amount of ship time required to obtain such an extensive in situ
data set, the CTE covers only 2 months of the year 2011. The CTE was scheduled between May and July to focus
on the processes responsible for the variability of SST and SSS during the development phase of the ACT.
The paper is structured as follows. After describing the general approach, we present the data and the
detailed methods in section 2. In section 3, we derive and combine the relevant contributions to the ML
heat and salinity budgets and compare the results with mean seasonal cycles. A conclusion is presented
and possible error sources are discussed in section 4.
2. Data and Methods
To determine the various components of the ML heat and salinity budgets, observations of various parame-
ters were required at adequate resolution, i.e., data for all heat and freshwater flux components between
Figure 1. Box edges (black lines) and SSTs from TMI (background colors) between 14 and 17 June 2011. ‘‘ACT’’ describes the area associ-
ated with the Atlantic cold tongue (ACT box), while the northern box is denoted ‘‘North.’’
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7884
atmosphere and ocean, microstructure data to quantify oceanic mixing processes, ocean velocities, and the
hydrography of the ML itself. The strategy pursued here, which is explained in more detail below, requires
complementing the in situ database with further products, such as satellite observations, reanalysis prod-
ucts for ocean-atmosphere heat and freshwater fluxes, surface and ML velocities, and the output of a high-
resolution assimilation model run. In this study, salinities are reported in practical salinity units (PSS-78).
2.1. Box-Averaging Strategy
In situ data collected during the CTE indicate that the ML within the ACT region is characterized by a rela-
tively homogenous water mass. Elevated meridional temperature gradients limit the ACT region to the
south between 3S and 4S and to the north between 1N and 3N. Maximum meridional temperature gra-
dients from satellite SST at 3 day resolution were used to define the meridional extent of the ACT (Figure 1).
The zonal boundaries of the ACT box were set to 23W and 10W based on the availability of ship and glider
data. To compare the distinct characteristics of the heat and salinity budgets of the ACT region, a second
box located to the north of the ACT between the northern ACT boundary and a fixed boundary at 8N (Fig-
ure 1) was defined. Boxes used for the model analysis followed exactly the same approach. As detailed in
Appendix A, all individual contributions to the ML heat and salinity budgets were calculated either from
individual profiles or from a regular 131grid and subsequently averaged in the two boxes.
2.2. Ship Data
During the two research cruises ‘‘MSM18/2’’ and ‘‘MSM18/3,’’ lasting from the 11 May to the 11 July, profiles
with a conductivity, temperature and depth (CTD) probe were acquired, as well as continuous observations of
the upper-ocean temperature and salinity with a thermosalinograph (TSG) (Figure 2). CTD profiling was per-
formed with a SeaBird 911 CTD rosette system and measured salinity was calibrated against bottle salinity
samples analyzed with a Guildline Autosal salinometer. The TSG data were recorded every minute using a Sea-
Bird 38/45 system with an intake located at 6.5 m depth in the front of the ship. The TSG observations were
calibrated against CTD data from 6 to 7 m depth and later considered as additional ML temperature (MLT)
and MLS observations for the box and time averaging and for the comparison with satellite observations.
Along with the CTD profiles, microstructure observations were performed at almost all stations. The micro-
structure data were collected using an MSS-90D profiler manufactured by Sea&Sun Technology. It was
equipped with two shear sensors, a fast temperature sensor, an acceleration sensor, and a tilt sensor, plus a
24oW 20
oW 16
oW 12
oW 8
oW
6oS
3oS
0o
3oN
6oN
24
24.5
24.5
24.5
25
25
25
25
25.5
25.
5
25.5
25.5
26
26
26
26
26
26
26.5
26.5
26.5
26.5
26.
5
26.5
27
27
27
27
27
27
27
27.5
27.5
27.5
27.5
27.5
27.5
28
28
28
28
28
28.5
28.5
28.5
28.5
28.5
33.5
34
34.5
35
35.5
36
36.5
ifm02 ifm05 ifm07 ifm08 ifm09 ifm11
Figure 2. Observations during the cold tongue experiment conducted between May and July 2011. Colored lines denote the glider tracks,
dashed black lines show cruise tracks, and black dots are CTD stations. Background colors show 3 month mean SSS from SMOS with con-
tour interval 0.1, while contour lines show 3 month mean SST from TMI in C. Contour interval is 0.5C.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7885
set of slower response standard CTD sensors. The data were sampled at a rate of 1024 Hz. A detailed
description of the probe is given in Prandke and Stips [1998]. From the observed velocity microstructure, it
is possible to derive the dissipation rate of turbulent kinetic energy (TKE), from which the diapycnal fluxes
of heat and salt can be estimated. The dissipation rate is calculated under the assumption of isotropic turbu-
lence from the shear wave number spectrum (Edu0=dz ). The spectrum is integrated between dynamically
adapted wave number limits with e57:5mÐkmax
kmin Edu0=dz kðÞdk to estimate the dissipation rate e(mis the kine-
matic viscosity of seawater). Due to the limited resolved wave number band, a variance loss correction is
applied according to the universal Nasmyth spectrum [Oakey, 1982]. The derivation of dissipation rates fol-
lowed here is described in detail in Schafstall et al. [2010] and Hummels et al. [2013].
Shipboard observations of atmospheric properties during CTE included measurements of the downward
shortwave radiation flux with a pyranometer and the downward longwave radiation flux with a pyrgeome-
ter every 2 s (for a description of the devices and the data processing see Kalisch and Macke [2012]). Precipi-
tation was monitored using an optical disdrometer [Großklaus et al., 1998] and the ship’s rain gauge [Hasse
et al., 1998]. A description of the analysis inferring precipitation from the ship’s rain gauge is given by Bumke
and Seltmann [2012]. The reflected shortwave radiation was computed according to Taylor et al. [1996],
while upward longwave radiation (F"
LW ) was calculated according to the Stefan-Boltzmann law,
F"
LW 50:97rSST4, where ris the Boltzmann constant, assuming an emissivity of 0.97 for the sea surface.
Turbulent heat fluxes and evaporation were computed from the ship’s weather station data by using the
parameterization of Bumke et al. [2014]. They estimated the bulk transfer coefficients for latent and sensible
heat and the drag coefficient with the inertial dissipation method and compared their fluxes with the fluxes
estimated with the latest version of the COARE algorithm [Fairall et al., 2003] (Matlab program codes
cor3_0af.m and cor3_0ah.m, available from ftp://ftp1.esrl.noaa.gov/users/cfairall/bulkalg/cor3_0/matlab3_0/
). We also calculated the turbulent fluxes and the evaporation with the COARE algorithm, which yielded
only marginal differences in the final fluxes, similar to the findings of Bumke et al. [2014].
2.3. Glider Data
Six autonomously operating Slocum electric gliders provided temperature and salinity profiles at approxi-
mately 3–4 km horizontal resolution and to maximum depths of 800 m. Altogether, 10 glider deployments,
mostly along meridional sections (Figure 2), were performed during the CTE, yielding in a total of about
5600 profiles. Thermal lag hysteresis in salinity calculations was corrected by applying the method of Garau
et al. [2011], where four correction parameters are determined by minimizing the area between two
temperature-salinity curves of successive CTD casts.
2.4. Auxiliary Data Sets
2.4.1. Hydrographic Data
Temperature and salinity profiles from Argo floats (provided by the U.S. Global Ocean Data Assimilation
Experiment (USGODAE)) and time series of temperature and salinity from the moored PIRATA buoys (pro-
vided by the Pacific Marine Environmental Laboratory (PMEL)) were used to further supplement the hydro-
graphic data set during the CTE as a part of a large hydrographic data set.
2.4.2. Atmospheric Data
Several atmospheric data products for the surface radiative and turbulent heat fluxes and the freshwater
flux were compared. In particular, the data from the ERA-Interim reanalysis [Dee et al., 2011] from the Euro-
pean Centre for Medium-Range Weather Forecasts (ECMWF) are available for the net shortwave radiation,
net longwave radiation, latent and sensible heat flux, evaporation, and precipitation. They provide 12 hourly
data at 0.75resolution. The second product that provides all relevant surface fluxes is the NCEP2 reanalysis
[Kanamitsu et al., 2002] from the National Center for Environmental Prediction (NCEP) with daily fields at
1.875resolution. All radiative and turbulent heat fluxes together with evaporation are also available from
TropFlux [Praveen Kumar et al., 2012] on a 1grid in the tropics (30S–30N) and at daily resolution. On the
same spatial and temporal resolution, latent heat flux and evaporation were taken from the Objectively
Analyzed air-sea fluxes (OAFlux) data from Woods Hole Oceanographic Institute [Yu et al., 2008]. All daily
averaged data sets were compared with daily averages of onboard in situ observations (for details see
Appendix B). Bias, standard deviation, and root-mean-square of the differences are shown in Table 1. The
best agreement between the different radiative or turbulent fluxes and the ship-based observations was
achieved with different data products. However, the best general agreement was found for the TropFlux
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7886
product and the turbulent heat fluxes, evaporation, as well as net surface shortwave and longwave radiation
are in the following taken from the TropFlux product.
Precipitation estimates were taken from the Advanced Microwave Scanning Radiometer (AMSR-E) onboard
the NASA Aqua spacecraft, from the TRMM Microwave Imager (TMI) onboard the Tropical Rainfall Measure-
ment Mission (TRMM) satellite, as well as from the Special Sensor Microwave Imager Sounder (SSMIS) F17
onboard the DMSP satellite. From all three products, daily averages at a quarter-degree resolution were stat-
istically compared against the direct shipboard observations with the method of Bumke et al. [2012]. The
analysis revealed by taking the statistical parameters into account, that the AMSR-E satellite precipitation
product was closest to the observations and it was finally chosen (for details see Appendix B).
2.4.3. SST and SSS Data
SST data were taken from the TMI onboard the TRMM satellite to calculate horizontal SST gradients required
to estimate the advective contribution to the heat budget. The 3 day means on a spatial grid of a quarter
degree were used for the comparison between satellite SSTs and in situ SST observations (Figure 3a; for
details see Appendix C). A standard deviation of the differences between satellite and in situ SST of 0.52C
was taken as the uncertainty for the satellite SSTs.
Horizontal SSS gradients were calculated based on the SMOS data. SMOS SSS measurements on a 131
grid and with 10 day composites have, compared to Argo SSS, a current bias of 0.3–0.4 [Boutin et al., 2012;
Reul et al., 2012]. Here we used the 3 day mean fields on a quarter-degree grid to compare the satellite SSSs
with in situ SSS observations (Figure 3b; for details see Appendix C). For the satellite SSS, a standard devia-
tion of the differences between satellite and in situ SSS of 0.34 was calculated, which is used as the uncer-
tainty for the satellite SSS.
Table 1. Bias, Standard Deviation (std), and Root-Mean-Square of the Differences (rmsd) of Shipboard In Situ Observations and Reanaly-
sis/Satellite Products for Evaporation (E), Latent Heat Flux (LHF), Sensible Heat Flux (SHF), Net Surface Shortwave Radiation (SSR), and
Net Surface Longwave Radiation (SLR)
a
ERA-Interim NCEP2 OAFlux TropFlux
Bias 6std rmsd Bias 6std rmsd Bias 6std rmsd Bias 6std rmsd
E20.6 60.7 0.9 20.7 61.1 1.5 0.1 60.7 0.7 20.3 60.7 0.8
LHF 218.7 621.0 26.3 219.2 631.5 40.2 0.2 619.3 19.8 27.2 621.6 21.5
SHF 22.9 64.1 5.4 2.9 64.5 5.2 0.1 62.7 2.7
SSR 4.5624.9 24.9 28.4660.2 68.9 8.6 634.4 34.6
SLR 28.3 68.0 10.5 21.4 615.1 15.0 23.7 69.6 9.6
a
Unit for E is mm d
21
while the unit for LHF, SHF, SSR, and SLR is Wm
22
.
20 22 24 26 28 30
20
22
24
26
28
30
(a)
In−situ SST [°C]
TMI SST [°C]
34 35 36 37
34
35
36
37
(b)
In−situ SSS
SMOS SSS
Figure 3. Scatterplot of (a) satellite SST with in situ SST and (b) satellite SSS with in situ SSS. Red dots in Figure 3b denote satellite grid
points where SMOS observations coincide with precipitation (rain rate0.1 mm h
21
) observations. The maximum distance between the
closest satellite grid point and in situ measurement is 1/6or 10 nm.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7887
2.4.4. Velocity Data
Surface velocities within the entire study region were required to estimate the advection terms. The Ocean
Surface Current Analysis Real-time (OSCAR) product is used, which is derived from sea level measurements,
wind stress, and SST data [Lagerloef et al., 1999]. This data set represents vertically averaged geostrophic
and Ekman velocities in the upper 30 m of the ocean [Bonjean and Lagerloef, 2002]. The filtered version of
this data set at a horizontal resolution of 131with a temporal resolution of 5 days is used here. Explana-
tion and validation of the OSCAR product as well as error estimates are described in Johnson et al. [2007]. A
comparison between OSCAR velocities and Moored Acoustic Doppler Current Profiler (mADCP) data at
equatorial moorings (23W and 10W) is shown in Figure 4 and the analysis is described in Appendix D.
Foltz et al. [2013] made a similar comparison between OSCAR velocities and moored velocities at 4N, 23W
and their estimates of the uncertainties for the zonal and meridional velocity were used as the uncertainty
in the northern box in this study. The comparison of OSCAR velocities with the mADCP velocities at the
equatorial moorings at 23W and 10W showed larger variability of the zonal velocity component (Figures
4a and 4b) for the OSCAR product compared to mADCP velocities. OSCAR meridional velocities are instead
weak (Figures 4c and 4d), indicating that the OSCAR product does not capture TIWs, which otherwise are
clearly identifiable in the meridional component of the mADCP data.
2.4.5. Mercator Assimilation Model
In addition to the observation-based data products, we used the ‘‘Mercator Global operational System
PSY2V4R2’’ model output in our analysis to estimate the horizontal advection terms and the entrainment.
The model output corresponds to a simulation that assimilates SST and sea level anomaly (SLA) fields, tem-
perature, and salinity profiles, and a mean dynamic topography. Three-dimensional fields of zonal and
meridional velocities, temperature, and salinity are provided as output. The horizontal resolution is 1/123
1/12(9 km at the equator; decreasing poleward) with daily fields. There are 36 vertical levels in the first
(b)
10°W
O N D J F M A M J J A S O N
(d)
10°W
20112010 |
−0.6
−0.4
−0.2
0
0.2
0.4 (a)
23°W
[m s
−1]
O N D J F M A M J J A S O N
−0.2
−0.15
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25 (c)
23°W
[m s
−1]
20112010 |
ADCP Mercator OSCAR
Figure 4. ML-averaged (a and b) zonal and (c and d) meridional velocities at the equator at 23W (Figures 4a and 4c) and at 10W (Figures 4b and 4d). The black line denotes the OSCAR
product and the red line is the 5 day running mean Mercator estimate. The blue line denotes the 5 day mean subsurface ADCP-data.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7888
1000 m. In the model, the vertical grid spacing changes from nearly 1 m close to the surface to about
150 m at 1000 m depth [Lellouche et al., 2013].
Mercator velocities were compared to the OSCAR and mADCP data (Figure 4; see Appendix D). In contrast
to OSCAR, the meridional velocities of Mercator capture TIWs (Figures 4c and 4d), which can have effects on
the heat budget through horizontal eddy heat advection [Foltz et al., 2003; Giordani et al., 2013; Jochum
et al., 2007; Peter et al., 2006]. However, particularly in the western cold tongue region, TIWs of the Mercator
assimilation model were partly out of phase compared to observations (Figure 4c). Zonal velocities from
OSCAR and the Mercator assimilation model showed some similarities, such as the same phase and ampli-
tude of intraseasonal to seasonal signals, but had large differences as well (Figures 4a and 4b). In particular,
the large westward OSCAR velocities during the CTE in the central cold tongue region (Figure 4b) are not
present in the model output.
2.4.6. Data for the Mean Seasonal Cycles of MLS
The seasonal salinity budgets were estimated at the locations of three PIRATA buoys at the equator, 10W
and 23W, and at 4N, 23W. A unique data set of microstructure shear and temperature profiles and CTD
salinity profiles was used to estimate the dissipation rates of TKE. The data set was collected during nine
cruises to the ACT region carried out in different seasons between 2005 and 2012. A detailed description of
the data set and postprocessing procedures are given in Hummels et al. [2013, 2014]. For this study, the
data set was supplemented by microstructure data from the R/V Maria S. Merian cruise MSM18/3. From this
data set, profiles were used in the latitude range 62and in the longitude range 60.3relative to the nomi-
nal locations of the three selected PIRATA buoys.
All available buoy and mooring data, the essential database for this part of the study, from January 1999 to
December 2012 (first dates used: 30 January 1999 for 0N, 10W; 7 March 1999 for 0N, 23W; 12 June 2006
for 4N, 23W) as well as several climatological products were used. Surface velocities were constructed
from a combination of the YOMAHA’07 data set [Lebedev et al., 2007] and available surface drifter trajecto-
ries [Lumpkin and Garzoli, 2005]. The YOMAHA’07 velocities were derived from Argo float trajectories and
provided by the Asia-Pacific Data Research Center and the International Pacific Research Center (APDRC/
IPRC). A detailed description of the wind-slip correction and the construction of the combined velocity
product are given in Perez et al. [2014].
The horizontal salinity gradients were constructed from the global monthly mean salinity data set from the
Japan Agency for Marine-Earth Science and Technology (JAMSTEC). This data set is derived from Argo float
observations which are binned to 131monthly means from January 2001 ongoing [Hosoda et al., 2008].
We used the data until December 2012. The periods of data coverage from satellite observations of SSS,
SMOS started in November 2009 and Aquarius in July 2011, are not long enough for a robust seasonal cycle.
Monthly means from AMSR-E precipitation (beginning in June 2002) and TropFlux evaporation (beginning
in January 1999), and 3 day averages of SMOS SSS (beginning in January 2010) and TMI SST (beginning in
January 2010), until December 2012 were used as well. The MLD climatology of de Boyer Mont
egut et al.
[2004] was implemented to derive the horizontal gradients of the MLD. The MLDs were interpolated on a 1
31grid and the gradients were estimated with central differences on that grid.
2.5. Methodology
2.5.1. Heat and Salinity Budgets
The ML heat balance can be expressed as follows [Foltz et al., 2003; Stevenson and Niiler, 1983]
qcph@T
@t52qcphurT1u0rT0

2qcpweDT1q01R:(1)
Tis the MLT, tis time, his the MLD, uis the ML-averaged horizontal velocity, u0and T0are deviations from
the temporal average (the temporal average is denoted with an overbar), DT5T2T2his the difference
between Tand the temperature at the base of the ML (T2h), and weis the entrainment velocity. qis the
density of the ML, cpthe specific heat capacity at constant pressure, and q0is the net heat flux through the
ocean’s surface corrected for the penetrative shortwave radiation through the ML base.
The local heat storage on the left-hand side of equation (1) is balanced by horizontal temperature advection
(divided into a mean and an eddy part), entrainment into the mixed layer, net surface heat flux, and a
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7889
residual term R. The residual represents the sum of all unresolved physical processes and the accumulation
of errors from the other terms. The net heat flux at the ocean’s surface is the sum of the net (incoming
minus reflected) shortwave radiation, corrected for the amount penetrating below the ML, the net long-
wave radiation, the latent heat flux, and the sensible heat flux.
Similarly, the balance for MLS is after, e.g., Delcroix and H
enin [1991] given by:
h@S
@t52hurS1u0rS0

2weDS1ðE2PÞS1R:(2)
Sis the MLS, S0is the deviation from the temporal average, Eis the evaporation, Pis the precipitation, and D
S5S2S2his the difference between Sand the salinity at the base of the ML (S2h). The first term on the
right-hand side of equation (2) describes the horizontal advection of salinity (divided into a mean and an
eddy part) and the second term represents salinity entrainment through the ML base. The third term is the
freshwater flux through the ocean’s surface, while Ragain represents the residual including the sum of all
unresolved physical processes and the accumulated errors from the other terms.
According to Stevenson and Niiler [1983], the entrainment velocity can be defined as
we5H@h
@t1w2h1urh

:(3)
The entrainment velocity thereafter is the sum of the local change in MLD with time, the vertical velocity
w2hat the ML base, and the horizontal advection of MLD (urh). Only the upward movement (entrain-
ment) in equation (3) was considered because downward movement (detrainment) does not affect the tem-
perature or the salinity in the ML. This constraint was implemented with the use of the Heaviside unit
function: HxðÞ5ð1;x0 and 0;x<0Þ.
The different terms of equations (1) and (2) were estimated for the two boxes described in section 2.1 for
the period of the CTE using 10 day time steps as well as for the PIRATA buoy sites at 4N and the equator at
23W and at the equator at 10W using monthly averaged data. The different methodologies used to deter-
mine the individual contributions are described in the following.
2.5.2. Methodology to Derive Heat and Salinity Budget Terms
This section describes the calculation of the individual quantities and terms from equations (1) and (2). For
all quantities, the calculation for the CTE is described at first and followed by the calculation for the mean
seasonal cycles at the PIRATA buoys.
2.5.2.1. Mixed Layer Depth
The ML can be defined as the surface layer of constant potential density and hence the depth where the
density starts to increase is the MLD. Another definition of the ML is a surface layer of constant temperature.
Here, the depth where the temperature starts decreasing is called isothermal layer depth (ILD). The ILD was
determined as the depth at which temperature is 0.5C, lower than the temperature averaged between 2
and 6 m depth. The MLD was defined as the depth where the potential density has increased equivalently
to a temperature decrease of 0.5C while salinity and pressure are held constant. The required potential
density increase was about 0.15 kg m
23
, slightly varying with SSS. We avoided effects of diurnal cycles in
MLD/ILD for the temporal averaging of fluxes through the ML base by using these definitions. Diurnal cycles
were present when using 0.2C ILD-criterion (corresponding to a MLD-criterion of 0.06 kg m
23
). No signifi-
cant difference between MLD and ILD was found in the high-resolution glider data at all locations. There-
fore, the simpler temperature criterion was chosen for our definition of the MLD.
To estimate the seasonal cycles, the daily mean MLDs were averaged for all days of the year, using all avail-
able data from January 1999 to December 2012. Afterward, monthly means were constructed. From these,
the temporal evolution of the MLD was calculated as well.
2.5.2.2. Horizontal Advection
For the CTE, horizontal advection was not separated into mean and eddy contribution, because the time
period of the CTE was too short. Here, total advection of heat and salinity were estimated (1) by combining
velocities from the OSCAR product with horizontal temperature and salinity gradients from satellites,
respectively, and (2) from the Mercator assimilation model output. Horizontal gradients of satellite SST and
SSS were calculated using central differences of the 1/4gridded data set that were subsequently averaged
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7890
onto the OSCAR-native 131grid. Finally, the 131advection terms were spatially averaged for the
two boxes described in section 2.1.
Velocities from the Mercator assimilation model were averaged in the upper 30 m using trapezoidal integra-
tion in order to be comparable to the OSCAR estimates. Advection of heat and salinity were then estimated
on the 1/1231/12grid and subsequently averaged for the two boxes described above. Finally, the
advective terms from the OSCAR/satellite product and from the model were temporally averaged over 10
days.
A comparison of the spatially averaged advection terms in the two boxes showed that zonal heat and salin-
ity advection determined from the model output was smaller compared to those determined from the
OSCAR/satellite product (not shown). The main reason for the difference is the generally weaker zonal
velocity of the Mercator assimilation model output compared to the OSCAR estimates for the CTE period.
This behavior is illustrated in Figures 4a and 4b for the equatorial PIRATA buoy sites at 23W and 10W rep-
resenting the ACT box. The meridional heat and salinity advection as obtained from the OSCAR/satellite
product and the model output shows smaller differences. Despite the strong difference of the meridional
velocities as obtained from the OSCAR product and the model output, the resulting contribution of meridio-
nal advection to ML heat or salinity changes is small for both products.
To obtain a mean seasonal cycle of horizontal salinity advection, mean and eddy advection were calculated
separately at the PIRATA buoy sites. For the mean salinity advection, monthly mean velocities were calcu-
lated on a 131grid from all available float and drifter data. The monthly mean horizontal gradients of
SSS from float observations were calculated with central differences on the same 131grid and afterward
averaged for the mean seasonal cycle. Note, that the salinity climatology captures only the time from Janu-
ary 2001 to December 2012.
Eddy salinity advection, (u0rS0), was estimated for the equatorial moorings by assuming a correlation
between temperature and salinity fluctuations. Using the eddy temperature advection (u0rT0), the eddy
salinity advection was calculated via
u0rS0u0rT0dS
dT:(4)
The eddy temperature advection is calculated indirectly as the residual of the mean horizontal advection,
estimated with the mean TMI SST and the aforementioned mean velocities [Hummels et al., 2014], and the
total horizontal advection, estimated by the difference between total time derivative and local time deriva-
tive of SST [Swenson and Hansen, 1999]. The total time derivative of SST is calculated from SST changes
along Lagrangian drifter trajectories. The local time derivative is estimated from monthly averaged TMI SSTs.
The regression of SSS and SST (dS
dT) was calculated on a monthly basis by using 3 years (2010–2012) of 3 day
averages of satellite SSS and SST observations in a box 2.532.5around the buoy locations and daily
PIRATA SSS and SST observations from 1999 to 2012. The regression coefficient for one climatological month
is calculated separately for the satellite and the buoy observations by taking the slope of the regression line
of all pairs of SSS and SST observations in the particular month in all years. Finally, the monthly mean of the
monthly satellite and buoy regression coefficients are used. The two independent estimates of the regres-
sion of SSS and SST are very similar with regard to their seasonal cycle and their annual means (Table 2).
2.5.2.3. Entrainment
Entrainment was calculated from the Mercator assimilation model output only. There were no observational
vertical velocity estimates available for the period of the CTE. The vertical velocity at the ML base was calcu-
lated using the continuity equation w2h5hðr  uÞ. Horizontal gradients of ML velocity and MLD were esti-
mated with central differences on the 1/1231/12model grid. Local changes in MLD were derived on a
daily basis.
Entrainment for the mean seasonal cycles at the PIRATA buoy locations was calculated from the MLD gra-
dients, the horizontal divergence of the monthly mean horizontal velocities and the local time derivative of
the MLD.
2.5.2.4. Surface Heat and Freshwater Fluxes
All atmospheric data sets were regridded linearly on a 131grid and for the surface freshwater flux com-
bined with box and time-averaged MLS and MLD (Appendix A). The penetrative shortwave radiation was
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7891
calculated following Wang and McPhaden [1999], assuming an
exponential decay of surface shortwave radiation with 25 m e-
folding depth for the North box and 15 m for the ACT box. The
smaller e-folding depth (i.e., stronger absorption of shortwave
radiation) in the ACT region follows from enhanced chlorophyll
concentrations near the equator [e.g., Grodsky et al., 2008].
For the seasonal cycles of the surface freshwater flux, monthly
means of all atmospheric data sets were used and averaged in
mean months of the year. These were combined with monthly
means of MLS and MLD estimated with the PIRATA buoy data.
2.5.2.5. Diapycnal Diffusivities and Fluxes
In regions where the stratification is dominated by temperature, the diapycnal diffusivities of heat and mass
are similar (Kh5Kq, e.g., Peters et al. [1988]). Using the observed dissipation rates, the diapycnal diffusivity of
mass was estimated following Osborn [1980]
Kq5CeN22:(5)
N2is the buoyancy frequency and Cis the turbulent mixing efficiency. Cis set constant to 0.2, which is
commonly used in several other studies [e.g., Moum et al., 1989; Hummels et al., 2013]. Further, the diapycnal
heat flux was estimated using
Jheat52qcpKq
@T
@z:(6)
For shear-driven turbulence, the diapycnal diffusivity of salt ðKSÞis equal to the diapycnal diffusivity of heat
[e.g., Osborn and Cox, 1972; Osborn, 1980; Schmitt et al., 2005] and, thus, equal to the diapycnal diffusivity of
mass (KS5Kh5Kq). With this assumption, the diapycnal salt flux was calculated using
Jsalt52Kq
@S
@z:(7)
Diapycnal diffusivities, vertical temperature gradients, and vertical salinity gradients were averaged vertically
in this study between 5 and 15 m below the ML base. The upper boundary was chosen to exclude ML values
from the average. Lien et al.[2008]andHummels et al. [2013] showed that the diapycnal heat flux is highly
divergent in the vertical and rapidly decreases below the ML base. Therefore, the average within the narrow
layer between 5 and 15 m below the ML base was used. Due to the limited amount of microstructure profiles
available and the large variability inherent in turbulent mixing in the ocean, diapycnal fluxes were averaged
for two periods: (1) the first half of the CTE (May until mid-June) and (2) the second half of the CTE (mid-June/
July). To these two periods will be referred later in section 3, when the ML budgets are described. For the sea-
sonal cycles, monthly means of the diapycnal salt flux were calculated. Uncertainties of the fluxes were esti-
mated from error propagation and boot strapping as detailed in Hummels et al. [2013].
2.5.2.6. Harmonic Fitting
Annual and semiannual harmonics were fitted to all monthly averaged data, which were used for the
description of the seasonal cycles. The total error of the fit (errtot ) combines the data error (errdata ) and the
error estimated as the standard deviation between the data and the fit (errfit) and is calculated with
errtot5ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
errdata21errfit 2
p. The data error consists of given uncertainties from the data provider and errors
from averaging of daily or monthly data into monthly climatologies using error propagation.
3. Results
3.1. The ACT in Boreal Summer 2011
3.1.1. Surface Observations of Temperature and Salinity
The long-term observations of SST and SSS at the PIRATA buoy sites at 23W and 10W on the equator can
be used to investigate the exact timing of ACT development during 2011 with respect to the climatological
cycles (Figure 5). At 23W (western part of ACT region), SST (Figure 5a) and SSS (Figure 5b) in 2011 align
well with the average seasonal cycle. In contrast, in the center of the cold tongue at 10W, the onset of the
cooling was approximately 1 week earlier and the cooling was stronger in 2011 compared to the
Table 2. Annual Mean and Annual Standard
Deviation of Regression Coefficients (in C
21
)
Between SSSs and SSTs
a
Satellite
(SMOS/TMI) PIRATA
0N, 10W20.25 60.11 20.21 60.14
0N, 23W20.17 60.09 20.18 60.17
4N, 23W20.12 60.13 20.27 60.35
a
Shown are the values for three mooring
sites for satellite data and PIRATA buoy data.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7892
climatology (Figure 5c). Similarly, SSS exhibited an earlier and stronger increase this year (Figure 5d). The
comparison shows that the time period of our experimental campaign (May-July 2011) was well chosen in
terms of studying the processes during cold tongue development as well as covering the entire cooling
period in the center of the ACT. However, a large SSS increase occurred before the CTE that was not covered
by our measurements.
The onset and spreading of the ACT in boreal summer 2011 is apparent in the monthly equatorial SST evo-
lution (Figures 6a, 6c, 6e, and 6g). The strongest cooling, resulting in minimum temperatures of less than
22C, was found at the equator at around 10W. In June and July, the negative SST anomaly expanded fur-
ther to the west- and southwest, but with a weaker intensity. The monthly satellite data showed a strong
increase in SSS in the EEA from April to May (Figures 6b and 6d) followed by a period of nearly constant SSS
(Figures 6f and 6h). Within the northern box (Figure 1), a reduction of SSS occurred from April through June,
which is associated with the northward migration of the ITCZ. From June to July, SSS increased in the south-
ern part of that region (2N–5N).
3.1.2. Vertical Structure of Temperature and Salinity
Subsurface hydrographic changes during ACT development at 10W were monitored with high vertical and
temporal resolution using glider (Figure 7) and moored temperature and salinity recorders. Surface cooling,
as evident from the satellite SSTs, was also clearly visible in the gliders’ CTD measurements (Figure 7a). In
addition, the high vertical resolution of those data revealed a shoaling of the MLD during the first month of
observations (Figure 7c).
The temporal evolution of the vertical salinity structure was more complex. At the beginning of the time
series, a pronounced salinity maximum was present below a rather fresh ML (Figure 7b). The salinity maxi-
mum is related to the eastward transport of saline water from the western Atlantic within the EUC occur-
ring during spring [Johns et al., 2014; Kolodziejczyk et al., 2014]. A strong increase in MLS was observed
22
24
26
28
30 (a) 23°W
[°C]
22
24
26
28
30 (c) 10°W
[°C]
Jan Mar May Jul Sep Nov
34.5
35
35.5
36
36.5 (b) 23°W
Jan Mar May Jul Sep Nov
34.5
35
35.5
36
36.5 (d) 10°W
Figure 5. Mean seasonal cycle (blue) of (a and c) SST and (b and d) SSS for the PIRATA buoy at 23W at the equator (Figures 5a and 5b) and 10W at the equator (Figures 5c and 5d)
based on averaging all data from beginning of the measuring period (SST/SSS at 23W: 7 March 1999; SST at 10W: 15 September 1997; SSS at 10W: 29 January 1999) to the end of
2012. The red line indicates SST and SSS in 2011, while the dashed vertical black lines mark the beginning and end of the CTE.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7893
with the onset of the cold tongue, while the subsurface salinity maximum was reduced. The simplest
explanation would be a vertical redistribution and mixing of salinity through exchange processes across
the ML base. However, advective processes played the dominant role as will be shown below. MLS
remained elevated during the further development of the cold tongue, whereas in the beginning of July
the subsurface salinity maximum reappeared. Typically, MLDs are shallowest in tropical upwelling regions
[e.g., de Boyer Mont
egut et al., 2007]. Indeed, at the survey site, MLD never exceeded 30 m, but a clear
diurnal cycle is visible when using a smaller temperature criterion (dT 50.2C) for the MLD than chosen
for the ML budgets, i.e., dT 50.5C (Figure 7c). The abrupt increase in MLT and decrease in MLS on 14
May (Figures 7a and 7b) was a remarkable event in the temperature and salinity time series, which coun-
teracted the trends expected from the cold tongue development. Satellite SST distributions from this
period (not shown) suggested that the anomaly was caused by the propagation of a TIW that moved the
SST front, here defined as the maximum meridional SST gradient at 10W, north of the ACT southward
(Figure 7d).
The glider surveys along meridional or zonal sections (Figure 2) exhibit a mixture of temporal and spatial
variability. One glider was assigned to profile along a meridional section at 15.5W between 2S and 2N
(Figure 8d). The glider crossed the SST front on 10 June, stayed north of the front, and crossed back on 18
June (Figure 8d). The first crossing of the front was clearly visible in the freshening of the ML in association
with increased temperatures (Figures 8a and 8b). North of the front, a diurnal cycle in MLD (using the 0.2C
criterion) is not evident (Figure 8c). The glider was close to the equator and crossed the EUC core with the
subsurface salinity maximum three times: at the beginning of the section, around the 10 June, and at the
Figure 6. Monthly mean fields of SST (a, c, e, and g) from TMI and SSS (b, d, f, and h) from SMOS. Figures 6a and 6b are April, Figures 6c and 6d are May, Figures 6e and 6f are June, and
Figures 6g and 6h are July.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7894
end of June (Figures 8b and 8d). North and south of the equator and away from the EUC, the subsurface
salinity maximum was weak or not present.
To further investigate the spatial and temporal distribution and evolution of the subsurface salinity
maximum, the glider and CTD data were supplemented with CTD data from a French PIRATA cruise
Figure 7. Time series of glider ifm02 of (a) temperature, (b) salinity, (c) MLD, and (d) latitudinal position (black). In Figures 7a and 7b, the
potential density surfaces 24.5 and 26.2 are denoted in black. In Figure 7d, the latitude of the SST front at 10W is denoted in red. The
glider profiled close to the PIRATA buoy at 10W all the time. The two gaps in the time series stem from recoveries for battery exchanges.
Figure 8. Same as Figure 7, except for glider ifm11 profiling along a meridional section at about 15.5W.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7895
(PIRATA-FR21) and Argo float profiles. The results revealed a general reversal of the upper-ocean vertical
salinity gradient (Figure 9). At the beginning of the CTE, the thermocline layer (TL; defined with the potential
density range: 24.5 rh26.2) was more saline than the ML in the entire equatorial ACT region (Figure
9a). This is due to the fact that the TL contains the EUC core that is advecting high-saline waters from the
west during this period. However, south of the equator, the ML was saltier than the TL during May and
beginning of June (Figures 9b–9d). This reversed vertical salinity gradient weakened later at the end of June
and the beginning of July, when MLS and the TL-salinity became equal in the ACT region (Figures 9e and
9f). Note, that during June and July, the upper boundary of the TL (rh524.5) reached the sea surface (Fig-
ures 7b and 8b). The described variability of the vertical salinity gradient is independent of the measure-
ment device and visible in glider profiles as well as CTD and Argo float profiles. Taking the mean seasonal
cycle of the salinity difference between ML and TL from Argo, it is obvious that the difference changes sign
in boreal summer (Figure 9g). While the mean seasonal cycle of the difference is a robust feature, its magni-
tude depends on the latitudinal boundaries used for averaging.
Figure 9. (a–f) Salinity difference between MLS and maximum salinity in the upper thermocline layer below the ML (24:5rh26:2or rMLD rh26:2if rMLD >24:5) from
glider profiles (filled circles), shipboard CTD profiles (open circles), and Argo float profiles (triangles) for the CTE. (g) Mean salinity difference between MLS and maximum salinity in the
upper thermocline layer (24:5rh26:2) from all Argo float profiles from the years 2000 to 2012 in boxes with the longitudinal boundaries 23W and 10W and the latitudinal boun-
daries 2S and 1N. Red bars denote the standard error of the mean of all profiles in the box within 1 month.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7896
3.2. Heat Budget
The contributions of the different processes (equation (1)) to the ML heat budget in the North box (Figures
10a and 10b) and the ACT box (Figures 10 c and 10d) are described in the following. Since this study
focuses on the cold tongue development, we start with the ACT box.
3.2.1. ACT Box
Absorbed shortwave radiation in the ACT ranged from 194 636 to 226 636 Wm
22
. The variability is caused
by the variability of cloud cover and ML thickness. The latent and sensible heat fluxes as well as the net sur-
face longwave radiation cooled the ML throughout the CTE (Figure 10c). The net surface longwave radiation
in the ACT box was nearly constant (ranging from 42 610 to 55610 Wm
22
) during the CTE due to the bal-
ance of outgoing and downward longwave radiation. Outgoing longwave radiation slightly weakened with
decreasing SSTs and the downward longwave radiation weakened with the reduction of clouds. The latent
heat flux ranged between a maximum of 110 621 Wm
22
and a minimum of 78 621 Wm
22
. The variability
was predominantly associated with varying winds. The magnitude of the sensible heat flux from the ocean to
the atmosphere was the smallest compared to the aforementioned heat fluxes with a mean of 263Wm
22
.
The resulting net surface heat flux warmed the ML during the whole CTE with a mean of 61643 Wm
22
.
The strong heat loss of the ACT ML in boreal summer can only be explained by ocean dynamics. Zonal heat
advection played an important role for ML cooling during the CTE with strongest cooling of 263 625
Wm
22
at the end of May, while meridional heat advection was a minor contributor to cooling with a maxi-
mum of 25611 Wm
22
. Similar results were obtained for the Mercator assimilation model, where
−200
−150
−100
−50
0
50
100
(b)
[W m−2]
ρcphT/t North sum North
−150
−100
−50
0
50
100
150
200
(a)
[W m−2]
05/15 05/22 05/29 06/05 06/12 06/19 06/26 07/03
−200
−150
−100
−50
0
50
100
(d)
2011
[W m−2]
ρcphT/t ACT sum ACT w/o sum ACT with
05/15 05/22 05/29 06/05 06/12 06/19 06/26 07/03
−150
−100
−50
0
50
100
150
200
(c)
2011
[W m−2]
u adv v adv entr SSR LHF SLR+SHF
Figure 10. (a and c) The contribution of each term to the ML heat budget and (b and d) the local heat tendency and the sum of the terms for the northern box (Figures 10a and 10b)
and for the ACT box (Figures 10c and 10d). The different contributions in Figures 10a and 10c are zonal (u adv) and meridional (v adv) heat advection, entrainment (entr), net surface
shortwave radiation corrected for the penetrative part (SSR), latent heat flux (LHF), and the sum of net surface longwave radiation (SLR) and sensible heat flux (SHF). The contribution of
diapycnal mixing in Figure 10c is shown with the red dots (each dot is representative for the averaging period of about 1 month). In Figure 10d, the sum without (sum ACT w/o; dashed-
dotted line) and with (sum ACT with; red dots) diapycnal mixing is shown. The black shadings in Figures 10b and 10d are the uncertainty of the observed heat changes. The grey shad-
ings in Figures 10b and 10d are the accumulation of all errors of the different processes.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7897
meridional heat advection showed a maximum contribution of 212 61Wm
22
. Entrainment consistently
cooled the ML with a mean 2861Wm
22
resulting from high vertical velocities in combination with a
weak positive vertical temperature gradient below the ML.
In accordance with the reduced SST, the ML heat content tendency in the ACT box was negative during the
entire CTE period (Figure 10d). The tendency was weakly negative during early May, but strongly negative
during the period end of May to early June, when the ML locally lost up to 2144 615 Wm
22
of heat. With-
out considering diapycnal mixing (Figure 10d), the mixed layer heat tendency cannot be explained by the
sum of processes described above.
The diapycnal heat flux determined from the microstructure data was elevated during the first period of the
CTE from the second half of May to the beginning of June. For this period, a mean flux of 2111 616 Wm
22
was estimated that reduced in magnitude to 249 69Wm
22
during the second half of the CTE (Figure 10c).
Unfortunately, microstructure observations were not continuously available for the entire CTE period within
the ACT box thus longer averaging periods had to be used to estimate representative diapycnal heat flux
contributions. The elevated diapycnal heat flux during the first period of the CTE resulted from elevated tur-
bulent eddy diffusivities that persisted despite increased upper-ocean stratification during this period.
The magnitude and temporal variability of the sum of the individual heat flux contributions including the
diapycnal heat flux (Figure 10d) agrees well with the magnitude and temporal variability of the heat content
tendency. This indicates that within the uncertainties, the heat budget in the ACT region was closed utiliz-
ing the above flux estimates for the sampled period.
3.2.2. Northern Box
The dominant processes contributing to the heat balance of the northern box differ from those dominating
the ACT heat balance. In particular, the net surface heat flux was comparably lower and even changed sign
during the CTE. Absorbed shortwave radiation ranged from 144 635 to 183 635 Wm
22
and was mainly
balanced by the heat loss due to the other atmospheric fluxes (sum ranging from 2142 624 to 2205 624
Wm
22
) (Figure 10a). However, the net surface heat flux warmed the ML in the beginning of the CTE with
8642 Wm
22
, while it cooled the ML during the rest of the experiment with a minimum of 229 642 Wm
22
at the end of the CTE. This was mainly caused by the increased cooling contribution of the latent heat flux
due to the increased wind.
Zonal heat advection significantly contributed to cooling of the ML only during the beginning of June
(256 668 Wm
22
), but was small during the rest of the CTE period. The contribution of meridional heat
advection to ML cooling in the northern box was in general small with a minimum value of 28649
Wm
22
. The high uncertainties for the advection terms result, on the one hand, from the uncertainties in
velocity, estimated through comparison of OSCAR velocities with velocities from moored measurements
and, on the other hand, from uncertainties of estimating the horizontal temperature gradients from the
satellite data. The contribution of entrainment ranged from 2162to2962Wm
22
during the whole
CTE and is not important for ML cooling. The diapycnal heat flux was not estimated for the northern box
due to the lack of data.
The ML heat content tendency was negative throughout the CTE (Figure 10b). The negative tendency was
largest at the end of May/beginning of June (270 647 Wm
22
) while it was smallest (219 624 Wm
22
)at
the end of June. During the whole CTE period and within the given uncertainties, the sum of the aforemen-
tioned individual flux terms balances the ML heat content tendency.
3.3. Salinity Budget
The contributions of the different processes (equation (2)) to the MLS budget in the North box (Figures 11a
and 11b) and in the ACT box (Figures 11c and 11d) are described in the following. As for the heat budget,
we start with the ACT box.
3.3.1. ACT Box
Evaporation in the ACT region increased MLS constantly at a rate between 0.16 60.03 and 0.20 60.05 per
month (month
21
), while precipitation reduced the MLS only weakly in the beginning of May (Figure 11c).
The contribution of precipitation to MLS changes during the development of the ACT was negligible, due
to the position of the ITCZ further to the north (Figure 11c). Hence, the difference E-P is at this location
dominated by evaporation.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7898
By far, the largest absolute contribution to the MLS budget was by zonal advection. In May and the begin-
ning of June 2011, negative salinity anomalies were transported into the ACT within the westward branches
of the SEC. Minimum salinity advection occurred in early June contributing to a freshening of 20.70 60.37
month
21
. However, zonal advection exhibited elevated variability throughout the CTE and almost disap-
peared in the middle of June when it contributed to a weak salinity increase of 0.01 60.01 month
21
.
Meridional salinity advection increased MLS in May and the beginning of June 2011 with a maximum contri-
bution of 0.11 60.20 month
21
(Figure 11c). During the rest of the CTE, the contribution of meridional
advection was weak. Entrainment also played a minor role for salinity changes in the ML during the CTE
with a maximum value of 0.03 60.01 month
21
. As vertical velocities used for the entrainment estimates in
the salinity and the heat budgets are the same, the minor role of entrainment in the salinity budget is due
to the small vertical salinity gradients below the ML as pointed out in section 3.1.
The diapycnal salt flux inferred from microstructure observations increased MLS by 0.1060.01 month
21
dur-
ing the first period of the CTE (Figure 11c). It decreased in June and locally partly changed sign in the end of
June/beginning of July around 10W, according to a local change in sign of the vertical salinity gradient (Fig-
ure 7b). However, the average diapycnal salt flux determined from all data collected between the second
week of June and the end of the CTE resulted in a very weak salt flux of 0.0160.01 month
21
(Figure 11c).
The MLS in the ACT box increased with the cold tongue onset in May 2011, with a maximum tendency of
1.24 60.65 month
21
(Figure 11d). This tendency reduced during the further expansion of the cold tongue
and was followed by a period of weak ML freshening. With the data sets available for this study, the salinity
content change during cold tongue development could not be fully balanced by the individual flux
−1
−0.5
0
0.5
1
1.5 (b)
[mth−1]
S/t North sum North
−0.6
−0.4
−0.2
0
0.2
0.4 (a)
[mth−1]
05/15 05/22 05/29 06/05 06/12 06/19 06/26 07/03
−1
−0.5
0
0.5
1
1.5 (d)
2011
[mth−1]
S/t ACT sum ACT w/o sum ACT with
05/15 05/22 05/29 06/05 06/12 06/19 06/26 07/03
−0.6
−0.4
−0.2
0
0.2
0.4 (c)
2011
[mth−1]
u adv v adv entr E P
Figure 11. (a and c) The contribution of each term to the MLS budget and (b and d) the local salinity tendency and the sum of the terms for the northern box (Figures 11a and 11b) and
for the ACT box (Figures 11c and 11d). The different contributions in Figures 11a and 11c are zonal (u adv) and meridional (v adv) salinity advection, entrainment (entr), evaporation (E),
and precipitation (P). The contribution of diapycnal mixing in Figure 11c is shown with the red dots (each dot is representative for the averaging period of about 1 month). In Figure
11d, the sum without (sum ACT w/o; dashed-dotted line) and with (sum ACT with; red dots) diapycnal mixing is shown. The black shadings in Figures 11b and 11d are the uncertainty of
the observed salinity changes and the gray shadings in Figures 11b and 11d are the accumulation of all errors of the different processes.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7899
contributions, although contributions due to diapycnal mixing were evaluated here. The latter accounted
for some of the salinity content increase observed during the beginning of the experiment, but an unre-
solved residual remained.
3.3.2. Northern Box
Evaporation in the northern box was similar to the evaporation in the ACT box and weakly increasing during
the CTE from 0.13 60.03 to 0.21 60.06 month
21
. In analogy to the latent heat flux, this increase was caused
by the increasing wind speed. The main source of freshwater in the northern box during the CTE was pre-
cipitation (Figure 11a). The ITCZ occupied parts of the northern box during the CTE, which led to an ele-
vated freshwater input of predominantly convective rainfall resulting in a maximum salinity decrease of
20.40 60.11 month
21
at the beginning of May. Hence, the difference E-P was negative during most parts
of the CTE and only changed sign in the second half of June.
Similar to the ACT box, zonal advection played a key role in salinity changes for the northern box (Figure
11a). During May, westward flow transported negative salinity anomalies into the northern box from the
east leading to a minimum of zonal salinity advection of 20.14 60.12 month
21
. Later in June, elevated
zonal salinity advection of up to 0.49 60.6 month
21
contributed to increase MLS. This is caused by positive
salinity advection with the nSEC, which was strong during June. However, this contribution is uncertain due
to the misrepresentation of TIWs in the OSCAR product.
Meridional salinity advection contributed to increase MLS during May and June except at the beginning of
the CTE in early May. Northward flow transported salty water from the cold tongue into the northern box
with a maximum contribution of 0.17 60.35 month
21
. Entrainment derived from model output was weak
during the CTE with a maximum contribution of 0.04 60.01 month
21
in mid-June. The diapycnal salt flux
was not estimated for the northern box due to the lack of data.
In the northern box, MLS tendency was positive during the whole CTE except for a weak freshening during
the beginning of the experiment (Figure 11b). The MLS tendency was balanced within the uncertainties by
the sum of precipitation, evaporation, horizontal advection, and entrainment during most of the CTE
period.
3.4. The CTE in the Seasonal Cycle
In order to incorporate the different contributions to the ML budgets inferred during the CTE into a broader
perspective, we compare the results obtained for the period of the CTE to the mean seasonal cycle of the
contributions to the ML budgets estimated at three PIRATA buoy locations within our study area. For the
salinity budget, the seasonal cycles of the individual contributions are estimated in the following at the
three PIRATA buoy sites at 23W and 10W on the equator, as well as 4N, 23W (Figure 12). The seasonal
ML heat budgets at these locations were already examined in various previous studies [Foltz et al., 2013,
2003; Hummels et al., 2013]. Hence, the results of the CTE concerning the ML heat budgets are compared to
the results of these previous studies as part of section 4.
3.4.1. Mean Seasonal Mixed Layer Salinity Budgets
3.4.1.1. 4N, 23W
The variability of precipitation dictates the mean seasonal cycle of MLS tendency at the PIRATA buoy site at
4N, 23W. It follows a semiannual cycle caused by the seasonal migration of the ITCZ (Figure 12a). During
May-July, the contribution weakens due to the northward migration of the ITCZ. However, although it is
reduced during this period, precipitation was the dominant contributor to MLS changes in the northern box
during the CTE (Figure 11a). Monthly mean evaporation is nearly constant over the year, but reduced com-
pared to evaporation at the two equatorial locations. Hence, the net surface freshwater flux at 4N, 23Wis
predominantly determined by the semiannual cycle of precipitation and is only positive in July and August,
when precipitation is strongly reduced.
The seasonal cycle of zonal advection at the buoy site is weak and follows the seasonal cycle of the NECC,
which strengthens from its minimum eastward velocity during boreal spring to maximum eastward veloc-
ities in July [e.g., Richardson and Reverdin, 1987; Goes et al., 2013]. The monthly mean zonal advection during
May-July at the buoy site is much weaker than suggested for the northern box during the CTE period in
2011. In particular, the variability of zonal advection included a change in sign during the CTE, which is not
captured in the seasonal estimate at the buoy location.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7900
Meridional advection has its maximum in late fall/early winter and is the main contributor for the increase
of the MLS this period (Figure 12a). Eddy salinity advection exhibits a weak semiannual cycle with a freshen-
ing contribution in April/May and from October to December. During the latter period, eddy salinity advec-
tion is elevated but negative, leading to a decrease of MLS content. The sum of the seasonal cycle of
meridional salinity advection and eddy salinity advection, dominated by the meridional eddy advection due
to the TIWs, at the mooring site is small in boreal summer, which is similar to the total meridional salinity
advection estimated during the CTE. The sum during May is negative, indicating negative total meridional
salinity advection, which was not determined from the data collected during the CTE period. During June
and July, the sum is positive, indicating a total meridional salinity advection similar to the results obtained
for the same period of the CTE.
Entrainment is weak throughout the year and exhibits a maximum in spring. During the CTE, the entrain-
ment in the northern box was similarly weak. The diapycnal salt flux is negligible throughout the year, at
least during the resolved periods. The sum of the contributing terms balances the observed salinity tend-
ency within the uncertainties over the entire year at the PIRATA buoy at 4N, 23W (Figure 12b). Although
the seasonal cycle was evaluated locally, the results generally agree with the findings during the CTE point-
ing toward the fact that the salinity variability observed during the CTE is typical for this season (Figures
11b and 12b).
3.4.1.2. 0N, 23W
From January to May, precipitation exceeds evaporation at the equatorial PIRATA buoy at 23W represent-
ing the western ACT region (Figure 12c). Later in the year, precipitation is negligible and nearly constant
evaporation yields a positive surface freshwater flux during that part of the year. Similarly, in 2011 during
Figure 12. (a, c, and e) Seasonal cycles of the contributing terms to the MLS budget and (b, d, and f) the comparison of the local salinity tendency and the sum of the contributing terms
at three PIRATA buoys. The different contributions in Figures 12a, 12c, and 12e are zonal (u adv), meridional (v adv), and eddy (eddy adv) salinity advection, evaporation (E), precipitation
(P), entrainment (entr), and diapycnal mixing (mix). Black dashed-dotted lines (right) are the sum without diapycnal mixing and red dots are the sum with diapycnal mixing. The gray
shadings in Figures 12b, 12d, and 12f are the accumulation of all errors of the different processes and the error of the salinity tendency.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7901
the CTE period from May to July, the influence of precipitation on the MLS budget was weak in the ACT
region and a positive freshwater flux due to excess evaporation with only weak variability was indicated.
This freshwater flux contributed to a MLS increase, comparable to the climatological freshwater flux from
May to July (Figure 12c).
The zonal salinity advection at the buoy site is negative throughout the year representing a significant con-
tribution to the total salinity budget. It is characterized by a weak semiannual cycle that peaks in boreal
summer and winter. The negative salinity advection during early boreal summer is consistent with the
results from the CTE period, but weaker in magnitude. Meridional salinity advection exhibits a maximum in
boreal winter and represents the largest positive flux contribution to the MLS budget. Eddy salinity advec-
tion reduces MLS content at 23W and is largest in the boreal winter months. At the end of spring, the con-
tribution is negligible but is again relevant during summer. The sum of the seasonal cycle of mean
meridional advection and eddy salinity advection at the buoy site is small in boreal summer. Due to the fact
that eddy variability at 23W is dominated by TIWs and the meridional gradients exceed the zonal ones,
eddy advection is presumed to consist mostly of the meridional eddy component. The small magnitude of
the total meridional heat advection agrees with the results from the CTE (Figure 12c), however, the sign dif-
fers to the CTE results, where weak positive meridional heat advection is found.
Within the seasonal cycle, entrainment at the PIRATA buoy at 23W has its maximum during May and June
(Figure 12c) when it contributes to a salinity increase. During the rest of the year, its contribution is weak.
Within the ACT and during the CTE period in 2011, the weak salinity difference between the ML and below
the ML resulted in weak entrainment contributions to MLS changes, albeit elevated entrainment velocities.
The diapycnal salt flux calculated from individual cruise data (section 2.5.3) exhibits elevated variability
within the seasonal cycle at 23W. Strongest diapycnal salt flux, leading to a MLS increase occur during
February-March. Additionally, diapycnal salt fluxes are elevated in June and November. In May and July, its
contribution at 23W is weak. The results for June and July are comparable to the CTE results from 2011,
when the diapycnal salt flux led to a MLS increase in June and had a negligible contribution in July.
The seasonal cycle of the salinity tendency at the equatorial PIRATA buoy at 23W is weak, but shows a pos-
itive tendency during spring, which reduces and even reverses toward July (Figure 12d). The weak variability
of the MLS tendency is generally captured by the sum of the contributing terms. In February and March, the
salinity-increasing contribution of diapycnal mixing decreases the imbalance between tendency and the
sum of fluxes. During this period, large freshening contributions result from zonal advection, eddy advec-
tion as well as precipitation while the diapycnal flux increases MLS together with meridional advection and
evaporation. Although the seasonal cycle was evaluated locally, the results generally agree with the findings
during the CTE pointing toward the fact that the salinity variability observed during the CTE is typical for
this season (Figures 11d and 12d). However, some flux contributions may vary locally within the ACT during
the seasonal cycle.
3.4.1.3. 0N, 10W
In winter and early spring, the mean seasonal cycles of evaporation and precipitation at the equatorial
PIRATA buoy site at 10W are comparable in magnitude, resulting in a weak surface freshwater flux (Figure
12e). During the rest of the year, evaporation exceeds precipitation, thus the surface freshwater flux contrib-
utes to increase MLS. This is comparable to the freshwater flux in the ACT region during the CTE period.
Additionally, comparable results of the surface freshwater flux were obtained by Da-Allada et al. [2013] for
the GG, suggesting that the surface freshwater flux is a large-scale phenomenon during this period.
Zonal advection is the largest contributing term to the mean seasonal cycle of the MLS balance at the equa-
tor at 10W. It acts to reduce MLS and is most pronounced from December to July. Later in the year, its con-
tribution weakens (Figure 12e). Although zonal advection at 23W is reduced compared to 10W, zonal
advection in the ACT region during the CTE period is of similar magnitude compared to 10W and also
shows a similar temporal evolution. This freshening contribution to the MLS balance can be explained by
negative salinity advection mainly associated with the westward current branches, the northern SEC (nSEC),
and the central SEC (cSEC). However, as will be discussed in section 4, the seasonal evolution of zonal salin-
ity advection is predominantly controlled by the seasonal evolution of zonal MLS gradients. In general, the
spring/early summer dominance of zonal advection in the MLS budget found here also agrees with recent
results from a model study of the MLS balance in the GG region reported by Da-Allada et al. [2013].
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7902
Similar to the results from 23W, the contribution of meridional advection to the MLS budget has a maxi-
mum in late fall/early winter. During this period, the meridional salinity advection represents the dominant
MLS source at 10W (Figure 12e). Magnitude and phase of eddy salinity advection at 10W is also very simi-
lar to 23W. It exhibits a semiannual cycle with minimum in December/January and July/August. The sum of
the seasonal cycle of meridional salinity advection and eddy salinity advection at the buoy site is again
small (cf. 23W) in boreal summer, which is similar to the total meridional salinity advection estimated dur-
ing the CTE. Also as for 23W, the sum during June and July is negative, indicating freshening due to meridi-
onal salinity advection, which was not found during the CTE.
The seasonal cycle of entrainment at the PIRATA buoy at 10W has its maximum in March and April contrib-
uting to the MLS increase during this period (Figure 12e). During the rest of the year, the contribution of
entrainment is weak. This is consistent with the estimate of entrainment during the CTE, when the contribu-
tion of entrainment was rather negligible due to weak salinity differences between the ML and below the
ML albeit rather strong entrainment velocities. Weak positive and negative entrainment contributions to
MLS changes were obtained due to changes in sign of the vertical salinity gradient below the ML.
The diapycnal salt flux at the 10W-PIRATA buoy increases the MLS in May and June, followed by a freshen-
ing contribution during July. Later in September and November, again a positive salt flux from the subsur-
face layer into the ML through diapycnal mixing was observed. The findings for June and July are
comparable with the CTE results that showed a positive diapycnal salt flux, leading to a MLS increase during
June, followed by a negligible contribution to the MLS budget in July 2011.
The observed MLS increase during May 2011 in the ACT region is identifiable in the seasonal cycle of the
salinity tendency at 10W (and 23W) as well (Figure 12f). This increase weakens during June and July in the
central ACT region. The MLS increase at the equatorial PIRATA buoy site at 10W is not explained by the
considered processes from equation (2). The remaining residual indicates either a missing source of salinity
or an overestimate of the freshening contributions (Figure 12f). Observed diapycnal mixing in May and
June provides a positive salt flux into the ML and thus reduces the residual, leading to a balanced salinity
budget within the uncertainties at least during June. From March to May, vertical salinity gradients between
the ML and the thermocline are largest (cf. Figure 9). It is thus likely that diapycnal mixing during this period
contributes to increase MLS and thus decrease the residual. However, so far no microstructure observations
from 10W during these months are available. The remaining imbalance between the sum of terms and the
observed salinity tendency at this location also coincides with periods of highly elevated zonal advection
(March to July) and periods of elevated meridional advection (September to December). Possibly, the
remaining imbalance is caused by an overestimation of these terms, which is either caused by overesti-
mated horizontal velocities or horizontal salinity gradients. As the same velocity product did not cause
imbalances in the ML heat budget at this location in another study [Hummels et al., 2014], it could be
argued that the zonal and meridional salinity gradients are still not sufficiently well resolved by Argo floats.
4. Summary and Conclusion
Within the present study, the physical processes responsible for MLT and MLS changes during cold tongue
development in 2011 have been investigated using an extensive set of in situ and satellite data, reanalysis
products, and assimilation model output. In contrast to other studies evaluating the ML heat and salinity
budget at individual locations or empirically defined boxes, the strategy pursued here was to evaluate the
individual contributions to the ML budgets in two boxes representing the western ACT region and a region
to the north of it. The boundary between the two boxes was defined by the temporally varying maximum
of the meridional SST gradient. In general, the results concerning the ML heat budget agree with previous
studies of the same region [Hummels et al., 2013; Jouanno et al., 2011a; Wade et al., 2011], despite the differ-
ent approaches regarding box-averaged or local budgets. The diapycnal heat flux stands out as the domi-
nant cooling term during ACT development, which has been inferred here from microstructure
observations distributed within the ACT box. This finding is in agreement with previous studies [Hummels
et al., 2013; Jouanno et al., 2011a]. It also gives confidence that the chosen region for averaging is adequate
to investigate the contributions to the ML budgets representative for the western ACT.
The MLS budget has to our knowledge not been investigated within this region before. MLS tendency is posi-
tive during ACT development. However, diapycnal mixing played only a minor role in the MLS budget. This is
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7903
due to a rather weak vertical salinity gradient in the western ACT during the period of the CTE, which partly
even changed sign. The horizontal salinity advection, especially its zonal component, represents the dominant
contribution to the MLS budget. Contrary to the MLT budget, the MLS budget is partly not closed within the
estimated uncertainties, although considering diapycnal flux contributions. This suggests the presence of
unaccounted errors in the MLS budget that might result from the used products or methodology. In summary,
tendencies and major flux contributions to MLT and MLS within the CTE period from May to July 2011 were:
ACT box
1. The heat content tendency of the ML was negative throughout the CTE period. The net surface heat flux
constantly warmed the ML. Dominant cooling terms were diapycnal mixing through the ML base and
zonal advection associated with the transport of cold water from the GG. The sum of these contributions
balanced the observed MLT changes within the uncertainties.
2. MLS tendency was positive at the beginning of the CTE in May and became close to zero in June and
July. Evaporation exceeded precipitation throughout the CTE. Horizontal salinity advection was the dom-
inant contributor to ML freshening from May to mid-June. Entrainment was negligible throughout the
CTE. The contribution of diapycnal mixing was relatively small. The sum of these contributions balanced
the observed MLS changes within the uncertainties in June and the beginning of July. A residual
between observed MLS changes and the sum of all contributions remained during May.
Northern box
1. Heat content tendency was negative during the CTE but considerably lower when compared to the ACT
box. Net surface heat flux was small, warming the ML at the beginning of the CTE and cooling the ML
afterward. Horizontal heat advection cooled the ML during the entire CTE, except at the end of June,
when advection weakly warmed the ML. Entrainment weakly cooled the ML throughout the CTE. The
sum of these contributions balanced the observed MLT changes within the uncertainties.
2. The MLS decreased during May and increased later in June and July. Precipitation was the main contrib-
utor to MLS changes during the beginning of the CTE in May, when zonal and meridional advection can-
celled out each other. In June and July, when precipitation and evaporation were of similar magnitude,
horizontal salinity advection, mainly the zonal component, was the main contributor to the MLS
increase, particularly in the beginning of June. The observed MLS tendency was balanced within the
uncertainties by the sum of resolved flux contributions.
To address the generality of the results obtained during the period of the CTE, the mean seasonal cycle of
the contributions to the MLS budget was evaluated at three PIRATA buoy locations, on the equator at 23W
and 10W and at 4N, 23W. Overall, the dominant flux contributions determined at the buoy positions for
the CTE period agreed well in magnitude and phase with the box averaged flux contributions. At the equa-
torial buoy site at 10W, zonal salinity advection is the dominant term contributing to a freshening of the
ML from December to July (Figure 12e) and the magnitude of the contribution is a factor of two larger in
the central ACT region at 10W compared to the western ACT region at 23W (Figure 12c). Precipitation and
meridional eddy advection also significantly contribute to a freshening of the ML within the ACT region
throughout the year. Meridional advection, entrainment, and evaporation contribute to increase MLS.
Finally, diapycnal mixing increases salinity predominantly from November to June and is negligible during
the other months of the year. The results emphasize that the MLS tendency is largely balanced by ocean
processes and to a lesser extent by the net surface freshwater flux (see Figure 12).
Furthermore, the evaluation of the mean seasonal cycles at the PIRATA buoy locations emphasized that
zonal salinity advection is the main contributor to the MLS budget in the western ACT during the first half
of the year. Two major branches of the SEC, the nSEC slightly north of the equator (mean position: 1Nat
10W and 2Nat23
W) and the cSEC south of the equator (mean position: 3S–4Sat10
W and 4S
at 23W) [e.g., Lumpkin and Garzoli, 2005; Brandt et al., 2006; Kolodziejczyk et al., 2009], provide negative
salinity advection. However, its contribution rapidly weakens toward the end of June. Due to the fact that
the two branches of the SEC are still present during this period, the strong weakening of zonal salinity
advection results from a decrease of the zonal salinity gradient. In fact, a large MLS increase in the EEA in
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7904
June and July 2011 was observed in the SMOS SSS (Figure 6). As evaporation increases MLS in the central
and EEA homogeneously, the elevated MLS in the EEA must have their origin in subsurface processes.
Recently, Kolodziejczyk et al. [2014] conjectured vertical diffusion toward the sea surface in the GG occurring
between May and August as the fate of the high-saline thermocline waters that are transported eastward
within the EUC during the spring period [Johns et al., 2014]. This requires enhanced diapycnal mixing in the
GG from late boreal spring through summer, which was indeed indicated by numerical simulations with a
general circulation model [Jouanno et al., 2011b].
Although the salinity flux into the ML associated with diapycnal mixing was weak during the CTE, it is in
general a salinity-increasing contributor to the MLS budget of the ACT. This is due to the fact that the EUC
core is generally associated with a subsurface salinity maximum and hence diapycnal mixing acts to
increase MLS. Surprisingly though on some stations in the central ACT region, diapycnal mixing partly con-
tributed to decrease MLS in boreal summer. These local freshening events are due to the local reversal of
the vertical salinity gradient below the ML. The question arises which processes are responsible for the
reversal of the vertical salinity gradient. According to the findings of Jouanno et al. [2011b] and as hypothe-
sized in recent studies by Kolodziejczyk et al. [2014] and Johns et al. [2014], a possible explanation is that a
part of the additional salt in the ML was previously entrained or mixed from the thermocline layer in the GG
(and parts of the central ACT) upward, simultaneously eroding the salinity maximum of the EUC. Evapora-
tion further increased the MLS, finally leading to a strong reduction or reversal of the vertical salinity gradi-
ent below the ML. Differing meridional displacements of existing meridional salinity gradients in the ML
and the EUC below due to TIWs might as well contribute to local changes in the vertical salinity gradient
below the ML, which is illustrated by the glider section at the PIRATA buoy at the equator, 10W.
Strong local differences of vertical gradients and diapycnal diffusivities at the base of the ML within the ACT box
were observed. Diapycnal fluxes calculated by using box averages of vertical gradients and diapycnal diffusivities
do not agree with the box average of locally calculated diapycnal fluxes due to a correlation of vertical gradients
and diapycnal diffusivities. Similarly, diapycnal mixing as conjectured from box-averaged budget residuals might
become erroneous when averaging over larger regions. Microstructure observations as they are used here are
still rather sparse. To get more reliable estimates on the variability in space and time of the diapycnal heat and
salt fluxes at the ML base, other observational platforms have to be used with higher spatial and/or temporal
resolution such as, e.g., gliders equipped with microstructure sensors or moored microstructure measurements.
For the calculation of the horizontal advection, surface or upper-ocean currents are necessary together with
horizontal gradients of MLT and MLS. To estimate box averages of the advection, a velocity product is
needed with a high temporal and spatial resolution within the entire box. In this study, the OSCAR product
is used, which is estimated from satellite observation and not directly constrained by observations. In partic-
ular, in the equatorial band, OSCAR velocities strongly deviate from subsurface ADCP velocities measured at
PIRATA locations (Figure 4) and also from velocities of the MERCATOR assimilation model. The discrepancy
in the velocity products results in a large uncertainty of the box-averaged horizontal advection for the CTE.
A general good agreement between in situ SSS and SMOS SSS was shown for the cold tongue region with
an estimated accuracy of satellite SSS similar to the previous studies [Boutin et al., 2013, 2012; Reul et al.,
2012]. In regions strongly affected by precipitation, substantial differences between SSS and MLS were
observed (Figure 3) that introduce additional uncertainty when using SSS for studying ML processes. How-
ever, the steadily improving SSS measurements are the basis for improving the MLS budget particularly by
better constraining individual contributions due to the different atmospheric and oceanic processes in
dynamically varying regions.
Together with several previous studies, the present study improved our understanding of the seasonal and
intraseasonal variability of temperature and salinity within the ML in the central equatorial Atlantic. The
interesting question regarding the contribution of different processes to interannual variations of ML prop-
erties and thus to a deterministic or stochastic behavior of climate relevant SST variability remains to be
addressed in future studies.
Appendix A: Box Clustering of the Data
Time series of glider observations were used to study the upper-ocean variability. The high resolution in
time (one profile in approximately 4 h) as well as in space (one profile every 3–4 km) made it possible to
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7905
examine the spatiotemporal variability in the central tropical Atlantic. To discriminate between the ACT and
the region north of it, two boxes were chosen in which data were averaged. The two boxes are the ACT box
characterized by strong cooling and the northern box north of the ACT box with weak cooling during cold
tongue development.
Here we used the meridional SST gradient to determine the northern and southern boundary of the ACT
box. The maximum SST gradient between 1S and 4N defines the northern boundary and the minimum
SST gradient between 5S and 2S the southern boundary of the ACT box. These boundaries vary in time. A
wave-like structure, associated with TIW propagation is weak in May and at the beginning of June in the
SST data. In addition, the undulations of the SST front increased in amplitude with the intensification of the
cooling. The calculation of the meridional boundaries of the ACT box was done separately with the satellite
SSTs and with the model SSTs, due to the different phases of TIW propagation in both products as visible in
the meridional velocities at two PIRATA buoy sites (Figure 4c). The zonal boundaries of the ACT box were
fixed and set to 23W and 10W. The northern box has a fixed northern boundary at 8N, a western bound-
ary at 23W and a northeastern boundary connecting 16W, 8N with 10W, 2N.
The time period of 10 days that was used for averaging all data in the boxes follows from the decorrelation
time scales for tropical Atlantic currents. Decorrelation time scale for the meridional currents is between 7–
10 days [Perez et al., 2014] and about 8–10 days for the zonal currents [Garraffo et al., 2001; Lumpkin et al.,
2002]. The 10 day period is also consistent with the Argo period and the representation of the OSCAR prod-
uct [Johnson et al., 2007]. The profiles in both boxes were averaged by bootstrapping with 1000 realizations.
The mean of all realizations was taken as the value for the box (for MLT, MLS, MLD, the temperature differ-
ence between MLT and the temperature below the ML base, the salinity difference of MLS and salinity
below the ML base). The error was computed as two standard deviations encompassing 95% of the data.
Uncertainties for dissipation rates of TKE were estimated by bootstrapping and the propagation of errors
was calculated with the procedure of Ferrari and Polzin [2005] for diapycnal diffusivities and accordingly for
the heat and salt fluxes [Schafstall et al., 2010].
Box-averaged lateral gradients of MLT and MLS were calculated by using satellite SSTs and SSSs. These
gridded products are available at 3 day time resolution. Precipitation, evaporation, net shortwave and long-
wave radiation, as well as turbulent heat fluxes (latent and sensible) through the ocean’s surface were aver-
aged over the two boxes. The uncertainty of the box average was determined by taking the error,
determined by the comparison with shipboard measurements, at every grid point into account. Errors for
the velocities were estimated by comparing satellite-derived velocities with in situ measurements.
Appendix B: Comparison of Atmospheric Data Sets for Heat and Freshwater Fluxes
With Observations
For the calculation of the surface radiative and turbulent heat fluxes and the freshwater flux, several prod-
ucts were compared with in situ shipboard measurements, which are a combination of onboard radiation
and rain measurement devices, a pyranometer/pyrgeometer system, and an optical disdrometer. The meth-
odology used to compare in situ data and satellite-derived precipitation is described by Bumke et al. [2012]
for the Baltic Sea and adapted here for the equatorial Atlantic. Due to the high spatial and temporal variabil-
ity of precipitation, a specific statistical analysis was used that compared in situ precipitation measurements
and satellite-derived data or reanalysis data. This analysis follows the recommendations given by the World
Meteorological Organization (WMO) for binary or dichotomous forecasts (a detailed description is provided
through the ‘‘WWRP/WGNE Joint Working Group on Forecast Verification Research’’ and online available at:
http://www.cawcr.gov.au/projects/verification/#Methods_for_dichotomous_forecasts). For comparison, we
used hourly data. These 1 h averages of measured rain rates were compared to hourly interpolated fields.
We allowed 65 km distance between ship and grid point for AMSR-E, TMI, and SSMIS. For ERA-Interim, we
used 120 km and for NCEP2, 240 km according to the data sets spatial resolution. The bias score close to 1
for AMSR-E and SSMIS indicates similar rain probabilities as observed, while rain probabilities for NCEP2 are
more than four times and that of ERA-Interim 14 times higher than observed. For TMI data, we estimated a
bias score of 0.7. The statistical analysis suggests that we have to take into account that precipitation is a
rare event. In this case, an estimate of the performance is the so-called threat score or critical success index
(CSI) instead of the correct proportion. The CSI is about 0.3 for the AMSR-E and SSMIS data compared to less
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7906
than 0.15 for NCEP2, ERA-Interim, and TMI. These results indicate that SSMIS and AMSR-E data give the
most reliable information about precipitation. This is supported by biases in average rain rates, which are of
the order of 215% to130% for SSMIS and AMSR-E, 270% for TMI, 1110% for NCEP2, and even 1250% for
ERA-Interim.
We computed daily mean heat fluxes and evaporation along the cruise track of the research vessel using mete-
orological data for the estimation of the turbulent heat fluxes and evaporation and the bulk transfer coefficients
of Bumke et al. [2014]. They estimated the coefficients after the inertial dissipation method and compared their
calculated fluxes with fluxes determined after the COARE algorithm [Fairall et al., 2003]. All considered atmos-
pheric products were interpolated on a regular 131grid and averaged over 24 h. The grid boxes on the 1
31grid covered by the vessel on 1 day were then averaged. The comparison of the statistical parameters
(bias, standard deviation, and rms difference) of the several products against the in situ shipboard measure-
ments (Table 1) indicate larger than observed latent heat fluxes into the atmosphere in the equatorial regions in
the reanalysis products. This bias was also found in a validation study of Kubota et al. [2003].
The heat fluxes and the evaporation from TropFlux and the precipitation from AMSR-E were chosen for fur-
ther calculations within this study.
Appendix C: Comparison of Satellite SST and SSS With Observations
All available surface data from glider, float, and CTD measurements as well as hourly averaged thermosali-
nograph observations were used to compare in situ and satellite data for the period from 7 May to 11 July
2011 (Figure 3). The maximum distance between the closest satellite grid point, from 3 day mean SST and
SSS satellite images, and in situ measurements is 1/6or 10 nm. Satellite and in situ SST data agree well
(Figure 3a) with a correlation coefficient of 0.98. We estimated (and corrected for the analysis) an offset of
20.17C (SST
TMI
– SST
in situ
) with a standard deviation of 0.52C. As expected for the SSS (Figure 3b), the cor-
relation was lower, with a correlation coefficient of about 0.77. The local nature of freshwater anomalies,
due to the lack of an air/sea feedback, is expected to be one reason, but also the differences in the observed
ocean volume contribute. For example, the dispersion of a freshwater anomaly due to a local rainfall event
will be very much confined to the upper few centimeters of the water column and may not be recognized
in the near-surface in situ data typically taken below 1 m depth [see also, e.g., Boutin et al., 2013]. This can
be evidenced when evaluating SMOS SSS observations coinciding (within 3 day averages) with precipitation
events (rain rate 0.1 mm h
21
) as observed by AMSR-E. These data points indicate a mean bias of MLS
compared to satellite SSS of 20.39 (SSS
SMOS
– SSS
in situ
). Since only 8% of the SSS data coincide with precipi-
tation data, the total bias reduces to 20.11. Hence, for all satellite SSS data points, an offset of 0.11 was
added and a standard deviation of 0.34 was applied.
Appendix D: Velocity Validation
Johnson et al. [2007] found significant correlations (0.5–0.8) between the OSCAR zonal velocities and zonal
velocities from moored current meters (MCMs: MCMs are a combination of mechanical current meters and
mADCPs), mADCPs, drifters, and shipboard current profilers for the near-equatorial region of the Pacific.
Meridional velocity is significantly less correlated compared to the correlation of the zonal velocity. The
Atlantic Ocean was also examined but not shown in the paper and showed similar correlations. Foltz et al.
[2012] found similar results at two mooring sites (4N and 12N) at 23W in the equatorial North Atlantic by
comparing 5 day averaged velocity data of OSCAR and MCMs. At 4N, they found a correlation of 0.8 for the
zonal velocity and no correlation for the meridional velocity.
For the equatorial mooring at 10W, we found correlations of 0.43 for the zonal velocity and 20.1 for the
meridional velocity by applying the same method as Johnson et al. [2007] with 5 day averaged data and
using almost 13 months of upward looking mADCP profiles. Data in the upper 10–12 m of these profiles
affected by surface reflections were discarded. To fill the gap of the upper 10–12 m, we extrapolated the
profiles to the surface assuming constant shear. Afterward, the upper 30 m were averaged via trapezoidal
integration.
The RMS of the differences was estimated with 0.21 m s
21
for the zonal velocity and 0.06 m s
21
for the
meridional velocity. The same comparison for the same time period was done at the equatorial mooring at
Acknowledgments
Most of the data used in this study,
obtained during research cruises, are
stored and available from June 2014
ongoing under https://portal.geomar.
de/web/guest/kdmi. Argo data were
collected and made freely available by
the International Argo Project and the
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7907
23W. A correlation between the zonal velocities of 0.66 and between the meridional velocities of 20.05
showed a slightly better correlation for the zonal component than at the equatorial mooring at 10W. The
RMS difference was estimated with 0.17 m s
21
. For the meridional velocities, a RMS of the differences of
0.03 m s
21
was calculated. Finally, the mean of the RMS of the differences at both sites was taken as the
uncertainty for the horizontal velocities in the ACT box, leading to 0.19 m s
21
for the zonal velocity and
0.04 m s
21
for the meridional velocity. The absence of correlation for the meridional velocity at both moor-
ing sites highlights the fact that the OSCAR velocities do not capture TIWs. In the northern box, the uncer-
tainty estimates of Foltz et al. [2013] for the mooring at 4N, 23W were chosen, with a zonal uncertainty of
0.10 m s
21
and a meridional uncertainty of 0.08 m s
21
. They compared monthly OSCAR velocities with
PIRATA currents (combination of ADCP and MCM) at three mooring sites (4N, 11.5N, and 20.5N) at 23W.
They excluded the equatorial mooring, for which we found a large discrepancy between zonal velocities
from OSCAR and moored observations.
References
Allen, M. R., and W. J. Ingram (2002), Constraints on future changes in climate and the hydrologic cycle, Nature,419(6903), 224–232, doi:
10.1038/nature01092.
Alory, G., C. Maes, T. Delcroix, N. Reul, and S. Illig (2012), Seasonal dynamics of sea surface salinity off Panama: The far Eastern Pacific fresh
pool, J. Geophys. Res.,117, C04028, doi:10.1029/2011JC007802.
Berger, M., et al. (2002), Measuring ocean salinity with ESA’s SMOS mission—Advancing the science, ESA Bull.,111, 113–121.
Bingham, F. M., G. R. Foltz, and M. J. McPhaden (2012), Characteristics of the seasonal cycle of surface layer salinity in the global ocean,
Ocean Sci.,8(5), 915–929, doi:10.5194/os-8-915-2012.
Bonjean, F., and G. S. E. Lagerloef (2002), Diagnostic model and analysis of the surface currents in the tropical Pacific Ocean, J. Phys. Ocean-
ogr.,32(10), 2938–2954, doi:10.1175/1520-0485(2002)032<2938:DMAAOT>2.0.CO;2.
Bourles, B., et al. (2008), The PIRATA program: History, accomplishments, and future directions, Bull. Am. Meteorol. Soc.,89(8), 1111–1125,
doi:10.1175/2008BAMS2462.1.
Boutin, J., N. Martin, Y. Xiaobin, J. Font, N. Reul, and P. Spurgeon (2012), First assessment of SMOS data over open ocean: Part II-sea surface
salinity, IEEE Trans. Geosci. Remote Sens.,50(5), 1662–1675, doi:10.1109/TGRS.2012.2184546.
Boutin, J., N. Martin, G. Reverdin, X. Yin, and F. Gaillard (2013), Sea surface freshening inferred from SMOS and ARGO salinity: Impact of
rain, Ocean Sci.,9(1), 183–192, doi:10.5194/os-9-183-2013.
Brandt, P., F. A. Schott, C. Provost, A. Kartavtseff, V. Hormann, B. Bourle
`s, and J. Fischer (2006), Circulation in the central equatorial Atlantic:
Mean and intraseasonal to seasonal variability, Geophys. Res. Lett.,33, L07609, doi:10.1029/ 2005GL025498.
Brandt, P., G. Caniaux, B. Bourle
`s, A. Lazar, M. Dengler, A. Funk, V. Hormann, H. Giordani, and F. Marin (2011), Equatorial upper-ocean
dynamics and their interaction with the West African monsoon, Atmos. Sci. Lett.,12(1), 24–30, doi:10.1002/asl.287.
Bumke, K., and J. Seltmann (2012), Analysis of measured drop size spectra over land and sea, ISRN Meteorol.,2012, 10 pp., doi:10.5402/
2012/296575.
Bumke, K., K. Fennig, A. Strehz, R. Mecking, and M. Schr
oder (2012), HOAPS precipitation validation with ship-borne rain gauge measure-
ments over the Baltic Sea, Tellus, Ser. A,64, 18486, 9 pp., doi:10.3402/tellusa.v64i0.18486.
Bumke, K., M. Schlundt, J. Kalisch, A. Macke, and H. Kleta (2014), Measured and parameterized energy fluxes estimated for Atlantic transects
of R/V Polarstern, J. Phys. Oceanogr.,44(2), 482–491, doi:10.1175/JPO-D-13-0152.1.
Caniaux, G., H. Giordani, J.-L. Redelsperger, F. Guichard, E. Key, and M. Wade (2011), Coupling between the Atlantic cold tongue and the
West African monsoon in boreal spring and summer, J. Geophys. Res.,116, C04003, doi:10.1029/2010JC006570.
Carton, J. A., and Z. Zhou (1997), Annual cycle of sea surface temperature in the tropical Atlantic Ocean, J. Geophys. Res.,102(C13), 27,813–
27,824, doi:10.1029/97JC02197.
Da-Allada, C. Y., G. Alory, Y. du Penhoat, E. Kestenare, F. Durand, and N. M. Hounkonnou (2013), Seasonal mixed-layer salinity balance in
the tropical Atlantic Ocean: Mean state and seasonal cycle, J. Geophys. Res. Oceans,118, 1–15, doi:10.1029/2012JC008357.
de Boyer Mont
egut, C., G. Madec, A. S. Fischer, A. Lazar, and D. Iudicone (2004), Mixed layer depth over the global ocean: An examination
of profile data and a profile-based climatology, J. Geophys. Res.,109, C12003, doi:10.1029/2004JC002378.
de Boyer Mont
egut, C., J. Mignot, A. Lazar, and S. Cravatte (2007), Control of salinity on the mixed layer depth in the world ocean: 1. Gen-
eral description, J. Geophys. Res.,112, C06011, doi:10.1029/2006JC003953.
Dee, D. P., et al. (2011), The ERA-Interim reanalysis: configuration and performance of the data assimilation system, Q. J. R. Meteorol. Soc.,
137(656), 553–597, doi:10.1002/qj.828.
Delcroix, T., and C. H
enin (1991), Seasonal and interannual variations of sea surface salinity in the tropical Pacific Ocean, J. Geophys. Res.,
96(C12), 22,135–22,150, doi:10.1029/91JC02124.
Deser, C., A. S. Phillips, and M. A. Alexander (2010), Twentieth century tropical sea surface temperature trends revisited, Geophys. Res. Lett.,
37, L10701, doi:10.1029/2010GL043321.
Dessier, A., and J. R. Donguy (1994), The sea surface salinity in the tropical Atlantic between 10S and 30N—Seasonal and interannual var-
iations (1977–1989), Deep Sea Res., Part I,41(1), 81–100, doi:10.1016/0967-0637(94)90027-2.
Duing, W., P. Hisard, E. Katz, J. Meincke, L. Miller, K. V. Moroshkin, G. Philander, A. A. Ribnikov, K. Voigt, and R. Weisberg (1975), Meanders
and long waves in the equatorial Atlantic, Nature,257(5524), 280–284.
Durack, P. J., and S. E. Wijffels (2010), Fifty-year trends in global ocean salinities and their relationship to broad-scale warming, J. Clim.,
23(16), 4342–4362, doi:10.1175/2010JCLI3377.1.
Fairall, C. W., E. F. Bradley, J. E. Hare, A. A. Grachev, and J. B. Edson (2003), Bulk parameterization of air-sea fluxes: Updates and verification
for the COARE algorithm, J. Clim.,16(4), 571–591, doi:10.1175/1520-0442(2003)016<0571:BPOASF>2.0.CO;2.
Ferrari, R., and K. L. Polzin (2005), Finescale structure of the T-S relation in the eastern North Atlantic, J. Phys. Oceanogr.,35(8), 1437–1454,
doi:10.1175/JPO2763.1.
Foltz, G. R., and M. J. McPhaden (2008), Seasonal mixed layer salinity balance of the tropical North Atlantic Ocean, J. Geophys. Res.,113,
C02013, doi:10.1029/2007JC004178.
national initiatives that contribute to it
(http://www.argo.net). Argo is a pilot
program of the Global Ocean
Observing System. PIRATA mooring
data were provided by the TAO Project
Office of NOAA/PMEL (http://www.
pmel.noaa.gov/tao/disdel/disdel.html)
and PIRATA cruise data from IRD/
LEGOS (http://www.brest.ird.fr/pirata/
pirata_cruises.php). OSCAR data were
downloaded from www.oscar.noaa.
gov. NCEP data were downloaded
from www.esrl.noaa.gov. ERA-Interim
data were downloaded from http://
data-portal.ecmwf.int/data/d/interim_
full_daily/. The drifter data were
downloaded from the Drifter Assembly
Center (DAC; www.aoml.noaa.gov/
phod/dac/). YOMAHA’07 data were
downloaded from http://apdrc.soest.
hawaii.edu/projects/yomaha/. The
JAMSTEC Argo salinity data were
downloaded from http://www.jamstec.
go.jp/ARGO/J_ARGOe.html. TMI and
SSMIS data were produced by Remote
Sensing Systems and sponsored by the
NASA Earth Science MEaSUREs
DISCOVER Project. AMSR-E data were
produced by Remote Sensing Systems
and sponsored by the NASA Earth
Science MEaSUREs DISCOVER Project
and the AMSR-E Science Team. TMI,
SSMIS, and AMSR-E data are available
at www.remss.com. The SMOS SSS
data were produced by the Barcelona
Expert Centre (www.smos-bec.icm.csic.
es), a joint initiative of the Spanish
Research Council (CSIC) and the
Technical University of Catalonia (UPC),
mainly founded by the Spanish
National Program on Space. The
TropFlux data are produced under a
collaboration between Laboratoire
d’Oc
eanographie: Exp
erimentation et
Approches Num
eriques (LOCEAN)
from Institut Pierre Simon Laplace
(IPSL, Paris, France) and National
Institute of Oceanography/CSIR (NIO,
Goa, India), and supported by Institut
de Recherche pour le D
eveloppement
(IRD, France). TropFlux relies on data
provided by the ECMWF Re-Analysis
interim (ERA-I) and ISCCP projects and
was downloaded from www.locean-
ipsl.upmc.fr/tropflux/overview.html.
OAFlux data were downloaded from
http://oaflux.whoi.edu/data.html. Mer-
cator model output was provided by
Mercator Ocean via contract 2011/SG/
CUTD/56 and downloaded from ftp://
ftp.mercator-ocean.fr. This research
was funded by the BMBF grants
Nordatlantik (03F0443B), RACE
(03F0651B), SOPRAN (03F0662A), and
SACUS (03G0837A) and by European
Union 7th Framework Programme (FP7
2007–2013) under grant agreement
284321 GROOM and 603521 PREFACE
Project. We thank Arne K
ortzinger, the
captain and crew of the R/V Maria S.
Merian as well as our technical group
for their help with the fieldwork.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7908
Foltz, G. R., S. A. Grodsky, J. A. Carton, and M. J. McPhaden (2003), Seasonal mixed layer heat budget of the tropical Atlantic Ocean, J. Geo-
phys. Res.,108(C5), 3146, doi:10.1029/2002JC001584.
Foltz, G. R., S. A. Grodsky, J. A. Carton, and M. J. McPhaden (2004), Seasonal salt budget of the northwestern tropical Atlantic Ocean along
38W, J. Geophys. Res.,109, C03052, doi:10.1029/2003JC002111.
Foltz, G. R., M. J. McPhaden, and R. Lumpkin (2012), A strong Atlantic meridional mode event in 2009: The role of mixed layer dynamics, J.
Clim.,25(1), 363–380, doi:10.1175/JCLI-D-11-00150.1.
Foltz, G. R., C. Schmid, and R. Lumpkin (2013), Seasonal cycle of the mixed layer heat budget in the northeastern tropical Atlantic Ocean, J.
Clim.,26(20), 8169–8188, doi:10.1175/JCLI-D-13-00037.1.
Font, J., et al. (2012), SMOS first data analysis for sea surface salinity determination, Int. J. Remote Sens.,34(9-10), 3654–3670, doi:10.1080/
01431161.2012.716541.
Garau, B., S. Ruiz, W. F. G. Zhang, A. Pascual, E. Heslop, J. Kerfoot, and J. Tintore (2011), Thermal lag correction on slocum CTD glider data, J.
Atmos. Oceanic Technol.,28(9), 1065–1071, doi:10.1175/JTECH-D-10-05030.1.
Garraffo, Z. D., A. J. Mariano, A. Griffa, C. Veneziani, and E. P. Chassignet (2001), Lagrangian data in a high-resolution numerical simulation
of the North Atlantic: I. Comparison with in situ drifter data, J. Mar. Syst.,29(1–4), 157–176, doi:10.1016/S0924-7963(01)00015-X.
Giordani, H., G. Caniaux, and A. Voldoire (2013), Intraseasonal mixed-layer heat budget in the equatorial Atlantic during the cold tongue
development in 2006, J. Geophys. Res. Oceans,118, 650–671, doi:10.1029/2012JC008280.
Goes, M., G. Goni, V. Hormann, and R. C. Perez (2013), Variability of the Atlantic off-equatorial eastward currents during 1993–2010 using a
synthetic method, J. Geophys. Res. Oceans,118, 3026–3045, doi:10.1002/jgrc.20186.
Gouriou, Y., and G. Reverdin (1992), Isopycnal and diapycnal circulation of the upper equatorial Atlantic Ocean in 1983–1984, J. Geophys.
Res.,97(C3), 3543–3572, doi:10.1029/91JC02935.
Grodsky, S. A., J. A. Carton, and C. R. McClain (2008), Variability of upwelling and chlorophyll in the equatorial Atlantic, Geophys. Res. Lett.,
35, L03610, doi:10.1029/2007GL032466.
Großklaus, M., K. Uhlig, and L. Hasse (1998), An optical disdrometer for use in high wind speeds, J. Atmos. Oceanic Technol.,15(4), 1051–
1059, doi:10.1175/1520-0426(1998)015<1051:AODFUI>2.0.CO;2.
Hall, A., and S. Manabe (1997), Can local linear stochastic theory explain sea surface temperature and salinity variability?, Clim. Dyn.,13(3),
167–180, doi:10.1007/s003820050158.
Hasse, L., M. Grossklaus, K. Uhlig, and P. Timm (1998), A ship rain gauge for use in high wind speeds, J. Atmos. Oceanic Technol.,15(2), 380–
386, doi:10.1175/1520-0426(1998)015<0380:ASRGFU>2.0.CO;2.
Hisard, P., and A. Morlie
`re (1973), La terminaison du contre courant
equatorial subsuperficiel Atlantique (courant de Lomonosov) dans le
golfe de Guin
ee, Cah. O. R. S. T. O. M., Ser. Oceanogr.,11(4), 455–464.
Hosoda, S., T. Ohira, and T. Nakamura (2008), A monthly mean dataset of global oceanic temperature and salinity derived from Argo float
observations, JAMSTEC Rep. Res. Div.,8, 47–59.
Hummels, R., M. Dengler, and B. Bourle
`s (2013), Seasonal and regional variability of upper ocean diapycnal heat flux in the Atlantic cold
tongue, Prog. Oceanogr.,111, 52–74, doi:10.1016/j.pocean.2012.11.001.
Hummels, R., M. Dengler, P. Brandt, and M. Schlundt (2014), Diapycnal heat flux and mixed layer heat budget within the Atlantic Cold
Tongue, Clim. Dyn., doi:10.1007/s00382-014-2339-6, in press.
Jochum, M., M. F. Cronin, W. S. Kessler, and D. Shea (2007), Observed horizontal temperature advection by tropical instability waves, Geo-
phys. Res. Lett.,34, L09604, doi:10.1029/2007GL029416.
Johns, W. E., P. Brandt, B. Bourles, A. Tantet, A. Papapostolou, and A. Houk (2014), Zonal structure and seasonal variability of the Atlantic
equatorial undercurrent, Clim. Dyn., 1–22, doi:10.1007/s00382-014-2136-2, in press.
Johnson, E. S., F. Bonjean, G. S. E. Lagerloef, J. T. Gunn, and G. T. Mitchum (2007), Validation and error analysis of OSCAR sea surface cur-
rents, J. Atmos. Oceanic Technol.,24(4), 688–701, doi:10.1175/JTECH1971.1.
Jouanno, J., F. Marin, Y. du Penhoat, J. Sheinbaum, and J.-M. Molines (2011a), Seasonal heat balance in the upper 100 m of the equatorial
Atlantic Ocean, J. Geophys. Res.,116, C09003, doi:10.1029/2010JC006912.
Jouanno, J., F. Marin, Y. du Penhoat, J.-M. Molines, and J. Sheinbaum (2011b), Seasonal modes of surface cooling in the Gulf of Guinea, J.
Phys. Oceanogr.,41(7), 1408–1416, doi:10.1175/JPO-D-11-031.1.
Kalisch, J., and A. Macke (2012), Radiative budget and cloud radiative effect over the Atlantic from ship-based observations, Atmos. Meas.
Tech.,5(10), 2391–2401, doi:10.5194/amt-5-2391-2012.
Kanamitsu, M., W. Ebisuzaki, J. Woollen, S. K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter (2002), NCEP-DOE AMIP-II reanalysis (R-2), Bull. Am.
Meteorol. Soc.,83(11), 1631–1643, doi:10.1175/BAMS-83-11-1631.
Kolodziejczyk, N., B. Bourle
`s, F. Marin, J. Grelet, and R. Chuchla (2009), Seasonal variability of the Equatorial Undercurrent at 10Was
inferred from recent in situ observations, J. Geophys. Res.,114, C06014, doi:10.1029/2008JC004976.
Kolodziejczyk, N., F. Marin, B. Bourle
`s, Y. Gouriou, and H. Berger (2014), Seasonal variability of the equatorial undercurrent termination and
associated salinity maximum in the Gulf of Guinea, Clim. Dyn., doi:10.1007/s00382-014-2107-7, in press.
Kubota, M., A. Kano, H. Muramatsu, and H. Tomita (2003), Intercomparison of various surface latent heat flux fields, J. Clim.,16(4), 670–678,
doi:10.1175/1520-0442(2003)016<0670:IOVSLH>2.0.CO;2.
Lagerloef, G. S. E., G. T. Mitchum, R. B. Lukas, and P. P. Niiler (1999), Tropical Pacific near-surface currents estimated from altimeter, wind,
and drifter data, J. Geophys. Res.,104(C10), 23,313–23,326, doi:10.1029/1999JC900197.
Lagerloef, G. S. E., et al. (2008), The Aquarius/SAC-D mission: Designed to meet the salinity remote-sensing challenge, Oceanography,21(1),
68–81, doi:10.5670/oceanog.2008.68.
Lebedev, K. V., H. Yoshinari, N. A. Maximenko, and P. W. Hacker (2007), YoMaHa’07: Velocity data assessed from trajectories of Argo floats
at parking level and at the sea surface, IPRC Tech. Note,4(2), 1–16.
Lee, T., G. Lagerloef, M. M. Gierach, H.-Y. Kao, S. Yueh, and K. Dohan (2012), Aquarius reveals salinity structure of tropical instability waves,
Geophys. Res. Lett.,39, L12610, doi:10.1029/2012GL052232.
Legeckis, R. (1977), Long waves in the Eastern Equatorial Pacific Ocean: A view from a geostationary satellite, Science,197(4309), 1179–
1181, doi:10.1126/science.197.4309.1179.
Lellouche, J. M., et al. (2013), Evaluation of global monitoring and forecasting systems at Mercator Oc
ean, Ocean Sci.,9(1), 57–81, doi:
10.5194/os-9-57-2013.
Lien, R. C., E. A. D’Asaro, and C. E. Menkes (2008), Modulation of equatorial turbulence by tropical instability waves, Geophys. Res. Lett.,35,
L24607, doi:10.1029/2008GL035860.
Lumpkin, R., and S. L. Garzoli (2005), Near-surface circulation in the Tropical Atlantic Ocean, Deep Sea Res., Part I,52(3), 495–518, doi:
10.1016/j.dsr.2004.09.001.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7909
Lumpkin, R., A. M. Treguier, and K. Speer (2002), Lagrangian eddy scales in the Northern Atlantic Ocean, J. Phys. Oceanogr.,32(9), 2425–
2440, doi:10.1175/1520-0485(2002)032<2425:LESITN>2.0.CO;2.
Moum, J. N., D. R. Caldwell, and C. A. Paulson (1989), Mixing in the equatorial surface layer and thermocline, J. Geophys. Res.,94(C2), 2005–
2021, doi:10.1029/JC094iC02p02005.
Oakey, N. S. (1982), Determination of the rate of dissipation of turbulent energy from simultaneous temperature and velocity shear micro-
structure measurements, J. Phys. Oceanogr.,12(3), 256–271, doi:10.1175/1520-0485(1982)012<0256:DOTROD>2.0.CO;2.
Osborn, T. R. (1980), Estimates of the local rate of vertical diffusion from dissipation measurements, J. Phys. Oceanogr.,10(1), 83–89, doi:
10.1175/1520-0485(1980)010<0083:EOTLRO>2.0.CO;2.
Osborn, T. R., and C. S. Cox (1972), Oceanic fine structure, Geophys. Fluid Dyn.,3(1), 321–345, doi:10.1080/03091927208236085.
Perez, R. C., V. Hormann, R. Lumpkin, P. Brandt, W. E. Johns, F. Hernandez, C. Schmid, and B. Bourle
`s (2014), Mean meridional currents in
the central and eastern equatorial Atlantic, Clim. Dyn., doi:10.1007/s00382-013-1968-5, in press.
Peter, A.-C., M. Le H
enaff, Y. du Penhoat, C. E. Menkes, F. Marin, J. Vialard, G. Caniaux, and A. Lazar (2006), A model study of the seasonal
mixed layer heat budget in the equatorial Atlantic, J. Geophys. Res.,111, C06014, doi:10.1029/2005JC003157.
Peters, H., M. C. Gregg, and J. M. Toole (1988), On the parameterization of equatorial turbulence, J. Geophys. Res.,93(C2), 1199–1218, doi:
10.1029/JC093iC02p01199.
Philander, S. G. H. (1978), Instabilities of zonal equatorial currents, 2, J. Geophys. Res.,83(C7), 3679–3682, doi:10.1029/JC083iC07p03679.
Philander, S. G. H., and R. C. Pacanowski (1981), The oceanic response to cross-equatorial winds (with application to coastal upwelling in
low latitudes), Tellus,33(2), 201–210, doi:10.1111/j.2153-3490.1981.tb01744.x.
Philander, S. G. H., D. Gu, G. Lambert, T. Li, D. Halpern, N. C. Lau, and R. C. Pacanowski (1996), Why the ITCZ is mostly north of the equator,
J. Clim.,9(12), 2958–2972, doi:10.1175/1520-0442(1996)009<2958:WTIIMN>2.0.CO;2.
Prandke, H., and A. Stips (1998), Test measurements with an operational microstructure-turbul ence profiler: Detection limit of dissipation
rates, Aquat. Sci.,60(3), 191–209, doi:10.1007/s000270050036.
Praveen Kumar, B., J. Vialard, M. Lengaigne, V. S. N. Murty, and M. J. McPhaden (2012), TropFlux: Air-sea fluxes for the global tropical
oceans—Description and evaluation, Clim. Dyn.,38(7-8), 1521–1543, doi:10.1007/s00382-011-1115-0.
Reul, N., J. Tenerelli, J. Boutin, B. Chapron, F. Paul, E. Brion, F. Gaillard, and O. Archer (2012), Overview of the first SMOS sea surface salinity
products. Part I: Quality assessment for the second half of 2010, IEEE Trans. Geosci. Remote Sens.,50(5), 1636–1647, doi:10.1109/
TGRS.2012.2188408.
Richardson, P. L., and G. Reverdin (1987), Seasonal cycle of velocity in the Atlantic North Equatorial Countercurrent as measured by surface
drifters, current meters, and ship drifts, J. Geophys. Res.,92(C4), 3691–3708, doi:10.1029/JC092iC04p03691.
Schafstall, J., M. Dengler, P. Brandt, and H. Bange (2010), Tidal-induced mixing and diapycnal nutrient fluxes in the Mauritanian upwelling
region, J. Geophys. Res.,115, C10014, doi:10.1029/2009JC005940.
Schmitt, R. W., J. R. Ledwell, E. T. Montgomery, K. L. Polzin, and J. M. Toole (2005), Enhanced diapycnal mixing by salt fingers in the thermo-
cline of the tropical Atlantic, Science,308(5722), 685–688, doi:10.1126/science.1108678.
Stevenson, J. W., and P. P. Niiler (1983), Upper ocean heat budget during the Hawaii-to-Tahiti Shuttle experiment, J. Phys. Oceanogr.,
13(10), 1894–1907, doi:10.1175/1520-0485(1983)013<1894:UOHBDT>2.0.CO;2.
Swenson, M. S., and D. V. Hansen (1999), Tropical Pacific ocean mixed layer heat budget: The Pacific cold tongue, J. Phys. Oceanogr.,29(1),
69–81, doi:10.1175/1520-0485(1999)029<0069:TPOMLH>2.0.CO;2.
Taylor, J. P., J. M. Edwards, M. D. Glew, P. Hignett, and A. Slingo (1996), Studies with a flexible new radiation code. II: Comparisons with air-
craft short-wave observations, Q. J. R. Meteorol. Soc.,122(532), 839–861, doi:10.1002/qj.49712253204.
Tzortzi, E., S. A. Josey, M. Srokosz, and C. Gommenginger (2013), Tropical Atlantic salinity variability: New insights from SMOS, Geophys. Res.
Lett.,40, 1–5, doi:10.1002/grl.50225.
von Schuckmann, K., P. Brandt, and C. Eden (2008), Generation of tropical instability waves in the Atlantic Ocean, J. Geophys. Res.,113,
C08034, doi:10.1029/2007JC004712.
Wade, M., G. Caniaux, and Y. du Penhoat (2011), Variability of the mixed layer heat budget in the eastern equatorial Atlantic during 2005–
2007 as inferred using Argo floats, J. Geophys. Res.,116, C08006, doi:10.1029/2010JC006683.
Wang, W. M., and M. J. McPhaden (1999), The surface-layer heat balance in the equatorial Pacific Ocean. Part I: Mean seasonal cycle, J.
Phys. Oceanogr.,29(8), 1812–1831, doi:10.1175/1520-0485(1999)029<1812:TSLHBI>2.0.CO;2.
Xie, S.-P., C. Deser, G. A. Vecchi, J. Ma, H. Teng, and A. T. Wittenberg (2010), Global warming pattern formation: Sea surface temperature
and rainfall, J. Clim.,23(4), 966–986, doi:10.1175/2009JCLI3329.1.
Yu, L., X. Jin, and R. A. Weller (2008), Multidecade global flux datasets from the Objectively Analyzed Air-sea Fluxes (OAFlux) project: Latent
and sensible heat fluxes, ocean evaporation, and related surface meteorological variables, OAFlux Project Tech. Rep. OA-2008-01,pp.1
64, Woods Hole Oceanogr. Inst, Woods Hole, Mass.
Journal of Geophysical Research: Oceans 10.1002/2014JC010021
SCHLUNDT ET AL. V
C2014. American Geophysical Union. All Rights Reserved. 7910
... e W is the entrainment velocity (Equation 3) at the base of the mixed layer. Only positive vertical velocities are considered and the detrainment (downward movement, negative velocities) is set to zero, since the water mass that flows out from the base of the mixed layer has the same properties as the water in the mixed layer, and does not affect the salinity in the MLS (Da-allada et al., 2013;Foltz et al., 2004;Ren & Riser, 2009;Ren et al., 2011;Schlundt et al., 2014). This constraint is setup with the use of the Heaviside step function: ...
... Indeed, the increased horizontal shear between the EUC and the westward flowing SEC may intensify vertical mixing within the superficial layer (Da-Allada et al., 2017), thus increasing the salinity of the surface layers. This period corresponds to the peak season of the Atlantic Cold Tongue characterized by the presence of cold and salty water in the GG (Brandt et al., 2011;Da-Allada et al., 2017;Schlundt et al., 2014). ...
Article
Full-text available
In the eastern Gulf of Guinea (GG), freshwater originated from rivers discharges into the ocean and high precipitation rate are key contributors to the upper ocean vertical density stratification, and play a key role in modulating local air‐sea interactions as well as biogeochemical cycle. Nevertheless, the dynamics of the GG freshwater plumes remain poorly documented because of the scarcity of historical, in situ observations and the lack of an ad hoc satellite‐based analysis in this region. Recent advances in remote sensing capabilities from the Soil Moisture and Ocean Salinity (SMOS) satellite mission offer unprecedented coverage and spatiotemporal resolution of Sea Surface Salinity (SSS) in the GG. Using SMOS SSS and available in situ measurements, the seasonal variability of freshwater plumes and associated physical mechanisms controlling their seasonal cycle are presented and analyzed. Freshwater plumes in the GG follow two dynamical regimes. They present maximum offshore extension during boreal winter and exhibit minimum signature during summer. In the northeastern GG, SSS variability is mainly explained by high precipitation rate and Niger River runoff during winter, while during late summer, SSS is mainly driven by horizontal advection. In contrast, southeast of GG, freshwater plumes are mainly supplied by Congo River runoff. From September to March, SSS variability is driven by zonal advection, with a major contribution from Ekman wind‐driven currents. During spring‐summer, the observed SSS increase is likely explained by entrainment and vertical mixing. SSS budget and freshwater advection processes are discussed in the context of the shallow stratification induced by freshwater.
... Ocean-atmosphere interaction is a major cause of climate variability and is affected by climate change through global warming; energy exchange through air-sea processes; and changes in planetary, regional, and local meteorological processes [1]. Anthropogenic greenhouse gas emissions have resulted in an energy imbalance of about 0.4-1 W/m 2 on Earth, most of which accumulates as heat in the ocean [2]. ...
... No significant difference between the MLD and ILD was observed in the high-resolution glider data at any site. Therefore, the simpler temperature criterion was chosen for our definition of MLD [1]. The mixed layer depth (MLD) was calculated using [32,33]. ...
Article
We observed interannual changes in the temperature and salinity of the surface layer of the Adriatic Sea when measured during the period 2005–2020. We observed non-stationarity and a positive linear trend in the series of mixed layer depth, heat storage, and potential energy anomalies. This non-stationarity was related to the climate regime that prevailed between 2011 and 2017. We observed significant changes in the interannual variability of salinity above and below the mixed layer depth and a positive difference in the surface barrier layer. In an effort to reconstruct the cause of this phenomenon, a multi-stage investigation was conducted. The first suspected culprit was the change in wind regime over the Mediterranean and Northeast Atlantic regions in September. Using the growing neural gas algorithm, September wind fields over the past 40 years were classified into nine distinct patterns. Further analysis of the CTD data indicated an increase in heat storage, a physical property of the Adriatic Sea known to be strongly influenced by the inflow of warm water masses controlled by the bimodal oscillating system (BiOS). The observed increase in salinity confirmed the assumption that BiOS activity affects heat storage. Unexpectedly, this analysis showed that an inverse vertical salinity profile was present during the summer months of 2015, 2017, and 2020, which can only be explained by salinity changes being a dominant factor. In addition, the aforementioned wind regime caused an increase in energy loss through latent energy dissipation, contributing to an even larger increase in salinity. While changes in the depth of the mixed layer in the Adriatic are usually due to temperature changes, this phenomenon was primarily caused by abrupt changes in salinity due to a combination of BiOS and local factors. This is the first record of such an event.
... Q net is the net air-sea flux (such that positive fluxes are directed into the ocean), and ρ and c p (1027kg m −3 and 3850J (kg C) −1 ) are the water density and specific heat capacity, respectively. The heat loss by the shortwave radiation that penetrates below the depth h is given by q(h) assuming an exponential decay of surface shortwave radiation with 25 m e-folding depth (Wang and McPhaden, 1999;Schlundt et al., 2014). The residual term accounts for vertical advection and horizontal and vertical diffusion. ...
Article
Full-text available
Marine heatwaves (MHWs) are extreme warming events that can result in significant damage to marine ecosystems and local economies. The primary drivers of these events have been frequently studied using an upper ocean heat budget. However, various surface mixed layer (SML) depths have been used with little attention paid to the impact of the depth chosen on heat budget term estimates. We analyse MHW drivers in two dynamically contrasting regions off the east coast of Australia (East Australian Current extension) and the west coast of New Zealand over a 30-year period (1985–2014, inclusive). We compare the magnitude of the air-sea heatflux and advection terms in a volume-averaged heat budget using three different SML depth estimates. We show that the SML depth over which the heat budget is calculated has direct consequences on the identification of MHW dominant drivers. The air-sea heatflux term is amplified when the SML depth is underestimated and dampened when overestimated. The variation in the magnitude of the advection term is dependent on the barotropic or baroclinic structure of the currents. We, also, show that the impact on MHW driver classification is both temporally and regionally dependent. Generally, a deep SML estimate results in more MHWs being classified as advection and less classified as air-sea heatflux-driven. However, during the cool months, a shallow estimate produces the opposite pattern and to a varying degree of intensity depending on the region's dynamics. Use of daily and spatially variable SML depth in a heat budget calculation allows the comparison between regions with different dynamics influencing the mixed layer depth. These results show that when using a heat budget approach to explore marine heatwaves over extended time and space (e.g., regions and seasons), it is imperative to consider the temporal and spatial variability in the SML depth.
... Q net is the net air-sea flux (such that positive fluxes are directed into the ocean), and  and p c (1027 3 kg m  and 3850 1 J kgC ( )  ) are the water density and specific heat capacity, respectively. The heat loss by the shortwave radiation that penetrates below the mixed layer is given by ( ) q h assuming an exponential decay of surface shortwave radiation with 25 m e-folding depth (Schlundt et al., 2014;Wang & McPhaden, 1999) and h is the daily varying mixed layer depth. Residual is the sum of vertical advection and horizontal and vertical diffusion. ...
Article
Full-text available
Plain Language Summary Marine heatwaves (MHWs) are extreme ocean warming events that can have devastating ecological impacts on surface and deep ocean communities, yet knowledge of what drives their depth is limited. Here, we identify drivers and characteristics of MHWs on the surface and at depth over the 1985–2014 period in four dynamically different regions of the Tasman Sea. We classify each event’s driver and relate it to duration, intensity, seasonality, and depth extent of anomalous warming. We show that MHWs in the Western Boundary Current (WBC) jet are predominantly driven by atmospheric forcing whilst in the WBC extension they are driven by ocean currents through advection. Anomalous advection‐driven MHWs are deepest (three times deeper), longest (four times longer) and more prevalent in autumn and winter. This contrasts with atmospherically forced MHWs, which have shallower depth ranges and exhibit a strong seasonal cycle mostly occurring in summer. We also show that the sub‐surface intensity extent is not detectable from surface data alone, underscoring the need for sustained sub‐surface information. Understanding the dynamical causes and sub‐surface structure of the ocean during MHWs is a step toward the prediction of these anomalously warm high impact events.
... e w is the entrainment velocity across the base of the mixed layer, following the approach of Stevenson and Niiler (1983), we compute the entrainment velocity as in Ren and Riser (2009): the entrainment (positive) velocity is considered to cool the mixed layer while the detrainment (negative) velocity plays no role. This is because the water that flows out from the base of the mixed layer has approximately the same characteristics as the water in the mixed layer and hence will not affect the temperature in the mixed layer (Nyadjro et al., 2012;Ren & Riser, 2009;Schlundt et al., 2014). ...
Article
Full-text available
Mesoscale eddies redistribute heat, salt, and nutrients in oceans. The South Atlantic Ocean (SA) is a basin that has active mesoscale eddies for which characteristics of the three-dimensional structure and its leading mechanism are complex but have yet been studied sufficiently. Here based on ocean reanalysis datasets we use a composite analysis approach to analyze the mixed layer anomalous heat budget and find distinct two types of spatial patterns: dipole and monopole – mainly present in the northern and southern regions of the SA, respectively. The dipole can be attributed to ocean horizontal advection, especially to the combined effect of eddy anomalous meridional current and meridional gradient of mean temperature. The monopole, on the other hand, is associated with complex contributions, for which zonal and meridional advections play opposite roles as cooling or heating around the eddies. At the eddy center, the vertical advection is non-negligible, especially the mean upwelling and vertical temperature gradient playing a vital role in the formation of a monopole. The analysis of eddy meridional heat transport shows that the stirring component is dominant, and poleward in most areas, especially at high latitudes. Such analysis on the leading mechanism of eddy-induced temperature anomaly could help improve our understanding on meso- and small-scale air-sea interactions and eddy- induced heat transport in the SA.
... ref. 23 ). In recent years, upper ocean turbulence data have become increasingly available in the equatorial Atlantic and Pacific, and have supported the hypothesis of elevated vertical diffusive ML heat loss in the central and eastern equatorial oceans 19,20,24,25 . The vertical diffusive ML heat flux is among the largest cooling terms in the ML budgets of the equatorial regions. ...
Article
Full-text available
The direct response of the tropical mixed layer to near-inertial waves (NIWs) has only rarely been observed. Here, we present upper-ocean turbulence data that provide evidence for a strongly elevated vertical diffusive heat flux across the base of the mixed layer in the presence of a NIW, thereby cooling the mixed layer at a rate of 244 W m⁻² over the 20 h of continuous measurements. We investigate the seasonal cycle of strong NIW events and find that despite their local intermittent nature, they occur preferentially during boreal summer, presumably associated with the passage of atmospheric African Easterly Waves. We illustrate the impact of these rare but intense NIW induced mixing events on the mixed layer heat balance, highlight their contribution to the seasonal evolution of sea surface temperature, and discuss their potential impact on biological productivity in the tropical North Atlantic.
Article
Full-text available
The equatorial retroflection of the North Brazil Current (NBC) into the Equatorial Undercurrent (EUC) and its posterior tropical recirculation is a major regulator for the returning limb of the Atlantic Meridional Overturning Circulation. Indeed, most surface and thermocline NBC waters retroflect at the equator all the way into the central and eastern Atlantic Ocean, before they recirculate back through the tropics to the western boundary. Here, we use cruise data in the western equatorial Atlantic during April 2010 and reanalysis time series for the equatorial and tropical waters in both hemispheres in order to explore the recirculation pathways and transport variability. During the 1998–2016 period, the annual‐mean EUC transports 15.1 ± 1.3 Sv at 32°W, with 2.8 ± 0.4 Sv from the North Atlantic and 11.4 ± 1.3 Sv from the South Atlantic. At 32°W most of the total EUC transport comes from the western boundary retroflection south of 3°N (7.2 ± 0.9 Sv), a substantial fraction retroflects north of 3°N (5.6 ± 0.4 Sv), and the remaining flow (2.3 Sv) joins through the interior basin. The South Atlantic subtropical waters feed the EUC at all thermocline depths while the North Atlantic and South Atlantic tropical waters do so at the surface and upper‐thermocline levels. The EUC transport at 32°W has a pronounced seasonality, with spring and fall maxima and a range of 8.8 Sv. The 18 yr of reanalysis data shows a weak yet significant correlation with an Atlantic Niño index, and also suggests an enhanced contribution from the South Atlantic tropical waters during 2008–2016 as compared with 1997–2007.
Article
Full-text available
By increasing Earth-atmosphere system albedo, Stratospheric Aerosol Geoengineering (SAG) using sulfur dioxide is an artificial potential means, with the goal to mitigate the global warming effects. In this study, we used the simulations from Geoengineering Large Ensemble project realized under the climate change scenario of Representative Concentration Pathway 8.5 (RCP8.5), to investigate the potential impact of SAG on the Sea Surface Temperature (SST) in Equatorial Atlantic Cold Tongue (EACT) and the physical processes driving these changes. Results reveal that in the EACT region, under RCP8.5, SST warms significantly (compared to present‐day climate) with a maximum of 1.7 °C in July, and this increase in SST is mainly due to the local processes related to the weakening of vertical mixing at the base of the mixed layer. This reduction of the vertical mixing is associated to the diminution of the vertical shear from July to April and to the increase of ocean stratification from May to June. However, under SAG, SST decreases significantly throughout the year (compared to present‐day climate) with a maximum cooling of − 0.4 °C in the cold tongue period (May–June). This SST cooling is mainly associated with the non-local processes related to intensification of the westerly equatorial Atlantic wind stress. Finally, results show that the use of SAG to offset all global warming under RCP8.5 results in a slight over compensation of SST in the EACT region.
Article
Full-text available
The Tropical Atlantic Ocean has recently been the source of enormous amounts of floating Sargassum macroalgae that have started to inundate shorelines in the Caribbean, the western coast of Africa and northern Brazil. It is still unclear, however, how the surface currents carry the Sargassum , largely restricted to the upper meter of the ocean, and whether observed surface drifter trajectories and hydrodynamical ocean models can be used to simulate its pathways. Here, we analyze a dataset of two types of surface drifters (38 in total), purposely deployed in the Tropical Atlantic Ocean in July, 2019. Twenty of the surface drifters were undrogued and reached only ∼8 cm into the water, while the other 18 were standard Surface Velocity Program (SVP) drifters that all had a drogue centered around 15 m depth. We show that the undrogued drifters separate more slowly than the drogued SVP drifters, likely because of the suppressed turbulence due to convergence in wind rows, which was stronger right at the surface than at 15 m depth. Undrogued drifters were also more likely to enter the Caribbean Sea. We also show that the novel Surface and Merged Ocean Currents (SMOC) product from the Copernicus Marine Environmental Service (CMEMS) does not clearly simulate one type of drifter better than the other, highlighting the need for further improvements in assimilated hydrodynamic models in the region, for a better understanding and forecasting of Sargassum drift in the Tropical Atlantic.
Article
Full-text available
We identify and analyze tropical instability waves (TIWs) in the equatorial Atlantic Ocean during 2010–2018 using satellite derived observations of sea surface salinity (SSS), sea surface temperature (SST), sea level anomaly, and Argo profiles. In particular, the weekly 50-km resolution SSS time series from the climate change initiative project provides an unprecedented opportunity to observe the salinity structure at a scale closer to the SST scale. We examine the relative contributions of SSS and SST to the horizontal surface density gradient on seasonal and interannual time scales and how they contribute to the TIW properties and energetics. For the central Atlantic TIWs, the maximum of the SST contribution to the density anomaly lags the SSS one by approximately one month. Argo vertical profiles indicate that temperature and salinity both significantly contribute to TIW-related density anomalies. In May–June, salinity contributes to 50% of the perturbation potential energy in the top 60 m, and between 30% and 45% from July to September. While variations in SST appear to be related to dynamic processes, the interannual variability of SSS is also influenced by precipitations. However, the two leading modes of variability in the region (Atlantic Meridional and Zonal modes) do not well explain at 1°N these interannual variations.
Technical Report
Full-text available
This variant of technical paper accompanies the release of public dataset YoMaHa'07 (the update of YoMaHa’05) and provides a brief description of its structure and the techniques used for its preparation. The dataset contains estimates of velocities of deep and surface currents obtained using data of the trajectories from Argo floats. It includes data from 4284 floats stored in nine Data Assembly Centers (DAC’s) worldwide and about 297,000 values of velocity. This is a 41% increase in the number of floats and 78% increase in the amount of data compared to YoMaHa’05. The data span the period from August 04, 1997 through May 15, 2007. Surface velocities are linearly regressed from float coordinates fixed by the ARGOS satellites. Deep velocities are estimates from floats displacements during each submerged phase of the cycle. Both surface and deep velocities are accompanied by error estimates, which are typically an order of magnitude smaller than velocity values. The paper describes data distribution in space, time and between the DAC’s as well as probability distributions of programmed float parameters and statistics of their displacements. Possible application of the dataset is illustrated on the example of the ensemble-mean velocity field.
Article
Full-text available
The seasonal variability of surface layer salinity (SLS), evaporation ( E ), precipitation ( P ), E-P , advection and vertical entrainment over the global ocean is examined using in situ salinity data, the National Centers for Environmental Prediction's Climate System Forecast Reanalysis and a number of other ancillary data. Seasonal amplitudes and phases are calculated using harmonic analysis and presented in all areas of the open ocean between 60° S and 60° N. Areas with large amplitude SLS seasonal variations include: the intertropical convergence zone (ITCZ) in the Atlantic, Pacific and Indian Oceans; western marginal seas of the Pacific; and the Arabian Sea. The median amplitude in areas that have statistically significant seasonal cycles of SLS is 0.19. Between about 60° S and 60° N, 37% of the ocean surface has a statistically significant seasonal cycle of SLS and 75% has a seasonal cycle of E-P . Phases of SLS have a bimodal distribution, with most areas in the Northern Hemisphere peaking in SLS in March/April and in the Southern Hemisphere in September/October. The seasonal cycle is also estimated for surface freshwater forcing using a mixed-layer depth climatology. With the exception of areas near the western boundaries of the North Atlantic and North Pacific, seasonal variability is dominated by precipitation. Surface freshwater forcing also has a bimodal distribution, with peaks in January and July, 1–2 months before the peaks of SLS. Seasonal amplitudes and phases calculated for horizontal advection show it to be important in the tropical oceans. Vertical entrainment, estimated from mixed-layer heaving, is largest in mid and high latitudes, with a seasonal cycle that peaks in late winter. The amplitudes and phases of SLS and surface fluxes compare well in a qualitative sense, suggesting that much of the variability in SLS is due to E-P . However, the amplitudes of SLS are somewhat different than would be expected and the peak of SLS comes typically about one month earlier than expected. The differences of the amplitudes of the two quantities is largest in such areas as the Amazon River plume, the Arabian Sea, the ITCZ and the eastern equatorial Pacific and Atlantic.
Article
Full-text available
Since December 2010, the MyOcean global analysis and forecasting system has consisted of the Mercator Océan NEMO global 1/4° configuration with a 1/12° nested model over the Atlantic and the Mediterranean. The open boundary data for the nested configuration come from the global 1/4° configuration at 20° S and 80° N. The data are assimilated by means of a reduced-order Kalman filter with a 3-D multivariate modal decomposition of the forecast error. It includes an adaptive-error estimate and a localization algorithm. A 3-D-Var scheme provides a correction for the slowly evolving large-scale biases in temperature and salinity. Altimeter data, satellite sea surface temperature and in situ temperature and salinity vertical profiles are jointly assimilated to estimate the initial conditions for numerical ocean forecasting. In addition to the quality control performed by data producers, the system carries out a proper quality control on temperature and salinity vertical profiles in order to minimise the risk of erroneous observed profiles being assimilated in the model. This paper describes the recent systems used by Mercator Océan and the validation procedure applied to current MyOcean systems as well as systems under development. The paper shows how refinements or adjustments to the system during the validation procedure affect its quality. Additionally, we show that quality checks (in situ, drifters) and data sources (satellite sea surface temperature) have as great an impact as the system design (model physics and assimilation parameters). The results of the scientific assessment are illustrated with diagnostics over the year 2010 mainly, assorted with time series over the 2007–2011 period. The validation procedure demonstrates the accuracy of MyOcean global products, whose quality is stable over time. All monitoring systems are close to altimetric observations with a forecast RMS difference of 7 cm. The update of the mean dynamic topography corrects local biases in the Indonesian Throughflow and in the western tropical Pacific. This improves also the subsurface currents at the Equator. The global systems give an accurate description of water masses almost everywhere. Between 0 and 500 m, departures from in situ observations rarely exceed 1 °C and 0.2 psu. The assimilation of an improved sea surface temperature product aims to better represent the sea ice concentration and the sea ice edge. The systems under development are still suffering from a drift which can only be detected by means of a 5-yr hindcast, preventing us from upgrading them in real time. This emphasizes the need to pursue research while building future systems for MyOcean2 forecasting.
Article
Full-text available
A model/data comparison was performed between simulated drifters from a high-resolution numerical simulation of the North Atlantic and a data set from in situ surface drifters. The comparison makes use of pseudo-Eulerian statistics such as mean velocity and eddy kinetic energy, and Lagrangian statistics such as integral time scales. The space and time distribution of the two data sets differ in the sense that the in situ drifters were released inhomogeneously in space and time while the simulated drifters were homogeneously seeded at the same time over a regular 1° grid. Despite this difference, the total data distributions computed over the complete data sets show some similarities that are mostly related to the large-scale pattern of Ekman divergence/convergence.
Article
Full-text available
Global ocean precipitation is an important part of the water cycle in the climate system. A number of efforts have been undertaken to acquire reliable estimates of precipitation over the oceans based on remote sensing and reanalysis modelling. However, validation of these data is still a challenging task, mainly due to a lack of suitable in situ measurements of precipitation over the oceans. In this study, validation of the satellite-based Hamburg Ocean Atmosphere Parameters and fluxes from Satellite data (HOAPS) climatology was conducted with in situ measurements by ship rain gauges over the Baltic Sea from 1995 to 1997. The ship rain gauge data are point-to-area collocated against the HOAPS data. By choosing suitable collocation parameters, a detection rate of up to about 70% is achieved. Investigation of the influence of the synoptic situation on the detectability shows that HOAPS performs better for stratiform than for convective precipitation. The number of collocated data is not sufficient to validate precipitation rates. Thus, precipitation rates were analysed by applying an interpolation scheme based on the Kriging method to both data sets. It was found that HOAPS underestimates precipitation by about 10%, taking into account that precipitation rates below 0.3 mm h(-1) cannot be detected from satellite information.
Article
Sensible and latent heat fluxes were estimated from turbulence measurements gathered during several Atlantic Ocean transects of the research vessel (R/V) Polarstern. The inertial dissipation method was used to analyze the data. Resulting bulk transfer coefficients were then applied to the data from the ship's meteorological system to get continuous time series of the heat fluxes. Combined with the measured downward solar and longwave radiation fluxes it allows for an estimate of the total energy budget at the air-sea interface. Comparing these parameterized energy fluxes to those based on the Coupled Ocean-Atmosphere Response Experiment, version 3 (COARE3.0), bulk flux algorithm shows very strong agreement.