Seasonal salt budget of the northwestern tropical Atlantic ...carton/pdfs/foltzetal04.pdf · Seasonal salt budget of the northwestern tropical Atlantic Ocean along 38oW Gregory R.

Seasonal salt budget of the northwestern tropical

Atlantic Ocean along 38oW

Gregory R. Foltz, Semyon A. Grodsky*, James A. Carton, and Michael J. McPhaden1

Draft

July 18, 2003

Department of Meteorology University of Maryland College Park, MD 20742 1NOAA/Pacific Marine Environmental Laboratory 7600 Sand Point Way NE Seattle, WA 98115

* corresponding author: [email protected]

Abstract

The western tropical Atlantic has a complex salt budget, with strong seasonal river

discharge, precipitation, evaporation, and the confluence of different water masses all

contributing. This paper addresses the atmospheric and oceanic causes of the seasonal budget of

mixed layer salinity based on direct observations. Data sets include up to five years (September

1997 - December 2002) of measurements from moored buoys of the Pilot Research Array in the

Tropical Atlantic (PIRATA), ship-intake salinity observations, and near-surface drifting buoys.

We analyze the mixed layer salt balance at four PIRATA mooring locations along the 38oW

(4oN, 8oN, 12oN, and 15oN). At 4oN precipitation is primarily semiannual. Here there are also

strong contributions from zonal and meridional advection. The sum of the three balances a strong

annual cycle of local storage. The strongest seasonal cycle of precipitation occurs at 8oN and is

balanced by horizontal advection and local storage. At 12oN the balance is mostly local, with

only a minor contribution from horizontal advection. At 15oN horizontal advection plays a key

role in balancing a semiannual cycle of local storage.

1. Introduction

Salinity affects the density of seawater, spatial gradients of which affect buoyancy,

turbulent mixing, and ocean currents. Recent numerical modeling studies indicate that realistic

three-dimensional salinity fields are required in order to accurately represent tropical ocean

dynamics and thermodynamics (Murtugudde and Busalacchi, 1998; Vialard and Delecluse,

1998). In support of these modeling results, a number of observational studies show that changes

in mixed layer salinity can dramatically affect currents and temperature in the tropics through the

formation of barrier layers (Lukas and Lindstrom, 1991; Roemmich et al., 1994; Pailler et al.,

2

1999). In this study we use in situ and satellite data to determine the oceanic and atmospheric

causes of the seasonal cycle of mixed layer salinity in the tropical Atlantic.

The flux of moisture at the ocean’s surface is given by the difference between

precipitation and evaporation. The annual mean surface moisture flux in the tropical Atlantic

reflects the mean position of the Intertropical Convergence Zone (ITCZ) and the subtropical

high-pressure systems. In the southern (south of 5°S) and northern (north of 10°N) tropical

Atlantic, annual mean evaporation dominates precipitation, while in the latitude band of the

ITCZ precipitation dominates, with a (zonally averaged) peak of 20 cm mo-1 near 5°N (Yoo and

Carton, 1990). Seasonal changes in the surface moisture flux result mainly from latitudinal

movements of the ITCZ.

The seasonal cycle of evaporation is primarily annual throughout most of the tropical

Atlantic, with increasing amplitude toward the subtropics. In the northern tropics, evaporation

peaks in boreal winter. In this season, the ITCZ is situated close to its southernmost latitude, with

increasing northeast trade winds and decreasing relative humidity northward from the equator. In

boreal summer and fall, the ITCZ is located near its northernmost latitude (10 - 12°N), resulting

in higher relative humidity, lower wind speed, and less evaporation north of 5°N. Between the

equator and 8°N the double passage of the ITCZ results in a significant semiannual harmonic,

with highest amplitude in the west (da Silva et al., 1994).

The seasonal cycle of precipitation in the tropical Atlantic results from meridional

movements of the ITCZ and zonal shifts of convergence within it. In April, a band of high

precipitation (> 20 cm mo-1) is situated west of 30°W between 5°S and 5°N, increasing westward

toward the coast of South America. Precipitation is also high (> 15 cm mo-1) at this time in the

Gulf of Guinea, increasing eastward along the equator toward the African coast. During May

3

through August the zonal band of high rainfall (> 20 cm mo-1) shifts northward and increases in

intensity to up to 60 cm mo-1 near the African coast. From September through March, the ITCZ

drifts southward from its northernmost latitude (~ 10-12°N during August), while the region of

high rainfall decreases in intensity and becomes increasingly concentrated in the western basin.

As a result, the seasonal cycle of precipitation acquires a significant semiannual harmonic near

5°N, with maxima in June and December.

In addition to the surface moisture flux, ocean dynamics can affect mixed layer salinity.

Near-surface currents in the tropical Atlantic are dominated by the westward South Equatorial

Current south of 5°N and the eastward North Equatorial Countercurrent centered near 5°N (see

Fig. 1). The intensity of these zonal currents fluctuates seasonally in response to the seasonal

movement of the ITCZ and associated changes in near-surface winds. Near the equator, the

South Equatorial Current is strongest in boreal summer, with speeds exceeding 50 cm s-1 in mid

basin. On and just north of the equator are found strong velocity and temperature fluctuations

associated with tropical instability waves, which we anticipate are also important in mixing

salinity. Near the western boundary, the South Equatorial Current feeds the North Brazil Current,

which flows continuously northward toward the Caribbean during boreal winter and spring. In

boreal summer and fall, part of the North Brazil Current curves back upon itself near 8°N, then

flows eastward to feed the North Equatorial Countercurrent. In the process, strong eddies form in

the western basin between the equator and 10°N (Richardson and Reverdin, 1987). These eddies

have little effect on sea surface temperature (SST) due to weak gradients of SST in the western

basin. However, we anticipate that they may have a significant effect on mixed layer salinity due

to Amazon River discharge and rainfall associated with the ITCZ.

4

The large-scale spatial distribution of sea surface salinity (SSS) in the tropical Atlantic

(see Fig. 1) to a large extent reflects the annual mean surface moisture flux. SSS is highest in the

subtropics (> 36 psu), where annual mean evaporation exceeds precipitation, and decreases to a

minimum near 5°N (< 35 psu in the east), where precipitation reaches a maximum.

Although the surface moisture flux contributes significantly to the annual mixed layer

salinity budget in the tropical Atlantic, horizontal advection also plays an important role. Indeed,

Johnson et al. (2002) found that a substantial portion of the annual mean freshwater flux in the

tropical oceans is balanced by oceanic transports within the surface layer (upper 32.5 m).

They found large contributions from both horizontal and vertical advection near the equator in

the eastern tropical Atlantic and from meridional advection in the northwest, where meridional

currents and salinity gradients are strong (see Fig. 1). Observational studies in the tropical

Pacific have also stressed the importance of horizontal advection in the mixed layer salinity

budget on seasonal to interannual time scales (e.g., Delcroix and Henin, 1991; Cronin and

McPhaden, 1998; Delcroix and Picaut, 1998; Henin et al., 1998).

Several observational studies have addressed the seasonal cycle of surface salinity in the

western basin. Dessier and Donguy (1994) show that the seasonal cycle is strongly influenced by

river discharge, advection by ocean currents, and local rainfall. In boreal winter a region of fresh

water (< 36 psu) extends northwestward along the coast of South America from the mouth of the

Amazon to 15°N. In boreal spring, as outflow from the Amazon increases, the fresh pool

expands outward from the coast to 45°W and northward into the Caribbean. By late boreal

summer and fall, coincident with the arrival of the ITCZ and the strengthening of the North

Equatorial Countercurrent, a band of fresh water extends eastward across the basin between 4°

and 10°N. Associated with this fresh pool is a significant barrier layer that is maintained at the

5

surface by eastward advection from the Amazon and at depth by southwestward advection of

high-salinity water subducted in the subtropics (Sprintall and Tomczak, 1992; Pailler et al.

1999). Further south the seasonal cycle of surface salinity acquires a significant semiannual

harmonic in the central basin.

Based on a limited number of observational studies in the tropical Atlantic, it is clear that

evaporation, precipitation, and horizontal advection all contribute to the observed seasonal cycle

of mixed layer salinity. The need to understand the seasonal cycle of mixed layer salinity has

been supported by modeling studies, which reveal that mixed layer salinity contributes

significantly to upper ocean dynamic height and potentially SST, through the formation of

barrier layers. In this study we examine the seasonal mixed layer salinity budget using new

comprehensive oceanic and surface meteorological data sets from four PIRATA mooring

locations in the western tropical Atlantic. Our study complements an earlier analysis of the

mixed layer heat balance using similar data sets at the PIRATA mooring sites (Foltz et al.,

2003).

2. Data and Methods

Applying the methodology of Stevenson and Niiler (1983) and Delcroix and Henin

(1991) to the mixed layer salt balance gives

( ) ( ) hh

e FSPEdzSvwSSvSvhtSh −

−

+−+⋅∇−∆−′∇⋅′+∇⋅−=∂∂

∫ )(ˆˆ0 rrr (1a)

⎟⎠⎞

⎜⎝⎛ ⋅∇+∂∂

= vhthHwe

r (1b)

The terms in (1a) represent, from left to right, local salt storage, horizontal advection (separated

into monthly mean and eddy terms), entrainment, vertical salinity/velocity covariance, surface

6

moisture flux, and turbulent diffusion at the base of the mixed layer. Entrainment velocity (1b) is

associated with a mass flux that crosses an isopycnal surface (H is the Heaviside unit function).

Here h is the depth of the mixed layer, S and vr are salinity and velocity, respectively, vertically

and monthly averaged from the surface to a depth of –h, S ′ and v ′r are deviations from their

monthly means (the overbar represents a monthy mean), and vS r represent deviations from the

vertical average, , E is evaporation, and P is precipitation. We have found that the

vertical salinity/velocity covariance term in (1a) is insignificant (< 0.7 mg m

hSSS −−=∆

-2 s-1 on a monthly

basis) at all locations considered in this study (based on climatological subsurface data from the

SODA reanalysis of Carton et al., 2000). We therefore proceed to neglect this term. We also

neglect vertical turbulent diffusion (last term on the right), although studies in the equatorial

Pacific (e.g., Hayes et al., 1991; Lien et al., 2002) indicate that this term may contribute

significantly along the equator.

The PIRATA mooring array (Servain et al., 1998) consists of 12 buoys. We focus on the

four westernmost moorings with record lengths exceeding three years (see Fig. 1). Deployed in

1997 to study ocean-atmosphere interactions, these Next Generation Autonomous Temperature

Line Acquisition System (ATLAS) buoys measure subsurface ocean temperature and salinity,

rainfall, and surface air temperature, relative humidity, and wind velocity. Ocean temperature is

measured at 11 depths between 1 and 500 m with 20 m spacing in the upper 140 m, while

salinity is measured only at 1, 20, 40, and 120 m. The 1 m temperature and salinity sensors

provide bulk measures of SST and SSS, respectively. Air temperature and relative humidity are

measured at a height of 3 m above sea level, while rainfall and wind velocity are measured at 3.5

and 4 m, respectively. The sampling interval is ten minutes for all variables except rainfall,

which is sampled at one-minute intervals. The instrument accuracies are: water temperature

7

within ± 0.01°C, salinity within ± 0.02 psu, wind speed ± 0.3 m s-1 or 3% (whichever is greater),

air temperature ± 0.2°C, relative humidity ± 3%, and rainfall ± 0.4 mm hr-1 (Freitag et al., 1994,

1999, 2001; Serra et al., 2001, Lake et al., 2003). Here we use both 10-minute and daily

averaged data, which are transmitted in near-real time via satellite by Service Argos.

All 10-minute PIRATA data used in this study (precipitation, air temperature, relative

humidity, SST, and wind speed) have been analyzed by PMEL/NOAA for quality control

purposes (Freitag et al., 1999). The quantity of 10-minute quality-controlled subsurface salinity

data is low and thus we use daily-averaged values based on real-time data streams, which are

available with minimal quality control (H. P. Freitag, personal communication, 2003). We have

carried out our own quality control, which includes subjective removal of uncharacteristic high-

frequency oscillations, removal of data with a significant bias (> 1 psu) with respect to the mean

of the remaining time series at that location, and elimination of data at multiple depth levels if

the difference in salinity between the levels rapidly changes (indicating instrument drift).

Precipitation is available directly from the moorings. In contrast, evaporation is estimated

using the bulk parameterization, )( qqWCE sea −= ρ , where E is evaporation rate, aρ is air

density, is the transfer coefficient, W is wind speed, q is the water vapor mixing ratio, and

is the interfacial water vapor mixing ratio, which is assumed to be proportional

to the saturation water vapor mixing ratio (the factor of 0.98 accounts for salinity effects). Tests

of this algorithm, developed from the Coupled Ocean-Atmosphere Response Experiment

(COARE) in the tropical west Pacific have revealed a negative bias of 0.16 cm mo

eC

( )ssats Tqq 98.0=

-1 (Fairall et

al., 1996). We use ten-minute measurements of air temperature, SST, wind speed, and relative

humidity to estimate evaporation and neglect both cool skin and warm layer effects (see Foltz et

al., 2003 for an estimate of their magnitudes).

8

Following the procedure of Foltz et al. (2003), we estimate the mixed layer depth as the

depth at which density is 0.15 kg m-3 greater than the nighttime surface value (a 0.15 kg m-3

density step approximately corresponds to a 0.5 oC temperature drop at SST=28 oC and SSS=36

psu). Throughout much of the ocean upper ocean, density is controlled primarily by temperature.

However, the western tropical Atlantic is affected by strong precipitation, river discharge, and

persistent salinity fronts, all of which influence the mixed layer stratification. At the two

southern locations (4°N and 8°N) the difference between the isothermal layer depth and the

mixed layer depth (defining the barrier layer thickness) is greatest in boreal fall (see Figs. 2c,

2d). These thick barrier layers (reaching 40 m at 4°N) develop in response to the eastward

extension of the low-salinity Amazon plume, driven by the seasonal strengthening of the North

Equatorial Countercurrent in late summer and fall (Carton, 1991; Pailler et al., 1999), and in

response to ITCZ rainfall that develops in summer through fall at 8°N and twice per year (winter

and late spring/early summer) at 4°N. The barrier layer widens to up to 20 m in boreal fall at

8°N. At 4oN a barrier layer persists throughout the year, thickening to ~30 m in early spring and

up to ~40 m in early fall. At the two northern mooring locations (12°N and 15°N) the barrier

layer widens in boreal winter, in phase with the strengthening of the northeasterly trade winds

(Figs. 2a, 2b) and supported by the northward wind-driven transport of lower salinity waters and

their interaction with the subtropical salinity front. At 12oN a moderate thickness (~ 10 m)

barrier layer is apparent during boreal fall and early winter. At 15oN the barrier layer widens to

~ 30 m in winter and gradually recedes during spring.

To calculate horizontal salt advection and entrainment velocity we first estimate the

seasonal cycle of near-surface (~ 15 m) horizontal velocity following the procedure of Grodsky

and Carton (2001) and Foltz et al. (2003). This method uses on quasi-Lagrangian drifter

9

velocity, ship-drift velocity, and satellite-based sea level and wind stress. We anticipate that

meridional velocity in the mixed layer is primarily the result of Ekman drift (since the trade

winds are mainly zonal in the western basin) that decreases with increasing depth. We therefore

apply a correction that assumes a linear decrease in meridional velocity from the observed value

at 15 m to zero at -h, following Foltz et al. (2003). No correction is applied for h < 15 m since

we cannot accurately estimate surface velocity needed for interpolation from 15 m to the surface.

Since we cannot estimate the vertical distribution of zonal velocity (we anticipate that it depends

strongly on horizontal pressure gradients, which we cannot calculate), we have not applied a

correction to zonal velocity estimates. We also use divergence of the velocity estimates, as well

as estimates of the time derivative of mixed layer depth based on PIRATA subsurface

temperature and salinity, to calculate we (see (1b)). Horizontal gradients of mixed layer depth are

estimated from a monthly climatology of White (1995).

Monthly climatological SSS gradients are obtained from the climatology of Dessier and

Donguy (1994), based on near-surface ship intake salinity measurements.

Since we are interested primarily in the seasonal cycle, we eliminate high-frequency

variability by fitting each term in (1a) to annual and semiannual harmonics using least squares

(Fourier) analysis (see Appendix for further details). The fitting procedure contains errors due to

a combination of missing PIRATA data coverage and interannual variability. In particular,

climatologies of different PIRATA variables may incorporate data from different time periods

(see Fig. 4). For this reason we display the number of daily PIRATA measurements that go into

each estimate at each location. Estimates of each term in (1) use a maximum of about 150

individual daily measurements per calendar month (since most PIRATA moorings have been

operational for about five years). Low counts (<< 150) indicate high uncertainty. Although the

10

data span a maximum of 5 years, they reveal a discernable seasonal cycle for all surface

variables, illustrated for 8oN in Fig. 4.

3. Results

In this section we examine the balance of terms in (1a), beginning with the northern

locations.

a) 15°N and 12°N

At these sites, surface fluxes contribute significantly to seasonal salt storage (Figs. 5, 6).

The surface moisture flux is mostly annual and is connected to the seasonal meridional march of

the ITCZ. Rainfall is significant during August – October, when the ITCZ is near its

northernmost position, while evaporation at this time is at a minimum. As a result, surface

salinity decreases at both locations during late boreal summer and early fall. In boreal winter the

ITCZ shifts southward (see Fig. 3d for the ITCZ position), resulting in decreased rainfall at 12oN

and 15oN. The southward ITCZ shift also results in a strengthening of the northeast trade winds

and a decrease in relative humidity that, in turn, lead to high rates of evaporation. These changes

in the surface flux cause surface salinity to increase at 12°N. At 15°N, ocean dynamics

counteract the effects of evaporation, leading to weak freshening in boreal winter.

At 12oN surface fluxes are primarily responsible for the seasonal cycle of the mixed layer

salinity, while ocean dynamics play a lesser role. The zonal gradient of salinity is weak, resulting

in a very weak seasonal cycle of zonal advection. Interestingly, there are seasonal variations in

the meridional salinity gradient that reflect the meridional march of the ITCZ. Indeed, the

meridional gradient at 12oN increases in boreal summer and fall (Fig. 3c). However, the mixed

11

layer is shallow (Fig. 2b) and the meridional currents are weak during this season, reducing the

contribution of horizontal advection to the seasonal salinity budget.

In contrast, at 15oN horizontal transport is an important factor in the annual and

semiannual signal of the salt budget. Meridional advection has a strong annual cycle at 15°N

(Fig. 5). Meridional velocity reaches a maximum during boreal winter (Fig. 3d), when the

westward northeast trade winds and resultant northward Ekman drift are strong. As at 12°N,

seasonal variations of the meridional salinity gradient at 15°N result from meridional movements

of the subtropical surface salinity front. The northward gradient is strongest during boreal winter,

when evaporation rates are high to the north and rainfall is abundant to the south (Fig. 3c).

During this season, there is a strong northward component of velocity and a deep mixed layer

(Fig. 2a), providing a significant source of freshening at 15°N.

Zonal transport at 15oN is weaker than meridional transport (see Fig. 5) due to weaker

zonal gradients of surface salinity. Zonal surface salinity gradients are negative (i.e., higher

salinity in the west) throughout most of the year, becoming positive only during boreal spring

and late fall. Hence, the westward currents provide a source of freshening that varies

semiannually. The decrease in negative zonal salt transport partially accounts for the positive rate

of storage in boreal spring (see Fig. 5).

Entrainment at 15°N as well as 12oN results from a deepening mixed layer in boreal fall

and winter that acts to increase mixed layer salinity.

b) 8°N and 4°N

We next consider the mixed layer salinity balance at two locations within the latitudinal

range of the ITCZ (8°N and 4°N; Figs. 7, 8). Rainfall at 8°N is dominated by the annual

12

harmonic, with a maximum in early fall when the ITCZ is farthest north. At 4°N a significant

semiannual harmonic appears, associated with the double passage of the ITCZ in May-June and

December. Seasonal variations of evaporation at these locations are somewhat weaker than those

to the north (12°N and 15°N) due to weaker wind speed and higher relative humidity (especially

at 4°N). Although precipitation is the most important factor affecting seasonal variations of

mixed layer salinity at 8°N and 4°N, horizontal transport is also important.

At 8°N the meridional gradient of salt is positive throughout the year, except during

boreal fall, and experiences substantial variability (see Fig. 3c). In fall the gradient becomes

negative in response to the presence of ITCZ rainfall to the north (see Fig. 3d for the ITCZ

latitude). As a result, meridional advection by prevailing northward currents freshens the mixed

layer in winter and spring but increases mixed layer salinity in fall (Fig. 7). At 4°N the

meridional gradient of salinity is negative throughout the year (Fig. 3c), reaching a maximum in

boreal summer, when both the ITCZ and the Amazon freshwater plume are situated to the north.

In this season, meridional velocity also reaches a maximum at 4°N (Fig. 3d), resulting in

significant northward salt advection.

Zonal advection also contributes significantly at 8°N and 4oN (see Figs. 7, 8). At 8°N

zonal currents strengthen during July – October, when the eastward North Equatorial

Countercurrent reaches a maximum (Fig. 3b) and the zonal gradient of salinity is generally

positive (Fig. 3a), resulting in freshening of ~3 mg m-2 s-1 (Fig. 7). At 4°N (see Fig. 3b) zonal

currents are westward during the first half of the year (due to the South Equatorial Current) and

eastward during the latter half (due to the North Equatorial Countercurrent). The zonal gradient

of salinity also varies seasonally, leading to a semiannual variation in zonal transport (Fig. 8).

Entrainment at these locations appears to be weak.

13

At 8°N discrepancies between the sum of terms and local salt storage rate indicate that all

terms have not been properly represented. We anticipate that horizontal eddy advection or

turbulent exchange across the bottom of mixed layer (the mixed layer is very shallow during the

second half of the year at 8oN; see Fig. 2c) may account for at least part of the large

discrepancies present during late summer and fall at 8°N.

4. Summary

This paper attempts to provide a comprehensive the mixed layer salt budget in the

western tropical Atlantic, a region of complex processes, including strong precipitation,

evaporation, river discharge, subducting water masses, and fluctuating mixed layer depths. Our

analysis exploits the availability of measurements from four PIRATA moorings and other in-situ

data. It follows the spirit of a similar study by Cronin and McPhaden (1998) in the western

equatorial Pacific. Our main results are as follows:

• At 15oN all terms contribute significantly to seasonal changes of the mixed layer

salinity. The negative tendency of salinity storage in boreal fall is due to the

presence of the ITCZ, which provides strong precipitation and weak evaporation.

Zonal transport by the westward currents varies semiannually, reflecting changes

in the zonal gradient of salinity. This component of transport accounts in part for

the local maxima in the salinity storage rate in both spring and late fall. Both

components of transport account for the freshening tendency in boreal winter.

• At 12oN local fluxes are most important in balancing local storage, with transport

playing a lesser role in regulating mixed layer salinity.

14

• At 8oN the freshening effects of precipitation and zonal advection are most

important during late summer and fall. The freshening due to eastward currents in

summer and early fall is due to eastward transport of waters partly diluted by

Amazon discharge.

• The mixed layer salt balance at 4ºN includes significant contributions from

semiannually varying precipitation and horizontal advection. Seasonal changes in

evaporation are less significant at this location. Meridional advection provides a

positive salt flux in summer, reflecting a strengthening of both the southward

salinity gradient and the northward wind-driven currents associated with the

seasonal northward shift of the ITCZ. Westward transport by the South Equatorial

Current in the first half of the year and eastward transport by the North Equatorial

Countercurrent in the second half result in a strong annual cycle of zonal salinity

advection, with peak eastward transport in late summer.

These results are in striking contrast to the mixed layer heat balance along 38ºW, where

horizontal advection makes minimal contributions (Foltz et al., 2003) due to weak surface

temperature gradients. Strong seasonal variations of horizontal salt transport are associated with

the meridional movement of the ITCZ and strong freshwater discharge from the Amazon River.

Thus, surface fluxes (and river discharge) indirectly affect the salt budget through the formation

of strong horizontal salinity gradients in the presence of horizontal currents. Such an indirect

connection appears to be much weaker in the mixed layer heat budget.

There are large uncertainties in our estimates of horizontal salt advection due to a lack of

direct velocity measurements as well as a lack of salinity surveys. There are also uncertainties in

the estimates of the seasonal cycle due to the relatively short mooring records. Despite these

15

uncertainties, it is clear that horizontal advection is an important component of the mixed layer

salt balance. In contrast, entrainment plays a more limited role. To improve our knowledge of the

seasonal salinity budget, it will be necessary to improve our estimates of horizontal salt transport

through the addition of current meters to the PIRATA buoys and satellite-based estimates of

surface salinity. Though subject to considerable uncertainties because of data limitations, our

results nonetheless provide a basis for evaluating interannual and decadal variability, which are

both linked to the annual cycle.

Acknowledgements

This work was supported by NOAA’s office of Oceanic and Atmospheric Research and Office of

Global Programs. The authors also gratefully acknowledge the support provided by the National

Science Foundation. We are grateful to the Drifter DAC of the GOOS Center at NOAA/AOML

for providing the drifter data set. Dr. Alain Dessier has made the SSS ship-intake archive

accessible via http://www.brest.ird.fr/sss/clim_atl.html. Quikscat wind velocity has been

obtained from the NASA/NOAA sponsored system Seaflux at JPL through the courtesy of W.

Timothy Liu and Wenqing Tang.

Appendix: Harmonic fitting and error estimates.

Terms in the seasonal salt budget are estimated in the following way. Let the matrix Y

contain the monthly mean state observations, while A is the annual and semiannual harmonic

filter, and X, containing the filtered state estimate, is obtained by minimizing the score function

)AX(YV)AX(Y 1T −−= −J

16

Here V is the observation error covariance matrix. We assume zero covariance between different

measurements (V is diagonal) and estimate the diagonal, Vobs, (that is the diagonal of V) as the

variance of daily observations in each monthly mean. This procedure gives the harmonics fit, Y ,

and the variance of the fit, Vfit. The total error shown in Figs. 5 – 8 is a sum of the error

associated with representing the observations by two harmonics and the error associated with

estimating those harmonics from our limited data set

fitobs)( VVY +=ε

Note that we have assumed that the accuracy of the PIRATA sensors is much higher than the

‘noise’ introduced by fluctuations due to high-frequency processes that are not resolved by our

analysis.

References

Carton, J.A., Effect of seasonal surface freshwater flux on sea surface temperature in the

Tropical Atlantic Ocean, J. Geophys. Res., 96, 12,593-12,598, 1991.

Carton, J.A., G. Chepurin, X.H. Cao, and B. Giese, A Simple Ocean Data Assimilation analysis

of the global upper ocean 1950-95. Part I: Methodology, J. Phys. Oceanogr., 30, 294-309,

2000.

Cronin, M.F., and M.J. McPhaden, Upper ocean salinity balance in the western equatorial

Pacific, J. Geophys. Res., 103, 27,567-87, 1998.

da Silva, A., A.C. Young, and S. Levitus, Atlas of Surface Marine Data 1994, Volume 1:

Algorithms and Procedures. NOAA Atlas NESDIS 6, U.S. Department of Commerce,

Washington, D.C., 1994.

17

Delcroix, T., and C. Henin, Seasonal and interannual variations of sea surface salinity in the

tropical Pacific Ocean, J. Geophys. Res., 96, 22,135-22,150, 1991.

Delcroix, T., and J. Picaut, Zonal displacement of the western equatorial Pacific “fresh pool,” J.

Geophys. Res., 103, 1087-1098, 1998.

Dessier, A., and J.R. Donguy, The sea surface salinity in the tropical Atlantic between 10°S and

30°N – seasonal and interannual variations (1977-1989), Deep Sea Res. I, 41, 81-100, 1994

Fairall, C.W., E.F. Bradley, D.P. Rogers, J.B. Edson, G.S. Young, Bulk parameterization of air-

sea fluxes for TOGA COARE. J. Geophys. Res., 101, 3747-3764, 1996.

Foltz, G.R., S.A. Grodsky, J.A. Carton, and M.J. McPhaden, Seasonal mixed layer heat budget

of the tropical Atlantic Ocean, J. Geophys. Res., 108(C5), 3146, doi:10.1029/2002JC001584,

2003.

Freitag, H.P., Y. Feng, L.J. Mangum, M.P. McPhaden, J. Neander, and L.D. Stratton, Calibration

procedures and instrumental accuracy estimates of TAO temperature, relative humidity and

radiation measurements. NOAA Tech. Memo. ERL PMEL-104, 32 pp., 1994.

Freitag, H.P., M.E. McCarty, C. Nosse, R. Lukas, M.J. McPhaden, and M.F. Cronin, COARE

Seacat data: Calibrations and quality control procedures. NOAA Tech. Memo. ERL PMEL-

115, 89 pp., 1999.

Freitag, H.P., M. O'Haleck, G.C. Thomas, and M.J. McPhaden, Calibration procedures and

instrumental accuracies for ATLAS wind measurements. NOAA. Tech. Memo. OAR PMEL-

119, NOAA/Pacific Marine Environmental Laboratory, Seattle, Washington, 20 pp., 2001.

Grodsky, S.A., and J.A. Carton, Intense surface currents in the tropical Pacific during 1996-

1998, J. Geophys. Res., 106, 16673-16684, 2001.

18

Hayes, S.P., P. Chang, and M.J. McPhaden, Variability of the sea surface temperature in the

eastern equatorial Pacific during 1986-1988, J. Geophys. Res., 96, 10553-10566, 1991.

Henin, C., Y. du Penhoat, and M. Ioualalen, Observations of sea surface salinity in the western

Pacific fresh pool: Large-scale changes in 1992-1995, J. Geophys. Res., 103, 7523-7536,

1998.

Johnson, E.S., G.S.E. Lagerloef, J.T. Gunn, and F. Bonjean, Salinity advection in the tropical

oceans compared to atmospheric freshwater forcing: a trial balance, J. Geophys. Res., 107,

art. no. 8014, 2002.

Lake, B. J., S.M. Noor, H.P. Freitag, and M.J. McPhaden, Calibration procedures and

instrumental accuracy estimates for ATLAS air temperature and relative humidity

measurements, NOAA Tech. Memo., in press, 2003.

Lien, R.-C., E.A. D’Asaro, and M.J. McPhaden, Internal waves and turbulence in the upper

central equatorial Pacific: lagrangian and eulerian observations, J. Phys. Oceanogr., 32,

2619-2639, 2002.

Lukas, R., and E. Lindstrom, The mixed layer of the western equatorial Pacific Ocean, J.

Geophys. Res., 96, suppl. S, 3343-3357, 1991.

Murtugudde R., and A.J. Busalacchi, Salinity effects in a tropical ocean model, J. Geophys. Res.,

103, 3283-3300, 1998.

Pailler, K., B. Bourles, and Y. Gouriou, The barrier layer in the western tropical Atlantic Ocean,

Geophys. Res. Lett., 26, 2069-2072, 1999.

Richardson, P.L., and G. Reverdin, Seasonal cycle of velocity in the Atlantic North Equatorial

Countercurrent as measured by surface drifters, current meters, and ship drifts, J. Geophys.

Res., 92, 3691-3708, 1987.

19

Roemmich, D., M. Morris, W.R. Young, and J.R. Donguy, Fresh equatorial jets, J. Phys.

Oceanogr., 24, 540-558, 1994.

Serra, Y.L., P. A'Hearn, H.P. Freitag, and M.J. McPhaden, ATLAS self-siphoning rain gauge

error estimates. J. Atmos. Ocean. Tech., 18, 1989-2002, 2001.

Servain, J., A.J. Busalacchi, M.J. McPhaden, A.D. Moura, G. Reverdin, M. Vianna, and S.E.

Zebiak, A Pilot Research Moored Array in the Tropical Atlantic (PIRATA), Bull. Amer.

Meteorol. Soc., 79, 2019-2031, 1998.

Sprintall, J., and M. Tomczak, Evidence of the barrier layer in the surface layer of the tropics, J.

Geophys. Res., 97, 7305-7316, 1992.

Stevenson, J.W., and P.P. Niiler, Upper ocean heat budget during the Hawaii-to-Tahiti shuttle

experiment, J. Phys. Oceanogr., 13, 1894-1907, 1983.

Vialard, J., and P. Delecluse, An OGCM study for the TOGA decade. Part I: Role of salinity in

the physics of the western Pacific fresh pool, J. Phys. Oceanogr., 28, 1071-1088, 1998.

White, W. B., Design of a global observing system for gyre-scale upper ocean temperature

variability, Progress in Oceanogr., 36, Pergamon, 169-217, 1995.

Yoo, J.-M., and J.A. Carton, Annual and interannual variation of the freshwater budget in the

tropical Atlantic Ocean and the Caribbean Sea, J. Phys. Oceanogr., 20, 831-845, 1990.

20

Figure captions

Figure 1 Locations of the PIRATA moored buoys used in this study (solid circles). Background contours and arrows are climatological winter-spring and summer-fall surface salinity (from the Dessier and Donguy (1994) analysis of ship-intake data) and near-surface ocean velocity of Grodsky and Carton (2001). Bold contour lines represent 1 psu intervals. Figure 2 Comparison of the isothermal mixed layer depth based on a 0.5oC criteron, ILD0.5, and the mixed layer depth based on a 0.15 kg m-3 density criteria, MLD0.15. The difference between the two is the barrier layer thickness. Figure 3 Seasonal variations of the sea surface salinity gradient (left panels) and near-surface currents (right panels) at 38oW. The bold line in (d) represents the position of the ITCZ (defined as the latitude of maximum QuikSCAT wind convergence). Figure 4 Daily PIRATA near-surface atmospheric and oceanic measurements at 8°N, 38°W during 1998 – 2002. Solid gray lines represent monthly mean World Ocean Atlas 2001 climatological surface salinity, TRMM Microwave Imager/Precipitation Radar rainfall, QuikSCAT near-surface wind speed, and NCEP/NCAR Reanalysis relative humidity. Figure 5 The mixed layer salt balance at 15°N, 38°W. Left panel shows individual contributions to the salt balance equation (1) in the form of evaporation, precipitation, entrainment, and mean zonal and meridional salt advection. Plots in lefthand panel show least squares fits of mean + annual and semiannual harmonics to monthly data. Righthand panel shows the sum of the terms in the lefthand panel and the actual mixed layer salt storage rate. Shading and cross-hatching in righthand panel indicate error estimates. Bars in righthand panels indicate number of days in each month for which all PIRATA-based terms in the lefthand panel are available (maximum of ~ 150 days for each month, corresponding to ~ 5 years of data: September 1997 – December 2002). Figure 6 As in Figure 5, but for 12°N, 38°W. Figure 7 As in Figure 5, but for 8°N, 38°W. Figure 8 As in Figure 5, but for 4°N, 38°W.

21

Figure 1

22

2 4 6 8 10 12

0

10

20

30

40

50

60

70

80

90

100

(a) 15N38WD

epth

, m

ILD0.5

MLD0.15

2 4 6 8 10 12

0

10

20

30

40

50

60

70

80

90

100

(b) 12N38W

2 4 6 8 10 12

0

10

20

30

40

50

60

70

80

90

100

(c) 08N38W

Dep

th, m

Month2 4 6 8 10 12

0

10

20

30

40

50

60

70

80

90

100

(d) 04N38W

Month Figure 2

23

Figure 3

24

Figure 4

25

Jan

M

ar

May

Ju

l

Sep

N

ov

−6

−4

−20246

15n

38w

mg m−2

s−1

mg m−2

s−1

Jan

M

ar

May

Ju

l

Sep

N

ov

−8

−6

−4

−20246

days

0100

sum

ρh∂S

/∂t

evap

prec

ipu

adv

v ad

ven

tr

Figure 5

26

Jan

M

ar

May

Ju

l

Sep

N

ov

−6

−4

−20246

12n

38w

mg m−2

s−1

mg m−2

s−1

Jan

M

ar

May

Ju

l

Sep

N

ov

−8

−6

−4

−20246

days

0100

sum

ρh∂S

/∂t

evap

prec

ipu

adv

v ad

ven

tr

Figure 6

27

Jan

M

ar

May

Ju

l

Sep

N

ov

−6

−4

−20246

8n38

w

mg m−2

s−1

mg m−2

s−1

Jan

M

ar

May

Ju

l

Sep

N

ov

−8

−6

−4

−20246

days

0100

sum

ρh∂S

/∂t

evap

prec

ipu

adv

v ad

ven

tr

Figure 7

28

Jan

M

ar

May

Ju

l

Sep

N

ov

−6

−4

−20246

4n38

wmg m

−2 s

−1

mg m−2

s−1

Jan

M

ar

May

Ju

l

Sep

N

ov

−8

−6

−4

−20246

days

0100

sum

ρh∂S

/∂t

evap

prec

ipu

adv

v ad

ven

tr

Figure 8

29